diff --git "a/Introducing-Bluetooth-LE-Audio-book.txt" "b/Introducing-Bluetooth-LE-Audio-book.txt"
new file mode 100644--- /dev/null
+++ "b/Introducing-Bluetooth-LE-Audio-book.txt"
@@ -0,0 +1,15085 @@
+Introducing
+Bluetooth® LE Audio
+A guide to the latest Bluetooth specifications and how they will
+change the way we design and use audio and telephony products.
+
+by Nick Hunn
+
+First published January 2022
+
+Paperback ISBN: 979-8-72-723725-0
+Hardback ISBN: 979-8-78-846477-0
+
+Copyright © 2022 Nick Hunn
+All rights reserved.
+The Bluetooth® word mark and logos are registered trademarks owned by Bluetooth SIG, Inc.
+and any use of such marks is with permission. Other trademarks and trade names are those
+of their respective owners.
+
+www bleaudio.com | www.nickhunn.com
+
+2
+
+Acknowledgements
+
+I would like to thank everyone who has helped towards the existence of this
+book. Without the work of many brilliant people contributing to the
+specifications, there would be nothing to write about. They are too numerous
+to list here, but they are recorded in each of the Bluetooth® documents. Trying
+to develop standards of this complexity results in many days and evenings of
+wide-ranging discussions, which have been both challenging and enjoyable. It
+has been a pleasure to work with so many people of passion who are always
+ready to spend time explaining the intricacies of wireless, audio and its
+application.
+For taking the time to read through the drafts and coming back with detailed
+comments, I would particularly like to thank Richard Einhorn, Kanji Kerai,
+Ken Kolderup, Mahendra Tailor, Jonathan Tanner and Martin Woolley. Their
+input, from a variety of different reader angles, has helped make this into a
+much better book than it would otherwise have been. I must also thank the
+Bluetooth SIG for their help and support.
+For giving me permission to use the image of Figure 1.8, and also for playing a
+pivotal role in kickstarting the whole world of hearables, many thanks to
+Nikolaj Hviid of Bragi.
+Finally, to my wife, Chris, for reading the endless drafts, as well as the final
+book from cover to cover multiple times, questioning everything, as well as
+putting up with the many hours I spent writing it.
+
+3
+
+4
+
+Contents
+Chapter 1.
+
+The background and heritage ................................................................................ 11
+
+1.1
+
+The hearing aid legacy...................................................................................................... 19
+
+1.2
+
+Limitations and proprietary extensions ......................................................................... 20
+
+1.3
+
+What’s in a hearable? ........................................................................................................ 24
+
+Chapter 2.
+
+The Bluetooth® LE Audio architecture ............................................................... 29
+
+2.1
+
+The use cases ..................................................................................................................... 29
+
+2.2
+
+The Bluetooth LE Audio architecture .......................................................................... 38
+
+2.3
+
+Talking about Bluetooth LE Audio ............................................................................... 48
+
+Chapter 3.
+
+New concepts in Bluetooth® LE Audio .............................................................. 51
+
+3.1
+
+Multi-profile by design..................................................................................................... 51
+
+3.2
+
+The Audio Sink led journey ............................................................................................ 51
+
+3.3
+
+Terminology ...................................................................................................................... 52
+
+3.4
+
+Context Types ................................................................................................................... 56
+
+3.5
+
+Availability ......................................................................................................................... 60
+
+3.6
+
+Audio Location ................................................................................................................. 60
+
+3.7
+
+Channel Allocation (multiplexing) ................................................................................. 61
+
+3.8
+
+Call Control ID - CCID .................................................................................................. 63
+
+3.9
+
+Coordinated Sets............................................................................................................... 64
+
+3.10
+
+Presentation Delay and serialisation of audio data...................................................... 65
+
+3.11
+
+Announcements ................................................................................................................ 71
+
+3.12
+
+Remote controls (Commanders) .................................................................................... 73
+
+Chapter 4.
+
+Isochronous Streams .............................................................................................. 75
+
+4.1
+
+Bluetooth LE Audio topologies ..................................................................................... 75
+
+4.2
+
+Isochronous Streams and Roles ..................................................................................... 77
+
+4.3
+
+Connected Isochronous Streams ................................................................................... 80
+
+4.4
+
+Broadcast Isochronous Streams ..................................................................................... 98
+
+4.5
+
+ISOAL – The Isochronous Adaptation Layer ........................................................... 122
+
+Chapter 5.
+
+LC3, latency and QoS ........................................................................................... 125
+
+5.1
+
+Introduction .................................................................................................................... 125
+
+5.2
+
+Codecs and latency ......................................................................................................... 126
+5
+
+5.3
+
+Classic Bluetooth codecs – their strengths and limitations...................................... 128
+
+5.4
+
+The LC3 codec ................................................................................................................ 131
+
+5.5
+
+LC3 latency ...................................................................................................................... 137
+
+5.6
+
+Quality of Service (QoS)................................................................................................ 138
+
+5.7
+
+Audio quality ................................................................................................................... 145
+
+5.8
+
+Multi-channel LC3 audio ............................................................................................... 146
+
+5.9
+
+Additional codecs ........................................................................................................... 151
+
+Chapter 6.
+6.1
+
+CSIPS – the Coordinated Set Identification Profile and Service ............................ 153
+
+6.2
+
+CAP – the Common Audio Profile ............................................................................. 156
+
+Chapter 7.
+
+Setting up Unicast Audio Streams ...................................................................... 163
+
+7.1
+
+PACS – the Published Audio Capabilities Service .................................................... 163
+
+7.2
+
+ASCS – the Audio Stream Control Service ................................................................ 174
+
+7.3
+
+BAP – the Basic Audio Profile ..................................................................................... 179
+
+7.4
+
+Configuring an ASE and a CIG ................................................................................... 182
+
+7.5
+
+Handling missing Acceptors ......................................................................................... 198
+
+7.6
+
+Preconfiguring CISes ..................................................................................................... 198
+
+7.7
+
+Who’s in charge? ............................................................................................................. 199
+
+Chapter 8.
+
+Setting up and using Broadcast Audio Streams ................................................ 201
+
+8.1
+
+Setting up a Broadcast Source ...................................................................................... 202
+
+8.2
+
+Starting a broadcast Audio Stream............................................................................... 203
+
+8.3
+
+Receiving broadcast Audio Streams ............................................................................ 211
+
+8.4
+
+The broadcast reception user experience ................................................................... 213
+
+8.5
+
+BASS – the Broadcast Audio Scan Service................................................................. 213
+
+8.6
+
+Commanders ................................................................................................................... 214
+
+8.7
+
+Broadcast_Codes ............................................................................................................ 222
+
+8.8
+
+Receiving Broadcast Audio Streams (with a Commander) ...................................... 223
+
+8.9
+
+Handovers between Broadcast and Unicast ............................................................... 228
+
+8.10
+
+Presentation Delay – setting values for broadcast..................................................... 229
+
+Chapter 9.
+
+6
+
+CAP and CSIPS ..................................................................................................... 153
+
+Telephony and Media Control ............................................................................ 231
+
+9.1
+
+Terminology and Generic TBS and MCS features.................................................... 232
+
+9.2
+
+Control topologies .......................................................................................................... 234
+
+9.3
+
+TBS and CCP .................................................................................................................. 235
+
+9.4
+
+MCS and MCP ................................................................................................................ 242
+
+Chapter 10.
+
+Volume, Audio Input and Microphone Control .............................................. 251
+
+10.1
+
+Volume and input control ............................................................................................. 251
+
+10.2
+
+Volume Control Service ................................................................................................ 253
+
+10.3
+
+Volume Offset Control Service.................................................................................... 256
+
+10.4
+
+Audio Input Control Service ........................................................................................ 257
+
+10.5
+
+Putting the volume controls together.......................................................................... 261
+
+10.6
+
+Microphone control ....................................................................................................... 262
+
+10.7
+
+A codicil on terminology ............................................................................................... 264
+
+Chapter 11.
+
+Top level Bluetooth® LE Audio profiles ........................................................... 267
+
+11.1
+
+HAPS the Hearing Access Profile and Service .......................................................... 268
+
+11.2
+
+TMAP – The Telephony and Media Audio Profile .................................................. 271
+
+11.3
+
+Public Broadcast Profile ................................................................................................ 274
+
+Chapter 12.
+
+Bluetooth® LE Audio applications ..................................................................... 277
+
+12.1
+
+Changing the way we acquire and consume audio .................................................... 278
+
+12.2
+
+Broadcast for all .............................................................................................................. 279
+
+12.3
+
+TVs and broadcast.......................................................................................................... 285
+
+12.4
+
+Phones and broadcast .................................................................................................... 288
+
+12.5
+
+Audio Sharing.................................................................................................................. 289
+
+12.6
+
+Personal communication ............................................................................................... 290
+
+12.7
+
+Market development and notes for developers ......................................................... 292
+
+Chapter 13.
+
+Glossary and concordances ................................................................................. 295
+
+13.1
+
+Abbreviations and initialisms ........................................................................................ 295
+
+13.2
+
+Bluetooth LE Audio specifications ............................................................................. 299
+
+13.3
+
+Procedures in Bluetooth LE Audio ............................................................................. 300
+
+13.4
+
+Bluetooth LE Audio characteristics ............................................................................ 304
+
+13.5
+
+Bluetooth LE Audio terms ........................................................................................... 306
+
+7
+
+8
+
+Introduction
+Back in the spring of 2013, I remember sitting in a conference room in Trondheim with
+representatives of the hearing aid industry as they explained to the Bluetooth Board of
+Directors why they should commit time and effort to develop a Bluetooth ® Low Energy
+specification that would support the streaming of audio. I’d been asked by the hearing aid
+companies if I would chair the working group to develop the new specifications. Everyone
+agreed it was a good idea and the two groups – the Bluetooth Special Interest Group (SIG),
+and EHIMA – the Hearing Instrument Manufacturer’s Association – the trade body
+representing the industry, signed a Memorandum of Understanding to start work on a new
+specification to support audio over Bluetooth Low Energy.
+At the time, we all thought it would be a fairly quick development – hearing aids didn’t need
+enormously high audio quality – their main concern in terms of Bluetooth technology was to
+minimise power consumption. What none of us had realised at the time was that the
+technology and use cases that had been developed by the hearing aid industry were quite a
+long way ahead of what the consumer audio market was currently doing. Although the longestablished telecoil system of inductive loops, which allowed broadcast audio to reach multiple
+hearing aids only provided limited quality audio, the connection topologies they supported
+were more complex than those provided by the existing Bluetooth A2DP and HFP audio
+profiles. In addition, the power management techniques and optimisations used in hearing
+aids gave battery lives that were an order of magnitude greater than those in similarly sized
+consumer products.
+Over the next twelve months, as we developed functional requirements documents, more and
+more of the traditional audio and silicon companies came to look at what we were doing, and
+decided that many of the features that we were proposing for hearing aids were equally
+applicable to their markets. In fact, they appeared to solve many of the limitations that existed
+in the current Bluetooth audio specifications. As a result, the requirements list grew and the
+small hearing aid project evolved into the largest single specification development that the
+Bluetooth SIG has ever done, culminating in what is now collectively known as Bluetooth LE
+Audio.
+It’s hard to believe that the journey has taken eight years. At the end of it we have produced
+two major revisions to the Core specification, introduced a completely new, high efficiency
+codec and released twenty-three profile and service specifications, along with accompanying
+documentation and Assigned Numbers documents, which between them contain around
+1,250 new pages.
+For anyone not involved with that eight-year journey, it’s a pretty formidable set of documents
+to start reading. The purpose of this book is to try and put those specifications into context,
+adding some of the history and rationale behind them, to help readers understand how the
+different parts interconnect. I’ve also provided some background information on the market
+9
+
+and what’s in a hearable, to help readers relate the specification to actual products. In the
+chapters which delve into detail, I’ve included references to the specific part of the
+specifications using their abbreviated name and section number, e.g. [BAP 3.5.1] for Section
+3.5.1 of the Basic Audio Profile. I’ve tried to limit the number of references, so that they don’t
+get in the way of the text. The glossaries and concordances in Chapter 13 should also help
+developers navigate their way around the documents.
+I wanted to get this information out as quickly as possible, so for the first time I’ve resorted
+to self-publishing, avoiding the lengthy delays I’ve experienced with publishing houses in the
+past. I have to thank Amazon for the ability to do that so easily. If you find this book helpful,
+I’d really appreciate it if you could write a review and tell your friends. If you think there are
+omissions or something is unclear, please drop me an email at nick@wifore.com. The
+advantage of self-publishing is that I can update the book far more readily than with a normal
+book. I’ll also try and answer comments and publish corrections at the book’s website at
+www.bleaudio.com.
+All of us involved in the specification development think that Bluetooth LE Audio gives us
+the tools to develop exciting new audio products and applications. I hope that this book helps
+to explain those new concepts and inspires you to develop new ideas. If so, the eight years
+that so many of us have spent working on it will have been time well spent.
+The bulk of this book will explain how these new specifications work, how they fit together
+and what you can do with them, but we’ll also take a glimpse at the future in the final chapter.
+Before jumping into the specifications, it’s useful to understand where we are today and what
+goes into a hearable device, to help understand how everything fits together. That’s the
+purpose of Chapter 1. If you want to get straight to the detail, skip to Chapter 2.
+A few of the Bluetooth LE Audio specifications are not yet adopted, so I’ve relied on prepublication drafts which the Bluetooth SIG has made public. The versions used for this
+edition are listed in Chapter 13.
+
+10
+
+Chapter 1 - The background and heritage
+
+Chapter 1. The background and heritage
+Since it was first announced in 1998, Bluetooth® technology has, arguably, grown to be the
+most successful two-way wireless standard in history. In the wireless standards business,
+success is normally counted as the number of chips which are sold each year. On that basis,
+Bluetooth is the winner, with around 4.5 billion chip shipments in 2020. Wi-Fi is close behind,
+with 4.2 billion, followed by 1.84 billion for all variants of GSM and 3GPP phones and a mere
+145 million for DECT.
+However, for much of its history, only a small number of those chips were actually used.
+When Bluetooth technology was first proposed, its developers identified four main use cases.
+Three of them were audio applications, focussing on simple telephony functions:
+•
+•
+•
+
+a straightforward wireless headset that was just an extension of your phone, defined
+in the Headset profile
+an intercom specification for use around the house and in business, and
+a new technology for cordless telephony, hoping to replace the proprietary analogue
+standards used in the US and the emerging DECT standard within Europe. Its
+intent was to combine the functions of cordless and cellular in a single phone.
+
+The fourth use case was called dial-up networking or DUN, which provided a means to
+connect your laptop to your GSM phone, using the phone as a modem to give you internet
+access wherever you were in the world. As is often the case with new technology standards,
+none of those four use cases really took off, despite some initial enthusiasm from PC and
+phone companies. Cordless telephony and intercom failed because they potentially took
+revenue away from mobile phone operators. Dial up networking worked, but at that point
+mobile phone tariffs for data were expensive, which encouraged people to use the new Wi-Fi
+standard instead. Headsets started to sell, but unless you were a taxi driver, you weren’t likely
+to buy one. It became clear that these particular use cases probably weren't the ones that were
+going to generate scale in the market, so the Bluetooth SIG started work on a host of other
+features, such as printing and object transfer, none of which attracted much more interest
+from consumers.
+What happened next is what all standards bodies hope for - Government regulations appeared
+which gave Bluetooth technology a better reason to exist.
+At the end of the 1990s, global mobile phone usage exploded as the falling price of both
+phones and phone contracts changed them from a business tool to a consumer essential.
+Mobile phone operators started to become High Street names, growing to become substantial
+businesses.
+
+11
+
+Section 1.1 - The hearing aid legacy
+
+Figure 1.1 Growth of global mobile phone subscriptions
+
+As phone usage increased, so did a concern about where they were used, as more and more
+road accidents were reported where drivers had been holding their phones and become
+distracted. Legislators around the world started to propose bans on the use of mobile phones
+whilst driving. For both the phone industry and the mobile operators this was a potential
+disaster. In the US, it was reported that almost a third of mobile subscription revenue (which
+at that time was based on the number and length of phone calls) came from calls made whilst
+driving. It was a golden egg that the industry could not afford to lose. To save that income,
+they proposed a compromise to the legislators, which is that safety could be restored if the
+driver didn’t need to hold their phone; instead, the phone call could be taken using a HandsFree solution, either built into the car itself, or by using a Bluetooth wireless headset.
+That was the spur that Bluetooth technology needed. With the new safety legislation coming
+into effect, phone manufacturers started putting Bluetooth into more and more of their phone
+models, rather than just the top end ones. The automotive industry began integrating
+Bluetooth technology into cars and worked with the Bluetooth SIG to develop the HandsFree profile for that use case. It was a turning point for Bluetooth technology. In 2003, only
+around 10% of mobile phones were sold which contained a Bluetooth chip – almost all topend phones. The following year it doubled. The number was to grow every year, as shown in
+Figure 1.2. By 2008, two thirds of all new mobile phones contained a Bluetooth chip.
+
+12
+
+Chapter 1 - The background and heritage
+
+Figure 1.2 Percentage of mobile phones containing a Bluetooth® technology chip
+
+Very few of those chips were actually used for the purposes that were intended. That’s
+obvious, as only around 14 million headsets were sold in the same year, and that number didn’t
+grow significantly in the following years. But it was the start of a “free ride” that saw 250
+million Bluetooth chips ship in 2005. Those volumes brought competition and manufacturing
+efficiencies, pushing the price down and starting a virtuous circle where the incremental cost
+of adding Bluetooth technology was negligible, at least in terms of hardware. The challenge
+was to find an application which was compelling enough to encourage more people to use it.
+Back in 1998, the year Bluetooth technology was announced, the music streaming services we
+know today, like Spotify and Apple Music, were still ten years away. Ironically, subscriptionbased music streaming wasn’t an unknown concept – it was over a century old. As far back
+as 1881, some telephone networks ran a service allowing you to listen remotely to live opera.
+But that concept had been trumped by a better organised industry model. The arrival of
+physical media in the form of wax discs and records, which you could buy and listen to
+whenever you wanted, killed that original streaming concept. Having discovered that they
+could own the artists, the recording industry spent the next hundred years doing everything
+they could to tighten copyright laws around the world to protect their stranglehold on how we
+listened to music. It was going to prove to be a long dominance. In 1998 we still bought our
+music in the form of physical recordings. LPs had been almost totally replaced by CDs, but
+around 20% of all the recorded music that we bought that year still came on cassette. Then
+things changed, thanks largely to the separate efforts of the Fraunhofer Institute for Integrated
+Circuits (IIS) in Germany and a small Californian startup called Napster.
+The Fraunhofer IIS is part of a wider research organisation and a centre of excellence in
+developing audio codecs – software that compresses audio files so that they are small enough
+to be wirelessly transmitted or stored as digital files. In 1993, they developed a new audio
+codec for the international MPEG-1 video compression standard. It was particularly efficient
+13
+
+Section 1.1 - The hearing aid legacy
+compared to other audio coding schemes, such as the one used for CDs, and quickly became
+the standard for audio files transmitted over the internet. It became known as MP3. Without
+it, we would have had to wait a lot longer for downloadable music.
+Instead, the early 2000s became the era of MP3 players and free, downloadable music. Napster
+burst onto the scene in 1999 and created the first real disruption that the recorded music
+industry had faced in its hundred year existence. They immediately took Napster to court for
+copyright infringement, as well as trying to prosecute individual users for downloading music.
+Consumers voted with their fingers, repeatedly clicking download buttons to get hold of more
+music. For the first time, music that you could listen to, whenever you wanted, was free. By
+2000, just a year after it burst onto the market, most of its users probably had more
+downloaded songs on their PCs than they had physical albums. However, the recording
+companies turned out to have the better lawyers and Napster lost – the initial attempt to make
+music free had failed. While the battle to kill off Napster had been dragging through the
+courts, the next stage of competition had already arrived in the form of Apple’s iTunes. It
+wasn’t free, but it was easy to use. Subscribers flocked to the new offering. Six months later,
+it became even easier, with the launch of the first iPod.
+Most of the engineers working on Bluetooth technology were likely to have been Napster and
+iTunes subscribers, not least to have something to listen to on long-haul flights to standards
+meetings. By the time the courts had decided Napster’s fate, work was already underway to
+add music streaming to Bluetooth in the form of the Advanced Audio Distribution Profile,
+better known as A2DP. Despite its far-from-understandable moniker, it was to become the
+most successful of all Bluetooth profile specifications and cement Bluetooth technology’s
+place as the standard for wireless audio.
+A2DP, along with its supporting specifications, was adopted in 2006. Companies like Nokia,
+who had been heavily involved in its development, expected it to be an instant success, but it
+proved to be slow in taking off. Consumers didn’t see the advantage in buying an expensive
+wireless set of headphones, with most using the free, corded earbuds which came with every
+handset, despite their often limited audio quality. Much to the industry’s surprise, the initial
+growth came not from headphones, but from Bluetooth speakers. That initial mobile music
+market was one where you took your music with you, but didn’t necessarily play it while you
+were mobile. Speakers grew into soundbars as TVs began to include A2DP, but the
+headphone market remained remarkably resistant to growth until a new service appeared –
+Spotify.
+Spotify (and its US precursor – Pandora) introduced a new business model. You could once
+again listen to music for free, as long as you accepted advertisements. If you didn’t want the
+interruptions, that was fine – you could get rid of them by paying a subscription. It neatly
+solved the copyright problem by generating revenue to pay licence fees, regardless of whether
+users took the subscription or the ad-supported route. It effectively replicated what Napster
+had done, but found a way to do it legally. It helped that Spotify’s launch coincided with the
+14
+
+Chapter 1 - The background and heritage
+first iPhone, where consumers began to realise that smartphones wouldn’t be used
+predominantly for phone calls. Instead, they’d spend most of their lives doing other stuff,
+especially as their appearance coincided with the advent of affordable mobile data plans which
+offered unlimited data usage. Mobile operators were quick to bundle Spotify with their phone
+contracts and users signed up. Most importantly, it was easy to use. iTunes was still a service
+where you bought a download and had to make a decision to play it. With Spotify you pressed
+a button and music appeared. It could be your own playlist, or that of a music blogger, a
+favourite DJ or an algorithm. Its great attraction was that it was frictionless – you pressed a
+button and music came out of your earbuds until you stopped it. You no longer needed to
+hold your phone; you could keep in it your pocket or bag, which made it ideal for listening on
+the move.
+At the point that you don’t need to hold your phone, an earbud cable becomes a nuisance. It
+was the nudge users needed to persuade them to start buying Bluetooth headphones. Branded
+Bluetooth headphones sales started to rise, as manufacturers saw customers cut the cable. By
+2016, Bluetooth headphones were outselling wired ones.
+
+Figure 1.3 The transition from wired to Bluetooth® headphones
+
+Figure 1.4 shows how much the Spotify experience helped drive that change. Over the ten
+years since 2006, when Spotify and the A2DP specification independently appeared on the
+market, the growth in Bluetooth headphones has almost exactly tracked that of Spotify
+subscribers.
+
+15
+
+Section 1.1 - The hearing aid legacy
+
+Figure 1.4 The growth of Spotify subscriptions and Bluetooth® headphones
+
+The growth in chip sales for wireless headphones drove innovation. By 2010, Bluetooth
+silicon companies were shipping over 1.5 billion chips per year and were looking for more
+ways to differentiate their products. The new Bluetooth Low Energy standard had just
+launched, but it wouldn’t be until the end of 2011 that it appeared in the iPhone 4s, and several
+more years before a new generation of wristbands started to use it. In the interim, Cambridge
+Silicon Radio (CSR, who were subsequently acquired by Qualcomm), had become the leading
+player in chips for Bluetooth audio devices, and were looking at how they might extend the
+functionality of the Bluetooth audio profiles.
+Why would they want to do that? The problem they saw was that both of the two main
+Bluetooth audio standards were written for very specific use cases, which hadn’t anticipated
+the future. The result of that is that they behave very differently. HFP concentrates on low
+latency, bidirectional, mono voice transmission, whereas A2DP supports high quality music
+streaming to a single device with no return audio path. Neither of those profiles were easily
+extensible, which limited what you could do with them. CSR were trying to push the
+boundaries of wireless audio. They had already been successful in developing a more efficient
+codec than the SBC codec mandated by the Bluetooth specifications, adding their own
+enhanced AptX codec into their chips. Now they decided to push further ahead and see if
+they could discover a way to allow the A2DP specification to be used to stream stereo to two
+independent earbuds.
+Their efforts succeeded and started a new era for Bluetooth audio, which would lead to
+massive growth. The consumer uptake of wireless earbuds meant that Bluetooth technology
+decoupled itself from Spotify’s growth and gained a new momentum of its own. CSR’s
+innovation was to develop a way for a single earbud to receive an A2DP stream and forward
+one channel of the stereo stream, together with timing information, to a second earbud. Using
+the timing information, the first earbud could delay rendering its audio channel until it knew
+the second earbud had received its audio data and was ready to render its stream. As far as
+16
+
+Chapter 1 - The background and heritage
+the user was concerned, it appeared as if each earbud was receiving its own left or right
+channel. It was not quite as simple as that to make it work, as we’ll see later, but it would
+become a game changer.
+The first movers were not the big companies. At the start of 2014, two companies in Europe
+– Earin in Sweden and Bragi in Germany kicked it all off with crowdfunding campaigns for
+stereo wireless earbuds. Earin’s offering was a pair of wireless earbuds that used the new
+chipsets to push the boundaries of what could be done. They’re still making iconic designs,
+launching their third generation at the start of 2021. However, it was Bragi’s Dash that really
+caught the imagination. In the spring of 2014, they broke the Kickstarter record for the highest
+funded campaign to date, raising over $3.3 million for their Dash wireless earbuds. It was an
+amazing piece of engineering, which promised to cram almost every feature you could think
+of into a pair of earbuds. It raised the bar of what you can do with something in your ear,
+opening up a whole new market sector for which I coined the name “hearables”. It was a
+massively ambitious concept, and to their credit, they managed to ship the Dash. Despite an
+enthusiastic following, it showed that it’s not enough to just integrate the technology – you
+need to find a use for it. Despite providing an SDK for software developers, the Dash failed
+to gain enough traction with consumers. In the following two years, other crowdfunded
+projects dreamt up even more complexity and managed to attract around $50 million dollars
+of funding. Many failed to deliver, coming to realise just how hard it is to cram that amount
+of technology into a small earbud. Hardware is hard. Most of these startups fell by the
+wayside, and even Bragi was forced to make the difficult decision to move out of hardware,
+selling its product range and concentrating on embedded operating systems for other
+mainstream audio products.
+Gradually, bigger names started to dip their toes into the hearables pond, utilizing their deeper
+resources to make these difficult products. Then, in September 2016, Apple launched its
+AirPods. Although initially derided by many journalists, consumers loved them. In just two
+years they became the fastest selling consumer product ever, and have sold over 250 million
+pairs since they were launched. Other brands rapidly followed on, with China’s chip vendors
+jumping on the bandwagon. Back in 2014, when Bragi and Earin were setting out the future
+of the market, there were only four or five silicon vendors making Bluetooth audio chipsets.
+Today there are over thirty. Despite supply chain problems, it is estimated that around 500
+million earbuds were shipped in 2020, with a prediction that the number will double by the
+end of 2022. Those figures are for earbuds based on the classic HFP and A2DP profiles. As
+new Bluetooth LE Audio products start to ship, with their additional functionality, broadcast
+features and enhanced battery life, the numbers are likely to exceed those predictions.
+However, wireless audio is not just about earbuds. Most of the growth in Bluetooth audio has
+been driven by music streaming, which in turn has been driven by the ready availability of
+content from services like Spotify, Apple Music and Amazon Prime music. The arrival of
+streaming video has been equally popular, with wireless earbuds becoming the device of choice
+for listening. Consumers like ease of use, and in most cases that translates into consuming
+17
+
+Section 1.1 - The hearing aid legacy
+content, not generating it themselves, although applications like TikTok are showing that if
+content generation can be made simple and amusing, users will produce and share their own.
+In this evolution, voice, mainly the preserve of voice calls, had looked as if it was becoming
+the poor relation. That changed when Amazon launched Alexa. Despite reservations,
+consumers started talking to the internet. Since then, voice recognition has seen a renaissance,
+to the point where most home appliances now want to talk to us, and for us to talk to them.
+Those applications are likely to grow with the introduction of the new features supported by
+Bluetooth LE Audio, which are covered in this book. With broadcast topologies and the
+ability to prioritise which devices can talk to you, it becomes possible to add extra functionality
+and new opportunities to voice to machine communication. It may be the saviour of the
+Smart Home industry.
+The speed of development surrounding the introduction of earbuds, and Apple’s Airpods in
+particular, has been amazing. We have seen many new companies developing Bluetooth audio
+chips, advances in miniature audio transducers and MEMS1 microphones, along with a
+massive growth in the number of companies providing advanced audio algorithms to enable
+features like active noise cancellation, echo cancellation, spatial sound and frequency
+balancing.
+Although the Hands-Free Profile and A2DP continue to serve us well, both were designed for
+a simple peer-to-peer topology. Referred to as Bluetooth Classic Audio, their design assumes
+a single connection between two devices, where each individual audio data packet is
+acknowledged. That doesn’t work if there are two separate devices that want to receive the
+audio stream. Because of that, today’s stereo earbuds depend on proprietary solutions which
+generally involve the addition of a second radio to communicate between the left and right
+earbuds to ensure that both of them play the audio at the right time. Another limitation is that
+the Hands-Free Profile was not designed for the wide range of cellular and Voice over IP
+telephony applications we use today. Having just these two, independent, dedicated audio
+profiles has led to multi-profile problems when users want to swap from one application to
+another. That’s before you add in new requirements which have emerged for concurrent voice
+control and shared audio.
+We’re all using wireless audio more frequently throughout our daily lives, whether that’s to
+talk to each other, or insulate ourselves from the outside world. To support this change in
+behaviour we need to think less about connections between individual devices that are
+normally used together, like a headset and the phone. Instead, we need to be far more aware
+of a wider audio ecosystem, where we have different devices that we might wear on our ears
+during the course of a day, constantly listening to and changing what they do with other
+
+MEMS stands for Microelectromechanical Systems, which in this case refers to microphones which
+have been made by etching physical structures into a silicon wafer. It’s a very efficient way of making
+small, highly accurate sensors.
+1
+
+18
+
+Chapter 1 - The background and heritage
+devices. For that to work seamlessly, control needs to become far more flexible. This is the
+background that led to the development of Bluetooth LE Audio.
+The Bluetooth SIG realised that their audio specifications needed to evolve and adapt, both
+to cope with current requirements and also to look to the increasingly diverse range of things
+we're doing with audio, as well as what we can expect to appear over the next 20 years. There
+are some very similar requirements that come up in many of the use cases. Designers want
+lower power. Not just for extended battery life, but to be able to support more processing for
+noise reduction and the other interesting audio algorithms that are emerging, such as detecting
+oncoming traffic or relevant conversation. For other applications, they want to reduce latency,
+particularly for gaming or listening to live conversations or broadcasts. There's also the neverending quest for higher audio quality. The Bluetooth LE Audio working groups developing
+the specifications have had the task of coming up with new standards that support these
+requirements, as well as all of the topologies that are envisioned, without having to rely on
+non-interoperable extensions. The aim is to allow the industry to move on from the position
+it’s in today, which relies on proprietary implementations, to one where you can mix and match
+devices from different manufacturers.
+
+1.1
+
+The hearing aid legacy
+
+It surprises many people to hear that a lot of this innovation was kick-started by the hearing
+aid industry. Hearing aids have needed to solve the issues of audio quality, latency, battery life
+and broadcast transmissions for many, many years. They are worn for an average of nine
+hours a day, so battery life is critical. During that time hearing aids are constantly amplifying
+and processing ambient sound so that the wearer can hear what is happening and being said
+around them. They typically include multiple microphones to allow audio processing
+algorithms to recognise and react to the local audio environment in order to filter out
+distracting sound. In public spaces, where the facility is available, they can connect to a system
+called telecoil, essentially induction loops, which are used in theatres, public transport and
+other public areas to hear audio and provide information. These are broadcast systems which
+can cope with hundreds of people within the transmission area of the telecoil, or allow private
+conversations using very small loops.
+Hearing aid users have always wanted to be able to connect to phones and other Bluetooth
+devices, but the power consumption of traditional HFP and A2DP solutions was a challenge.
+In 2013, Apple launched a proprietary solution based on the Bluetooth Low Energy
+specification, adding an audio stream which could connect to special Bluetooth LE chips in
+hearing aids. It was licensed to hearing aid manufacturers, and appreciated by consumers, but
+it only worked with iPhones and was unidirectional.
+Although welcoming the development, the hearing aid industry was concerned that an Applespecific solution was not inclusive. They wanted a global standard which would work with
+any phone or TV, and which could also replace the ageing telecoil specification, which dated
+back to the 1950s. In 2013, representatives of all of the major hearing aid companies sat down
+19
+
+Section 1.2 - Limitations and proprietary extensions
+with Bluetooth SIG’s Board, and came up with a joint agreement to provide resources to help
+develop a new low power Bluetooth standard for audio to bring interoperability to the hearing
+aid ecosystem. Fairly soon after the development work began, many consumer audio
+companies started looking at the hearing aid use cases and realised that they were equally
+applicable to the consumer market. Although the audio quality requirements for hearing aids
+were less stringent (as their users have hearing loss), the use cases, which combined ambient
+audio, Bluetooth audio and broadcast infrastructure, were far more advanced than the ones
+currently covered by HFP and A2DP. They had the potential to solve many of the known
+problems with the current audio specifications.
+As more and more companies got involved, the project expanded. Over the eight years that
+the work has taken, the Bluetooth LE Audio initiative has evolved into the largest specification
+development project that the Bluetooth SIG has ever done. The resulting specifications cover
+every layer of the Bluetooth standard, and consist of over 1,250 pages of text in new and
+updated documents, most of which have now been adopted, or are in the process of being
+adopted.
+
+1.2
+1.2.1
+
+Limitations and proprietary extensions
+Apple’s Made for iPhone (Mfi) for hearing devices and ASHA
+
+In 2014, Apple launched its own proprietary Bluetooth Low Energy solution for hearing aids,
+which it licensed to hearing aid manufacturers. Developed in conjunction with one of its
+silicon partners, it added extensions to the Bluetooth LE protocols to allow unidirectional
+transmission of data between a phone and one or two hearing aids. An app on the phone
+allowed the user to select which hearing aid to connect to, as well as allowing them to set
+volume (either independently, or as a pair) and to select a variety of presets, which apply preconfigured settings on the hearing aids to cope with different acoustic environments. The Mfi
+hearing devices solution worked on iPhone 5 phones and iPad (4th generation) devices and
+subsequent products.
+One of the popular features that Mfi supports is “Live Listen”, which allows the iPhone or
+iPad to be used as a remote microphone. For hearing aid wearers, this lets them place their
+phone on a table to pick up and stream a conversation. Remote microphones are useful
+accessories for hearing aids and the Live Listen feature provides this without the need to buy
+an additional device.
+Early in 2021, Apple announced that their Mfi for hearing devices would be upgraded to allow
+bidirectional audio, bringing Hands-Free capability.
+Apple’s motives weren’t entirely altruistic. Accessibility regulations in many countries forced
+them to include a telecoil in their phones, which adds cost and constrains the physical design.
+They hoped that regulators would accept a Bluetooth solution as an alternative. Nokia had a
+similar desire and was active in supporting the development of the Bluetooth LE Audio
+20
+
+Chapter 1 - The background and heritage
+specification before they withdrew from making handsets.
+Apple’s Mfi for hearing solution only works with iPhone and iPads, leaving Android owners
+with no solution. In parallel with the Bluetooth LE Audio development, Google and hearing
+aid manufacturer GN Resound worked together to develop an open Bluetooth LE
+specification called ASHA (Audio Streaming for Hearing Aids) which is a software solution
+which works on Android 10 and above. It provides a different proprietary extension to
+Bluetooth LE to support unidirectional streaming from any compliant Android device to an
+ASHA hearing aid.
+ASHA has provided a welcome fill-in for Android users in the gap before Bluetooth LE Audio
+starts to appear in phones. It’s likely that hearing aid manufacturers who currently support
+Mfi for hearing aids or ASHA will continue to do so. They will extend their support by adding
+Bluetooth LE Audio to provide the widest choice for consumers. At the end of the day, it is
+just another protocol, which means more firmware. However, it is likely that most new
+development will move towards the globally interoperable Bluetooth LE Audio standard.
+
+1.2.2
+
+True Wireless
+
+A practical solution to send a stereo stream to two earbuds was developed by Cambridge
+Silicon Radio around 2013. After they were acquired by Qualcomm it was rebranded as
+TrueWireless, which is the phrase that most of the world now uses for stereo earbuds. As
+competitors looked at the success of Apple’s Airpods, which use Apple’s own, proprietary
+chipset, Qualcomm provided a viable alternative for every other company that wanted to
+develop a competing earbud. The name True Wireless Stereo or TWS quickly became
+established and applied to almost every new product, regardless of whose chip was in it.
+The main obstacle to using A2DP with separate earbuds is that it was designed for single
+point-to-point communications. The Bluetooth SIG did provide guidance for how streams
+could be sent to multiple Bluetooth LE Audio sinks in a white paper titled “White paper on
+usage of multiple headphones” back in 2008, but it avoided the question of synchronisation,
+assuming that each audio sink could make up its own mind about when to render the stream.
+For earbuds, as soon as the left and right streams move out of sync, you get a very unpleasant
+sensation of sound moving around inside your head. The white paper approach also relied on
+modifications to the A2DP specification at the audio source. That’s a problem, as it means
+that earbuds or speakers that didn’t adopt them wouldn’t be compatible. To be successful in
+the market there needed to be a solution which would work with any audio source, which
+effectively meant that the solutions needed to be implemented solely in the audio sinks – the
+earbuds.
+
+21
+
+Section 1.2 - Limitations and proprietary extensions
+The solution that CSR came up with is known as the replay or forwarding approach, as shown
+in Figure 1.5.
+
+Figure 1.5 Replay scheme for TWS
+
+Before they connect to the phone, the two earbuds pair with each other (which may be done
+at manufacture). One is set to be the primary device, the other as a secondary. To receive an
+A2DP stream, the primary device pairs with the audio source, appearing as a single device
+receiving a stereo stream. When audio packets arrive, it decodes the left and right channels
+and, in the example shown above, relays the left channel directly to the secondary earbud.
+This may be possible using Bluetooth technology, but if the earbuds have small antennas, this
+may be problematic, as the head absorbs the 2.4GHz signal very well. Many companies
+discovered this when they first tested their devices outside. Indoors, reflections from walls
+and ceilings will generally ensure that the Bluetooth signal gets through. Outside, with no
+reflecting surfaces to help, they may not. To compensate for this, a second radio is typically
+added which gets around the absorption problem. The most popular option is Near Field
+Magnetic Induction (NFMI), which is an efficient low power solution for short range audio
+transfer. Other chip vendors have used similar Low Band Retransmission (LBRT) schemes.
+As the primary device is in charge of the timing for the audio relay, it knows exactly when the
+secondary earbud will render it. It needs to delay the rendering of its audio stream to match.
+That is one of the issues with the relay approach, as the relay process and buffering adds to
+the latency of the audio connection. For most applications it is unlikely to be noticed by the
+user, not least because A2DP latency will normally dominate.
+The relay approach also works for HFP, although in most cases, only the primary device is
+used for the voice return path.
+22
+
+Chapter 1 - The background and heritage
+Volume and content control, such as answering a call or pausing music still uses the
+Audio/Video Remote Control Profile (AVRCP) over the connection to the primary earbud.
+The second radio, when present, can also be used to relay user interface and AVRCP controls,
+such as pause, play and volume, between the primary and secondary device.
+More recently, companies have started moving to a sniffing approach, illustrated in Figure 1.6
+
+Figure 1.6 The sniffing approach for TWS
+
+Here, the added latency of the relay approach is removed by both earbuds listening to the
+primary A2DP stream. Only one of the earbuds (the primary device) acknowledges receipt of
+the audio data to the phone. Normally, the secondary device wouldn’t know where to find
+the audio stream, or be able to decode the audio data. In this case, the primary device provides
+it with that information, either through a Bluetooth link or a proprietary sub-GHz link. That
+link is configured by the manufacturer, so there’s no chance of other devices being able to
+pick up the A2DP stream.
+As both earbuds receive the same A2DP stream directly, this is potentially a far more robust
+scheme, particularly if implemented as a Bluetooth only solution. Where there is no relaying
+of the audio stream, the latency is appreciably better.
+Both of these schemes require clever extensions at a fairly low level of the Bluetooth stacks in
+the earbuds, so have largely been the domain of chip vendors. Apple’s success has spurred
+competing silicon providers to innovate, with the result that around a dozen different True
+Wireless schemes are now in existence, built on variants of these two techniques, with some
+including additional features like primary swapping, so that the two earbuds can change roles
+if one loses the link.
+23
+
+Section 1.3 - What’s in a hearable?
+The problem with these proprietary approaches is that they can fragment the market. Apple,
+Qualcomm and others have all filed patents for their solutions, which results in their
+competitors spending time and effort looking at how to get around these patents rather than
+innovating to drive the market forward. It also means that it is almost impossible to mix
+earbuds from two different manufacturers, as they’re likely to use different TWS schemes.
+That may not be an issue for consumer TWS earbuds, where they are always bought as a paired
+set, but it is for hearing aids, where different types may be needed for left and right ears. It
+makes it difficult to extend the scheme to speakers, where surround sound systems often
+combine speakers from different manufacturers. Although proprietary solutions can spur
+market growth in its early stages, they rarely help its long term development. Which is where
+Bluetooth LE Audio comes in.
+
+1.2.3
+
+Shared listening
+
+Sharing audio isn’t just about sending signals to a pair of earbuds and hearing aids – it’s also
+about extending the number of people who can listen to the signal. Whilst public installations
+can cater for large numbers of people listening simultaneously, there’s a more personal
+application where you want to share your music with a friend. The White Paper on multiple
+headphones (referred to above) sets out how shared listening with one other person is possible
+with A2DP, but the solution it suggests never seems to have made it into products. Instead,
+this segment of the market has largely been owned by Tempow – a Paris based Bluetooth
+software company. They have developed a phone based audio operating system which they
+call Dual-A2DP, which generates a separate left and right synchronised stream, allowing two
+separate pairs of earbuds to render the streams at the same time. Although this is useful, it is
+limited and has only been taken up by a few manufacturers. Bluetooth LE Audio goes beyond
+this by providing a scalable solution to transition from one to many listeners.
+
+1.3
+
+What’s in a hearable?
+
+The relatively recent arrival of wireless earbuds was limited by the availability of chips which
+could provide a way to support separate left and right stereo streams. But that’s not the only
+reason. As all of the original startup companies who entered this market discovered, packing
+all of the functionality that you need to reproduce decent audio into something as small as an
+earbud is hard. In fact, it’s very hard. Adding Bluetooth technology to that, along with any
+additional sensors to support it, becomes incredibly hard. Only around 25% of the
+crowdfunded hearable companies ever managed to ship a product. Even companies like
+Apple had to commit almost five years of development to bring about the AirPods. They
+even resorted to designing their own Bluetooth chips to get the performance which made
+AirPods such a success. That’s something that only the largest earbud companies – Apple and
+Huawei, have had the resources to do.
+It’s a story that hearing aid manufacturers know well, as they’ve been perfecting the
+miniaturization of hearing aids for many years. They also work with customised chips to
+optimise performance, resulting in devices which run for days on small, primary zinc-air
+24
+
+Chapter 1 - The background and heritage
+batteries. There are a surprising number of elements in a Bluetooth hearing aid, as shown in
+Figure 1.7, which represents a fairly basic design. It’s a very similar architecture to an earbud.
+
+Figure 1.7 Architecture of a simple Bluetooth® hearing aid
+
+When you take a hearing aid or earbud to pieces, the first surprise most people encounter is
+the number of microphones in them. To perform active noise cancellation, you need one
+microphone monitoring the ambient sound and a second one in the ear canal. If you want to
+pick up the user’s voice to send over the Bluetooth link to their phone, there will normally be
+a bone conduction microphone helping to pick up and isolate the spoken voice from the
+ambient. That’s the minimum. However, it’s common to include additional microphones to
+generate a beam-forming array to improve directionality. Other audio algorithms may be
+included to detect the type of environment and further enhance the user’s voice. The latter is
+important, as the microphones in earbuds and hearing aids aren’t located as close to the mouth
+as designers would want them to be – they may often be behind the ear.
+We’re about to see new regulations come into effect to measure the sound level in the ear
+canal and warn users about potential hearing damage, which will probably lead to even more
+internal microphones. The good news is that there has been a lot of development in MEMS
+microphones in the last few years, partly driven by the demands of voice assistants. These
+innovations are enhancing directionality and beam steering, so that they can follow you around
+the room. The advantage of MEMS over traditional microphone structures is that you can
+integrate digital signal processors into the microphone itself, which reduces size and power
+consumption. It’s not unusual to find four or more microphones in the latest hearables. These
+inputs need to be mixed and dispatched to the different functions of the hearing aid.
+If the device contains both Bluetooth technology and telecoil receivers, the appropriate audio
+needs to be routed from them. For earbuds which send control signals or audio streams to a
+second one, these all need to be routed via a sub-GHz radio (normally NFMI), whilst the
+primary signal is buffered to synchronise the rendering time at both earbuds.
+25
+
+Section 1.3 - What’s in a hearable?
+At the output, audio transducers have become smaller, with traditional open coil structures
+being challenged by micro-miniature balanced armature transducers and audio valves. The
+market is also seeing the appearance of MEMS speakers, offering high sound levels. Whilst
+they are probably more suited to headphones at this stage, the technology is likely to migrate
+to earbuds.
+Looking at the new use cases that are being made possible with Bluetooth LE Audio, this
+processing can become very complex. A noise cancelling earbud receiving a Bluetooth stream
+and allowing voice commands to be transmitted at the same time will need to separate the
+voice component from the ambient sound (and apply echo cancellation) before transmitting
+it, whilst at the same time suppressing the ambient sound that is mixed with the incoming
+Bluetooth signal. If the incoming Bluetooth stream is being broadcast from the same source
+as the ambient, such as when you’re in a theatre, conference room, or watching TV, then this
+all needs to be done whilst maintaining an overall latency of around 30 msecs. That is
+challenging.
+Managing all of this needs an application processor, which controls the specialised audio
+processing blocks. To minimise power and latency in hearing aids, these are normally
+implemented in hardware rather than in a general purpose Digital Signal Processor (DSP).
+The application processor will also normally control the user interface, where commands may
+come from buttons or capacitive sensors on the hearing aid, via the Bluetooth interface, or
+from a remote control, which may be using either a Bluetooth technology or proprietary subGHz radio link. Most devices include an optical sensor to detect when it is removed from the
+ear, so that it can be placed into a sleep state. Finally, there is a battery and power management
+function. Many hearing aids still use zinc-air batteries, which provide a better power density
+than rechargeable batteries. That provides two main advantages – the battery life is longer,
+and they weigh less. The weight is important if you’re wearing a hearing aid all day, which is
+why most modern hearing aids weigh less than 2 grams – around half the weight of an AirPod.
+That’s just the electronics. Fitting all of this into an earbud or hearing aid is a further challenge,
+pushing the limits of flexible circuitry and multi-layer packaging. Designers need to make it
+comfortable, ensure it doesn’t fall out of your ear, but also accomplish all of this whilst
+maintaining a good auditory path, so that neither the fit, nor the electronics packed into it,
+affects the quality of the sound. You also need to consider the question of whether the hearing
+aid occludes, i.e., whether it blocks your ear to stop ambient sound, or is open to allow you to
+hear both. Until recently it was felt that occlusion was necessary if you wanted to have
+effective ambient noise cancellation, but a few recent innovations suggest that may no longer
+be the case. Completely blocking the ear canal can cause issues with a build-up of humidity,
+especially for prolonged wearing, so we will probably see more designs taking the open
+direction.
+None of this is easy, which is one of the reasons that hearing aids remain expensive. However,
+a growing number of chip companies are offering reference designs which have already done
+26
+
+Chapter 1 - The background and heritage
+much of the work. Along with partner programs with specialist consultancies offering design
+expertise to reduce the time to market, this is resulting in the consumer earbud market growing
+at a phenomenal pace. But we’re still just at the beginning of the possibilities for what you
+can put in your ear.
+The most ambitious hearable which has shipped so far was Bragi’s Dash. As well as providing
+true wireless stereo, they decided to add an internal MP3 player and flash storage, allowing you
+to leave your phone at home while you’re out running or in the gym, yet still be able to listen
+to your stored music directly from the earbuds.
+
+Figure 1.8 The Dash earbud from Bragi - the first real hearable
+
+Recognising that the ear is the best place on the body for most physiological sensors, Bragi
+equipped the Dash with a plethora of sensors and features: a total of nine degrees of freedom
+movement sensing with an accelerometer, magnetometer and gyroscope, a thermometer, a
+heart rate monitor and a pulse oximeter. They produced a very nice graphic showing all of
+the elements with their relative sizes, which is represented in Figure 1.8. It was a stunning
+achievement, but sadly, it was more than the market wanted at that time. As fitness wristband
+manufacturers discovered, it’s surprisingly difficult to keep customers engaged with their
+health data. Unlike music, where they just devour someone else’s content, health data needs
+a lot of analytics to turn it into a compelling story for the user.
+
+27
+
+Section 1.3 - What’s in a hearable?
+That’s a Catch 22 that is illustrated in Figure 1.9. You need to capture a lot of data before you
+have enough to turn into anything valuable. During that time, you need to employ some very
+expensive data scientists to try and develop some compelling feedback, pay for the ongoing
+cost of cloud storage and analytics, as well as app updates for every new version of phone
+operating system. There is no guarantee that it will ever provide feedback which is compelling
+enough for the user to pay a monthly subscription to support the ongoing development costs.
+Without that insight, users give up, which is why so many fitness bands are now sitting at the
+back of bedroom drawers. It doesn’t help that few companies making these devices have a
+data analytics business background, rather than a pure hardware business model.
+
+Figure 1.9 The Catch 22 of health and fitness data
+
+The difficulty of developing an insight business model is why almost all of the 500 million
+earbuds which shipped in 2020 concentrated on just one thing – playing content, where the
+user has a choice of multiple, mature services to choose from. I suspect that in time we will
+see sensors return to hearables, as the ear is the best place for a wearable device to measure
+biometrics. It’s stable, it doesn’t move much, it’s close to blood flow and provides a good site
+to measure core temperature. It’s everything that the wrist isn’t. But the industry has learnt
+the lesson that data is also hard, which means that these sensors are likely to appear as minor
+features, allowing companies with the analytic resources time to develop that compelling
+feedback. It’s what we’ve seen Apple do with the Watch. It’s a long slow business.
+Fortunately for earbuds, there is already a compelling reason for consumers to buy them,
+which means that hearables can be a useful platform to experiment with other things.
+All of this translates to a vast amount of excitement in the market. Earbuds are the fastest
+growing consumer product ever and the pace shows no sign of slowing. For the rest of this
+book, we’ll look at how Bluetooth LE Audio can add to the excitement and increase that rate
+of growth.
+
+28
+
+Chapter 2 - The Bluetooth® LE Audio architecture
+
+Chapter 2. The Bluetooth® LE Audio architecture
+Bluetooth specification development follows a well-defined process. It starts off with a New
+Work Proposal, which develops use cases and assesses the market need for any new feature.
+The New Work Proposal is usually generated by a small study group consisting of a few
+companies who want the feature and is then shared and appraised by any others who are
+interested in it. At that point, other Bluetooth SIG members are asked if they’re interested in
+helping develop and prototype it, in order to see if there’s enough critical mass for it to happen.
+Once that level of commitment has been demonstrated, the Bluetooth SIG Board of Directors
+reviews it and assigns it to a group to convert the initial proposal into a set of requirements
+and to put more flesh on the use cases. Those requirements are reviewed to make sure they
+fit into the current architecture of Bluetooth technology without breaking it, and then
+development begins. Once the specification is deemed to be more or less complete,
+implementation teams from multiple member companies develop prototypes which are tested
+against each other in Interoperability Test Events, which check that the features work and
+meet the original requirements. This also provides a good check that the specifications are
+understandable and unambiguous. Any remaining problems get addressed at that stage, and
+once that’s done and everything is shown to work, the specifications are adopted and
+published. Only at that stage are companies allowed to make products, qualify them and start
+selling them.
+Although we always try to avoid specification creep during the process, we almost always fail,
+so new features tend to get added to the original ones. That’s been particularly true in the
+evolution of Bluetooth LE Audio, as it’s evolved from being a moderately simple solution for
+hearing aids, into its current form, which provides the toolkit for the next twenty years of
+Bluetooth audio products. To help understand why we have ended up with over twenty new
+specifications, it useful to look at that journey from the original use cases to see how the final
+architecture was determined
+
+2.1
+
+The use cases
+
+In the initial years of Bluetooth LE Audio development, we saw four main waves of use cases
+and requirements drive its evolution. It started off with a set of use cases which came from
+the hearing aid industry. These were focused on topology, power consumption and latency.
+
+29
+
+Section 2.1 - The use cases
+
+2.1.1
+
+The hearing aid use cases
+
+The topologies for hearing aids were a major step forward from what the Bluetooth Classic
+Audio profiles do, so we’ll start with them.
+2.1.1.1
+Basic telephony
+Figure 2.1. shows the two telephony use cases for hearing aids, allowing hearing aids to
+connect to phones. It’s an important requirement, as holding a phone next to a hearing aid
+in your ear often causes interference .
+
+Figure 2.1 Basic Hearing Aid topologies
+
+The simplest topology, on the left, is an audio stream from a phone to a hearing aid, which
+allows a return stream, aimed primarily at telephony. It can be configured to use the
+microphone on the hearing aid for capturing return speech, or the user can speak into their
+phone. That’s no different from what Hands-Free Profile (HFP) does. But from the
+beginning, the hearing aid requirements had the concept that the two directions of the audio
+stream were independent and could be configured by the application. In other words, the
+stream from the phone to the hearing aid, and the return stream from hearing aid to phone
+would be configured and controlled separately, so that either could be turned on or off. The
+topology on the right of Figure 2.1 moves beyond anything that A2DP or HFP can do. Here
+the phone sends a separate left and right audio stream to left and right hearing aids and then
+adds the complexity of optional return streams from each of the hearing aid microphones.
+That introduces a second step beyond anything that Bluetooth Classic Audio profiles can
+manage, requiring separate synchronised streams to two independent audio devices.
+2.1.1.2
+Low latency audio from a TV
+An interesting extension of the requirement arises from the fact that hearing aids may
+continue to receive ambient sound as well as the Bluetooth audio stream. Many hearing aids
+do not occlude the ear (occlude is the industry term for blocking the ear, like an earplug),
+which means that the wearer always hears a mix of ambient and amplified sound. As the
+processing delay within a hearing aid is minimal – less than a few milliseconds, this doesn’t
+present a problem. However, it become a problem in a situation like that of Figure 2.2,
+where some of the wearer’s family is listening to the sound through the TV’s speakers whilst
+the hearing aid user hears a mix of the ambient sound from the TV as well the same audio
+stream through their Bluetooth connection.
+30
+
+Chapter 2 - The Bluetooth® LE Audio architecture
+
+Figure 2.2 Bluetooth® LE Audio streaming with ambient sound
+
+If the delay between the two audio signals is much more than 30 – 40 milliseconds, it begins
+to add echo, making the sound more difficult to interpret, which is the opposite of what a
+hearing aid should be doing. 30 – 40 milliseconds is a much tighter latency than most
+existing A2DP solutions can provide, so this introduced a new requirement of very low
+latency.
+Although the bandwidth requirements for hearing aids are relatively modest, with a
+bandwidth of 7kHz sufficient for mono speech and 11kHz for stereo music, these could not
+be easily met with the existing Bluetooth codecs whilst achieving that latency requirement.
+That led to a separate investigation to scope the performance requirements for a suitable
+codec, leading to the incorporation of the LC3 codec, which we’ll cover in Chapter 5.
+2.1.1.3
+
+Adding more users
+
+Figure 2.3 Adding in multiple listeners
+
+Hearing loss may run in families, and is often linked with age, so it’s common for there to be
+more than one person in a household who wears hearing aids. Therefore, the new topology
+needed to support multiple hearing aid wearers. Figure 2.3 illustrates that use case for two
+people, both of whom should experience the same latency.
+2.1.1.4
+
+Adding more listeners to support larger areas
+
+The topology should also be scalable, so that multiple people can listen, as in a classroom or
+care home. That requirement lies on a spectrum which extends to the provision of a
+broadcast replacement for the current telecoil induction loops. This required a Bluetooth
+broadcast transmitter which could broadcast mono or stereo audio streams capable of being
+received by any number of hearing aids which are within range, as shown in Figure 2.4.
+31
+
+Section 2.1 - The use cases
+
+Figure 2.4 A broadcast topology to replace telecoil infrastructure
+
+Figure 2.4 also recognises the fact that some people have hearing loss in only one ear, whereas
+others have hearing loss in both, (which may often be different levels of hearing loss). That
+means that it should be possible to broadcast stereo signals at the same time as mono signals.
+It also highlights the fact that a user may wear hearing aids from two different companies to
+cope with those differences, or a hearing aid in one ear and a consumer earbud in the other.
+2.1.1.5
+
+Coordinating left and right hearing aids
+
+Whatever the combination, it should be possible to treat a pair of hearing aids as a single set
+of devices, so that both connect to the same audio source and common features like volume
+control work on both of them in a consistent manner. This introduced the concept of
+coordination, where different devices which may come from different manufacturers would
+accept control commands at the same time and interpret them in the same way.
+2.1.1.6
+
+Help with finding broadcasts and encryption
+
+With telecoil, users have only one option to obtain a signal - turn their telecoil receiver on,
+which picks up audio from the induction loop that surrounds them, or turn it off. Only one
+telecoil signal can be present in an area, so you don’t have to choose which signal you want.
+On the other hand, that means you can’t do things like broadcast multiple languages at the
+same time.
+With Bluetooth, multiple broadcast transmitters can operate in the same area. That has
+obvious advantages, but introduces two new problems - how do you pick up the right audio
+stream, and how do you prevent others from listening in to a private conversation?
+To help choose the correct stream, it’s important that users can find out information about
+what they are, so they can jump straight to their preferred choice. The richness of that
+experience will obviously differ depending upon how the search for the broadcast streams is
+implemented, either on the hearing aid, or on a phone or remote control, but the
+specification needs to cover all of those possibilities. Many public broadcasts wouldn’t need
+32
+
+Chapter 2 - The Bluetooth® LE Audio architecture
+to be private as they reinforce public audio announcements, but others would. In
+environments like the home, you wouldn’t want to pick up your neighbour’s TV. Therefore,
+it is important that the audio streams can be encrypted, requiring the ability to distribute
+encryption keys to authorised users. That process has to be secure, but easy to do.
+As well as the low latency, emulating current hearing aid usage added some other constraints.
+Where the user wears two hearing aids, regardless of whether they are receiving the same
+mono, or stereo audio streams, they need to render the audio within 25µs of each other to
+ensure that the audio image stays stable. That’s equally true for stereo earbuds, but is
+challenging when the left and right devices may come from different manufacturers.
+2.1.1.7
+
+Practical requirements
+
+Hearing aids are very small, which means they have very limited space for buttons. They are
+worn by all ages, but some older wearers have limited manual dexterity, so it’s important that
+controls for adjusting volume and making connections can be implemented on other
+devices, which are easier to use. That may be the audio source, typically the user’s phone,
+but it’s common for hearing aid users to have small keyfob like remote controls as well.
+These have the advantage that they work instantly. If you want to reduce the volume of
+your hearing aid, you just press the volume or mute button; you don’t need to enable the
+phone, find the hearing aid app and control it from there. That can take too long and is not
+a user experience that most hearing aid users appreciate. They need a volume and mute
+control method which is quick and convenient, otherwise they’ll take their hearing aid out of
+their ear, which is not the desired behaviour.
+There is another hearing aid requirement around volume, which is that the volume level
+(actually the gain) should be implemented on the hearing aid. The rationale for this is if the
+audio streams are transmitted at line level2 you get the maximum dynamic range to work with.
+For a hearing aid, which is processing the sound, it is important that the incoming signal
+provides the best possible signal to noise ratio, particularly if it is being mixed with an audio
+stream from ambient microphones. If the gain of the audio is reduced at the source, it results
+in a lower signal to noise ratio.
+An important difference between hearing aids and earbuds or headphones is that hearing aids
+are worn most of the time and are constantly active, amplifying and adapting the ambient
+sound to help the wearer hear more clearly. Users don’t regularly take them off and pop them
+back in a charging case. The typical time a pair of hearing aids is worn each day is around nine
+and a half hours, although some users may wear them for fifteen hours or more. That’s very
+
+Line level is an audio engineering term referring to a standardised output level which is fed into an
+amplifier. In its general sense, it refers to a signal which has been set so that maximum volume of the
+audio volume would correspond to the full range of the output signal, i.e., it makes full use of the
+available dynamic range.
+2
+
+33
+
+Section 2.1 - The use cases
+different from earbuds and headphones, which are only worn when the user is about to make
+or take a phone call, or listen to audio. Earbud manufacturers have been very clever with the
+design of their charging cases to encourage users to regularly recharge their earbuds during the
+course of a day, giving the impression of a much greater battery life. Hearing aids don’t have
+that option, so designers need to do everything they can to minimise power consumption.
+One of the things that takes up power is looking for other devices and maintaining background
+connections. Earbuds get clear signals about when to do this – it’s when they’re taken out of
+the charging box. Most also contain optical sensors to detect when they are in the ear, so they
+can go back to sleep if they’re on your desk. Hearing aids don’t get the same, unambiguous
+signal to start a Bluetooth connection as they’re always on, constantly working as hearing aids.
+That implies that they need to maintain ongoing Bluetooth connections with other devices
+while they’re waiting for something to happen. These can be low duty-cycle connections, but
+not too low, otherwise the hearing aid might miss an incoming call, or take too long to respond
+to a music streaming app being started. Because a hearing aid may connect to multiple
+different devices, for example a TV, a phone, or even a doorbell, connections like this would
+drain too much power, so there was a requirement for a new mechanism to allow them to
+make fast connections with a range of different products, without killing the battery life.
+
+2.1.2
+
+Core requirements to support the hearing aid use cases
+
+Having defined the requirements for topology and connections, it became obvious that a
+significant number of new features needed to be added to the Core specification to support
+them. This led to a second round of work to determine how best to meet the hearing aid
+requirements in the Core.
+The first part of the process was an analysis of whether the new features could be supported
+by extending the existing Bluetooth audio specifications rather than introducing a new audio
+streaming capability into Bluetooth Low Energy. If that were possible, it would have provided
+backwards compatibility with current audio profiles. The conclusion, similar to an analysis
+that had been performed when Bluetooth LE was first developed, was that it would involve
+too many compromises and that it would be better to do a “clean sheet” design on top of the
+Core 4.1 Low Energy specification.
+The proposal for the Core was to implement a new feature called Isochronous3 Channels
+which could carry audio streams in Bluetooth LE, alongside an existing ACL4 channel. The
+ACL channel would be used to configure, set up and control the streams, as well as carrying
+
+Isochronous means a sequence of events which are repeated at equal time intervals.
+ACL channels are Asynchronous Connection-oriented Logical transports which are the used for
+control requests and responses in Bluetooth LE GATT transactions. They carry all of the control
+information in Bluetooth LE Audio (with the sole exception of Broadcast Control subevents) and
+form the control plane. An ACL link must always be present during the life of a CIS.
+3
+4
+
+34
+
+Chapter 2 - The Bluetooth® LE Audio architecture
+more generic control information, such as volume, telephony and media control. The
+Isochronous Channels could support unidirectional or bidirectional Audio Streams, and
+multiple Isochronous Channels could be set up with multiple devices. This separated out the
+audio data and control planes, which makes Bluetooth LE Audio far more flexible.
+It was important that the audio connections were robust, which meant they needed to support
+multiple retransmissions, to cope with the fact that some transmissions might suffer from
+interference. For unicast streams, there is an ACK/NACK acknowledgement scheme, so that
+retransmissions could stop once the transmitter knew that data had been received. For
+broadcast, where there is no feedback, the source would need to unconditionally retransmit
+the audio packets. During the investigation of robustness, it became apparent that the
+frequency hopping scheme5 used to protect LE devices against interference could be
+improved, so that was added as another requirement.
+Broadcast required some new concepts, particularly in terms of how devices could find a
+broadcast without the presence of a connection. Bluetooth LE uses advertisements to let
+devices announce their presence. Devices wanting to make a connection scan for these
+advertisements, then connect to the device they discover to obtain the details of what it
+supports, how to connect – including information on when it’s transmitting, what its hopping
+sequence is and what it does. With the requirements of Bluetooth LE Audio, that requires a
+lot more information than can be fitted into a normal Bluetooth LE advertisement. To
+overcome this limitation, the Core added a new feature of Extended Advertisements (EA) and
+Periodic Advertising trains (PA) which allow this information to be carried in data packets on
+general data channels which are not normally used for advertising. To accompany this, it
+added new procedures for a receiving device to use this information to determine where the
+broadcast audio streams were located and synchronise to them.
+The requirement that an external device can help find a Broadcast stream added a requirement
+that it could then inform the receiver of how to connect to that stream – essentially an ability
+for the receiver to ask for directions from a remote control and be told where to go. That’s
+accomplished by a Core feature called PAST – Periodic Advertising Synchronisation Transfer,
+which is key to making broadcast acquisition simple. PAST is a really useful feature for hearing
+aids, as scanning takes a lot of power. Minimizing scanning in is a useful feature to help
+prolong the battery life of a hearing aid.
+The hearing aid requirements also resulted in a few other features being added to the core
+requirements, primarily around performance and power saving. The first was an ability for
+
+Bluetooth uses an adaptive frequency hopping (AFH) scheme to avoid interference, constantly
+changing the frequency channel a device uses to transmit data, based on an analysis of current
+interference. The sequence of channels to use is called the hopping scheme, which needs to be
+conveyed from the Central Device to all of its Peripherals.
+5
+
+35
+
+Section 2.1 - The use cases
+the new codec to be implemented in the Host or in the Controller. The latter makes it easier
+for hardware implementations, which are generally more power efficient. The second was to
+put constraints on the maximum time a transmission or reception needed to last, which
+impacted the design of the packet structure within Isochronous Channels. The reason behind
+this is that many hearing aids use primary, zinc-air batteries, because of their high power
+density. However, this battery chemistry relies on limiting current spikes and high-power
+current draw. Failing to observe these restrictions results in a very significant reduction of
+battery life. Meeting them shaped the overall design of the Isochronous Channels.
+Two final additions to the Core requirements, which came in fairly late in the development,
+were the introduction of the Isochronous Adaptation Layer (ISOAL) and the Enhanced
+Attribute Protocol (EATT).
+ISOAL allows devices to convert Service Data Units (SDUs) from the upper layer to
+differently sized Protocol Data Units (PDUs) at the Link Layer and vice versa. The reason
+this is needed is to cope with devices which may be using the recommended timing settings
+for the new LC3 codec, which is optimised for 10ms frames, alongside connections to older
+Bluetooth devices which run at a 7.5ms timing interval.
+EATT is an enhancement to the standard Attribute Protocol (ATT) of Bluetooth LE to allow
+multiple instances of the ATT protocol to run concurrently.
+The Extended Advertising features were adopted in the Core 5.1 release, with Isochronous
+Channels, EATT and ISOAL in the more recent Core 5.2 release, paving the way for all of the
+other Bluetooth LE Audio specifications to be built on top of them.
+
+2.1.3
+
+Doing everything that HFP and A2DP can do
+
+As the consumer electronics industry began to recognise the potential of the Bluetooth LE
+Audio features, which addressed many of the problems they had identified over the years, they
+made a pragmatic request for a third round of requirements, to ensure that Bluetooth LE
+Audio would be able to do everything that A2DP and HFP could do. They made the point
+that nobody would want to use Bluetooth LE Audio instead of Bluetooth Classic Audio if the
+user experience was worse.
+These requirements increased the performance requirements on the new codec and introduced
+a far more complex set of requirements for media and telephony control. The original hearing
+aid requirements included quite limited control functionality for interactions with phones,
+assuming that most users would directly control the more complex features on their phone or
+TV, not least because hearing aids have such a limited user interface. Many consumer audio
+products are larger, so don’t have that limitation. As a result, new telephony and media control
+requirements were added to allow much more sophisticated control.
+
+36
+
+Chapter 2 - The Bluetooth® LE Audio architecture
+
+2.1.4
+
+Evolving beyond HFP and A2DP
+
+The fourth round of requirements was a reflection that audio and telephony applications have
+outpaced HFP and A2DP. Many calls today are VoIP6 and it’s common to have multiple
+different calls arriving on a single device – whether that’s a laptop, tablet or phone. Bluetooth
+technology needed a better way of handling calls from multiple different bearers. Similarly,
+A2DP hadn’t anticipated streaming, and the search requirements that come with it, as it was
+written at a time when users owned local copies of music and rarely did anything more
+complex than selecting local files. Today, products needed much more sophisticated media
+control. They also needed to be able to support voice commands without interrupting a music
+stream.
+The complexity in today’s phone and conferencing apps where users handle multiple types of
+call, along with audio streaming, means that they make more frequent transitions between
+devices and applications. The inherent difference in architecture between HFP and A2DP has
+always made that difficult, resulting in a set of best practice rules which make up the MultiProfile specification for HFP and A2DP. The new Bluetooth LE Audio architecture was
+going to have to go beyond that and incorporate multi-profile support by design, with robust
+and interoperable transitions between devices and applications, as well as between unicast and
+broadcast.
+As more people in the consumer space started to understand how telecoil and the broadcast
+features of hearing aids worked, they began to realise that broadcast might have some very
+interesting mass consumer applications. At the top of their list was the realisation that it could
+be used for sharing music. That could be friends sharing music from their phones, silent
+discos or “silent” background music in coffee shops and public spaces. Public broadcast
+installations, such as those designed to provide travel information for hearing aid wearers,
+would now be accessible to everyone with a Bluetooth headset. The concept of Audio Sharing,
+which we’ll examine in more detail in Chapter 12, was born.
+Potential new use cases started to proliferate. If we could synchronise stereo channels for
+two earbuds, why not for surround sound? Companies were keen to make sure that it
+supported smart watches and wristbands, which could act as remote controls, or even be
+audio sources with embedded MP3 players. The low latencies were exciting for the gaming
+community. Microwave ovens could tell you when your dinner’s cooked (you can tell that
+idea came from an engineer). The number of use cases continued to grow as companies saw
+how it could benefit their customers, their product strategies and affect the future use of
+voice and music.
+
+Voice over Internet Protocol, which is used for voice in most PC and mobile OTT (Over The Top)
+telephony applications.
+6
+
+37
+
+Section 2.2 - The Bluetooth LE Audio architecture
+The number of features has meant it has taken a long time to complete the specification.
+What is gratifying is that most of the new use cases which have been raised in the last few
+years have not needed us to go back and reopen the specifications – we’ve found that they
+were already supported by the features that had been defined. That suggests the Bluetooth
+LE Audio standards have been well designed and are able to support the audio applications
+we have today as well as new audio applications which are yet to come.
+
+2.2
+
+The Bluetooth LE Audio architecture
+
+The Bluetooth LE Audio architecture has been built up in layers, as has every other Bluetooth
+specification before it. This is illustrated in Figure 2.5, which shows the main new specification
+blocks relating to Bluetooth LE Audio, (with key existing ones greyed out or dotted).
+At the bottom we have the Core, which contains the radio and Link Layer, (collectively known
+as the Controller). It is responsible for sending Bluetooth packets over the air. On top of that
+is the Host, which has the task of telling the Core what to do for any specific application. The
+separation between the Controller and Host is historic, reflecting the days when a Bluetooth
+radio would be sold in a USB stick or a PCMCIA card, with the Host implemented as a
+software application on a PC. Today both Host and Controller are often integrated into a
+single chip.
+In the Host, there is a new structure called the Generic Audio Framework or GAF. This is an
+audio middleware, which contains all of the functions which are considered to be generic, i.e.,
+features which are likely to be used by more than one audio application. The Core and the
+GAF are the heart of Bluetooth LE Audio. They provide great flexibility. Finally, at the top
+of the stack, we have what are termed “top level” profiles, which add application specific
+information to the GAF specifications.
+
+Figure 2.5 The Bluetooth® LE Audio architecture
+
+38
+
+Chapter 2 - The Bluetooth® LE Audio architecture
+It is perfectly possible to build interoperable Bluetooth LE Audio applications using just the
+GAF specifications. The individual specifications within it have been defined to ensure a base
+level of interoperability, which would enable any two Bluetooth LE Audio devices to transfer
+audio between them. The top level profile specifications largely add features specific to a
+particular type of audio application, mandating features which GAF only defines as optional,
+and adding application-specific functionality. The intention is that the top level profiles are
+relatively simple, building on features from within the GAF.
+At first glance, the Bluetooth LE Audio architecture looks complex, as we've ended up with
+23 different specifications inside the Generic Audio Framework, as well as an expanded Core
+and the new LC3 codec. But there's a logic to this. Each specification is trying to encapsulate
+the specific elements of the way you set up and control different aspects of an audio stream.
+In the rest of this chapter, I'll briefly explain each one and how they fit together. Then, in the
+rest of the book, we'll look at how each individual specification works and how they interact.
+
+2.2.1
+
+Profiles and Services
+
+All of the specifications within the GAF are classified as Profiles or Services using the standard
+Bluetooth LE GATT model depicted in Figure 2.6.
+
+.
+Figure 2.6 The Bluetooth® LE Profile and Service model
+
+In Bluetooth LE, Profiles and Services can be considered as Clients and Servers. The Service
+is implemented where the state7 resides, whereas Profile specifications describe how the state
+behaves and includes procedures to manage it. Service specifications define one or more
+characteristics which can represent individual features or the states of a state machine. They
+can also be control points, which cause transitions between the states of a state machine.
+Profiles act on these characteristics, reading or writing them and being notified whenever the
+values change. Multiple devices, each acting as Clients, can operate on a Server.
+
+7
+
+State refers to the value of a parameter or the current position within a state machine.
+
+39
+
+Section 2.2 - The Bluetooth LE Audio architecture
+Traditionally, in classic Bluetooth profiles (which didn’t have a corresponding service), there
+was a simple one-to-one relationship with only one Client and one Server, with everything
+described in a Profile specification. In Bluetooth LE Audio, the many-to-one topology is
+much more common, particularly in features like volume control and the selection of a
+Broadcast Source, where a user may have multiple devices implementing the Profile
+specification and acting as a Client. In most cases, these act on a first come, first served basis.
+The number of different control profiles that can be used in Bluetooth LE Audio drove the
+EATT enhancement to the Core. Profiles and Services communicate using the Attribute
+Protocol (ATT), but ATT assumes that only one command is occurring at a time. If more
+than one is happening, the second command can be delayed, because ATT is a blocking
+protocol. To get around this, the Extended Attribute Protocol (EATT) was added in Core 5.2
+release, allowing multiple ATT instances to operate at the same time.
+Figure 2.7 provides an overview of the Bluetooth LE Audio architecture, putting a name, or
+more precisely, a set of letters, to all of the 18 specifications which make up the GAF, along
+with the four in the current top level profiles. The dotted boxes indicate sets of profiles and
+services which work together. In most cases there is a one-to-one relationship of a profile and
+a service, although in the case of the Basic Audio Profile (BAP)8 and the Voice Control Profile
+(VCP), one profile can operate on three different services. The Public Broadcast Profile (PBP)
+is an anomaly, as it’s a profile without a service, but that’s one of the consequences of
+broadcast, as you cannot have a traditional Client-Server interaction when there is no
+connection.
+
+Figure 2.7 Overview of the Bluetooth LE Audio Specifications
+
+If the acronym or initialism ends with a “P” it’s a profile. If it ends with an “S”, it’s a service. If it
+ends with PS, it normally refers to a combination of separate Profile and Service documents.
+8
+
+40
+
+Chapter 2 - The Bluetooth® LE Audio architecture
+
+2.2.2
+
+The Generic Audio Framework
+
+We can now look at the constituent parts of the GAF. There is a significant amount of
+interaction between the various specifications, which makes it difficult to draw a clear
+hierarchy or set of relationships between them, but they can be broadly separated into four
+functional groups, arranged as in Figure 2.8.
+
+Figure 2.8 Functional grouping of specifications within the Generic Audio Framework
+
+This grouping is largely for the sake of explanation. In real implementations of Bluetooth LE
+Audio, most of these specifications interact with each other to a greater or lesser degree. It’s
+perfectly possible to make working products with just a few of them, but to design richly
+featured, interoperable products, the majority of them will be required.
+
+2.2.3
+
+Stream configuration and management – BAPS
+
+Starting at the bottom of Figure 2.8, we have a group of four specifications which are
+collectively known as the BAPS specifications. These four specifications form the foundation
+of the Generic Audio Framework. At their core is BAP – the Basic Audio Profile, which is
+used to set up and manage unicast and broadcast Audio Streams. As a profile, it works with
+three services:
+•
+•
+•
+
+PACS – the Published Audio Capabilities Service, which exposes the capabilities of
+a device,
+ASCS - the Audio Stream Control Service, which defines the state machine for
+setting up and maintaining unicast audio streams, and
+BASS - the Broadcast Audio Scan Service which defines the procedures for
+discovering and connecting to broadcast audio streams and distributing the
+broadcast encryption keys.
+41
+
+Section 2.2 - The Bluetooth LE Audio architecture
+Between them, they are responsible for the way we set up the underlying Isochronous
+Channels which carry the audio data. They also define a standard set of codec configurations
+for LC3 and a corresponding range of Quality of Service (QoS) settings for use with broadcast
+and unicast applications.
+State machines for each individual Isochronous Channel are defined for both unicast and
+broadcast, both of which move the audio stream through a configured state to a streaming
+state, as illustrated in the simplified state machine of Figure.2.9.
+
+Figure.2.9. The simplified Isochronous Channel state machine
+
+For unicast, the state machine is defined in the ASCS specification. The state resides within
+individual audio endpoints in the Server, with the Client control defined in BAP. For
+broadcast, where there is no connection between the transmitter and receiver, the concept of
+a Client-Server model becomes a little tenuous. As a result, a state machine is only defined for
+the transmitter and is solely under the control of its local application. With broadcast, the
+receiver needs to detect the presence of a stream and then receive it, but has no way of
+affecting its state.
+Multiple unicast or broadcast Isochronous Channels are bound together in Groups (which
+we explore in Chapter 4). BAP defines how these groups, and their constituent Isochronous
+Channels are put together for both broadcast and unicast streams.
+You can make a Bluetooth LE Audio product with just three of these specifications; BAP,
+ASCS and PACS for unicast and BAP alone for broadcast (although you’ll need to add PACS
+and BASS if you want to use a phone or remote control to help find the broadcasts). It would
+be quite a limited device in terms of functionality – just setting up an audio stream, using it to
+transmit audio and stopping it. However, by being able to do this, the BAPS set of
+specifications provide a base level of interoperability for all Bluetooth LE Audio devices. If
+two Bluetooth LE Audio devices have different top level profiles, they should still be able to
+set up an audio stream using BAP. It may have restricted functionality, but should provide an
+acceptable level of performance, removing the issue of multi-profile incompatibility that is
+present in Bluetooth Classic Audio, where devices with no common audio profile would not
+work together.
+
+42
+
+Chapter 2 - The Bluetooth® LE Audio architecture
+
+2.2.4
+
+Rendering and capture control
+
+Having set up a stream, users want to control the volume, both of the audio streams being
+rendered in their ears and the pick-up of microphones.
+Volume is a surprisingly difficult topic, as there are multiple places where the volume can be
+adjusted – on the source device, on the hearing aid, earbud or speaker, or on another “remote
+control” device, which could be a smartwatch or a separate Controller. In Bluetooth LE
+Audio, the final gain of the volume is performed in the hearing aid, earbud or speaker, and
+not on the incoming audio stream (although top level profiles may require that as well). With
+that assumption, the Volume Control Profile (VCP) defines how a Client manages the gain on
+the Audio Sink device. The state of that gain is defined in the Volume Control Service (VCS),
+with one instance of VCS on each audio sink. The Volume can be expressed as absolute or
+relative, and can also be muted.
+Where there are multiple Audio Streams, as with earbuds and hearing aids, a second service is
+required. VOCS - the Volume Offset Control Service, effectively acts as a balance control9,
+allowing the relative volume of multiple devices to be adjusted. These may be rendered on
+different devices, such as separate left and right earbuds or speakers, or on a single device, like
+a pair of headphones or a soundbar. The Audio Input Control Service (AICS) acknowledges
+the fact that most devices have the ability to support a number of different audio streams, as
+illustrated in Figure 2.10. AICS provides the ability to control multiple different inputs that
+can be mixed together and rendered within your earbud or speaker. illustrates how these three
+services could be used in a soundbar which has a Bluetooth, HDMI and microphone input.
+
+Figure 2.10 Audio Input Control Service (AICS), Volume Control Service (VCS) and Volume Offset Control
+Service (VOCS)
+
+“Balance” is the usual consumer nomenclature. Audio engineers would probably consider it to be a
+mixing control.
+9
+
+43
+
+Section 2.2 - The Bluetooth LE Audio architecture
+For a hearing aid, the inputs might be a Bluetooth stream, a microphone providing an ambient
+audio stream, and a telecoil antenna receiving a stream from an audio loop. At any point in
+time, the wearer may want to hear a combination of these different inputs. AICS supports
+that flexibility.
+An important feature of the volume services is that they notify any changes back to Client
+devices running the Voice Control Profile. This ensures that all potential Controllers are kept
+up to date with any changes to their state, regardless of whether that occurred over a Bluetooth
+link or from a local volume control. This ensures that they all have a synchronised knowledge
+of the volume state, so that the user can make changes from any one of them, without any
+unexpected effects from their having an outdated knowledge of the current state of the volume
+level.
+A complementary pair of specifications, MICP and MICS, the Microphone Control Profile
+and Service, are responsible for controlling the microphones that reside within hearing aids
+and earbuds. Nowadays, these devices typically contain multiple microphones. Hearing aids
+listen to both ambient sound (their primary function), as well as audio that's being received
+over Bluetooth. As earbuds get more sophisticated, we’re increasingly seeing similar ambient
+sound capabilities being built into them, with a growing popularity for some degree of
+transparency.
+MICP works in conjunction with AICS and MICS to control the overall gain and muting of
+multiple microphones. They are normally used for control of the captured audio that is
+destined for a Bluetooth stream, but can be used more widely. Figure 2.11 illustrates their use.
+
+Figure 2.11 The Audio Input Control Service (AICS) used with the Microphone Control Service (MICS)
+
+44
+
+Chapter 2 - The Bluetooth® LE Audio architecture
+
+2.2.5
+
+Content control
+
+Having specified how streams are set up and managed and how volume and microphone input
+is handled, we come to content control. The content we listen to is generated outside of the
+Bluetooth specifications – it may be streaming music, live TV, a phone call or a video
+conference. What the content control specifications do is to allow control in terms of starting,
+stopping, answering, pausing and selecting the streams. These are the types of control which
+were embedded into HFP and the Audio/Video Remote Control Profile (AVRCP), which
+accompanies A2DP. In Bluetooth LE Audio they are separated out into two sets of
+specifications – one for telephony in all of its forms, and the other for media. The key
+differentiator is that telephony is about the state of the call or calls, which normally reflects
+the state of the telephony service, whereas media control acts on the state of the stream –
+when and how it’s played and how it’s selected. Because these are decoupled from the audio
+streams, they can now be used to help control transitions, such as pausing music playback
+when you accept a phone call and restoring it when the call is finished. For both of these pairs
+of specifications, the Service resides on the primary audio source – typically the phone, PC,
+tablet or TV, whereas the Profile is implemented on the receiving device, such as the hearing
+aid or earbud. As with the rendering and capturing controls, multiple devices can act as
+Clients, so telephony and media state can be controlled from a smartwatch as well as from an
+earbud.
+The Media Control Service (MCS) resides on the source of audio media and reflects the state
+of the audio stream. The state machine allows a Client using the Media Control Profile (MCP)
+to transition each media source through Playing, Paused and Seeking states. At its simplest, it
+allows an earbud to control Play and Stop. However, MCS goes far beyond that, providing all
+of the features which users expect from content players today. It also provides higher level
+functions, where a user can search for tracks, modify the playing order, set up groups and
+adjust the playback speed. It defines metadata structures which can be used to identify the
+tracks and uses the existing Object Transfer Service (OTS) to allow a Client to perform media
+searches on the Server, or more typically the application behind it. All of this means that a
+suitably complex device running the Media Control Profile can recreate the controls of a music
+player.
+Telephony control is handled in a similar way using the Telephone Bearer Service (TBS), which
+resides on the device involved in the call (typically the phone, PC or laptop), with the
+complementary Call Control Profile (CCP) controlling the call by writing to the state machine
+in the TBS instance. TBS and CCP have expanded past the limitations of Hands-Free Profile
+to accommodate the fact that we now use telephony in many different forms. It's no longer
+just traditional circuit switched10 and cellular bearers, but PC and web-based communication
+
+Circuit switched is a definition derived from original telephone topology, where end-to-end
+connections were made by physically switching, using banks of relays. The term endured until IP
+10
+
+45
+
+Section 2.2 - The Bluetooth LE Audio architecture
+and conferencing applications, using multiple, different types of bearer service. TBS exposes
+the state of the call using a generic state machine. It supports multiple calls, call handling and
+joining, caller ID, inband and out of band ringtone selection and exposes call information,
+such as signal strength.
+Both TBS and MCS acknowledge the fact that there may be multiple sources of media and
+multiple different call applications on the Server devices. To accommodate this, both can be
+instantiated multiple times – once for each instance of an application. This allows a Client
+with the complementary profile to control each application separately. Alternatively, a single
+instance of the service can be used, with the media or call device using its specific
+implementation to direct the profile commands to the correct application. The single instance
+variants of TBS and MCS are known as the Generic Telephone Bearer Service (GTBS) and
+Generic Media Control Service (GMCS) and are included in the TBS and MCS specifications
+respectively. We’ll look at these in more detail in Chapter 9.
+
+2.2.6
+
+Transition and coordination control
+
+Next, we come to the Transition and Coordination Control specifications. Their purpose is
+to glue the other specifications together, providing a way for the top level profiles to call down
+to them without having to concern themselves with the fine detail of setting things up.
+One of the major enhancements in Isochronous Channels is the ability to stream audio to
+multiple different devices and render it at exactly the same time. The most common
+application of this is in streaming stereo music to left and right earbuds, speakers or hearing
+aids. The topology and syncnronisation of rendering are handled in the Core and BAP, but
+ensuring that control operations occur together, whether that’s changing volume or
+transitioning between connections is not. That’s where the Coordinated Set Identification
+Profile (CSIP) and Coordinated Set Identification Service (CSIS) come in.
+Where two or more Bluetooth LE Audio devices are expected to be used together, they are
+called a Coordinated Set and can be associated with each other by use of the Coordinated Set
+Identification Service. This allows other profiles, in particular CAP, to treat them as a single
+entity. It introduces the concepts of Lock and Rank to ensure that when there is a transition
+between audio connections, whether that’s to a new unicast or broadcast stream, the members
+of the set always react together. This prevents a new connection only being applied to a subset
+of the devices in the set, such as a TV connecting to your right earbud, while your phone
+connects to your left. Devices designed to be members of Coordinated Sets are generally
+configured as set members during manufacture.
+
+routing technologies appeared.
+
+46
+
+Chapter 2 - The Bluetooth® LE Audio architecture
+Multiple devices which are not configured as members of a Coordinated Set can still be used
+in GAF as an ad-hoc set. In this case they need to be individually configured by the
+application. It means that they do not benefit from the locking feature of CSIS, which could
+result in different connections to members of the ad-hoc set.
+CAP – the Common Audio Profile, introduces the Commander role, which brings together
+the features which can be used for remote control of Bluetooth LE Audio Streams. The
+Commander is a major change from anything we’ve seen in previous Bluetooth specifications,
+allowing the design of ubiquitous, distributed remote control for audio. It is particularly useful
+for encrypted broadcasts, where it provides a way to convert broadcast transmissions into a
+private listening experience. We’ll explore that in more detail in Chapter 8.
+CAP uses CSIS and CSIP to tie devices together and ensure that procedures are applied to
+both. It also introduces the concept of Context Types and Content Control IDs which allow
+applications to make decisions about stream setup and control based on a knowledge of the
+controlling devices, the use cases for the audio data and which applications are available. This
+is used to inform transitions between different streams, whether that is prompted by different
+applications on a device or a request from a different device for an audio connection. A lot of
+this functionality is based on new concepts, which have been introduced in Bluetooth LE
+Audio. These are explained in more detail in Chapter 3.
+
+2.2.7
+
+Top level Profiles
+
+Finally, on top of the GAF specifications, we have top level profiles which provide additional
+requirements for specific audio use cases. The first of these are the Hearing Access Profile
+and Service (HAP and HAS), which cover applications for the hearing aid ecosystem; the
+Telephony and Media Audio Profile (TMAP)11, which specifies the use of higher quality codec
+settings and more complex media and telephony control, and the Public Broadcast Profile
+(PBP), which helps users select globally interoperable broadcast streams. Public Broadcast
+Profile is anomalous in having no accompanying service, but that’s a consequence of the nature
+of broadcasts, where there is no connection for any Client-Server interaction.
+
+2.2.8
+
+The Low Complexity Communications Codec (LC3)
+
+Although not a part of the GAF, the Bluetooth LE Audio releases include a new, efficient
+codec called LC3 which is the mandated codec for Bluetooth LE Audio Streams. This
+provides excellent performance for telephony speech, wideband and super-wideband speech
+and high-quality audio and is the mandated codec within BAP. Every Bluetooth LE Audio
+product has to support the LC3 codec to ensure interoperability, but additional and proprietary
+codecs can be added if manufacturers require. LC3 encodes audio into single streams, so
+
+There is a corresponding, minimalist TMAS specification, which is included within the TMAP
+specification.
+11
+
+47
+
+Section 2.3 - Talking about Bluetooth LE Audio
+stereo is encoded as separate left and right streams. This means that GAF can configure a
+unicast stream to an earbud to carry only the audio which that earbud requires. A broadcast
+transmitter sending music would normally include both left and right Audio Streams in its
+broadcast. Individual devices would only need to receive and decode the data relevant to the
+stream which they want to render.
+
+2.3
+
+Talking about Bluetooth LE Audio
+
+Writing over twenty specifications has generated a lot of new abbreviations, including ones
+for the name of each individual specification. Over time, the working groups have come up
+with different ways of referring to these – sometimes as acronyms (where you pronounce them
+as words), sometimes as initialisms (where you just pronounce the letters, and occasionally as
+a mix of both. The following table captures the current pronunciations of the most commonly
+used ones, as a help to anyone talking about Bluetooth LE Audio.
+
+2.3.1
+
+Acronyms
+
+The following abbreviations are all pronounced as words, or a combination of letter and word:
+Abbreviation Meaning
+AICS
+ASE
+ATT
+BAP
+BAPS
+BASE
+BASS
+BIG
+BIS
+CAP
+CAS
+CIG
+CIS
+CSIP
+CSIS
+CSIPS
+EATT
+GAF
+GAP
+GATT
+HAP
+HARC
+48
+
+Audio Input Control Service
+Audio Stream Endpoint
+Attribute Protocol
+Basic Audio Profile
+The set of BAP, ASCS, BASS and PACS
+Broadcast Audio Source Endpoint
+Broadcast Audio Scan Service
+Broadcast Isochronous Group
+Broadcast Isochronous Stream
+Common Audio Profile
+Common Audio Service
+Connected Isochronous Group
+Connected Isochronous Stream
+Coordinated Set Identification Profile
+Coordinated Set Identification Service
+The set of CSIP and CSIS
+Enhanced ATT
+Generic Audio Framework
+Generic Access Profile
+Generic Attribute Profile
+Hearing Access Profile
+Hearing Aid Remote Controller
+
+Pronunciation
+aches (as in aches and pains)
+ase (rhymes with case)
+at
+bap (rhymes with tap)
+baps
+base (rhymes with case)
+rhymes with mass, not mace
+big
+biss
+cap (rhymes with tap)
+cas (rhymes with mass)
+sig
+sis
+see - sip
+see - sis
+see - sips
+ee - at
+gaffe
+gap (rhymes with tap)
+gat (rhymes with cat)
+hap (rhymes with tap)
+hark
+
+Chapter 2 - The Bluetooth® LE Audio architecture
+Abbreviation Meaning
+HAS
+HAUC
+INAP
+
+PHY
+QoS
+RAP
+SIRK
+
+Hearing Access Service
+Hearing Aid Unicast Client
+Immediate Need for Audio related
+Peripheral
+Logical Link Control and Adaptation
+protocol
+Microphone Control Profile
+Microphone Control Service
+Published Audio Capabilities
+Published Audio Capabilities Service
+Periodic Advertising Sync Transfer
+Public Broadcast Audio Stream
+Announcement
+physical layer
+Quality of Service
+Ready for Audio related Peripheral
+Set Identity Resolving Key
+
+TMAP
+TMAS
+VOCS
+
+Telephony and Media Audio Profile
+Telephony and Media Audio Service
+Volume Offset Control Service
+
+L2CAP
+MICP
+MICS
+PAC
+PACS
+PAST
+PBAS
+
+Pronunciation
+hass (rhymes with mass)
+hawk
+eye - nap
+el - two - cap
+mick - pee
+mick - ess
+pack
+packs
+past (rhymes with mast)
+pee - bass (bass rhymes with
+mass)
+fy (rhymes with fly)
+kwos
+rap
+sirk (like the first syllable of
+circus)
+tee - map
+tee - mas
+vocks
+
+Table 2.1 Pronunciation guide for Bluetooth® LE Audio acronyms
+
+2.3.2
+
+Initialisms
+
+The rest are simply pronounced as the individual letters that make up the abbreviation, e.g.,
+CCP is pronounced “see – see – pee” and LC3 is “el – see – three”.
+The most common initialisms in Bluetooth LE Audio are: ACL, AD, ASCS, BMR, BMS,
+BR/EDR, CCID, CCP, CG, CSS, CT, CTKD, EA, FT, GMCS, GSS, GTBS, HA, HCI, IA,
+IAC, IAS, INAP, IRC, IRK, LC3, MCP, MCS, MTU, NSE, PA, PBA, PBK, PBP, PBS, PDU,
+PTO, RFU, RTN, SDP, SDU, TBS, UI, UMR, UMS, UUID, VCP and VCS. These are all
+explained in the glossary.
+
+2.3.3
+
+Irregular pronunciations
+
+OOB is an oddity, as it is always spoken in full, i.e., “out of band”, although the abbreviation is
+generally used in text.
+
+49
+
+Section 2.3 - Talking about Bluetooth LE Audio
+-oOoThat was a whistle-stop tour through the main parts of the Generic Audio Framework. We
+will see even more specifications appear in GAF in the future, but for now, the ones described
+above emulate and expand on the audio functionality that Bluetooth has today with the classic
+audio profiles.
+In the remainder of the book, we’ll see how BAPS (BAP, BASS, ASCS and PACS) form the
+core of everything that you do with Bluetooth LE Audio. The other specifications add
+usability and functionality, while CAP glues it all together. As a first step, we’ll look at some
+of the new concepts which have been necessary to address the Bluetooth LE Audio
+requirements.
+
+50
+
+Chapter 3 - New concepts in Bluetooth® LE Audio
+
+Chapter 3. New concepts in Bluetooth® LE Audio
+To provide designers with the flexibility they need, the new Bluetooth LE Audio specifications
+have introduced some important new concepts. In this chapter, we’ll look at what they do
+and why they are needed. Because these features are closely integrated into the specifications,
+some of the descriptions below will become clearer as we dive down into the detail of the
+Core and the GAF. However, it’s useful to introduce them at this stage, as they pervade much
+of what follows.
+
+3.1
+
+Multi-profile by design
+
+As we’ve seen, one of the challenges faced by Bluetooth Classic Audio has been the multiprofile issue, where users change between streaming music, making phone calls and using
+voice recognition. As the complexity of audio use cases continues to increase, that problem
+won’t improve. The two successful classic Bluetooth audio profiles – Hands-Free Profile
+(HFP) and music streaming (A2DP), were designed at a time where users were assumed to use
+one or the other, starting new, independent sessions as they moved from phone calls to music.
+It soon became apparent that these were not independent functions, but that phone users
+would expect one use case to interrupt the other. Over the years, the industry has developed
+rules and methods to address this problem, culminating in the publication of the Multi Profile
+Specification (MPS) in 2013, which essentially collects sets of rules for common transitions
+between these profiles.
+The Multi Profile Specification is not particularly flexible and didn’t foresee the emergence of
+new use cases, such as voice recognition, voice assistants and interruptions from GPS satnavs.
+Nor are the underlying specifications particularly versatile. Although HFP has gone through
+multiple versions and spawned around twenty complementary specifications addressing
+specific aspects of car to phone interaction, it still struggles to keep up with non-cellular VoIP
+telephony applications. Neither it, nor A2DP, have sufficiently addressed the changing nature
+of audio, where users connect to multiple different audio sources during the course of a day,
+and might also own multiple different headsets and earbuds.
+From the start, Bluetooth LE Audio was developed with the philosophy that it would support
+“multi-profile by design”. It recognised that users might have multiple headsets and audio
+sources, constantly changing connections in an ad-hoc manner. The addition of broadcast
+audio makes this even more important, as the projected uptake of broadcast in venues, shops,
+leisure facilities and travel would extend the number of different connections a user is likely
+to make each day. It was obviously going to be important to provide tools to make the user
+experience seamless.
+
+3.2
+
+The Audio Sink led journey
+
+Alongside these demands came a realization that the phone would not necessarily remain the
+centre of the audio universe, which had been a tacit assumption throughout the evolution of
+51
+
+Section 3.3 - Terminology
+the Bluetooth Classic Audio specifications. As an increasing number of different audio
+sources became available, it was important to consider how a user would connect their earbuds
+or hearing aids through the course of the day. This turned the perceived view of a phonecentric audio world on its head, replacing it with the concept of an Audio Sink12 led journey.
+It’s an approach where the phone is no longer the arbiter of what you listen to. Instead, it
+assumes that users are increasingly likely to wear a pair of hearing aids or earbuds throughout
+the day, constantly changing their audio source. It may start with an alarm clock, followed by
+asking your voice assistant to turn your radio on, information from a bus stop, voice control
+for your computer, receiving phone calls, interruptions from the doorbell and relaxing in front
+of the TV when you get home, until the microwave tells you that your dinner is ready. This
+conceptual turn-around regarding who is in control takes a lot from the experience of hearing
+aid wearers, who typically wear their hearing aids for most of the day. For a large part of that
+time, the hearing aid works without any Bluetooth link, just listening to and reinforcing the
+sound around the wearer. But the wearer can connect it to a wide number of infrastructure
+audio sources – supermarket checkout machines, public transport information, theatres and
+auditoria and TVs, as well as phones and music players if they support Bluetooth Classic
+Audio. It’s functionality that many more consumers are likely to use once it becomes widely
+available.
+Phones will, of course, remain a major source of audio, but increasingly, so are PCs, laptops,
+TVs, tablets, Hi-Fi and voice assistants. They’re also being joined by a range of smart home
+devices from doorbells to ovens. At the same time, consumers are buying more headsets. It’s
+not uncommon for someone to own a pair of sleep buds, a set of earbuds, a pair of stereo
+headphones and a soundbar or Bluetooth speakers. As the potential combinations multiply,
+it’s vital that the user interfaces can accommodate this range of different connections and the
+transitions between them.
+
+3.3
+
+Terminology
+
+Different parts of the Bluetooth LE Audio specifications use different names for the
+transmitting and receiving devices and what they do. Each specification refers to these as
+Roles, so by the time we’ve gone from the Core to a top level profile we will have accumulated
+at least four different Roles for each device. There is a very good reason for this, as each step
+up the stack results in more specificity within those Roles, with the result that devices can
+implement specific combinations to fulfil a targeted function. But they can get confusing.
+
+In this case, the Audio Sink is the general term for what you have in your ear to listen to audio
+streams. It sidesteps the fact that it can be an Audio Source as well. The point is that the device in
+your ear can be more involved in decisions.
+12
+
+52
+
+Chapter 3 - New concepts in Bluetooth® LE Audio
+The Core refers to Central and Peripheral devices, where Central devices send commands and
+Peripherals respond. In BAP they’re called Clients and Server roles. Moving up a layer, CAP
+defines them as Initiator and Acceptor roles. The Initiator role exists in a Central device,
+which is responsible for setting up, scheduling and managing the Isochronous Streams.
+Acceptors are the devices that participate in these streams – typically hearing aids, speakers
+and earbuds. There is always one Initiator, but there can be multiple Acceptors. The different
+nomenclature is shown is Figure 3.1.
+
+Figure 3.1 Bluetooth® LE Audio roles
+
+I'm going to use the Initiator and Acceptor names most of time (even though they’re
+technically Roles), because I think they best explain the way that Bluetooth LE Audio works.
+Both Initiators and Acceptors can transmit and receive audio streams at the same time, and
+they can both contain multiple Bluetooth LE Audio Sinks and Sources. The important thing
+to remember is that the Initiator is the device that performs the configuration and scheduling
+of the Isochronous Streams, and the Acceptor is the device that accepts those streams. That
+concept applies both in unicast and broadcast. There’s one other abbreviation I’ll make. As
+only Initiators can broadcast Audio Streams, I’ll save a few words and refer to Initiators acting
+as Broadcast Sources as Broadcasters, unless there’s a need to be explicit.
+At this point, it's worth having a slight diversion to look at some of the terminology which is
+used within Bluetooth LE Audio. This will introduce some features we haven’t discussed yet,
+which we’ll cover in detail in Chapter 4, when we get to Isochronous Streams.
+Throughout the specifications, there are a lot of different names and phrases describing
+channels and streams. They have some very distinct and sometimes slightly different meanings
+depending on which of the specifications you're looking at, but I’ve tried to capture the main
+ones in Figure 3.2.
+
+Figure 3.2 Bluetooth® LE Audio terminology
+
+53
+
+Section 3.3 - Terminology
+The diagram shows the conceptual transmission of audio data from left to right. At the very
+left, we have audio coming in, which is typically an analogue signal. It’s shown as Audio IN.
+At the other end we see audio being rendered at Audio OUT, which may be on a speaker or
+an earbud. The audio at these two points is called the Audio Channel (both ends are the same
+Audio Channel), which is equivalent to what would be carried if it were a wired connection
+between an MP3 player and a speaker. An Audio Channel is defined in BAP as a unidirectional
+flow of audio into the input of your Bluetooth device and then out of the other end. (In
+practice, that “out” will normally be directly into your headphone transducer or a speaker.)
+The details of the input and output of the audio channel are implementation specific and
+outside the Bluetooth specifications. What the audio is and how the audio is handled after the
+wireless transmission and decoding is outside the scope of the specification13.
+How that audio input is carried to the audio output is what is defined inside the Bluetooth
+specifications. They are represented by the dotted box in Figure 3.2. The audio is encoded
+and decoded using the new LC3 codec, unless an implementation needs to use a specific,
+additional codec. Other codecs can be used, but all devices must support some basic LC3
+configurations, which are defined in BAP, to ensure interoperability. The encoder produces
+encoded audio data, which goes into the payloads for the Isochronous Streams.
+Once the audio data is encoded, it is placed into SDUs (Service Data Units), which the Core
+translates into PDUs (Protocol Data Units) which are transmitted to the receiving device.
+Once a PDU is received, it is reconstructed as an SDU for delivery to the decoder. The
+Isochronous Stream is the term used to describe the transport of SDUs, encapsulated in
+PDUs, from encoder output to the decoder input. It includes retransmissions and any
+buffering required to synchronise multiple Isochronous Streams. The flow of encoded audio
+data over an Isochronous Stream is defined in BAP as an Audio Stream, and, like an Audio
+Channel, is always unidirectional. The relationship of the different terms are illustrated in
+Figure 3.3.
+
+Figure 3.3 Representation of streaming terms
+
+With the exception of the Presentation Delay, which states when it is rendered. This is explained in
+Section 3.10.
+13
+
+54
+
+Chapter 3 - New concepts in Bluetooth® LE Audio
+The Core places Isochronous Streams serving the same application together into an
+Isochronous Group. For unicast streams, it’s called a Connected Isochronous Group (CIG),
+which contains one or more Connected Isochronous Streams (CIS). For Broadcast, it’s a
+Broadcast Isochronous Group (BIG). Where more than one Isochronous Stream is present
+in an Isochronous Group, carrying audio data in the same direction, the Isochronous Streams
+are expected to have a time relationship to each other at the application layer. By time
+relationship, we mean Audio Channels which are expected to be rendered or captured at the
+same time. A typical application is rendering a left and a right stereo stream into two separate
+earbuds, with one or two return streams from their microphones, which highlights the fact
+that a CIG or BIG can contain Isochronous Streams which go to multiple devices.
+Bluetooth LE Audio has been designed to be very flexible, which means that sometimes there
+are different ways of doing things. Taking one example, if we look at a simple connection
+from a phone to a headset, we need one audio stream going in each direction. One way to do
+that is to establish separate CISes for each direction.
+
+Figure 3.4 Two separate CISes in a CIG
+
+In Figure 3.4 we can see two CISes, an outgoing one (CIS 0) and an incoming one (CIS 1),
+both contained within the same CIG. We can optimise that by using a single CIS, which is
+what's shown in Figure 3.5, which shows a single bidirectional CIS, carrying audio data in both
+directions in the same CIG. We’ll see exactly how that works in the next chapter.
+
+Figure 3.5 A bidirectional CIS in a CIG
+
+Both of these options are allowed; it is up to the application to decide which is the most
+appropriate for its use. In most cases, the optimization of bidirectionality would be the option
+of choice, as it saves airtime.
+
+55
+
+Section 3.4 - Context Types
+
+Figure 3.6 A unidirectional and bidirectional CIS for two Acceptors, e.g., a phone supporting a telephone call with two
+hearing aids.
+
+Figure 3.6 shows a typical application, with a CIG containing two CISes, which might connect
+a phone to a pair of earbuds, but where the return microphone is only implemented in one of
+the earbuds. It shows CIS 0, which is a bidirectional CIS carrying audio from a phone
+conversation out to the earbud, and audio from the earbud’s microphone back to the phone.
+CIS 1 carries the audio stream from the phone to the other earbud. It is up to the application
+to determine whether it is a stereo stream that's being sent or mono. In the latter case, the
+same mono audio data is sent separately over CIS 0 and CIS 1 to the two earbuds.
+
+3.4
+
+Context Types
+
+To make decisions about which audio streams they want to connect to, devices need to know
+more about what those streams contain or what they’re for. Context Types have been
+introduced to describe the current use case or use cases associated with an audio stream. Their
+values are defined in the Bluetooth Assigned Numbers document for Generic Audio. Context
+Types can be used by both Initiators and Acceptors to indicate what type of activity or
+connection they want to participate in, and are applicable to both unicast and broadcast.
+With Bluetooth Classic Audio profiles, the conversation between a Central and Peripheral
+device is basically “I want to make an audio connection”, with no more information about
+what it is. As the HFP and A2DP profiles are essentially single purpose profiles, that’s not a
+problem, but in Bluetooth LE Audio, where the audio stream could be used for a ringtone,
+voice recognition, playing music, providing satnav instructions, or a host of other applications,
+it’s useful to know a little more about the intended reason for requesting a stream. That’s
+where Context Types come in.
+As we’ll see later, Context Types are used as optional metadata during the unicast stream
+configuration process. An Acceptor can expose which Context Types it is prepared to accept
+56
+
+Chapter 3 - New concepts in Bluetooth® LE Audio
+at any point in time. For example, if a hearing aid wearer is having a private (non-Bluetooth)
+conversation which they don’t want interrupted, they can set their hearing aids to be
+unavailable for a stream associated with a «Ringtone» Context Type. That means that the
+hearing aids will silently reject an incoming call. That’s more universal than putting their phone
+on silent, as by using Context Types they will reject an incoming call from any phone they
+have connected, or a VoIP call from any other connected device. They could also set the
+«Ringtone» Context Type while they are in a call, to prevent any other call interrupting the
+current one. This can be done on a per-device basis, meaning you could restrict incoming
+calls to one specific phone, providing a powerful method for an Acceptor to control which
+devices can request an audio stream.
+An Initiator uses the Context Type when it is attempting to establish an audio stream,
+informing the Acceptor of the associated use case. If that Context Type is set to be unavailable
+on the Acceptor, the configuration and stream establishment process is terminated. This
+happens on the ACL link before an Isochronous Stream is set up, which means that these
+decisions can take place in parallel to an existing audio stream without disturbing it, regardless
+of whether the request is from the device currently providing the stream, or another Initiator
+wanting to establish or replace an existing stream. It doesn’t matter whether an existing audio
+stream is unicast or broadcast, which means that a hearing aid user listening to their TV using
+a broadcast Audio Stream can use Context Types to prevent their current stream being
+disturbed by any phone call.
+There are currently twelve Context Types defined in the Generic Audio Assigned Numbers
+document, which are exposed in a two-octet bitfield of Context Types, as shown in Table 3.1,
+with each bit representing a Context Type. This allows multiple values to be used at any one
+time.
+Bit
+
+Context Type
+
+0
+
+Unspecified
+
+1
+
+Conversational
+
+2
+
+Media
+
+3
+
+Game
+
+4
+
+Instructional
+
+Description
+Any type of audio use case which is not explicitly supported by
+another Context Type on a device.
+Conversation between humans, typically voice calls, which can
+be of any form, e.g., landline, cellular, VoIP, PTT, etc.
+Audio content. Typically, this is one way, such as radio, TV or
+music playback. It is the same type of content as is handled by
+A2DP.
+Audio associated with gaming, which may be a mix of sound
+effects, music, and conversation, normally with low latency
+demands.
+Instructional information, such as satnav directions,
+announcements or user guidance, which often have a higher
+priority than other use cases
+57
+
+Section 3.4 - Context Types
+Bit
+
+Context Type
+
+5
+
+Voice
+assistants
+
+6
+
+Live
+
+7
+
+Sound effects
+
+8
+
+Notifications
+
+9
+
+Ringtone
+
+10
+
+Alerts
+
+11
+
+Emergency
+alarm
+RFU
+
+12 15
+
+Description
+Man-machine communication and voice recognition, other than
+what is covered by instructional. It is implied that this is in the
+form of speech.
+Live audio, where both the Bluetooth Audio Stream and the
+ambient sound is likely to be perceived at the same time,
+implying latency constraints.
+Sounds such as keyboard clicks, touch feedback and other
+application specific sounds.
+Attention seeking sounds, such as announcing the arrival of a
+message.
+Notification of an incoming call in the form of an inband audio
+stream. The Ringtone Context Type is not applied to an out of
+band ringtone, which is signalled using CCP and TBS.
+Machine generated notifications of events. These may range
+from critical battery alerts, doorbells, stopwatch and clock alarms
+to an alert about the completion of a cycle from a kitchen
+appliance or white goods.
+A high priority alarm, such as a smoke or fire alarm.
+Not yet allocated.
+
+Table 3.1 Currently defined Context Types
+
+Most of these are obvious, but two of the Context Types are worth special attention:
+«Ringtone» and «Unspecified»14.
+The «Ringtone» Context Type is used to announce an incoming phone call, but only where
+there is an inband15 ringtone which requires an audio stream to be set up. If the user was
+already receiving an audio stream from a different device to the one with the incoming phone
+call, this could be problematic. To signal the incoming call to the user without dropping the
+current audio stream would require at least one of the earbuds to set up a separate unicast
+stream from the second Initiator. In practice, many Acceptors are unlikely to have the
+resources to support concurrent streams from different Initiators at the same time. The
+Acceptor could drop the current audio stream to switch to the inband ringtone, but if they
+then reject the call, they’d need to restore the original Audio Stream. In most cases, a better
+
+When Context Types are used in text, they are enclosed in guillemets, i.e. « and ».
+An inband ringtone is one where the sound is carried in an Audio Stream from the phone, so you
+can hear your customised ringtone or message. In contrast, an out of band ringtone is a sound
+generated locally in the Acceptor, so only needs a control signal, not the presence of an Audio Stream.
+14
+15
+
+58
+
+Chapter 3 - New concepts in Bluetooth® LE Audio
+user experience is likely to result from providing an out of band ring tone, which is generated
+by the earbuds using CCP and TBS. The user can hear this mixed into their existing audio
+stream and decide whether to accept or reject the call. If they reject the call, they can continue
+listening to their original stream. In most cases, «Ringtone» is best used to manage whether
+devices are allowed to interrupt the current audio application with incoming phone calls. We
+will cover how out of band ringtones are handled in Chapter 9.
+The «Unspecified» Context Type is a catch-all category. Every Acceptor has to support the
+«Unspecified» Context Type, but doesn’t need to make it available. When an Acceptor does
+set the «Unspecified» Context Type as available, it is saying that it will accept any Context Type
+other than ones it has specifically said it will not support. As implementers get used to Context
+Types, some will use this to allow the Central device (typically a phone) to be in charge of the
+audio use case in much the same way it is for A2DP and HFP. An Acceptor that just sets
+«Unspecified» as available is effectively allowing the phone to take full control of what it is
+sent. We’ll discuss the concepts of Supported and Available in more detail in Chapter 7.
+All of the Context Types are independent of each other. The use of any one of them does
+not imply or require that another one needs to be supported, unless that requirement is
+imposed by a top level profile. As an example, support for the «Ringtone» Context Type does
+not generally imply or require support for the «Conversational» Context Type, as «Ringtone»
+could be used by itself for an extension bell to alert someone with hearing loss of an incoming
+phone call on a land-line phone.
+Context Types are used by both Initiators and Acceptors to provide information about the
+use case intended for a stream, allowing them to make a decision about whether to accept a
+stream. To accomplish this, they are used in a range of characteristics and LTV16 metadata
+structures. You’ll find them used in this way in the following places:
+•
+•
+•
+•
+
+Supported_Audio_Contexts characteristic [PACS 3.6]
+Available_Audio_Contexts characteristic [PACS 3.5] (also used in a Server
+announcement – [BAP 3.5.3])
+Preferred Audio Contexts LTV structure (metadata in a PAC record) (see also [BAP
+4.3.3])
+Streaming Audio Contexts LTV structure (metadata used by an Initiator to label an
+audio stream) [BAP 5.6.3, 5.6.4, 3.7.2.2]
+
+An Audio Stream can be associated with more than one Context Type, although the intention
+is that the Context Type value represents the current use case. The Streaming Audio Contexts
+metadata has procedures to allow a device to update the bitfield values as the use case changes.
+
+LTVs are triplet structures, consisting of a statement of the Length of the structure, its Type and
+the parameter Values in that order.
+16
+
+59
+
+Section 3.5 - Availability
+This is typically used where an established stream is used for multiple use cases. An example
+would be where an Audio Source mixes in audio from different applications, such as a satnav
+message that could interrupt music. In this case, it is efficient to continue to use the same
+stream, assuming its QoS parameters are suitable, and just update the current Context Type
+value.
+
+3.5
+
+Availability
+
+Bluetooth LE Audio supports far more possibilities and combinations than Bluetooth Classic
+Audio. To enable devices to make informed choices about what they’re doing, they not only
+need the ability to tell each other about the use case they’re engaging in (which is the reason
+for the Context Types), but also which ones they’re interested in participating in in the future.
+This is where Availability comes in.
+As we’ve seen above, Context Types are used by Acceptors to signal whether they can take
+part in a use case. That’s accomplished in two ways. An Acceptor uses the
+Available_Audio_Contexts characteristic defined in PACS to state which of its Supported
+Audio Context Types can currently be used to establish an Audio Stream. Audio Sources,
+whether Initiators or Acceptors, also use the Streaming_Audio_Contexts LTV structure in the
+metadata of their codec configurations, to inform an Audio Sink of the use case(s) that are
+associated with an Audio Stream.
+Bluetooth LE Audio devices wanting to establish unicast Audio Streams can also use Context
+Types to signal their availability before they even start an Audio Stream by including the
+Streaming_Audio_Contexts LTV structure in their advertising PDUs. In a similar way,
+Initiators, acting as Broadcasters, include this in the metadata section of their Periodic
+Advertisements, so that Broadcast Sinks and Broadcast Assistants can filter out the use cases
+they want from those they are not interested in receiving.
+For Broadcasters, you should note that if the Streaming_Audio_Contexts field is not present
+in a Codec ID’s metadata (which we’ll get to in Chapter 4), it will be interpreted as meaning
+that the only Supported_Audio_Contexts value is «Unspecified», so every Broadcast Sink can
+synchronise to it, unless they have specifically set «Unspecified» as non-available.
+
+3.6
+
+Audio Location
+
+With every previous Bluetooth audio specification there was a single Bluetooth LE Audio
+source and a single Bluetooth LE Audio sink. The audio stream was sent as mono or stereo
+and it was left to the audio sink to interpret how it was rendered. Bluetooth LE Audio is
+intended to address applications with multiple speakers or earbuds. It has the ability to
+optimise airtime for each Audio Sink, by only sending it the audio stream it needs. In general,
+for a pair of earbuds, the left earbud only receives the left audio stream, and the right earbud
+only receives the right audio stream. That means that each Audio Sink can minimise the time
+that its receiver is on, thereby reducing its power consumption.
+60
+
+Chapter 3 - New concepts in Bluetooth® LE Audio
+To accomplish this, devices need to know what spatial information they are meant to receive,
+for example, a left or a right stream from a stereo input. They do this by specifying an Audio
+Location. These are defined in the Bluetooth Generic Audio Assigned Numbers and follow
+the categorization of CTA-861-G’s Table 34 codes17 for speaker placement. The values are
+expressed as bits in a four-octet wide bitfield. The most common Audio Locations are shown
+in Table 3.1.
+Audio Location
+
+Value (bitmap)
+
+Front Left
+Front Right
+Front Centre
+Low Frequency Effects 1 (Front Woofer)
+Back Left
+Back Right
+Prohibited
+
+0x00000001 (bit 1)
+0x00000002 (bit 2)
+0x00000004 (bit 3)
+0x00000008 (bit 4)
+0x00000010 (bit 5)
+0x00000020 (bit 6)
+0x00000000
+
+Table 3.2 Common Audio Location values
+
+Every Acceptor which can receive an audio stream must set at least one Audio Location –
+leaving the entire bitfield blank is not allowed. An Audio Location is normally set at
+manufacture, but in some cases it may be changeable by the user – for instance, a speaker
+could have an application or a physical switch to set it to Front Left or Front Right.
+Note that mono is not a location, as mono is a property of the stream, not the physical
+rendering device. A single channel speaker would normally set both the Front Left and Front
+Right audio locations. An Initiator would determine the number of streams an Acceptor
+supports, along with the number of speakers available and make a decision of whether to send
+a downmixed stereo stream (i.e., mono), or a stereo stream that it assumes the speaker could
+downmix to mono. Broadcasters would denote a mono stream by labelling it as Front Left
+and Front Right. We’ll look further at exactly how Audio Locations are used in Chapter 5,
+but it brings us on to Channel Allocation.
+
+3.7
+
+Channel Allocation (multiplexing)
+
+You always knew what was being transported in the Bluetooth Classic Audio profiles. HFP
+carries a mono audio stream, and A2DP uses the four channel modes of SBC codec to transmit
+mono, dual channel, stereo or joint stereo encoded streams. Bluetooth LE Audio is a lot more
+flexible, allowing a CIS or BIS to contain one or more channels multiplexed into a single
+packet, limited only by the available bandwidth.
+
+17
+
+https://archive.org/stream/CTA-861-G/CTA-861-G_djvu.txt
+
+61
+
+Section 3.7 - Channel Allocation (multiplexing)
+The reason for this approach is that the LC3, which is the mandatory codec for all Bluetooth
+LE Audio implementations is a single channel codec. This means that it encodes each audio
+channel separately into discrete frames of a fixed length. With SBC, joint stereo coding, which
+combines both left and right audio channels into a single encoded stream was popular because
+it was more efficient than two separate left and right channels. That was due to its ability to
+encode differences between the two input audio channels. The more efficient design of LC3
+means that there is very little advantage in joint stereo coding compared to encoding each
+channel separately and then concatenating the individual encoded frames. This approach
+allows more than two channels to be encoded and grouped together.
+However, this meant that a mechanism needed to be introduced to package together multiple
+encoded frames for multiple channels, which is performed using Channel Allocation [BAP
+Section 4.2].
+
+Figure 3.7 Example of multiplexing multiple Bluetooth® LE Audio Channels
+
+Figure 3.7 shows an example of how this works for a five-channel surround-sound system.
+Five audio input channels are separately encoded using LC3 and the encoded frames are then
+arranged into a media packet which is transmitted as a single isochronous PDU. The LC3
+codec frames are always arranged in ascending order of the Published Audio Capability Audio
+Location associated with each audio channel, using the Assigned Numbers which were
+summarised in Table 3.2. So, in this case, they will be ordered as Front Left (0x0000000001),
+Front Right (0x0000000002), Front Centre (0x0000000004), Back Left (0x0000000010) and
+Back Right (0x0000000020).
+The Media Packet contains only encoded audio frames. It can be expanded to include multiple
+blocks of encoded audio frames, each of which include one frame for each of the Audio
+Locations, as shown in Figure 3.8. CF_N1 refers to frames from the first sample of each of
+the incoming audio channels; CF_N2 to those from the next samplings of those audio
+channels.
+
+62
+
+Chapter 3 - New concepts in Bluetooth® LE Audio
+
+Figure 3.8 Media Packet containing two blocks of five Audio Channels
+
+Using blocks may look efficient, but it comes with some important caveats. It results in larger
+packets, which are more vulnerable to interference. It also increases the latency, as the
+Controller needs to wait for multiple frames to be sampled before it can start transmission. If
+multiple blocks are added to the multiple Isochronous Channel features of Burst Number or
+Pre-Transmission Offset, which we’ll come to in the next chapter, it very quickly results in
+latencies that can be hundreds of milliseconds. There are some occasions where that may be
+useful, but not in the general audio applications we use today.
+There is no header information associated with the media packet. Instead, the number of
+Audio Channels that can be supported are specified in the Audio_Channel_Location LTV
+which is included in the Codec Specific Capabilities LTV in the PAC records, with the value
+for each Isochronous Stream being set by the Initiator during the stream configuration
+process. The number of blocks being used is configured at the same time using the
+Codec_Frame_Blocks_Per_SDU LTV structure.
+We’ll look at these in more detail in Chapter 5 when we discuss the LC3 and Quality of Service.
+
+3.8
+
+Call Control ID - CCID
+
+A consequence of separating the control and data planes in Bluetooth LE Audio is that there
+is no longer any direct relationship between a content control signal, such as phone control or
+media control, and the audio stream. That adds flexibility to Bluetooth LE Audio, but results
+in a little more complexity. An Initiator might have multiple applications that are running
+concurrently, where a user may want to associate control with a specific one of those
+applications. Think of the case of two separate telephony calls, such as a cellular call and a
+concurrent Teams meeting, where the user may want to put one call on hold, or terminate
+one, whilst retaining the other. To cope with this situation, the Content Control ID has been
+introduced, which associates a Content Control service instance with a specific unicast or
+broadcast stream.
+The statement that CCIDs can be used for broadcast might seem contradictory, but highlights
+the fact that although broadcast streams don’t need an ACL connection, there are many
+applications where one might be present. For example, if you transition from using unicast to
+listen to a music stream on your phone to broadcast, so that you can share it with your friends,
+you would still expect to be able to control the media player. In this case you would keep the
+ACL link alive and associate the media controls with the broadcast stream.
+63
+
+Section 3.9 - Coordinated Sets
+The Content Control ID characteristic is defined in Section 3.45 of the GATT Specification
+Supplement18 as having a value that uniquely identifies an instance of a service that either
+controls or provides status information on an audio-related feature. It is a single octet integer
+which provides a unique identifier across all instances of Content Control services on a device.
+When an Audio Stream contains content which is controlled by a content control service, it
+includes the CCID in a list of such services in the Audio Stream’s metadata to tell both
+Acceptors and Commanders where they can find the correct service. We’ll look at
+Commanders in a moment in Section 3.12.
+CCIDs are currently only used for the Telephone Bearer Service and the Media Control
+Service. They do not apply to rendering or capture control.
+
+3.9
+
+Coordinated Sets
+
+Despite the fact that we’ve only had them for a few years, we’re already so familiar with the
+concept of TWS earbuds, that most people forget that the Bluetooth Classic Audio
+specifications don’t cover the way they work. As explained in Chapter 1, they all rely on
+proprietary extensions from silicon chip companies. The design of Isochronous Channels
+rectifies that, allowing an Initiator to send separate Audio Streams to multiple Acceptors, along
+with a common reference point at which all Acceptors know they have received the audio data
+and can start decoding it. However, the flexibility of Bluetooth LE Audio, including the new
+use cases coming from broadcast, raised the need for a way to link Acceptors together as sets
+of devices, which can be treated as a single entity.
+The concept of a Coordinated Set addresses that need. It allows devices to expose the fact
+that they are part of a group of devices which together support a common use case and should
+be acted on as an entity. Whilst most people will immediately think of a pair of earbuds or
+hearing aids, that set could equally be a pair of speakers or a set of surround-sound speakers.
+A Coordinated Set of earbuds should be represented as a single device, so that anything that
+happens to one, happens to the other, although how that happens is down to the
+implementation. When you adjust the volume for a pair of earbuds, the volume of both should
+change at the same time19, and if you decide to listen to a different Audio Source, that change
+should occur simultaneously on both left and right earbuds. You do not want a user
+experience where your left earbud is listening to your TV, whilst the right earbud is streaming
+music from your phone.
+Coordination is handled by the Coordinated Set Identification Profile and Service (CSIP and
+CSIS), which are referenced from within CAP. The main feature of CSIS and CSIP, aside
+
+This is a document that describes generic features used within Bluetooth Low Energy.
+The Volume Control profile has the flexibility to allow these to be changed independently if the
+user prefers.
+18
+19
+
+64
+
+Chapter 3 - New concepts in Bluetooth® LE Audio
+from identifying devices as members of a Coordinated Set, is to provide a Lock function. This
+ensures that when an Initiator interacts with one of the members, the others can be locked,
+preventing any other Initiator from interacting with other members of that set.
+The need for such a Lock is that ear-worn devices, such as hearing aids and earbuds, have a
+problem with communicating directly with each other using Bluetooth technology. That’s
+because the human head is remarkably good at attenuating 2.4GHz signals. If an earbud is
+small, which means that its antenna will also be small, it is unlikely that a transmission from a
+device in one ear would be received by its partner in the other ear. Many current earbuds get
+around this limitation by including a different, lower frequency radio within the earbud for
+ear-to-ear communication, which is not significantly attenuated by the head, typically using
+Near Field Magnetic Induction (NFMI). This second radio adds cost and takes up space, but
+removing it and relying on the 2.4GHz Bluetooth link between the earbuds raises the risk that
+different Initiators could send conflicting commands to left and right earbuds.
+With CSIP and CSIS, if you accept a phone call on your left earbud, the Initiator would set
+the Lock on both earbuds, and transition your right earbud to the same stream before releasing
+the Lock. The Lock feature allows these interactions to be managed without the need for the
+members of the Coordinated Set to talk to each other.
+If members of a Coordinated Set do have another radio connection which can penetrate the
+head, the Hearing Access Service has a feature which will signal that this radio can be used to
+convey information to the other hearing aid. At the moment this feature is limited to
+information on preset settings, but may be used for other features in the future.
+Generally, members of a Coordinated Set are configured at manufacture and shipped as a pair,
+but membership can be set to be written as well as read, allowing for later configuration, or
+the replacement of faulty or lost units.
+
+3.10
+
+Presentation Delay and serialisation of audio data
+
+Supporting two earbuds brings us to another issue that Bluetooth LE Audio had to solve,
+which is ensuring that the sound at both the left and the right ear is rendered at exactly the
+same time. In the past, audio data was sent to a single device, which knew how to extract the
+left and right signals and render them at the same time. With Bluetooth LE Audio, CISes send
+data to different destinations serially, using different transmission slots. Although the
+incoming Audio Channels present data to the Initiator at the same point in time, the encoded
+packets arrive at the Acceptors one after the other. They may be further delayed by
+retransmissions. Figure 3.9 illustrates this by adding examples of left and right packets going
+to two acceptors onto the CIG diagram of Figure 3.6
+
+65
+
+Section 3.10 - Presentation Delay and serialisation of audio data
+
+Figure 3.9 The serial transmission of audio data
+
+This serialisation causes a problem. The human head is remarkably good at detecting a
+difference in the arrival time of sound between your left and right ears, and uses that difference
+to estimate the direction from which the sound is coming.
+
+Figure 3.10 Effect of head rotation on sound arrival
+
+66
+
+Chapter 3 - New concepts in Bluetooth® LE Audio
+Figure 3.10 illustrates that if you're two metres away from an audio source, and you rotate your
+head by just 10 degrees, that equates to just over a 70 µs difference in the arrival time of the
+sound. If there's a variation in the rendering time between your left and right ear, your brain
+interprets this as the sound source moving. If that difference is much more than 25
+microseconds and changes regularly, you start to get an unpleasant effect, where it feels as if
+the sound is moving around within your head. To prevent that, it's important to have a
+synchronisation technique to ensure that the left and right earbuds always render their
+respective audio data at exactly the same time.
+We can’t rely on any Bluetooth communication between two earbuds. As we’ve said before,
+the human head is very efficient at attenuating a 2.4GHz signal, as it contains a lot of water.
+If you have small earbuds, which fit neatly in the ear canal, there is no guarantee that they will
+be able to communicate with each other.
+There are two parts to the Bluetooth LE Audio solution. First, both earbuds need to know a
+common synchronisation point, which is the point in time at which every Acceptor can
+guarantee that every other Acceptor has had every possible chance to receive a transmitted
+packet, whether it’s a unicast or a broadcast stream. This point in time has to be provided by
+the Initiator, as it is the only device which knows how many attempts it will take to send
+packets to all of the Acceptors. (Remember that the Acceptors generally can’t talk to each
+other and are probably unaware of each other’s existence.)
+In most cases, the Acceptors will have received their audio data packets earlier than that
+common synchronisation point, as in audio applications data packets are scheduled to be
+retransmitted multiple times to maximise the chance of reception. That is because of the
+problem of drop-outs in an audio stream, as a result of missing packets. These are particularly
+annoying artefacts for the listener, so retransmissions are used to help improve the robustness
+of the signal. This means that every Acceptor needs to include enough buffering to store
+packets from the earliest possible arrival time – the time when their packet is first transmitted,
+until the common Synchronisation Point. We’ll learn more about the Synchronisation Point
+in Chapter 4.
+However, you can’t render the audio at the Synchronisation Point, as it’s still encoded.
+Between the Synchronisation Point and the final rendering point the data needs to be decoded,
+and any additional audio processing, such as Packet Loss Concealment20 (PLC), active noise
+cancellation (ANC), or hearing aid audio adjustments performed, before it can finally be
+rendered.
+
+Packet Loss Concealment is a technique that attempts to recreate missing or corrupted packets of
+audio data, generally based on what the previous packets contained. It is an attempt to avoid audible
+artefacts when an audio stream is disrupted.
+20
+
+67
+
+Section 3.10 - Presentation Delay and serialisation of audio data
+The time needed for that may vary between different Acceptors. Whilst you expect a pair of
+hearing aids, speakers or earbuds supplied as a pair from one manufacturer to be designed to
+have the same processing time for each earbud, it could be different if the Acceptors come
+from different manufacturers, or even if a firmware update is applied to one, but not the other.
+For this reason, the Bluetooth LE Audio specification includes the concept of Presentation
+Delay. Presentation Delay is an Initiator defined value which specifies the time in
+microseconds after the Synchronisation Point where the audio is to be rendered in every
+Acceptor. This is illustrated in Figure 3.11. The “C>P” suffix for the SDU Synchronisation
+Point refers to the Central to Peripheral direction (Initiator to Acceptor). LL refers to the
+Link Layer in the Controller, which is where the data being sent over the air is received.
+
+Figure 3.11 Presentation Delay for rendering on an Acceptor
+
+The Presentation Delay for rendering may also include a Host application dependent element
+to deliberately increase the time until rendering. This is a commonly used technique for audio
+streams linked to video, where it can be used to delay the rendering point to compensate for
+lip-synch issues. With public broadcast applications, where there may be multiple broadcast
+transmitters covering a large auditorium or stadium, Presentation Delay may also be used to
+set specific delays to compensate for differing distances between the broadcast transmitters
+serving each audience group and the audio source. Sound travels 343m every second, so it can
+take half a second for sound to propagate across a large stadium. For this reason, venues often
+apply delays to speakers to help synchronise the sound in different sections of the venue. The
+same effect can be achieved using Presentation Delay with multiple Bluetooth LE Audio
+broadcast transmitters to bring the audience closer to the origin of the sound.
+Every Acceptor contains values for the minimum and maximum Presentation Delay it can
+support. The minimum represents the shortest time in which it can decode the received codec
+packets and perform any audio processing before it renders the sound; the maximum reflects
+the longest amount of buffering it can add to that. These are read by an Initiator during
+configuration, which must respect the maximum and minimum values for all Acceptors within
+each stream it is transmitting. That means it cannot set a value greater than the lowest
+68
+
+Chapter 3 - New concepts in Bluetooth® LE Audio
+Presentation_Delay_Max value of any of the Acceptors, or lower than the highest value of
+Presentation_Delay_Min of any of them. Acceptors may also expose their preferred value for
+Presentation Delay, which an Initiator should attempt to use, unless an application on the
+Initiator sets a specific value, as it might for live applications or to compensate for lip-synch.
+It is not expected that Presentation Delay would be exposed to device users receiving the
+audio, as it is always set by the application on the Initiator
+Presentation Delay was designed to provide a common rendering point for multiple
+Acceptors. It supports the situation where the Acceptors might not be aware of each other’s
+existence, such as a user with hearing loss in one ear, who would wear a hearing aid and a
+single earbud. As the same Presentation Delay is applied to both, both devices would render
+at the same time. In most applications, the value for Presentation Delay should be as small as
+possible. Supporting higher values of Presentation Delay increases the burden on the
+Acceptor’s resources, as it needs to buffer the decoded audio stream for longer.
+For Broadcast, when there is no connection between and Initiator and an Acceptor, the
+Initiator needs to make a judgement of what value of Presentation Delay will be acceptable to
+all potential Broadcast Sinks, based solely on its application. In general, values above 40ms
+(which every Acceptor must support) should be avoided as they may introduce echo if the
+ambient sound is also present, unless they are being used specifically to accommodate this in
+very large venues. The specifications do not define what a Broadcast Sink should do if the
+Presentation Delay falls outside the range it can support.
+For both unicast and broadcast, a top level profile may specify a specific value for Presentation
+Delay, particularly if they are supporting low latency applications. For example, where an
+Audio Stream also has ambient sound present, they may require that the «Live» Context Type
+is used, along with a Presentation Delay setting of not more than 20ms (for HAP or TMAP
+support).
+Presentation Delay is also applied to audio capture, where audio data is travelling from an
+Acceptor to an Initiator (denoted in the specs as P>C, or Peripheral to Central). Here, it
+represents the time from the point that audio is captured, then subsequently processed,
+sampled and encoded, to the reference point where the first packet of the first Isochronous
+Stream could be transmitted, as shown in Figure 3.12.
+
+69
+
+Section 3.10 - Presentation Delay and serialisation of audio data
+
+Figure 3.12 Presentation Delay for audio capture
+
+Using Presentation Delay in audio capture ensures that every microphone or other audio
+transducer captures the sound at exactly the same point in time. It defines an interval during
+which every Acceptor can prepare its audio packets for transmission, with the first
+transmission occurring at the end of the Presentation Delay. All other audio data sources will
+then transmit their encoded packets in their allocated transmission slots. This is desirable
+when a phone wants to combine the microphone data from left and right earbuds, as it can
+use the knowledge of the common capture times to combine them. Rather than simple mixing
+based on capture time, an Initiator will normally use this knowledge to inform digital stitching
+techniques to align the multiple streams before running noise cancellation algorithms, but that
+is outside the Bluetooth specification.
+Whilst the most common use of Presentation Delay for capture in unicast audio sources is the
+case of one microphone in each earbud or hearing aid, it is equally valid for single devices,
+which have multiple microphones. That includes stereo microphones and stereo headsets
+with a microphone in each can. In these devices an implementation could encode the two
+microphone signals into a single codec frame using channel allocation for multiplexing (see
+Section 3.7 below), or transmit them separately using Presentation Delay to ensure that the
+capture is aligned. In both cases, a value for Presentation Delay needs to be set to cover the
+audio processing and encoding time.
+Where microphones are spaced further apart, the received data will not reflect the difference
+in audio path between their positions. The Initiator may need to adjust the incoming streams
+if it wants to restore that information.
+It is important to understand that Presentation Delay is only applied at the Acceptor, regardless
+of whether it is acting as an Audio Sink or an Audio Source. In many cases an Acceptor will
+be acting as both. Typically, the values of Presentation Delay will be different for each
+direction to cope with the fact that encoding audio data takes longer than decoding it. We will
+look in more detail at how Presentation Delay is used to influence latency and robustness in
+Chapter 4.
+
+70
+
+Chapter 3 - New concepts in Bluetooth® LE Audio
+
+3.11
+
+Announcements
+
+As part of the philosophy of giving more autonomy to an Acceptor, the Bluetooth LE Audio
+Specifications allow them to transmit Announcements, informing Initiators that they are
+available to receive or transmit audio data, by using a Service Data AD Type21. An Acceptor
+can decide whether to use a Targeted Announcement, where it is connectable and requesting
+a connection, or a General Announcement, where it is simply saying that it is available, but
+not initiating a specific connection. The LTV structure used in the AD Type field of an
+Announcement is shown below in Table 3.3.
+Field
+
+Size
+Description
+(Octets)
+
+Value
+
+Length
+Type
+ASCS UUID22
+Announcement Type
+
+1
+1
+2
+1
+
+Available_Audio_Contexts
+
+4
+
+Metadata Length
+
+1
+
+Metadata
+
+varies
+
+Length of Type and Value fields
+Service Data UUID (16 bit)
+0x0184E
+0x00 = General Announcement
+0x01 = Targeted Announcement
+Available_Audio_Contexts characteristic
+(from ASCS)
+≥ 1 if there is additional metadata,
+otherwise 0
+Metadata in LTV format
+
+Table 3.3 AD Values for Targeted and General Announcements
+
+The Common Audio Profile (CAP) describes how Initiators or Commanders (see Section 3.12
+below) react to Announcements by defining two specific modes, which helps explain when to
+use each. These are:
+•
+•
+
+INAP - the Immediate Need for Audio related Peripheral mode (INAP), and
+RAP – the Ready for Audio related Peripheral mode (RAP).
+
+INAP refers to the case where an Initiator or Commander wants to make a connection, usually
+due to a user action, and needs to determine which Acceptors are available to connect to. If
+it discovers an available Acceptor, it should connect, or, if it has discovered more than one,
+present the user with a range of available Acceptors to let them make that choice. The Initiator
+or Commander would normally commence scanning at a higher rate to find Acceptors when
+
+AD Data Types are structures used in Bluetooth advertisements to provide information about a
+device or its capabilities. They are defined the Core Specification Supplement (CSS).
+22 A UUID is a universally unique identifier, defined in the Bluetooth 16-bit UUIDs Assigned
+Numbers document, and used to identify a particular service or characteristic.
+21
+
+71
+
+Section 3.11 - Announcements
+operating in the INAP mode.
+In contrast, when in RAP mode, the Initiator scans at a lower rate, but will respond to a
+Targeted Announcement from an Acceptor by connecting.
+Acceptors should only use Targeted Announcements for a limited period of time, when they
+require an immediate connection.
+There is a further subtlety in Announcements, which is that they may or may not include a
+Context Type value. If they are used in relation to an Audio Stream, they should include it,
+and are called BAP Announcements [BAP 3.5.3], which contains an Available Audio Context
+field. If they are used for control purposes or Scan Delegation, they do not. They are then
+called CAP Announcements [CAP 8.1.1]
+Broadcasters use Announcements within their Extended Advertisements to inform Broadcast
+receivers and Broadcast Assistants (see Section 3.12) that they have a broadcast stream
+available. These take two forms:
+
+3.11.1
+
+Broadcast Audio Announcements
+
+Broadcast Audio Announcements inform any scanning device that a Broadcaster is
+transmitting a group of one or more broadcast Audio Streams. The LTV structure used for
+Broadcast Audio Announcements is shown below in Table 3.4.
+Field
+
+Size
+Description
+(Octets)
+
+Value
+
+Length
+Type
+Broadcast Audio
+Announcement Service UUID
+Broadcast_ID
+Supplementary Announcement
+Service UUIDs
+
+1
+1
+2
+
+Length of Type and Value fields
+Service Data UUID (16 bit)
+0x01852
+
+3
+
+A random ID, fixed for the life of the
+Broadcast Isochronous Group
+Optional UUIDs defined by top level
+profiles
+
+varies
+
+Table 3.4 AD Values for Targeted and General Announcements
+
+3.11.2
+
+Basic Audio Announcement
+
+The confusingly similarly titled Basic Audio Announcement is used in the Periodic Advertising
+train by a Broadcast Source to expose the broadcast Audio Stream parameters through the
+Broadcast Audio Stream Endpoint structure (BASE). Its use is described in Chapter 8.
+
+72
+
+Chapter 3 - New concepts in Bluetooth® LE Audio
+
+3.12
+
+Remote controls (Commanders)
+
+Bluetooth has always been a good candidate for remote control devices, but these are almost
+all simple remote controls, which are essentially a Bluetooth replacement for traditional infrared remote controllers. Hearing aids and earbuds make remote control an attractive option,
+as these devices are so small that they don’t have room for many buttons, and even where they
+do, manipulating a button that you can’t see (because it’s on the side of your head or behind
+your ear) is a poor user experience. As most earbuds are used with phones, laptops or PCs,
+and the application generating the audio stream generally contains the controls that you use to
+pause it, answer a call or change volume, that’s not a major issue. However, hearing aid users
+also need to control the volume of their hearing aids when they’re just being used to amplify
+ambient sound.
+To make it easier than fumbling for a button on the hearing aid, the industry has developed
+simple, keyfob-like remote controls that allow a user to easily adjust the volume or change the
+audio processing algorithms to suit their environments (these are known as preset settings).
+Even where a hearing aid contains Bluetooth technology, most of the time a user won’t have
+an active Bluetooth link enabled, and getting your phone out of a pocket or bag, followed by
+finding the appropriate app is an inconvenient way to adjust volume. If you are troubled by a
+loud noise, it’s quicker and easier to take your hearing aids out, which is not a good user
+experience.
+That is just the beginning of the problem. The new broadcast capabilities of Bluetooth LE
+Audio will result in far more public infrastructure, where a user will need to navigate through
+multiple different broadcasts and decide which to receive. Not only is that difficult to do on
+a hearing aid or earbud, it involves relatively power-hungry scanning. To address these issues,
+the concept of a separate device was developed, which could provide volume control, discover
+and display available Broadcast Sources, discover keys for encrypted broadcasts, and allow
+hearing aid users to change their presets. It’s also possible to use it to answer calls or control
+a media player. These devices take the Commander role, which is defined in CAP, but can
+also use the Broadcast Assistant role from BAP for discovering broadcasts, as well as being
+Content Control Clients. Most top level profiles introduce further names for additional Roles
+that these devices can assume. For the rest of the document, I’ll use the Commander
+terminology, unless it’s for the specific subset of a Broadcast Assistant.
+The Commander is an important new addition to Bluetooth topology. The role can be
+implemented on a phone, either as a stand-alone application, or part of a telephony or audio
+streaming app. It can also be implemented in any device which has a Bluetooth connection
+to an Acceptor or a Coordinated Sets of Acceptors. That means you can implement it in
+dedicated remote controls, smart watches, wrist bands and even battery cases for hearing aids
+and earbuds. Commanders can have a display to show textual information about broadcasts,
+to help you select them, or just volume buttons. Commanders work on a first-come, firstserved basis, so you can have multiple Commanders, allowing you to use whichever comes to
+73
+
+Section 3.12 - Remote controls (Commanders)
+hand first when you want to change volume or mute your hearing aids. Because all of the
+functions are explicitly specified, it also means that multiple devices can implement them
+interoperably. In Chapter 12, we’ll look at some of the ways in which they are likely to change
+the way we use Bluetooth audio devices.
+-oOoHaving covered these new concepts, we can now dive into the specifications to see how they
+work and how they enable new use cases for Bluetooth LE Audio.
+
+74
+
+Chapter 4 - Isochronous Streams
+
+Chapter 4. Isochronous Streams
+If you've worked with Bluetooth® applications in the past, you've probably concentrated on
+the profiles and barely looked at the Core specification. That's possible because the Bluetooth
+Classic Audio profiles use well-defined transport configurations tied in with Core settings, so
+that there is not much need to understand what's happening underneath the profile or its
+associated protocol. With Bluetooth LE Audio profiles, that changes, as you have more
+potential to affect how the Core works than with any of the Bluetooth Classic Audio profiles.
+In order to try and make the most flexible system possible, which would cope not only with
+today's audio requirements, but also the ones we haven't even thought about yet, the
+specifications had to allow a much greater degree of flexibility. To achieve that, a fundamental,
+architectural decision was made to split the audio plane and the control plane. What that
+means is that new isochronous physical channels have been defined, to carry the audio streams.
+These sit alongside, but are separate to the existing ACL links of Bluetooth LE. The
+isochronous physical channels in the Core let you build up a number of isochronous audio
+streams, which are capable of transporting all types of audio, from very low speech quality, up
+to incredibly high music quality. With one exception, which we'll see later on, Isochronous
+Streams contain no control information – they are purely for carrying audio. The
+accompanying ACL channels are used to set up the Isochronous Streams, turn them on, turn
+them off, add volume and media control, along with all of the other features that we need,
+using the standard GATT23 procedures that are part of Bluetooth Low Energy.
+In order to provide the flexibility that is needed for different latencies, different audio quality
+and different levels of robustness, developers need to be able to control the way these
+Isochronous Streams are configured. That’s done quite high up in the profile stack of the
+Generic Audio Framework. It means that when you start working on Bluetooth LE Audio
+applications, even if you’re just using the top level profiles, you still need to know a fair amount
+about how the underlying Isochronous Streams work. That's different from what you would
+have experienced in most previous Bluetooth applications. To help understand the overall
+architecture of Bluetooth LE Audio, we need to look at how those Isochronous Streams were
+developed, what they do, and how to use them.
+
+4.1
+
+Bluetooth LE Audio topologies
+
+Up until now the Bluetooth specification has largely been concerned with peer-to-peer
+connections: a Central24 device makes a connection to a Peripheral device, and they exchange
+
+The Generic Attribute Profile defines the procedures which are used with the Attribute Protocol in
+Bluetooth Low Energy.
+24 From December 2020, the Bluetooth SIG, like many other standards organisations, implemented a
+policy of replacing words which are considered to have negative connotations. That means that the
+specifications no longer use the traditional engineering terminology of Master and Slave when
+23
+
+75
+
+Section 4.1 - Bluetooth LE Audio topologies
+data. It’s a very constrained topology. Various companies have developed proprietary
+extensions to add flexibility, as we’ve seen with True Wireless Stereo earbuds, but Isochronous
+Streams were developed to cope with a much wider range of topologies than even these
+proprietary solutions could provide. As well as connecting a mobile phone to a pair of
+headphones or a single speaker, Bluetooth LE Audio needed the ability to send separate left
+and right signals to a left earbud and a right earbud. It also needed to be able to send the same
+information to more than one set of earbuds and scale up the number of devices and streams.
+There are two types of Isochronous Stream – unicast and broadcast. Unicast connections,
+known as Connected Isochronous Streams (CIS), are the closest to existing Bluetooth audio
+use cases. “Connected” in this sense means that they are transferring audio data between two
+devices, with an acknowledgement scheme between the two devices to provide flow control.
+Connected Isochronous Streams have an ACL control channel that is up and running
+throughout the lifetime of the CIS which is carrying the audio data.
+A very similar structure of Broadcast Isochronous Streams (BIS) is used for broadcast, but
+there’s a major difference. With broadcast, a device that transmits the Isochronous Streams
+has no knowledge of how many devices may be out there receiving the audio. There's no
+connection between devices and no need for an ACL link. At its simplest, broadcast is purely
+promiscuous. However, it is possible to add control links to broadcast. At the Core level,
+there is a clear distinction between Connected and Broadcast Isochronous Streams, which is
+based on whether data is acknowledged or not. However, many applications will switch
+between unicast and broadcast to fulfil different use cases, without the user being aware of
+what is happening. But for the time being, we’ll concentrate on the basics.
+Broadcast allows multiple devices to hear the same thing, in the same way as FM radio or
+broadcast TV. The requirements for Bluetooth LE Audio broadcast capability were initially
+driven by the hearing aid application of telecoils, where multiple people wearing hearing aids
+in a public venue can listen to the same signal. Telecoils are relatively low audio quality, used
+mostly for speech. With the higher quality possible with Bluetooth LE Audio, along with a
+significantly lower installation cost, the industry envisaged a much wider range of applications
+and usage. Traditional telecoil locations like conference centres, theatres and places of
+worship; public information, such as flight announcements, train departures and bus times
+would become accessible to everyone with a headset and earbud, not just people wearing
+hearing aids. Broadcast is also applicable for more personal applications, where a group of
+people can listen to the same TV, or share music from their mobile phones. That last example
+shows how Bluetooth LE Audio applications can invisibly switch the underlying protocols
+back and forth. If you are streaming music from your phone to your earbuds, it’s probably
+
+describing communications, but have replaced them with Central and Peripheral. A full list of these
+changes can be found at www.bluetooth.com/languagemapping/Appropriate-Language-MappingTable.
+
+76
+
+Chapter 4 - Isochronous Streams
+using a Connected Isochronous Stream. When your friends come along and you ask them
+“Do you want to listen to this too?”, your music sharing application will switch your phone
+from a private, unicast connection to an encrypted broadcast connection, so that you can all
+hear the same music, whether that’s on earbuds, hearing aids, headphones or speakers. At the
+application layer, listening to the music and sharing it with any number of other people should
+be seamless – the users don’t need to know about broadcast or unicast. But there will be a lot
+going on underneath during that transition.
+The building blocks for these different use cases are essentially the same. There are some
+differences between the Connected Isochronous Streams that are used for unicast use cases
+and Broadcast Isochronous Streams, which are used for the broadcast use cases, but the
+underlying principles are very similar. We’re now ready to see how they're all put together,
+which means diving into the Core.
+
+4.2
+
+Isochronous Streams and Roles
+
+The Isochronous Streams feature in the Core 5.2 release is a fundamentally new concept within
+Bluetooth Low Energy. If you're familiar with Hands-Free profile or A2DP, you'll know that
+they have quite a constrained topology. HFP has a bidirectional one-to-one link, typically
+between a phone and a headset or Hands-Free device. It has two roles: an Audio Gateway,
+and a Hands-Free device. A2DP is an even simpler unicast link, specifying a Source device
+that generates audio and a Sink device, which can be your headphone, speakers, amplifier or a
+recording device, which receives that audio.
+
+Figure 4.1 Bluetooth Classic Audio topologies
+
+Bluetooth LE Audio is built on the fundamental asymmetry that exists within the Bluetooth
+LE specification, where one device - the Central device, is responsible for setting up and
+controlling the Isochronous Streams. The Central can connect to a number of Peripheral
+devices, which use those Isochronous Streams to send and receive audio data. The asymmetry
+means that the Peripheral devices can be much lower power. For CISes, they have a say in
+how the Isochronous Streams are configured, which will affect the audio quality, latency and
+their battery life, giving them more control over the audio streams than with Bluetooth Classic
+Audio profiles. For BISes, the Central makes all of the decisions, with the Peripheral deciding
+77
+
+Section 4.2 - Isochronous Streams and Roles
+which Broadcast Isochronous Streams it wants to receive.
+Repeating what we said in the terminology overview in Section 3.3, as we move up the stack
+of the Bluetooth LE Audio specifications, we'll come across a number of different names for
+the roles which devices perform. In the Core, they are defined as Central and Peripheral
+devices. In the BAPS set of specifications they’re called Clients and Servers, and in CAP they
+become Initiators and Acceptors. The Initiator role always exists in a Central device which is
+responsible for scheduling the Isochronous Streams. Acceptors are the devices that participate
+in these streams. There is always one Initiator, but there can be multiple Acceptors.
+
+Figure 4.2 Bluetooth® LE Audio roles
+
+When we move up into the top level profiles, there’s an avalanche of new role names, with
+Senders, Broadcasters and Receivers. I'm going to ignore all of those and use the Initiator and
+Acceptor names for most of time (even though they’re technically roles), because I think they
+best explain the way that Bluetooth LE Audio Streams work. When there’s no Audio Stream
+involved, which is the case with the Control specifications, I’ll drop back to using Client and
+Server.
+One thing I’d like to point out at this stage is that other than for broadcast, either device can
+act as an audio source, generating audio data, or as an audio sink, receiving that data. Both
+Initiators and Acceptors can be sources and sinks at the same time and they can each contain
+multiple Bluetooth LE Audio sinks and sources. The concept of who generates the audio and
+who receives and renders it is orthogonal to the concept of the Initiator and the Acceptor.
+The important thing to remember is that the Initiator is the device that is responsible for
+working out the timing of every transmission of audio data that is sent; that task is called the
+scheduling. The Acceptor is the device that accepts those streams. That concept applies both
+in unicast and broadcast. An Acceptor can also generate audio data, such as capturing your
+voice from your headset’s microphone, but the Initiator is responsible for telling it when it
+needs to send that data back.
+As the Initiator role is far more complex than the Acceptor role, Initiators are normally devices
+like phones, TVs and tablets which have bigger batteries and more resources. The scheduling
+has to take account of other demands on the Initiator’s radio, which may include other
+Bluetooth connections, and often Wi-Fi as well. That’s a complex operation which is handled
+by the chip designers. But, as we’ll see later on, the Bluetooth LE Audio profiles give
+applications a fair amount of scope to influence that scheduling, which is why developers need
+to have a clear idea about how Isochronous Streams work.
+78
+
+Chapter 4 - Isochronous Streams
+For unicast Bluetooth LE Audio, we have a lot of flexibility in terms of topologies, which are
+depicted in Figure 4.3. We can replicate the same topologies that we had in HFP or A2DP,
+where we have a single device - typically your phone, and a single Peripheral device, such as
+your headset, with an audio link between them. Moving up from that, Bluetooth technology
+can now support an Initiator that talks to two or more Acceptors. The main application for
+that is to allow your phone to talk to a left and a right pair of earbuds or hearing aids. They
+no longer need to be from the same manufacturer, as Bluetooth LE Audio is an interoperable
+standard.
+We can extend this by adding additional unicast streams to support more than one pair of
+earbuds, or to connect multiple speakers to support surround sound systems, with the example
+of Figure 4.3 showing the addition of a central woofer unit.
+
+Figure 4.3 Unicast audio topologies
+
+In theory, the specification can support up to 31 separate unicast Isochronous Streams, which
+could connect 31 different devices. That's not actually realistic for audio, as we start to run
+out of bandwidth after three or four streams. The reason for the limit of 31 streams in the
+Core specification is that the Isochronous Streams feature was designed to support many
+different time critical applications – not just audio. Some of those need far less bandwidth
+than audio does. When we look at the LC3 codec, we’ll discover some of the trade-offs we
+have to make between latency, audio quality and robustness, which place a limit on the number
+of audio streams we can actually support.
+That airtime limitation on the number of streams that unicast can support is one of the reasons
+to use broadcast. As Figure 4.4 shows, a single Broadcaster can talk to multiple Acceptors,
+which will often be configured as pairs of devices, receiving either a single mono channel or
+separate left and right audio channels. Depending on the audio quality (which typically sets
+the sampling rate and hence the airtime usage and maximum number of streams), a
+Broadcaster may be able to offer greater functionality, such as transmitting simultaneous audio
+streams in different languages. These are the trade-offs that need to be understood when
+designing a Bluetooth LE Audio application, which we’ll cover in more detail in the following
+chapters.
+
+79
+
+Section 4.3 - Connected Isochronous Streams
+
+Figure 4.4 Broadcast audio topologies
+
+4.3
+
+Connected Isochronous Streams
+
+To understand the Core Isochronous features, we’ll start with unicast and Connected
+Isochronous Streams, which are known as CISes. Their structure is quite complex, but it’s
+built on some very simple principles. I’ll start by describing how Connected Isochronous
+Streams were designed, to explain the component parts and how they work, then look at how
+Broadcast is different. That gives you the foundations to move up into the Generic Audio
+Framework, where we put the Isochronous Streams to work.
+
+4.3.1
+
+The CIS structure and timings
+
+When you design for digital audio, you generally have the constraint that you’re going to be
+sampling the incoming audio at a standard, consistent rate. Once the incoming audio is
+sampled and encoded, it’s sent down to the Bluetooth transmitter to send to the receiving
+device. The system is repetitive, the audio data is time bounded and transmissions have a
+constant interval between them which is called the Isochronous Interval or the ISO_Interval.
+The start of each Isochronous Interval in a CIS is called its Anchor Point. As Figure 4.5
+shows, the transmission starts off with an Initiator sending a packet containing audio data (D)
+to an Acceptor. When the Acceptor receives it, it sends back an acknowledgement, and that
+process is repeated on a regular basis. The third data packet in Figure 4.5 has no
+acknowledgement, so the Initiator would presume it had not been received.
+
+80
+
+Chapter 4 - Isochronous Streams
+
+Figure 4.5 Simplified unidirectional audio transfer
+
+Most modern codecs are optimised to run at a frame rate of 10 milliseconds, i.e., they sample
+10 milliseconds of audio at a time, which provides a good compromise between audio quality
+and latency. That's the preferred setting for LC3, which is the mandatory codec for Bluetooth
+LE Audio. Unless I specify otherwise, we'll be assuming a 10 millisecond sampling interval is
+used throughout this book, so the Isochronous Intervals will always be 10ms, or multiples of
+10ms.
+4.3.1.1
+
+Isochronous Payloads
+
+The structure of the data in the PDU of the air interface packet (D) shown in Figure 4.5, which
+is sent between devices is very simple, and is shown in Figure 4.6.
+
+Figure 4.6 Bluetooth® LE Link Layer packet format
+
+The encoded ISO PDU is preceded by a preamble and Access Address, and followed by a
+CRC. These add 10 or 14 octets when transmitting over an LE 1M PHY25, (depending on
+
+PHY refers to the Physical layer, and specifically the choice of symbol rate, which corresponds to
+the number of bits which can be transmitted each second. Currently most Bluetooth LE products use
+a symbol rate of 1 million symbols per second. In contrast, most Bluetooth LE Audio products will
+use the enhanced symbol rate of 2 million symbols per second, as that allows higher quality codec
+parameters to be used. The trade-off is a slight reduction in range.
+25
+
+81
+
+Section 4.3 - Connected Isochronous Streams
+whether there is a Message Integrity Check (MIC)26 included with the ISO PDU payload), and
+11 or 15 octets when using a LE 2M PHY.
+For a CIS, the ISO PDU is called the CIS PDU, and its structure is shown in Figure 4.7.
+
+Figure 4.7 ISO PDU format and header for a CIS
+
+The ISO PDU has a header, followed by a payload of up to 251 octets. If the audio needs to
+be encrypted, there's an optional MIC at the end of that packet. In the CIS PDU header, there
+are five control elements:
+•
+•
+•
+•
+
+4.3.1.2
+
+the LLID (Link Layer ID), which indicates whether it is framed or unframed
+NESN and SN, the Next Expected Sequence Number and Sequence Number,
+which are used for acknowledgments and flow control at the Link Layer level,
+CIE, the Close Isochronous Event bit, and
+NPI. The Null Payload Indicator, which indicates the payload is a null PDU,
+identifying that there is no data to send.
+Subevents and retransmissions
+
+We’ll cover the control bits in the ISO PDU header as they become relevant to the explanation
+of how CISes work. Before that, we need to look at the structure of a Connected Isochronous
+Stream. We've already talked about the Isochronous Interval, which is the time between
+successive Anchor Points of a CIS. The Anchor Point is the point where the first packet of a
+CIS is transmitted by the Initiator and the start of each successive Isochronous Interval.
+Within a CIS, we can retransmit the CIS PDU if required, as the CIS structure supports
+multiple Subevents. Each Subevent starts with the transmission from the Initiator and ends
+with the final point of the expected response from an Acceptor. All of the Subevents within
+a CIS form a CIS event, which starts at the Anchor Point of the CIS and finishes at the
+reception of the last transmitted bit received from the Acceptor.
+
+Message Integrity Check. A value calculated from the payload contents to detect if it has been
+corrupted.
+26
+
+82
+
+Chapter 4 - Isochronous Streams
+
+Figure 4.8 Events, Subevents and Sub_Intervals
+
+The time between successive Subevents is defined as the Sub_Interval spacing, which is the
+maximum duration of the Subevent for that CIS, plus the inter-frame spacing, which is defined
+as being 150µs. The Sub-Interval spacing is determined when the CIS is configured and does
+not change for the lifetime of the CIS.
+4.3.1.3
+Frequency hopping
+An important reason for defining Subevents is that Bluetooth LE Audio changes the
+transmission channel on every Subevent, as illustrated in Figure 4.9, to protect against
+interference.
+
+Figure 4.9 Frequency hopping per Subevent
+
+The Core 5.2 specification introduced a new channel selection algorithm, which is more
+efficient than the one in Core 4.0. This is applied to each Subevent. If a Subevent is not
+transmitted, then the channel hopping scheme assumes that it has been and moves to the next
+frequency channel for the following Subevent.
+4.3.1.4
+
+Closing a Subevent
+
+When we looked at the isochronous PDU header, we saw that there's a Close Isochronous
+Event bit – the CIE. That's used by the Initiator to signify that it has received an
+acknowledgment from an Acceptor confirming that its packet was successfully received by the
+Acceptor, so that it will stop further retransmissions of that particular PDU. In Figure 4.10,
+the Initiator has sent out its first PDU and had an acknowledgment back, so it then sends a
+83
+
+Section 4.3 - Connected Isochronous Streams
+packet with the CIE bit set to 1, to tell the Acceptor that there will be no further transmissions,
+as it is closing the event. It would not normally bother to include the audio data in that
+transmission, so it would also set the Null Packet Indicator bit, allowing it to shorten the
+transmitted packet. (Figure 4.10 shows the duration of the original CIS Event for
+comparison.)
+
+Figure 4.10 Closing a CIS event early with the CIE bit
+
+This allows the Acceptor to go to sleep until the next Isochronous Interval. The Initiator can
+use the time to do other things. As many Initiators will also be interacting with other Bluetooth
+devices, and possibly sharing their radio and antenna with Wi-Fi, that can be useful. For the
+Acceptor, turning its receiver off until it's ready to do something again can bring a significant
+power saving.
+If the Acceptor does not receive the header with the CIE bit, it will continue to listen for data
+in each scheduled Subevent, but will not receive any packets from the Initiator to respond to.
+
+4.3.2
+
+Controlling audio quality and robustness
+
+Having covered the basic timing of transmissions in a CIS, we can now look at the parameters
+which are used to control the quality of the audio and the robustness of the link. For many
+audio applications, latency is important. For some applications, such as when you are listening
+to a live stream, it is important to minimise the latency, particularly if you can hear the ambient
+sounds as well. On the other hand, if you're streaming music through your phone and can't
+hear or see the source, latency doesn't matter that much. Other applications, such as gaming,
+have different priorities, and if you’re listening to audio while watching a film, lipsync becomes
+important.
+The Basic Audio Profile (BAP) can set many of the parameters which affect audio quality and
+latency, by using the LE_Set_CIG_Parameters HCI command. We’ll look at how it makes
+those choices in Chapter 7, but for now, we need to understand how the Isochronous Channel
+structure can be configured. The key items that an application can request in order to influence
+the latency and robustness are:
+84
+
+Chapter 4 - Isochronous Streams
+•
+•
+•
+
+The Maximum Transport Latency, which sets the maximum time that an Initiator
+can spend transmitting the PDUs for a particular CIS.
+The Maximum SDU size for both directions of the CIS
+The SDU interval for both directions
+
+The Maximum Transport Latency affects the overall latency, although it is only one element
+of it. Whilst many applications will want to minimise latency, in the real world of wireless you
+also need to address the inherent fragility of a wireless link where packets can be lost. To
+ensure sufficient robustness, (which translates into rendered voice and music streams without
+drop-outs, clicks and silence), we need to use a variety of techniques to help ensure that audio
+data gets through in an acceptable timeframe.
+4.3.2.1
+
+Flush Timeout and Number of Subevents
+
+The three parameters listed above are inputs to the Controller. The Controller takes them and
+uses them to calculate three parameters that affect the robustness of the Bluetooth LE Audio
+link for that CIS, which are:
+•
+
+•
+•
+
+•
+
+NSE - the Number of Subevents. This specifies the number of Subevents which
+will be scheduled in each Isochronous Interval. They are used for the initial
+transmission of a CIS
+PDU and its subsequent retransmissions. They may not all be used, but it is a fixed
+number that are scheduled.
+FT – the Flush Timeout. The Flush Timeout defines how many consecutive
+Isochronous Intervals can be used to transmit a PDU before it is discarded. The
+point at which it is no longer transmitted is called the Flush Point.
+BN – The Burst Number, which is the number of payloads supplied for
+transmission in each CIS event.
+
+These can be quite difficult to grasp, so it’s useful to look at some simple examples.
+The Number of Subevents (NSE) is the most straightforward of the three. It is simply the
+number of opportunities to transmit an Isochronous PDU which are available within each
+Isochronous Interval. In the simplest example, where only one PDU is supplied for
+transmission in each Isochronous Interval, the PDU will be transmitted in the first Subevent,
+and can then be retransmitted a maximum of (NSE-1) times in the same Isochronous
+Interval. If it is a unidirectional CIS, once the PDU’s transmission is acknowledged by the
+Acceptor, the Controller can set the Close Isochronous Event (CIE) bit in the header of its
+next transmission (which can have a null PDU payload), and any remaining Subevents in
+that CIS Event become free airtime for other radio applications.
+That is the case where Flush Timeout is 1, as the Flush Point then coincides with the end of
+the CIS Event for the Isochronous Interval. This simple case is illustrated in Figure 4.11. For
+the sake of clarity, the following examples only involve one Acceptor.
+85
+
+Section 4.3 - Connected Isochronous Streams
+
+Figure 4.11 A unidirectional CIS with NSE = 4 and FT = 1
+
+In this example, NSE is set to 4, so there are four opportunities for each packet to be
+transmitted. In the first CIS Event, none of the four attempts are successful, so packet P0 is
+flushed, The Acceptor will need to try to reconstruct it using some form of Packet Loss
+Correction.
+The second packet (P1), succeeds after the second attempt, after which the Initiator closes the
+Event. The third packet (P2) succeeds first time.
+If the Flush Timeout is increased, then the transmission of a packet can continue over more
+Isochronous Intervals. Figure 4.12 illustrates an example where NSE remains at 4, but the
+Flush Timeout is increased to 3.
+
+Figure 4.12 An example of FT = 3 and NSE = 4
+
+This illustrates a problem if you are only using the two parameters – NSE and FT. It allows
+a packet to dominate the transmission slots until it reaches its flush point. In Figure 4.12, the
+86
+
+Chapter 4 - Isochronous Streams
+first payload, P0, which is having problems getting through, occupies all of the Subevents in
+the first three Isochronous Intervals, leaving P1 and P2 waiting until after the Flush Point FP0.
+Although this situation should not be common, it can leave subsequent payloads more
+exposed until enough of them get through to bring the system back into equilibrium.
+4.3.2.2
+
+Burst Number
+
+The way to address this problem is to allow more than one payload to be transmitted in a
+single Isochronous Interval, so that Subevents in each Isochronous Interval can be shared
+between more than one PDU and not be used exclusively by one of them. This is made
+possible by taking advantage of Burst Number (BN), which is the number of payloads supplied
+for transmission in a CIS event. In the examples above, only one packet has been delivered
+in a CIS Event, which is the situation where the cadence of SDU and PDU generation is the
+same as the Isochronous Interval – meaning that one encoded 10ms audio frame becomes
+available for each 10ms Isochronous Interval. If we want to make use of BN and we’re
+continuing to sample the incoming Audio Channel every 10ms, we need to increase the
+Isochronous Interval to a multiple of that, so that we have more packets available in each
+Isochronous Interval. It means that by addressing one problem, we are potentially creating
+another, which is increasing latency.
+
+Figure 4.13 The effect of Burst Number = 2, with NSE = 4 and FT = 1
+
+In Figure 4.13, the Isochronous Interval has been doubled, allowing two packets to be supplied
+in each interval. The combination of the 10ms codec frame and 20ms Isochronous Interval
+means the Initiator has two PDUs available to transmit within each Isochronous Interval. A
+consequence of this is that there are now two flush points in each Isochronous Interval. In
+our simple example, each Flush Point occurs after two Subevents. For more complex
+combinations of FT and NSE parameters, the location of the Flush Points can be calculated
+from the equations in the Core [Vol 8, Part B, 4.5.13.5].
+87
+
+Section 4.3 - Connected Isochronous Streams
+Returning to Figure 4.13, we see the Initiator sending packet P0 in the first Subevent, which
+is acknowledged, so the Controller immediately moves on to transmit packet P1, transmitting
+it three times before it is acknowledged.
+In the second Isochronous Interval, packet P2 is transmitted, but the Acceptor sends NACKs
+to indicate errors in the received packet. As the Flush Timeout is set to 1 and BN=2, the
+Flush Point for packet P2 is in the same Isochronous Interval, coming after two further
+transmission attempts, neither of which have been successfully received by the Acceptor.
+At that point, the Initiator starts transmitting P3, although once again, in this example, the
+Acceptor responds with NACKs to indicate a problem with the packets it received. After two
+attempts, P3 is flushed by the Initiator.
+In the final Isochronous Interval of Figure 4.13, P4 is successfully transmitted, leaving three
+opportunities for P5, all of which are unsuccessful. In each case, the Initiator will attempt to
+transmit the PDUs in every available Subevent before that PDU’s Flush Point. After each
+acknowledged transmission it will move on to the next available packet or, if there are no
+further packets available, close the Isochronous Event.
+In this example, P2, P3 and P5 would be discarded. In real life, we’d expect a much better
+rate of acknowledgement – this is just an example to illustrate the principle.
+As a final example, Figure 4.14, shows the effect of increasing the Flush Timeout to 2, with a
+Burst Number of 2, which gives the opportunity to transmit in two consecutive Isochronous
+Intervals. In this example, no packets are flushed.
+
+Figure 4.14 Example of NSE=4, BN=2 and FT=2
+
+Using Flush Timeout and Burst Number can be very useful to provide more retransmission
+opportunities, spanning multiple Isochronous Intervals. They are particularly useful if you're
+88
+
+Chapter 4 - Isochronous Streams
+in a noisy environment. However, they have an effect on latency. Every increment of Flush
+Timeout increases latency as the retransmissions are spread across more Isochronous
+Intervals, whilst Burst Number increases the duration of the Isochronous Intervals. Note that
+Burst Number is confined to multiple payloads arriving in an Isochronous Interval. It can’t
+be sued to solve the problem of a single payload hogging transmission opportunities which
+we saw in Figure 4.12.
+These parameters cannot by set directly by the Host. It is limited to setting values for the
+Maximum Transport Latency, the Maximum SDU size and the SDU interval BN, NSE and
+FT are then calculated in the Controller, which takes into account any other radio requirements
+in the chip. It is, however, very useful to have an understanding of the potential effects these
+parameters have on the Isochronous Channel structure. Table 3.19 of TMAP provides
+examples of how a Controller might interpret the Host values for a CIS to suit different
+operating conditions, such as prioritising airtime for coexistence or minimising latency. The
+exact allocation of parameters is always down the algorithms in the scheduler, which will be
+set by the chip supplier.
+
+4.3.3
+
+Framing
+
+The other parameter in a CIS PDU header which needs to be understood is the pair of LLID
+(Link Layer ID) bits, which indicates whether the CIS is framed or unframed. Unframed refers
+to the case where a PDU consists of one or more complete codec frames. It is used where
+the Isochronous Interval is an integer multiple of the codec frame length. In contrast, framed
+is where you have a mismatch between the codec frame length and the Isochronous Interval,
+which results in a codec frame being segmented across multiple SDUs. This starts to get
+complex, but is important where an Initiator may need to support Bluetooth connections
+which have different timings. We'll revisit that when we look at a feature called ISOAL, which
+is the Isochronous Adaptation Layer, which has been designed to cope with this mismatch.
+(The LLID bits also indicate when there is no ISO PDU, which is different from the NPI in
+the ISO PDU header.)
+
+4.3.4
+
+Multiple CISes
+
+Having covered all of the features of a single, unidirectional CIS, the next step is to add more
+of them. The most common application for this is when an Initiator is sending audio to a left
+and a right earbud. In this case, the Initiator will set up separate Isochronous Streams with
+two different Acceptors. In Figure 4.15 we can see that it's a straightforward extension of
+what we've seen before - the Initiator transmits and receives an acknowledgment from the
+first Acceptor, then repeats this for the second Acceptor.
+An important point to note is that although the left and right Audio Channels are sampled at
+the same time, the ISO PDUs are sent serially.
+
+89
+
+Section 4.3 - Connected Isochronous Streams
+
+Figure 4.15 Timings for CISes to two Acceptors
+
+Note that where we have more than one CIS, they always have the same Isochronous Interval.
+Their Anchor Points are different, as data is sent serially, but for each CIS their Anchor Points
+are the same ISO Interval apart. Each Anchor Point represents the point of the first
+transmission from an Initiator to an Acceptor for that CIS. As Figure 4.16 shows, each CIS
+has an associated ACL link. The ACL link always needs to be present because it is used to set
+up the CIS and control it. If there are multiple CISes between an Initiator and an Acceptor,
+they can share the same ACL. If the ACL is lost for any reason, any associated CISes are
+terminated. The application then needs to decide what to do with any remaining CISes
+established with other Acceptors within that CIG.
+
+Figure 4.16 Multiple CISes and associated ACL links
+
+When an Initiator is scheduling two or more audio channels, there are two options for how
+they are transmitted, regardless of whether they are connected to one or multiple Acceptors.
+The obvious approach is to transmit them sequentially so that you send all of the data on CIS
+90
+
+Chapter 4 - Isochronous Streams
+0, and then all the data on CIS 1, continuing until the Initiator has worked through all of the
+packets for both CISes. This is illustrated in Figure 4.17, where we have an NSE of 3. It
+shows all of the Subevents within CIS 0 being transmitted before the Subevents for CIS 1.
+
+Figure 4.17 Sequential arrangement of two CISes
+
+The disadvantage of this approach is that if you have a reliable connection, where the first
+transmission of the packet is acknowledged, you end up with gaps between each CIS. In
+devices like phones, which are often sharing the Bluetooth and Wi-Fi radios, that's wasted
+airtime that could be used for something else. To address this, the Controller can choose to
+interleave CISes as an alternative approach, as shown in Figure 4.18.
+
+Figure 4.18 Interleaved arrangement of two CISes
+
+In this case, each Subevent for CIS 0 is followed by a Subevent for CIS 1, and so on for the
+remaining CISes. The diagram repeats the example of Figure 4.17, having an NSE of 3, and
+shows CIS 0 transmitting and receiving its first Subevent, followed by CIS 1. If both
+Acceptors receive those first transmissions and acknowledge them, they can close their CIS
+91
+
+Section 4.3 - Connected Isochronous Streams
+Events. It results in a much greater gap after their transmissions, freeing up airtime which can
+be used for other purposes.
+
+4.3.5
+
+Bidirectional CISes
+
+The multiple, unidirectional connected Isochronous Streams we’ve defined above replicate the
+use case of A2DP. For earbuds and many other audio applications, we want to get data back
+as well, requiring bidirectionality. One approach would be to set up a second CIS in the
+opposite direction, so that we would use one CIS to transmit from phone to earbud and the
+other to transmit from earbud to the phone. However, that’s inefficient. The Core
+specification provides an optimisation by adding return data into the acknowledgement
+packets. It's still a single CIS, but it's now being used for two separate Audio Streams.
+
+Figure 4.19 Example of a bidirectional CIS for the left earbud and a unidirectional CIS for a right earbud
+
+In the example shown in Figure 4.19, the Initiator has set up individual CISes with a left
+Acceptor and a right Acceptor. Both receive data from the Initiator, but the left one also
+sends data from its microphone back to the phone. The same principles we've seen before
+apply. Data is sent by the Initiator, the Acceptor immediately responds to acknowledge receipt
+(or not) and if it has data, includes it in the return CIS PDU. If that return PDU gets back
+with a good CRC, the Initiator will acknowledge it, close the event, and transmit data to the
+right Acceptor CIS.
+The acknowledgements in both directions use the NESN and SN bits in the ISO PDU header,
+flipping the bits when a packet is acknowledged. In Figure 4.19, all of the transmitted packets
+have the same ISO PDU format. Those marked “Ack” contain no audio data, so would have
+the NPI bit set to one to indicate they have a null CIS PDU. The data packets from the left
+Acceptor will be an identical format to the Initiator’s “L” and “R” packets, but will contain
+data from the Acceptor’s microphone. The Initiator’s “Ack” of the left Acceptor’s data would
+include the CIE bit, occurs before it transmits the data to the right Acceptor shows that it is
+transmitting the CISes in sequential mode. These “Acks” would also have the CIE bit set.
+92
+
+Chapter 4 - Isochronous Streams
+The value of NSE applies to both directions in a CIS, as the same Subevents are used for
+transporting PDUs both to and from the Acceptor. Once the payload in one direction has
+been acknowledged, future transmissions can replace the payload of that PDU with a null
+payload. (That has no effect on the timing of the Subevents, but saves a small amount of
+power.)
+Only the Initiator can set the CIE bit in the ISO PDU header in a bidirectional CIS to
+indicate that they are not ging to transmit any further payload data in the current CIS Event.
+It should not close the CIS Event unless the payloads for both directions have been
+acknowledged. If an Initiator has had its packet acknowledged, but has not received an
+Acceptor’s packet, it should continue to transmit null PDU packets in every available
+Subevent, to give the Acceptor opportunities to retransmit.
+The two directions in a bidirectional CIS can have different properties, such as codec and
+PHY and may even be used by different applications. Even some of the structural parameters
+can be different for the two directions. FT and BN can be different, as can the Max_PDU
+and Max_SDU values, which also means that the SDU_Intervals can be different, although
+one needs to be an integer multiple of the other. However, the ISO_Interval, Packing,
+Framing and NSE must be the same.
+
+4.3.6
+
+Synchronisation in a CIS
+
+Once we add a second Acceptor, both Acceptors act independently of each other, but under
+the Initiator’s control. If they are a Coordinated Set, they will know the other one exists,
+because they know how many members there are in the Coordinated Set. However, your left
+ear bud and right ear bud don't necessarily know the other one is present, turned on or
+receiving data. A user may take one out to share with a friend, or the battery in one may die.
+Even if they have a means of communicating, there is no guarantee that they will always be in
+contact with each other. To cope with this and enable them to render or capture audio streams
+at precisely the same time, the Core Isochronous Channels design includes a synchronisation
+method which allows microsecond level accuracy of rendering, without an Acceptor needing
+to have any knowledge of the presence (or otherwise) of any other Acceptors. Figure 4.20
+illustrates how this is done.
+Within a CIG, CIS Events are scheduled for each of the CISes in that CIG. For a pair of
+earbuds that will be two CIS Events - one for the left earbud one for the right earbud. It could
+be more, such as with multi-channel sound systems. These CIS events make up a CIG event.
+In Figure 4.20, for the sake of simplicity, the multiple CISes are shown as sequential. They
+could equally be interleaved.
+
+93
+
+Section 4.3 - Connected Isochronous Streams
+
+Figure 4.20 Synchronisation of multiple CISes
+
+The Core defines a CIG reference point which may be coincident with, but cannot be later
+than the Anchor Point of the first CIS. Typically, it will occur slightly before that. It also
+defines a CIG synchronisation point which occurs after the last possible receive event for the
+last CIS event within that CIG. At that point, the Initiator knows that every Acceptor will
+have had every possible opportunity to receive every PDU that has been sent to it and time to
+return any data for the Initiator.
+In order to reconstruct this common point, every device is informed of its individual CIS Sync
+Delay for every CIS which it supports, which is calculated from the Instant27 for the ACL link
+associated with that CIS. This allows it to calculate the CIG synchronisation point, which is a
+common timestamp across every device. With this knowledge, each device can determine
+when it should start to decode the audio. BAP adds a feature called Presentation Delay, which
+tells the Acceptor when to render the decoded stream. As we move up the layers and start to
+dig into the detail of the basic concepts of QoS and latency, we'll see how Presentation Delay
+is used to add flexibility to the rendering time.
+If you have devices with microphones, the same synchronisation problem exists, but in this
+case you need to coordinate the point at which audio is captured by an Acceptor. If it uses
+the uplink of a bidirectional CIS, it will use the same CIG synchronization point, but is likely
+to require a different value for Presentation Delay to the one used to render the downlink
+
+An Instant is a timing reference, normally the start of a transmitted packet, which is used for
+synchronisation. Its use is defined in the Core Vol 6, Part B, Sect 2.4.2.29.
+27
+
+94
+
+Chapter 4 - Isochronous Streams
+audio data from the Initiator, as it needs extra time to capture and encode the audio stream.
+The synchronisation timing is defined with a microsecond accuracy. This means that the
+Controllers of all of the Acceptors within a CIG have a timestamp assigned which will be
+within a few microseconds of each other. That needs to be conveyed to the application layer
+that renders the audio streams, although the method of doing that is down to the
+implementation. However, the specification means that left and right earbuds can present the
+two audio streams to the ears around an order of magnitude closer together than the brain can
+discern.
+
+4.3.7
+
+The CIG state machine
+
+Having covered all of the features that make up Connected Isochronous Streams, we can look
+at how to put them together to configure and establish a Connected Isochronous Group.
+
+Figure 4.21 The CIG state machine
+
+There's a fairly straightforward state machine used to configure and establish a CIG and its
+constituent CISes, which is shown in Figure 4.21. It contains three states:
+•
+
+28
+
+The Configurable CIG state, which is where all of the CISes that make up a CIG
+are defined. That's done through a set of HCI commands28 called the LE Set CIG
+
+HCI (Host Controller Interface) commands are defined to provide the link between the Host stack
+
+95
+
+Section 4.3 - Connected Isochronous Streams
+
+•
+
+•
+
+Parameters command. Once these are sent to the Controller, the Initiator has all
+the information that it needs to work out the scheduling and to optimise its airtime
+usage.
+The Active CIG state, in which one or more of the configured CISes is enabled.
+The CIG moves to the active state by enabling at least one of its configured CISes
+using the LE Create CIS HCI command. It doesn't need to enable all of the
+configured CISes. Once the CIG is in the Active state, it can disconnect individual
+CISes, and it can use the LE Create CIS command to enable other CISes which it
+has previously configured. In theory, that allows it to swap CISes whilst that CIG is
+active. This can be useful if there is only an occasional need for a CIS, such as a
+return path for voice commands. The CIS could be turned on and off depending
+on when it is needed. Once a CIG is in the active state, you cannot configure
+additional CISes. That can only be done by disconnecting all CISes, deactivating
+the CIG and starting again.
+The Inactive CIG state. A CIG moves to the Inactive state by disconnecting all of
+its established CISes. It cannot change the configuration of any CIS in this state,
+nor add new ones, but it can return to the Active state by using the LE Create CIS
+HCI command to re-establish a CIS. This allows a CIG to be reactivated without
+the need for the Controller to go back to the start and reschedule one or more of its
+configured CISes.
+
+When all of the established CISes have been disconnected, the CIG can be removed by issuing
+the LE Remove CIG HCI command.
+
+4.3.8
+
+HCI commands for CISes
+
+The Core specification defines five HCI commands which are used to inform the Controller
+about each CIG. The first is the LE Set CIG Parameters command [Vol 4, Part E, 7.8.97],
+which creates a CIG and configures an array of CISes within it. Each CIS within the CIG can
+be unidirectional or bidirectional. Each one can use a different PHY for each direction.
+The first use of the LE Set CIG Parameters command creates the CIG, moving it to the
+Configured CIG state, with the number of CISes defined in the array. The command can be
+used multiple times, to add more CISes to the CIG, or modify CISes which have already been
+defined. Each CIS is assigned a Connection handle.
+Each time the command is issued, the Controller should attempt to calculate a schedule for
+all of the CISes in the CIG, which meets the requirements specified by the HCI parameters,
+
+and the Controller. They are used for prototype testing and provide an interoperable interface where
+the Host stack and Controller silicon come from different manufacturers. Generally, operating
+systems do not make them available, providing a higher level API for developers.
+
+96
+
+Chapter 4 - Isochronous Streams
+and, if successful, will confirm this with an HCI Complete Command. Note that two of the
+parameters sent by the Host – Packing and RTN (the number of retransmissions), are only
+recommendations. The Controller should try to accommodate them when working out the
+schedule, but can ignore them, or attempt a best-case schedule that is guided by them. The
+Link Layer parameters of Burst Number, Flush Timeout and NSE are determined by the
+Controller’s scheduling algorithm and cannot be explicitly set by a higher layer specification.
+Some of the profile specifications provide recommendations that suggest optimal settings for
+specific use cases, but it is up to a specific chip implementation as to whether these are
+accessible by an application, or are part of a scheduler’s choice.
+This mismatch between the HCI parameters sent from the Host and those set by the
+Controller may seem odd, but there is a good reason, which is to prevent a top level profile or
+application from trying to prioritise the overall radio operation. Many Bluetooth chips include
+Wi-Fi, and they typically share an antenna. Hence the Controller needs to work out how to
+fit in the demands of both. The Bluetooth technology implementation may also be running
+multiple profiles. There is no reason a smart phone can’t receive a Bluetooth LE Audio
+broadcast stream and then retransmit it to a pair of hearing aids as a pair of Connected
+Isochronous Streams (other than the physical constraints of available airtime and processing
+resources). However, at a profile level, none of these applications may know that the other
+exists. If each one could dictate the exact values of BN, FT and NSE, it would be very likely
+that they would be unable to coexist with the constraints of other applications. That’s why
+the HCI commands prevent this level of control, allowing the Controller to work out how
+best to fit everything together. The scheduling algorithms are not accessible to application
+developers �� they are a critical part of the Bluetooth chip. However, it’s important to
+understand why you can’t dictate what you want and to be aware of the dangers of overspecifying the needs of your audio application.
+Once all of the CISes have been configured, the LE Create CIS command is used to create
+each CIS which is required, which causes the CIG to move to the Active CIG state. Not all
+CISes need to be created at this point. A CIS can be disconnected, and another CIS created
+at any point, which may be useful when an application decides to move from using a
+microphone in an earbud to the microphone on a smartphone. However, this does assume
+that the Controller managed to schedule all of the CISes.
+The CIG remains in the Active CIG state until the last CIS is disconnected. At this point it
+moves to the Inactive CIG state. It can either be permanently removed with the LE Remove
+CIG command, or returned to the Active CIG state by using the LE Create Command again
+on at least one of the CISes.
+As each CIS is created by the Initiator, each Acceptor will be informed of the connection
+parameters via Link Layer commands, leading to an HCI LE CIS Request event [7.7.65.26]
+being sent to its Host. If it accepts the request, it will respond with an HCI LE Accept CIS
+Request Command [7.8.101], leading to both Initiator and Acceptor Hosts receiving an HCI
+97
+
+Section 4.4 - Broadcast Isochronous Streams
+LE CIS Established event [7.7.65.25]. When we get to the ASCS, we’ll see how these work
+with the Isochronous Stream state machines.
+That completes a review of all of the component parts of a CIG and CIS and how we put
+unicast Connected Isochronous Streams together. Now, we'll look at the analogues for
+broadcast where we have to do the same thing, but without the luxury of having a connection
+between the two devices to allow them to negotiate what they're doing.
+
+4.4
+
+Broadcast Isochronous Streams
+
+Broadcast audio is a totally new concept for Bluetooth technology. In the past, audio has
+always been streamed directly from one device to another device, as we saw in the previous
+chapters. That concept has now been extended with Connected Isochronous Streams, so that
+we can stream audio to multiple devices. But those devices are still connected and every packet
+that is correctly received is acknowledged. With broadcast, there is no connection29. Any
+device which is within range of a broadcast transmitter can pick up and render the broadcast
+Audio Streams. The big difference from Bluetooth Classic Audio profiles is that it is one-way
+– there are no acknowledgements. That means that it is technically possible to build a
+Broadcast Source which does not contain a receiver, but which only transmits. Those
+broadcasts can be received by any and every Broadcast Sink which is within range.
+There is a practical advantage in having no acknowledgements, which is that it generally results
+in a much greater range. That occurs because the link budget over a connection is often very
+asymmetric. A broadcast transmitter is often mains powered, so can easily transmit at the
+maximum allowed power. That gives it a good link budget to an earbud. However, an earbud
+is unlikely to be transmitting acknowledgements back at anything more than 0dBm (1mW),
+both to conserve its battery and because its small antenna is probably going to have a gain
+below 0dB. That means that the link budget for the earbud to receive a broadcast transmission
+may be 10 – 20dB higher than for the return acknowledgement path. It is the latter which
+determines the maximum range for a CIS. For public coverage, a Broadcaster can cover a
+considerable area – far more than can be achieved with a connected Isochronous Stream
+transmission.
+
+As we will see later on, there can, and often will be, a connection. But that enhances the way that
+broadcast works – it doesn’t affect the structure of a BIS or BIG, so we’ll ignore it for the time being.
+29
+
+98
+
+Chapter 4 - Isochronous Streams
+
+4.4.1
+
+The BIS structure
+
+The definition of the structure of Broadcast Isochronous Streams and Groups is very similar
+to what we've just looked at with Connected Isochronous Streams.
+Figure 4.22 shows that Broadcast Isochronous Streams have the same basic PDU structure of
+header, payload and an optional MIC if you want encryption. However, the header for the BIS
+is a lot simpler, as it does not need the SN and NESN flow control bits that are in the CIS
+PDU header. It starts with the same two Link Layer ID (LLID) bits. Those indicate whether
+the PDU is framed or unframed. Then come CSSN and CSTF, which are used to signal the
+presence of a Control Subevent. CSSN is the Control Subevent Sequence Number and CSTF
+is the Control Subevent Transmission Flag, signalling that a Control Subevent is present in
+that BIS Event. These are new and don’t exist in Connected Isochronous Streams. Within a
+BIG, there is a Control Subevent which can be used to provide control information to every
+Acceptor. We need these in broadcast, as there’s no ACL to inform Acceptors of things like
+a change in the hopping sequence. The only other information in the header is the length of
+the payload.
+
+Figure 4.22 Isochronous PDU and header for Broadcast
+
+If we look at the structure of a Broadcast Isochronous Stream in Figure 4.23, it’s also very
+familiar. The fundamental difference between a BIS and a CIS is that there are no
+acknowledgments. A BIS is composed purely of Subevents which contain data, sent by the
+Initiator. There is no bidirectional data – everything is transmitted from the Broadcast Source.
+As before, we see each Subevent is transmitted on a different frequency channel.
+
+99
+
+Section 4.4 - Broadcast Isochronous Streams
+
+Figure 4.23 Structure of a Broadcast Isochronous Stream
+
+A notable difference between a BIG (a Broadcast Isochronous Group, made up of one or
+more BISes) and a CIG, is the potential existence of a Control Subevent, which is shown
+dotted in Figure 4.23. When it occurs, (which in in BIS Events where the header of the ISO
+PDU contains the CSTF flag), the Control Subevent is transmitted after the final Subevent of
+the last BIS, and provides an opportunity for control information to be sent to every device
+which is receiving a broadcast stream. We’ll cover what that does in Section 4.4.3.
+As Figure 4.23 shows, a Broadcast Isochronous Stream has the same basic elements of an
+Isochronous Interval, and Anchor Points for BIS and BIG Events. As there is no
+acknowledgement or return packet, a Sub_Interval time is defined, which is the time between
+the start of consecutive Subevents within a single BIS. It is also the time between the start of
+the final Subevent of the last BIS and a Control Subevent, if one is present.
+If the BIG contains more than one BIS, each BIS is separated by a BIS_Spacing. By adjusting
+the values of the Sub_Interval and the BIS_Spacing, the BISes can be arranged in either a
+Sequential or Interleaved fashion, which is illustrated in Figure 4.24 and Figure 4.25. If the
+BIS_Spacing is bigger than the Sub_Interval, they will be arranged sequentially. If it’s smaller,
+they will be interleaved.
+
+100
+
+Chapter 4 - Isochronous Streams
+
+Figure 4.24 Sequential BISes
+
+Both diagrams show two BISes, each of which have NSE set to two (i.e., two Subevents each).
+There are some important observations that can be made from these figures. The first is that
+the Control Subevent is never counted in the NSE – it is a totally separate Subevent. The
+second is that when the Control Subevent is present it does not affect any of the other packets
+or the Anchor Points. It is scheduled immediately after the last Subevent of the last BIS. But
+it does extend the BIG Event for the Isochronous Intervals which contain a Control Subevent.
+
+Figure 4.25 Interleaved BISes
+
+101
+
+Section 4.4 - Broadcast Isochronous Streams
+Whereas a CIS can stop transmitting audio packets once a device has received them, a
+Broadcast Source has no idea whether or not they have been received, so it has to repeat every
+transmission. However, as soon as an Acceptor has received a packet, it can turn its radio off
+and wait for the next scheduled BIS. This again highlights the asymmetry that we have within
+Bluetooth LE. Because the Broadcast Source is always transmitting, it typically has a much
+higher power drain than the Acceptors, which are likely to be earbuds. In general, that means
+that Broadcast Sources need bigger batteries, or a permanent power source. As they are
+typically phones or infrastructure transmitters in public places, that’s net generally an issue.
+However, it should be kept in mind if broadcast transmission is being embedded into small
+devices like watches or wristbands.
+Unlike CISes, all BISes have the same timing structure. Whereas the structure of each
+individual CISes is normally tailored to its codec configuration, optimising the Subevents for
+the PDU size, every BIS Subevent is the same length, which must fit the largest PDU that is
+being transmitted. That constraint is because the overall BIS timing structure is defined in the
+BIGInfo, which is included in the Periodic Advertisements. The BIGInfo structure is limited
+in size, so doesn’t have the granularity to specify different timings for different BISes – it
+specifies a “one size fits all”. It means that if you want to broadcast one channel at a 48kHz
+sampling rate and a second one at 24kHz, the smaller 24kHz channel would still be allocated
+BIS Subevent timings for the larger 48kHz packets, which wastes airtime. If that causes a
+problem, an alternative is to transmit the packets in separate BIGs. In this case, that would
+mean one BIG for the 48kHz sampled packets and another for the 24kHz sampled ones. The
+downside it that this adds complexity and doubles the amount of advertising, as each needs its
+own advertising set. Implementers need to decide which works best for their specific
+application.
+
+4.4.2
+
+Robustness in a BIS
+
+The way the parameters of a BIS are defined is a little different, as without acknowledgments,
+we need to employ some different strategies to maximise the chance of a transmission being
+received.
+A BIS still has a Burst Number (BN) and a Number of Subevents (NSE), defined in the same
+way as that for CISes. To compensate for the lack of acknowledgements, the Core introduces
+the concept of a Group Count within Isochronous Intervals, which is the ratio of NSE to
+Burst Number (NSE/BN). Group Count is used to determine the way that retransmissions
+are arranged in a BIS, allocating them to Groups, which are used to add diversity into the
+transmission scheme. (These Groups have nothing to do with Broadcast Isochronous Groups
+– it is an unfortunate use of the same word for two totally different concepts.)
+
+102
+
+Chapter 4 - Isochronous Streams
+
+Figure 4.26 The effect of Group Count
+
+Figure 4.26 shows two examples of a BIS, each of which contains six Subevents, so NSE = 6.
+In the first case, we have set a Burst Number of 3. That means that each of the three payloads
+which are available for that BIS event are sent twice. In the second example, on the right,
+there are the same number of Subevents, but with a Burst Number of 2, so we have three
+retransmissions of the two payloads. Each individual Subevent is given a group number (g)
+that goes from 0 up to (Group Count-1). So, in the left-hand diagram, we have a Group Count
+of 2, comprising group 0 and group 1. In the right-hand example, there are three groups: group
+0, group 1 and group 2.
+An important thing to note is that although these two examples look the same, they will have
+different Isochronous Intervals. Assuming that our frames are 10ms, the example on the left
+will have an Isochronous Interval of 20ms, as it contains two payloads – P0 and P1. The
+example on the right will have an Isochronous Interval of 30ms, as there are three payloads,
+P0, P1 and P2.
+Group Count works with two other, new parameters:
+•
+•
+
+Immediate Repetition Count (IRC), which defines the number of Groups which
+carry data associated with the current event, and
+Pre-Transmission Offset (PTO). The concept of Pre-transmission is to allow early
+transmission of packets in the hope that a receiving device can receive audio data
+packets early, allowing it to power down, and then be ready to render them at the
+point that the final transmission opportunity occurs.
+
+These parameters, which are calculated by the scheduler in the Controller, replace Flush
+Timeout. For a CIS, Flush Timeout is essentially an end-stop, terminating transmissions of a
+PDU. In most cases, it’s never needed, because as soon as an Initiator receives an
+acknowledgment that its data has been received, it can close the event and stop transmitting.
+103
+
+Section 4.4 - Broadcast Isochronous Streams
+A Broadcaster can’t do that – it has to keep on transmitting. Therefore, it’s advantageous to
+maximise the diversity of transmissions of packets to give every Acceptor the best chance of
+finding a packet. The sooner they can do that, the sooner they can go to sleep until the next
+one arrives.
+Pre-Transmission Offset works by introducing the concept of Subevents associated with a
+current event as well as Subevents associated with future events. This allocation of Subevents
+to PDUs is even more complex than the scheme for CIS, so the best was to explain it is to go
+through a series of examples showing the effect of changing the values. This is all worked out
+by the Controller and is not accessible to an application, but it helps to have an understanding
+of it when you set the HCI parameters to configure your streams.
+To illustrate the concept of future Subevents, Figure 4.27, demonstrates a BIS event where we
+have an NSE of eight Subevents; the first five of which are associated with the current BIS
+event, the last three of which are used for data from future BIS events.
+
+Figure 4.27 The concept of future BIS events
+
+We can now put everything together with some more examples.
+These group numbers (g), which we first saw in Figure 4.26, along with the Immediate
+Repetition Count (IRC), determine which payloads are sent in each transmission slot according
+to the following rules:
+•
+•
+
+104
+
+If g < IRC, group g shall contain the data associated with the current BIS event.
+If g ≥ IRC, group g shall contain the data associated with the future BIS event that
+is PTO × (g - IRC + 1) BIS events after the current BIS event.
+
+Chapter 4 - Isochronous Streams
+We'll now look at a few examples which illustrate how these different parameters result in a
+variety of different retransmission schemes.
+
+Figure 4.28 Pretransmissions with NSE=4, BN=1, IRC=3 and PTO=1
+
+In Figure 4.28, we have four Subevents with an NSE of 4. The Burst Number of 1 means
+one new payload is provided within each BIS event (so the Isochronous Interval for this
+example is 10ms). The Immediate Repeat Count is three, which means that the data associated
+with each event is transmitted in the first three slots. The conditions shown above show that
+the last transmission in that BIS event comes from the next BIS event. With this scheme,
+within every BIS event there are always three transmissions of the current data plus one
+transmission of the data from the next event.
+In case it seems like we’ve invented time travel, we haven’t. We’ve just bent our definitions
+slightly. Event x can’t happen until we have both the p0 and p1 PDUs available. So p1 is
+really the current data in Event x. This demonstrates that as soon as PTO is greater than 0,
+the latency starts to increase, as the Sync refence point cannot occur until after the last possible
+transmission of data, which for p0 is in Event x+1.
+
+105
+
+Section 4.4 - Broadcast Isochronous Streams
+
+Figure 4.29 Pretransmissions with NSE=5, BN=1, IRC=3 and PTO=1
+
+Figure 4.29 shows what happens when we increase the number of Subevents to 5. Here, the
+IRC remains at 3 but as NSE is 5, that provides two Subevents which are associated with data
+from future BIS events. These are used to pre-transmit a packet from event x+1 and event
+x+2. That means that transmissions are being spread across more Isochronous Intervals, with
+data coming from three consecutive BIS Events. That helps to provide robustness against
+bursts of wide-band interference which may last more than one Isochronous Interval, but it is
+at the expense of greater latency, which has grown by another 10ms.
+
+Figure 4.30 Pretransmissions with NSE=5, BN=1, IRC=3 and PTO=2
+
+106
+
+Chapter 4 - Isochronous Streams
+Figure 4.30 looks at the case where we are spreading even further, by increasing the PreTransmission Offset to 2. This results in the inclusion of data from an x+2 BIS event and an
+x+4 BIS event, effectively spreading the audio data transmissions over five events and adding
+a total of 40ms to the latency compared with what it would be with PTO = 0. In this figure,
+we're only showing the arrows for group 3 and group 4 in that first BIS Event x, otherwise
+the figure becomes very cluttered and difficult to read.
+
+Figure 4.31 Pretransmissions with NSE=6, BN=2, IRC=2 and PTO=2
+
+Finally, in Figure 4.31, we’ve set NSE to 6, Burst Number and IRC to 2, and the PreTransmission Offset is also 2. With these settings, each BIS event includes the two packets
+for that “current” event transmitted twice, one after the other, as we saw in Figure 4.26, with
+the final two Subevents used for packets two events in the future. Because BN is 2, the
+Isochronous Interval will be 20ms. As the PTO is 2, p4 and p5 come from two Isochronous
+Intervals in the future, so the latency is 50ms greater than the simple example of Figure 4.28.
+These BIS parameters allow for some very flexible transmission schemes in order to cope with
+different requirements in terms of latency and robustness. In the next chapter we’ll see how
+they can be used for different broadcast situations.
+In conclusion, both Flush Timeout and Pre-Transmission Offset increase latency. For a CIS,
+Flush Timeout helps to share battery savings between both Initiator and Acceptor, because
+the use of acknowledgements means that events can be closed. With a BIS, where there are
+no acknowledgements, a broadcast transmitter has to transmit at every Subevent. PTO
+effectively gives all of the power savings to the Acceptor, by providing a diversity of
+transmission that gives it the best chance of acquiring a packet.
+
+107
+
+Section 4.4 - Broadcast Isochronous Streams
+
+4.4.3
+
+The Control Subevent
+
+The Control Subevent [Core, Vol 6, B, 4.4.6.7] does not need to be included within every BIG
+event. Typically, Control Subevents are used for items such as channel map updates, so only
+occur occasionally, when that information needs to be provided to all of the devices that are
+receiving the broadcast Audio Streams. The advantage of using Control Subevents is that
+devices no longer need to scan to find out the basic BIG information, such as the hopping
+channels that are being used. This minimises the amount of work that a receiver has to do to
+ensure that it stays synchronised with a BIG and the BISes within it. Without them, a
+Broadcast Sink would need to continue to receive the Periodic Advertising train and
+periodically examine the BIGInfo to discover any changes.
+A Control Subevent is transmitted in six consecutive BIG events. It may then be transmitted
+in any subsequent BIG event, but only one control event can be transmitted at a time. Every
+BIS header for a BIS Subevent in a BIG event which includes a Control Subevent must have
+the Control Subevent Transmission Flag (CSTF) set to 1 to signal the presence of a Control
+Subevent in that BIG event. Its Control Subevent Sequence Number (CSSN) will tell a
+receiver if it is the same as one which they have already received. Control events use the same
+frequency hopping scheme as every other Subevent, taking the index of the first BIS event in
+the same BIG event.
+
+4.4.4
+
+BIG synchronisation
+
+As with Connected Isochronous Streams, individual receiving devices, typically a pair of
+earbuds or hearing aids, don't necessarily know about each other's existence. To keep their
+audio in synchronisation they need to use information that's included within the BIG to
+understand when they need to render it. To enable them to do this, a BIG Synchronisation
+Point is defined, which coincides with the end of the transmission of the audio data. Because
+we have no acknowledgments to take into account in broadcast, that BIG Synchronisation
+Point is located exactly at the end of the last BIS transmission of the last BIS in a BIG.
+Normally, that will be coincident with the end of the BIG event. However, for the case when
+there is a Control Subevent present, the BIG Synchronisation Point remains at the end of the
+last BIS Subevent transmission, whilst the BIG event extends to the end of the Control
+Subevent, as shown in Figure 4.32.
+
+108
+
+Chapter 4 - Isochronous Streams
+
+Figure 4.32 BIG synchronisation
+
+At the BIG Synchronisation Point, every device that is listening to any of the Broadcast
+Isochronous Streams within that BIG will know that every other device has received its data;
+hence the BIG Synchronisation Point is a fixed point in time at which they can apply the
+Presentation Delay. This is defined higher up the stack and dictates the point at which audio
+needs to be rendered. The important point here is that audio is not rendered at the BIG
+Synchronisation Point – it’s rendered at the end of the Presentation Delay, which commences
+at the BIG Synchronisation Point. As broadcasting consists only of transmissions from a
+Broadcast Source, there is no concept of Presentation Delay applied to captured data coming
+back from an Acceptor.
+
+Figure 4.33 The application of Presentation Delay in a BIG
+
+The derivation of the BIG Synchronisation Point is also a little different for broadcast. Unlike
+unicast, every BIS has the same basic timing parameters – that’s because they need to be
+defined in the BIGInfo and there is not enough space to allow different settings for individual
+BISes. This means that the BIG Synchronisation Point is a fixed interval from the BIG Anchor
+109
+
+Section 4.4 - Broadcast Isochronous Streams
+Point, both of which are known by every Acceptor, as they’re in the BIGInfo. (In contrast,
+each CIS uses a CIS_Sync_Delay based on its own Anchor Point, as it doesn’t know the
+relative timings of any other CISes and can’t assume they are the same as its own).
+
+4.4.5
+
+HCI Commands for BISes
+
+As with Connected Isochronous Streams, a Host application can define the SDU interval, the
+maximum SDU size and the Maximum Transport Latency, which is the maximum time
+allowed for the transmission of a BIS data PDU. AS with a CIG, it is up to the scheduler in
+the Controller to use these to guide it in defining the actual Link Layer parameters, i.e., NSE,
+BN, IRC and PTO.
+Setting up a BIG is much simpler than a CIS, largely because there is no communication with
+any of the receiving devices, so only two HCI commands are required. LE_Create_BIG
+configures and creates a BIG, with a Num_BIS number of BISes within it, and the
+LE_Terminate_BIG command ends and removes it. There are no options to add or remove
+BISes – everything is done in a single operation. Both commands apply only to an Initiator
+which is transmitting one or more BIGs.
+There are two similar commands for an Acceptor which wants to receive one or more BISes
+from within a BIG, a process which is called Synchronising with a BIS. They are the
+LE_BIG_Create_Sync Command and LE_BIG_Terminate_Sync Command. But before we
+get to those, we need to look at how an Acceptor can find a Broadcaster and connect to it.
+
+4.4.6
+
+Finding Broadcast Audio Streams
+
+When we looked at connected Isochronous Streams, we didn't talk about how the devices
+initiated the connection. The reason is that devices using Connected Isochronous Streams
+connect in exactly the same way as every other Bluetooth LE device, using the normal
+advertising and scanning procedures. They pair, bond, set up ACL links, discover each other's
+features and then get on with setting up the streams. If they are a Coordinated Set, CAP
+procedures ensure that all members of that set have the same procedures applied to them.
+With broadcast Audio Streams, none of that pairing and negotiation process happens, because
+there is no connection. Instead, receiving devices need to find a way to discover what is being
+transmitted, when it is being transmitted, and work out how to synchronise to it.
+The method for doing this is called Extended Advertising and was added to Core 5.1.
+Returning to basics for a moment, Figure 4.34 shows the channels which are used for
+Bluetooth LE. For Bluetooth LE, the 2.4 GHz spectrum is divided into 40 channels, each 2
+MHz wide. Three of these are reserved for advertising at fixed frequencies - channels 37, 38
+and 39 and are known as the Primary Advertising Channels.
+
+110
+
+Chapter 4 - Isochronous Streams
+
+Figure 4.34 Bluetooth® LE channels
+
+The position of these three channels was chosen to avoid the most commonly used parts of
+the Wi-Fi spectrum. In between the three advertising channels are 37 general purpose
+channels, numbered from 0 to 36.
+Broadcasters need to provide a fair amount of information if other devices are going to be
+able to receive their transmissions. Not only do they need to tell the receivers where the
+transmissions are located in terms of hopping channels, but they also need to provide all of
+the details about the structure of the BIG and its constituent BISes, which we have just
+covered. In many situations, there are likely to be multiple Broadcast transmitters within range
+of an Acceptor, particularly where they're being used in public places, and hence multiple
+broadcast Isochronous Streams to choose from. Acceptors need to be able to differentiate
+between them, implying that the broadcast advertisements contain information about the
+content of each broadcast stream. All of this requires a large amount of information to be
+conveyed by the Broadcaster. It is too much to fit into the three primary advertising channels,
+i.e., channels 37, 38 and 39, without the risk of overloading them.
+To overcome this limitation, the Extended Advertising scheme of Core 5.1 allow advertising
+information to be moved to the general-purpose channels - channels 0 to 36. To accomplish
+this, a Broadcaster includes an Auxiliary Pointer (AuxPtr) in its primary advertisements, which
+informs devices that further advertising information is available in the general-purpose
+channels (which now serve as Secondary Advertising channels). The Auxiliary Pointer
+provides the information that scanning devices need to determine where these secondary
+advertisements are located. At this point a scanner doesn’t know whether the AuxPtr is for a
+broadcast. That only becomes apparent when it finds and reads the Extended Advertisement.
+Figure 4.35 illustrates the basic structure of this scheme.
+
+111
+
+Section 4.4 - Broadcast Isochronous Streams
+
+Figure 4.35 Extended advertising for a BIG
+
+The Extended Advertising process is built up using three types of advertising PDU:
+•
+
+•
+
+•
+
+ADV_EXT_IND: Extended Advertising Indications, which use the primary
+advertising channels. Each BIG requires its own advertising set and
+ADV_EXT_IND . The headers of each ADV_EXT_IND include the advertising
+address (AdvA), the ADI (which includes the Set ID for this set of Extended
+Advertisements) , and an Auxiliary Pointer to the:
+AUX_ADV_IND: Auxiliary Advertising Indications. These are transmitted on the
+37 general purpose channels. They contain the Advertising Data specific to a BIG.
+The AUX_ADV_IND also include a SyncInfo field, which provides information
+for scanners on how to locate the:
+AUX_SYNC_IND: Auxiliary Synchronisation Information. These advertisements
+contain two important pieces of information. The first is the Additional Controller
+Advertising Data field (ACAD). This is information which is required by the
+Controller of the receiving device, describing the structure of the BIG and its
+constituent BISes. The second is information for the Host of the receiving device,
+which is contained in the AdvData field. This contains the Basic Audio
+Announcement Service UUID and the BASE, which contains a detailed definition
+of the streams in the BIG, including the codec configuration and metadata which
+describes the use cases and content of the audio streams in a user readable format.
+
+The presence of the Auxiliary Pointer (AuxPtr) in an ADV_EXT_IND packet indicates that
+some or all of the advertisement data can be found in an Extended Advertisement. However,
+the amount of information that's needed to describe and locate the broadcasting stream is still
+bigger than what normally fits into an Extended Advertising packet. To accommodate this,
+the Extended Advertisement could chain further packets by including an AuxPtr to an
+Auxiliary AUX_CHAIN_IND PDU, but it doesn’t. Instead, it sets up a Periodic Advertising
+train, which includes all of the information about a BIG and its BISes within an
+AUX_SYNC_IND PDU. The AUX_SYNC_IND packets in a Periodic Advertising train are
+112
+
+Chapter 4 - Isochronous Streams
+transmitted regularly, so once a scanning device has found them, it can continue tracking them,
+without having to refer back to the Primary or Extended Advertisements. The Broadcaster
+can even turn off its Extended Advertisements, but keep the Periodic Advertising train
+running.
+The presence of this Periodic Advertising train is indicated by a Syncinfo field in the
+AUX_ADV_IND PDU of the Extended Advertisement, which provides information on
+where to find it. Should the amount of information for the Periodic Advertising train be too
+large to fit into a single PDU, the AUX_ADV_IND can also use an AdvPtr to point to an
+Auxiliary AUX_CHAIN_IND PDU. In general, the AUX_CHAIN_IND PDUs are not
+needed.
+The data needed to locate and receive a broadcast Audio Stream starts in the Extended
+Advertisements. The AUX_ADV_IND contains the Broadcast Audio Announcement
+Service UUID, which identifies the Extended Advertisements as belonging to a specific
+broadcast Audio Stream, with a 3 octet Broadcast_ID which identifies it. The Broadcast_ID
+remains static for the life of the BIG. The AUX_EXT_IND may include additional Service
+UUIDs from top level profiles. These allow a scanning device an early opportunity to filter
+the different broadcasts it finds, so that it doesn’t need to synchronise with the PA and decode
+the information it carries for broadcasts that are not of interest to it. That feature is particularly
+useful to help choose public broadcasts, which are defined in the Public Broadcast Profile
+(PBP), which defines the Public Broadcast Announcement UUID.
+Having decided it likes the look of the BIG, a scanner will synchronise to the Periodic
+Advertising train. These advertisements use the general-purpose channels 0 to 36. Like
+Extended Advertisements, they can be chained to include more data, if required, using another
+AuxPtr. For broadcast audio, they contain two vital pieces of information.
+The first is the Additional Controller Advertising Data field, the ACAD. The ACAD contains
+advertising data structures. These are data structures containing an Advertising Data (AD)
+type and its corresponding data. All devices which are actively broadcasting will use the ACAD
+to send the BIGInfo structure, which provides all of the information needed by a receiving
+device to detect the broadcast and understand the BIS timings. This is the information which
+would be delivered to Acceptors at the Link Layer in a CIG, but with no direct connection in
+broadcast, that’s not possible. Hence this alternative method.
+
+113
+
+Section 4.4 - Broadcast Isochronous Streams
+
+Figure 4.36 The BIGInfo structure
+
+Figure 4.36 shows the structure of the BIGInfo. It starts with the BIG_Offset, which informs
+devices exactly where they can find the BIG in relationship to this AUX_SYNC_IND PDU.
+Following that, the BIGInfo describes the structure of that BIG:
+•
+•
+•
+•
+•
+
+The fundamental spacing – the ISO_Interval, the Sub_Interval and the BIS Spacing
+The number of BISes in the BIG (Num_BIS) and their features - NSE, Burst
+Number, PTO, IRC and Framing,
+The packet information - Max_PDU, SDU_Interval, Max_SDU and
+bisPayloadCount.
+The physical channel access information, in the form of the SeedAccessAddress,
+the BaseCRCInit, the Channel Map (ChM), and the PHY, as well as the
+The Group Initialization Vector (GIV) and the Group Session Key Derivation
+(GSKD) to allow encrypted broadcast streams to be decoded. If these last two
+fields are present, a scanning device knows that the broadcast is encrypted, which
+means it will need to acquire the Broadcast_Code (described below) to decrypt the
+audio streams.
+
+The presence of the GIV and GSKD is an easy way to tell whether a BIG contains encrypted
+BISes. If they are absent, the BIGInfo is shorter, indicating that the BISes are not encrypted.
+Before we leave BIGInfo, there is one important thing to repeat, which is that all of the BISes
+have exactly the same parameters. This is in contrast to CIGs, where multiple CISes can have
+different parameters. In a BIG, one setting applies to every BIS that is being used, which has
+to be specified to meet the maximum requirements of the different BISes.
+114
+
+Chapter 4 - Isochronous Streams
+
+4.4.7
+
+Synchronising to a Broadcast Audio Stream
+
+Having acquired the BIGInfo, devices can use this information to find out where those
+BISes are. The BIG_Offset, together with the BIG_Offset_Units tells receivers where the
+Anchor Point of the first BIS will be located after the start of the BIGInfo packet
+(remember that with Broadcast, there is no ACL to provide a timing instant, as there is with
+a CIS). An Acceptor can receive any number of BISes within a BIG. In Bluetooth LE
+Audio, this is called synchronising with a BIG or a BIS. Figure 4.37 shows how this
+information is used.
+
+Figure 4.37 Synchronisation timing for a BIG
+
+In most real life situations, a Broadcast Sink would not want to synchronise with all of the
+BISes. An earbud or hearing aid would normally only want to synchronise to the left or the
+right Audio Streams of a BIG which included a stereo stream, whereas a pair of headphones
+would want to synchronise to both left and right channels. To determine which BIS or BISes
+to connect to, a scanning device needs to look at a second important piece of information
+which is included in the Periodic Advertising AUX_SYNC_IND PDUs. This is the Broadcast
+Audio Source Endpoint structure (BASE), which is included in the AdvData field of each
+AUX_SYNC_IND PDU, following the Basic Audio Announcement Service UUID.
+
+4.4.8
+
+BASE – the Broadcast Audio Source Endpoint structure
+
+To understand the BASE, we need to move up into the Basic Audio Profile, which is the first
+profile that we'll encounter within the Generic Audio Framework (GAF). It is the key profile
+in terms of configuring and managing Audio Streams, both for Connected Isochronous
+Streams and Broadcast Isochronous Streams.
+BIGInfo describes the structure of the BIS, telling devices how many BISes it contains and
+where to find them, but that is just the practical information. It doesn’t provide any
+information about what is in the broadcast Audio Stream. The BASE provides that detail,
+telling devices what audio information is contained in each of the BISes and how it is
+configured. Where more than one Broadcaster is within range, that’s vital if a receiver is going
+to be able to choose which one it listens to. The BASE consists of three levels of information,
+115
+
+Section 4.4 - Broadcast Isochronous Streams
+as shown in Figure 4.38. We will revisit this in more detail when we look at how to set up a
+Broadcast Stream.
+
+Figure 4.38 Simplified BASE structure
+
+The highest level – Level 1, provides information which is valid for every BIS in the BIG – a
+single Presentation Delay and the number of Subgroups that the constituent BISes are divided
+into. These Subgroups contain BISes which have common features, which may be the way
+they are encoded, or the language of the Audio Streams.
+Each Subgroup is described in more detail at Level 2 of the BASE, where the codec
+information is provided for the BISes in that Subgroup, as well as stream metadata in the form
+of LTV structures. Much of this is data strings in human readable form, such as programme
+information, and it’s recommended that the metadata includes the ProgramInfo LTV as a
+minimum, so that a scanning device which is capable of displaying it can use it to help a user
+select between different audio streams. This is also where a language LTV would be provided.
+Finally, Level 3 of the BASE provides specific information for each BIS. This includes its
+BIS_Index, which identifies the order in which it appears within the BIG, as well as its
+Channel_Allocation, which identifies the Audio Locations it represents, such as Left or Right.
+Level 3 can include a different set of codec information for specific BISes, which will override
+the Subgroup value of Level 2. This is only likely to be used where one member of the BIG
+uses a different setting from the other BISes.
+The information contained in the BASE allows a Broadcast Sink to determine whether it wants
+to receive a BIG and which of the BISes contained within that BIG it would like to synchronise
+to, as well as telling it where that specific BIS is located within the overall BIG. Once it has
+parsed the BIGInfo and BASE, the device has all of the information it needs to synchronise
+to any of the BISes
+In most cases, a Broadcast Source will only transmit a single BIG, but it can transmit multiple
+BIGs. This might occur with public information broadcasts, where different types of messages
+116
+
+Chapter 4 - Isochronous Streams
+might be contained in different BIGs. For example, an airport might use one BIG for flight
+announcements, a second for general security announcements and a third, encrypted one for
+staff announcements. In this case, each BIG will have its own set of primary advertisements,
+each with a different Advertising Set ID (SID), along with its own extended advertising train
+containing its specific Broadcast_ID, BIGInfo and BASE (see Figure 4.39).
+
+Figure 4.39 Advertising for multiple BIGs
+
+The SID should not change often – it is typically static until a device goes through a power
+cycle and is often static for life. The Broadcast_ID is static for the life of the BIG. Broadcast
+Sinks can take advantage of these IDs to reconnect to a Broadcaster that they know if they
+recognise the SID and Broadcast_ID.
+
+4.4.9
+
+Helping to find broadcasts
+
+The expectation is that broadcast audio will be used in a very different way to unicast audio.
+The hearing aid industry is keen that Bluetooth LE Audio broadcast is used to complement
+and extend current telecoil infrastructure, making installation more cost effective for venues.
+Today, most telecoil usage is for sound reinforcement, specifically for people wearing hearing
+aids. In the future, as the same broadcast transmissions can be picked up by consumer earbuds
+and headphones, far more public audio information services are likely to be deployed. It’s
+also likely that Bluetooth LE Audio broadcast will become standard in TVs, both in public
+locations, such as gyms and bars, as well as at home. The Bluetooth SIG’s Audio Sharing
+program is promoting broadcast as a solution for shared music in cafes and for personal music.
+
+117
+
+Section 4.4 - Broadcast Isochronous Streams
+If the market develops in this way, users will often find themselves within range of multiple
+Broadcast Sources, which will mean they need to choose which to receive. A first level of
+selection can be provided by profile UUIDs in AUX_ADV_IND packets, but devices will still
+need to detect and parse the BASE information for each BIG to make a decision. (In the early
+days, it may be possible to select the first BIG you find, then press a button to move to the
+next, but that is not a scalable user experience for the long term.) This presents product
+designers with a problem, as devices like earbuds have no room for a display and often barely
+room for any buttons. In addition, the process of scanning is relatively power hungry –
+constant scanning would have a noticeable impact to an earbud or hearing aid’s battery life.
+To get around this limitation, the Bluetooth LE Audio specifications have introduced the
+concept of a Commander – a role which performs the scanning operation, allows a user to
+select a broadcast and then instruct a receiving device to synchronise to that selected BIS or
+BISes. The Commander role can be implemented as part of an app in a mobile phone, in a
+smart watch, an earbud case or a dedicated remote-control device. In fact, any Bluetooth LE
+device which has a connection with a Broadcast Sink can act as a Commander. Devices which
+implement the Commander role are called Broadcast Assistants and are defined in the
+Broadcast Audio Scanning Service (BASS). They can be integrated into devices like TVs to
+help automate the connection to earbuds and headphones.
+
+4.4.10
+
+Periodic Advertising Synchronisation Transfer – PAST30
+
+A Broadcast Assistant scans for advertisements which indicate the presence of Extended
+Advertisements in exactly the same way as any other scanning device. Once it discovers them,
+it can synchronise to the associated Periodic Advertising train, which contains a Broadcast
+Audio Announcement Service UUID and then discover the accompanying BIGInfo and
+BASE structure. It may apply filters to its scan before reading and parsing the BASE metadata
+of those it finds, after which it provides the list of available broadcast streams to the user,
+generally by displaying the human readable ProgramInfo.
+Once the user has made their selection, the Broadcast Assistant will use the PAST procedure
+to provide the Broadcast Receiver with the information it needs to find the relevant Periodic
+Advertising train. This allows the Broadcast Receiver to jump straight to those advertising
+packets, acquire the BIGInfo and BASE and synchronise to the appropriate BISes without
+having to expend energy in scanning. This process is called Periodic Advertising
+Synchronization Transfer or PAST.
+The level of information provided by a Broadcast Assistant to the user is entirely down to the
+implementation. A phone app could list every Broadcaster within range; it could limit the
+
+The Core does not use the acronym PAST for Periodic Advertising Synchronisation Transfer, so if
+you’re searching the Core, you need to use the full name. The PAST acronym is introduced in BAP.
+30
+
+118
+
+Chapter 4 - Isochronous Streams
+amount of information displayed based on user preferences, or use preconfigured settings
+which it had read from a set of earbuds. The Broadcast Assistant could equally be a button
+on a watch or fitness band which selects the last known synchronised stream from the list of
+Broadcast Sources it has found. Broadcast Assistants can also include applications to obtain
+the required Broadcast_Code to decrypt private, encrypted broadcasts, either by making a
+Bluetooth connection to the Broadcaster or a proxy, or by using an out of band (OOB)
+method.
+
+4.4.11
+
+Broadcast_Code
+
+Encrypted streams are not new in Bluetooth technology; security and confidentiality have
+always been an important part of the specifications. However, implementing encryption when
+there is no connection between the transmitting and receiving devices, as is the case with the
+broadcast streams in Bluetooth LE Audio, introduces a new problem, which is how to cope
+with encryption.
+Many broadcast streams will not be encrypted; they will be broadcast openly, so that anyone
+can pick them up. However, that generates a few issues. Firstly, anyone who is within range
+can receive them. As Bluetooth technology is quite efficient in penetrating walls, that can be
+an issue in meeting rooms, hotel rooms and even homes, where someone in a neighbouring
+room could inadvertently pick up the audio from your TV. That could range from annoying
+to embarrassing, depending on what the content is. In a business environment, it may well be
+confidential, so adding encryption is vital. Basic confidentiality is inherent within telecoil
+systems, as the audio transmission can only be picked up within the confines of the induction
+coil. To provide the same level of authentication in Bluetooth LE Audio broadcasts, the audio
+data needs to be encrypted, which requires the receiver to obtain the decryption key, which is
+known as the Broadcast_Code.
+Devices looking for broadcasts can detect whether a broadcast Audio Stream is encrypted by
+examining the length of the BIGInfo in the periodic advertising chain. If the stream is
+encrypted, the packet will include an additional 24 octets, containing a Group Initialisation
+Vector (GIV) and a Group Session Key Diversifier (GSKD). Scanners will detect their
+presence and, in conjunction with other metadata, determine whether or not to synchronise
+with such a stream, depending on whether they are able to retrieve the Broadcast_Code.
+Although broadcast does not need a Broadcast Source and Sink to be paired, in many cases
+there will be an ACL connection present. That may sound strange, but it’s an arrangement
+that allows the number of encrypted connections to be scaled far beyond what is possible with
+unicast, without hitting an airtime limit. This is expected to be the way most domestic TVs
+will work, so that neighbours can’t hear what you are listening to. In this case the users would
+be paired to the TV, using the features of BASS to obtain the Broadcast_Code. The
+Broadcast_Code is a property of the Host application. The Host provides it to the Controller
+when it is setting up the BISes, and it can equally supply it to trusted devices. In public spaces,
+conference rooms and hotels, it is likely that an out-of-band method would be used, probably
+119
+
+Section 4.4 - Broadcast Isochronous Streams
+similar to the printed information which is used to connect to a Wi-Fi access point.
+Broadcast_Codes can also be obtained through other out of band methods, such as scanning
+QR codes, tapping an NFC terminal, or even being included with a downloadable theatre
+ticket. We’ll explore these configurations in more detail in Chapter 12
+Note that the metadata used to describe broadcasts within advertising packets is not encrypted.
+Accordingly, care should be taken to ensure that it does not contain confidential or potentially
+embarrassing information.
+
+4.4.12
+
+Broadcast topologies
+
+The features included in Bluetooth LE Audio provide a wide range of options for finding
+broadcasts. We will look at them in more detail in later chapters, but the figures below show
+some of the common ones. In most cases, they show a single “Broadcast Information”
+stream, which encompasses all of the advertising information
+
+Figure 4.40 Direct synchronisation from headphones
+
+The simplest case, where a pair of headphones does its own scanning to find and select a
+broadcast audio stream, is shown in Figure 4.40.
+
+Figure 4.41 Using a phone as a Broadcast Sink
+
+Figure 4.41, is essentially the same, but points out the Broadcast Sink could also be a phone.
+Here, it can scan for Broadcasters, display the available choice of broadcast Audio Streams to
+the user and then render the audio to a pair of wired headphones. It could equally transmit
+the streams using Bluetooth (either Bluetooth Classic Audio or Bluetooth LE Audio), although
+that is unlikely to be an efficient use of airtime.
+
+120
+
+Chapter 4 - Isochronous Streams
+
+Figure 4.42 Using a Broadcast Assistant and PAST
+
+Figure 4.42 shows what is expected to be the most common use case, where a device like a
+phone, watch or remote control acts as a Broadcast Assistant to scan and present the choice
+of broadcast Audio Streams to the user, then uses PAST to allow the user’s headphone or
+earbuds to connect to the selected stream. As we will see, the Broadcast Assistant can also be
+used for volume control and mute, allowing a user to find, select and control the rendering of
+a broadcast Audio Stream. It is illustrated in Figure 4.43.
+Although there is no connection between a Broadcast Source and a Broadcast Sink for the
+Audio Stream, devices can use ACL connections to help synchronise to the BIS, so that a TV
+could automatically synchronise to an Audio Stream when you enter a room. In this case, the
+Broadcast Source would normally also contain a Broadcast Assistant, which would be paired
+to a user’s headset. When the user chooses to connect to the Broadcast Assistant, it would
+use PAST to provide details of how to synchronise to the BISes and BASS to transfer the
+Broadcast_Code.
+
+Figure 4.43 Collocation of a Broadcast Source with a Broadcast Assistant
+
+This is one of the topologies that forms the basis of Audio Sharing, where a number of friends
+can share music from one phone. Chapter 12 explains these options in greater detail.
+
+121
+
+Section 4.5 - ISOAL – The Isochronous Adaptation Layer
+
+4.5
+
+ISOAL – The Isochronous Adaptation Layer
+
+ISOAL [Core Vol 6, Part G] is one of the most complicated aspects of the Core Isochronous
+Streams feature. The good news is that it’s taken care of for you in the Bluetooth chips, but a
+knowledge of why it’s necessary and what it does is useful. ISOAL is used for both broadcast
+and unicast Audio Streams.
+ISOAL exists to solve the problem of what happens when there is a mismatch between frame
+sizes. The most common reason for this is that an Acceptor only supports a 10ms frame size,
+but an Initiator needs to use 7.5ms frames, as it is running connections with other Bluetooth
+Classic devices, such as older mice and keyboards, which are only capable of running at a
+7.5ms timing interval31. That will change in time, as new peripheral devices become more
+flexible, but until then, ISOAL provides a means to accommodate them while the whole
+Bluetooth ecosystem migrates to a 10ms timing interval.
+The ISOAL layer sits in the data path between the codec and the Link Layer. Encoded SDUs
+from the codec may be delivered via the HCI if the codec is implemented in the Host, or
+through a proprietary interface (regardless of where the codec is situated). As Figure 4.44
+shows, ISOAL provides fragmentation and recombination or segmentation and reassembly
+and is responsible for sending the encoded audio data in either framed or unframed PDUs.
+
+Figure 4.44 Architecture of the Isochronous Adaptation Layer (ISOAL)
+
+Classic Bluetooth is based on a 2.5ms interval, but many applications standardised on 7.5ms, which
+causes this problem for dual-mode devices.
+31
+
+122
+
+Chapter 4 - Isochronous Streams
+Depending on the setting provided by the Host layer, the Controller can decide whether to
+use framed or unframed PDUs. For unframed PDUs, if an SDU can fit within a single PDU,
+the Isochronous Adaptation Manager may fragment them, but sends them without a
+segmentation header. That’s the most efficient, lowest latency way to transport Isochronous
+data. If the SDU is larger than the Maximum PDU size, it will be segmented and sent in
+multiple unframed SDUs. The recombination or reassembly processes reassemble them into
+SDUs. Unframed PDUs can only be used when the ISO_Interval is equal to or is an integer
+multiple of the SDU_Interval, which is itself equal to or an integer multiple of the sampling
+frame. This means that generation of the SDUs need to be synchronised with the transport
+timing, so that they don’t drift with respect to each other. Otherwise, you need to use framed
+SDUs.
+For framed SDUs, the Isochronous Adaptation Manager adds a segmentation header and an
+optional Time_Offset. The Time_Offset allows multiple SDUs to be segmented across
+PDUs, providing a reference time that maintains an association between the SDU generation
+and transport timing. If you need more detail, the full specification of ISOAL is contained in
+Vol 6, Part G, Section 6 of the Core.
+The main aspect of ISOAL that designers need to be aware of is that it contains the set of
+equations which define the Transport_Delay. Transport_Delay is the time between an SDU
+being presented for transmission and the point where it is ready for decoding at the
+appropriate Synchronisation Reference.
+For framed SDUs, the CIG transport latencies are:
+Transport_Latency = CIG_Sync_Delay + FT × ISO_Interval + SDU_Interval
+Separate calculations need to be made using the respective values of CIG_Sync_Delay, FT
+and SDU_Interval for the Central to Peripheral and Peripheral to Central directions.
+For a BIG using framed SDUs,
+Transport_Latency = BIG_Sync_Delay + (PTO × (NSE÷BN–IRC)) × ISO_Interval
++ ISO_Interval + SDU_Interval
+For unframed SDUs, the calculations are slightly different. For a CIG:
+Transport_Latency = CIG_Sync_Delay + FT × ISO_Interval - SDU_Interval
+Again, you need to use the respective values of CIG_Sync_Delay, FT and SDU_Interval for
+the Central to Peripheral and Peripheral to Central directions.
+
+123
+
+Section 4.5 - ISOAL – The Isochronous Adaptation Layer
+For an unframed BIG,
+Transport_Latency = BIG_Sync_Delay + (PTO × (NSE÷BN-IRC) + 1) ×
+ISO_Interval - SDU_Interval
+--oOo-That concludes the basics of Isochronous Streams. Now we need to look at the LC3 and see
+how QoS choices influence robustness and latency.
+
+124
+
+Chapter 5 - LC3, latency and QoS
+
+Chapter 5. LC3, latency and QoS
+5.1
+
+Introduction
+
+Two of the most debated aspect of wireless audio, particularly amongst audiophiles, are audio
+quality and latency. In its early years, Bluetooth technology was often criticised for both,
+although in most cases the audio quality probably had more to do with the state of the
+transducers rendering the audio than anything to do with the Bluetooth specifications.
+Transmitting audio over any wireless connection involves compromises. Interference is a fact
+of life, which means that some of the audio data will be lost. That results in gaps in the audio
+unless you take measures to add redundancy. Typically, that involves transmitting the audio
+packets more than once, so that there are multiple chances that one of them will get through.
+However, to be able to do that, you have to be able to compress the audio, so that you have
+time to transmit multiple copies. That’s done using codecs (which is a portmanteau word for
+coder and decoder).
+The coder takes in an analogue signal, digitises it and compresses the digital data, so that it can
+be transmitted in a shorter time than the length of the original sample. This means that it can
+be transmitted multiple times before the next sample is taken. If the first transmission is lost
+or corrupted, the following retransmission can be used in its place. The decoder, in the
+receiving device, decodes the received data, expanding it to regenerate the original audio signal.
+As it takes time to perform the encoding and decoding, this results in a delay between the
+original signal and the reconstituted signal coming out of the decoder.
+Audio codecs are a relatively recent invention. The first hundred years of audio transmission,
+from the 1860’s phonoautograph, through radio broadcasts, vinyl records and magnetic tape,
+all worked with the original audio signal. If there was interference from the weather, a scratch
+on a record or wax cylinder, or stretching on a tape, the sound was lost or distorted. This
+changed with the introduction of CDs, which were made possible by the development of pulse
+code modulation (PCM), which converts an analogue signal to a digital signal.
+PCM works by sampling the audio signal at a higher frequency than we can hear (44,100 times
+per second for a CD), converting each sample into a digital value. Decoding performs this
+operation in reverse, using a digital to audio decoder to restore the analogue signal. The more
+bits in each sample, the closer the output audio will be to the original input. CDs, along with
+most audio codecs, use 16 bit samples. Where the sampling rate and the bits per sample are
+sufficiently high, the human ear can’t detect the difference. However, the file sizes for a pure
+PCM digital file are large, as there is no compression involved. Sampling at 44.1kHz and
+16bits generates 800kbits every second, so a five minute song in mono is around 26MB, or
+52MB for stereo. That’s what limits a standard CD to around an hour of music.
+The arrival of the MP3 audio codec, developed by the Fraunhofer institute, transformed the
+distribution of digital music. It uses a technique called perceptual coding (sometimes also
+125
+
+Section 5.2 - Codecs and latency
+called psychoacoustic modelling) which compares the audio stream with a knowledge of what
+a human ear can actually hear. That may be a high frequency sound which is above the range
+most people can hear, so can be encoded with less data, or a held note, where an encoder can
+indicate that you just need to repeat the previous sample or encode a difference from it. By
+applying these methods, it is possible to significantly reduce the size of a digitised audio file.
+MP3 typically reduces the size of a digitised music file by between 25% and 95%, depending
+on the content. Few listeners were worried about any slight loss of quality from this process,
+feeling that the increased convenience far outweighed any noticeable effect on the music.
+The reduction in the size of music files led to the creation of music sharing services like
+Napster and the appearance of MP3 players. It also fired the starting pistol for the
+development of streaming services and wireless audio transmission, as the reduced file sizes
+meant that there was plenty of time to retransmit the compressed audio packets, helping to
+cope with any interruptions to transmission.
+
+5.2
+
+Codecs and latency
+
+One downside to the use of codecs is that they add latency to the signal. This is a delay
+between the arrival of the original analogue signal at a transmitter and the rendering of the
+reconstituted signal at the receiver.
+
+Figure 5.1 The elements of latency in an audio transmission
+
+Figure 5.1 shows the elements that make up that latency. First, the audio is sampled.
+Perceptual coding requires a codec to look at multiple, consecutive samples, as a lot of the
+opportunities for compression come from identifying periods of repeated sound (or lack of
+sound). This means that most codecs need to capture sufficient, successive samples to have
+enough data to characterise these changes. This period of sampling is called a frame. Different
+encoding techniques use different frame lengths, but it’s almost always a fixed duration. If it’s
+too short, the limited number of samples starts to reduce the efficiency of the codec, as it
+doesn’t have enough information to apply the perceptual coding techniques, which impacts
+the quality. On the other hand, if the frame sizes grow, the quality improves, but the latency
+increases, as the codec has to wait longer to collect each frame of audio data.
+
+126
+
+Chapter 5 - LC3, latency and QoS
+
+Figure 5.2 The sweet spot for audio codec frame size
+
+Figure 5.2 illustrates the trade-off. This will vary from codec to codec, depending on how
+they perform the compression, but for a general purpose codec which can be used for both
+voice and music, the industry has found that there is a sweet spot for the frame length of
+around 10ms, which gives good quality at a reasonable latency.
+There is another trade-off, which is the amount of processing power that you need to run the
+codec, which is known as the complexity. As you try to squeeze more audio quality out of the
+codec, you need a faster processor, which starts to reduce the battery life. That may not be a
+problem on a phone or PC, but if you are encoding the microphone input of a hearing aid or
+earbud, it’s a very serious problem.
+Returning to Figure 5.1 and the general principles of wireless audio transmission, once the
+audio frame has been encoded, the radio will transmit it to the receiving device. The
+transmission is normally quick compared to the encoding, but if the protocol contains
+retransmission opportunities, you need to allow for these before you start decoding. The
+duration between the start of transmitting the first time, to the end of the last transmission
+being received is called the Transport Delay and can range from a few milliseconds to several
+tens of milliseconds. (You can start decoding as soon as you receive the first packet, but if
+you do you will need to buffer it. That’s because the output audio stream needs to be
+reconstructed to have no gaps, so it must be delayed until every opportunity for a
+retransmission has passed, to cope with the instances when a packet needs the maximum
+number of retransmissions to get through. Otherwise packets which arrive early will be
+rendered early, while others won’t.)
+Finally, after the encoded audio data has been received, it needs to be decoded and then
+converted back to analogue form to be rendered. Decoding is normally quicker than encoding
+and doesn’t have a frame delay, as the decoder expands the output frame automatically. It
+127
+
+Section 5.3 - Classic Bluetooth codecs – their strengths and limitations
+generally uses far less power than encoding, as most codecs are designed for use cases where
+a file is encoded once at production, then decoded many times (as when you’re streaming
+music from a central server), so there is an inherent asymmetry in the design.
+
+5.3
+
+Classic Bluetooth codecs – their strengths and limitations
+
+The existing Bluetooth audio profiles were both developed with specific requirements for their
+individual use cases, with different codecs optimised for each, as shown in Figure 5.3. The
+original HFP specification was designed to use a CVSD (Continuous Variable Slope Delta
+modulation) coding method, which is a low latency codec, widely used in telephony
+applications.
+
+Figure 5.3 Performance of HFP and A2DP profiles
+
+CVSD was one of the first methods for digitising and compressing voice. It samples rapidly
+- typically at 64,000 samples per second, but only captures the difference between the current
+sample and the preceding one. This means that it is frameless and has a comparatively short
+sampling and encoding delay. Similarly, the output decode can be performed quickly. The
+trade-off is that the quality is limited and because there is no compression it is effectively realtime, with no opportunity for retransmission.
+Later versions of HFP include mSBC - a modified version of the SBC codec specified in
+A2DP, to support wideband speech. mSBC is effectively a cut down version of SBC, with a
+limited sampling frequency for a single, monaural stream. Being a frame-based codec, it
+increases the latency, resulting in typical overall delays of around 30ms. These put HFP in the
+low latency, low to medium quality quadrant of Figure 5.3.
+In contrast, A2DP was designed for high quality music. It mandates the SBC (Sub Band
+Coding) Codec, which is a frame-based codec with fairly basic psychoacoustic modelling. It
+can produce very good audio quality, which is close to the limit of what an experienced listener
+can detect, compared to the original audio stream. The A2DP specification also allows the
+128
+
+Chapter 5 - LC3, latency and QoS
+use of alternative32 codecs which were developed by external companies or standards groups
+– a selection which includes AAC33 (which is used by Apple in most of their Bluetooth
+products), MP3 and ATRAC34, as well as an option for companies to use proprietary codecs.
+A number of these have become popular, of which the best known is the AptX range from
+Qualcomm. Almost all of these codecs have longer latencies.
+In Figure 5.3, A2DP is in the top right quadrant of the diagram, with a long latency. That is
+partly driven by the desire for using retransmissions to try to make the audio more robust.
+Whereas listeners used to accept glitches like scratches on records, they appear to be far less
+tolerant of the occasional “pop” or dropout in an audio stream. The simplest solution is to
+add more retransmissions and buffering, but that means that wireless music streaming typically
+has a latency of 100 – 200 ms, even if you’re streaming from a file on your phone or computer.
+Although the codec isn’t involved in this delay, the knowledge that it happens means that
+codec designers haven’t generally concentrated on improving latency, unless it’s for a specific
+application like gaming.
+Although a 100 – 200ms delay sounds excessive, for most music applications it’s not a
+problem. When streaming music, whether from a music player or an internet service, the user
+has nothing to indicate whether they’re listening in real time or not. As long as the music
+stream starts within a second of them pressing the Play button, and the music stream is
+continuous, without annoying interruptions, they’re happy. However, when the audio is a
+soundtrack for a video, they may notice a lip-synch problem, as a 200ms delay between seeing
+someone talking and hearing their voice looks wrong. Phone and TV manufacturers can
+address this by delaying the video to compensate for any audio delay. The Audio/Video
+Distribution Transport Protocol (AVDTP), which underlies the A2DP profile, contains a
+Delay Reporting feature which allows audio source devices to ask the receivers what the
+latency will be in the audio path. Knowing this, TVs and phones can delay the video, so that
+both sound and picture are synchronised. However, many TVs and earbuds have limited
+memory for audio or video buffering, so latencies above a few hundred milliseconds may be
+problematic.
+Even short audio delays can become a problem where a user can hear both the Bluetooth
+audio and also the original, ambient source of the sound. This has long been recognised by
+the hearing aid industry, where users are listening to live sound via a telecoil system in a theatre
+or cinema, but can also hear the ambient sound. The same problem occurs at home where a
+family watching the TV includes some members who are wearing hearing aids with support
+for wireless transmission and some who are not. Telecoil induction loops, which are currently
+used for these applications, are analogue, so exhibit virtually no delay. Moving to Bluetooth
+
+Referred to as “optional” in older specifications and “additional” in more recent ones.
+AAC is the Advanced Audio Codec
+34 ATRAC is the Adaptive Transform Acoustic Coding codec, developed by Sony
+32
+33
+
+129
+
+Section 5.3 - Classic Bluetooth codecs – their strengths and limitations
+requires a codec that is able to cover much more of the Quality / Latency spectrum than SBC.
+
+Figure 5.4 LC3 - a more efficient codec
+
+During the Bluetooth LE Audio development, it became apparent that the current Bluetooth
+codecs would struggle to meet the requirements. Not only were they limited in their quality
+and latency trade-offs, but SBC is not as efficient as earbud and hearing aid designers would
+like. It has a relatively low complexity, but takes up too much airtime, which has a major effect
+on the battery life of an earbud. That’s a problem for hearing aids which run on small zincair batteries. These are sensitive to both peak current and the length of a current burst for
+reception or transmission (Bluetooth chips can often consume more current receiving than
+transmitting.) If operating limits are exceeded for these batteries, their life can be drastically
+reduced. To address these limitations, the Bluetooth SIG went on a codec hunt, which
+resulted in the inclusion of LC3.
+Bluetooth LE Audio allows manufacturers to use other codecs, but LC3 is mandatory for all
+devices. The reason for this is to ensure interoperability, as every Audio Source and every
+Audio Sink has to support it. The full specification for the codec is published and falls under
+the Bluetooth RANDZ35 license, so anybody can write their own implementation and
+incorporate it into their Bluetooth product, as long as those products pass the Bluetooth
+Qualification process. Given its quality, that’s a powerful incentive to use it.
+
+A RANDZ licence is a Reasonable and Non-Discriminatory Zero fee license, which is how the
+Bluetooth IP is licensed. That doesn’t mean there are no conditions, but they are not arduous.
+They’re explained at the Bluetooth website.
+35
+
+130
+
+Chapter 5 - LC3, latency and QoS
+
+5.4
+
+The LC3 codec
+
+The LC3 is one of the most advanced audio codecs available today, providing enormous
+flexibility and covering everything from voice to high quality audio. Anyone can develop
+their own implementation of LC3 for a Bluetooth LE Audio product, although, given the
+specialised nature of writing and optimising a codec, very few people are ever likely to do
+that. Because of that, I’ll only provide an overview of what it does and how it works.
+There’s a more detailed introduction in the LC3 specification, along with around two
+hundred pages of specification detail for anyone who fancies doing their own
+implementation36.
+The LC3 specification is among the most successful attempts so far to cover the full range
+of audio quality and latency requirements for wireless audio in a single codec. It is optimised
+for a frame size of 10ms, and it is expected that all new applications, in particular public
+broadcast applications, will use the mandatory 10ms frames. It also works with a 7.5ms
+frame size to provide compatibility with Bluetooth Classic Audio applications which run
+with 7.5ms intervals (corresponding to EV-3 SCO packets). It also supports an extended
+10.88ms frame, to provide legacy 44.1kHz sampling. This is reduced to 8.163ms for a 7.5ms
+based system which needs to support 44.1kHz. However, these are specific variants to
+support legacy, or combined Bluetooth Classic Audio / Bluetooth LE Audio
+implementations.
+Feature
+
+Supported Range
+
+Frame Duration
+
+10ms (10.88 @ 44.1kHz sampling)
+7.5ms (8.163 @ 44.1kHz sampling)
+8kHz, 16kHz, 24kHz, 32kHz, 44.1kHz and 48kHz
+20 – 400 bytes per frame for each audio channel. The
+bitrate used is specified or recommended by the Bluetooth
+LE Audio profiles.
+16, 24 and 32. (The algorithm allows most intermediate
+values, but these are the recommended ones.)
+Unlimited by the specification. In practice, limited by the
+profile, implementation resources and airtime.
+
+Supported Sampling Rate
+Supported bitrates
+
+Supported bits per audio
+sample
+Number of audio channels
+Table 5.1 LC3 features
+
+Table 5.1 reproduces the key parameters from Table 3-1 of the LC3 specification.
+
+For those looking for a less complex explanation, there is an introductory video at
+www.bit.ly/LC3video.
+36
+
+131
+
+Section 5.4 - The LC3 codec
+The Bluetooth SIG has commissioned extensive audio quality testing from independent test
+labs to quantify the subjective performance of the LC3 codec. These show that at all sample
+rates, the audio quality exceeds that of SBC at the same sample rate, and provides equivalent
+or better audio quality at half the bitrate. The practical benefit is that the total size of LC3
+encoded packets is around half of the size of those for SBC for the same audio stream.
+Implementers can use that to their advantage, as it reduces the total airtime for transmission,
+saving battery life, or they can use it to increase the audio quality. It gives them more scope
+to play with parameters, particularly in power constrained devices like earbuds and hearing
+aids. The power saving also allows scope for adding extra functionality into earbuds, such as
+more advanced audio algorithms or physiological sensors, whilst retaining a long battery life.
+
+5.4.1
+
+The LC3 encoder
+
+Figure 5.5 provides a high-level view of the LC3 encoder.
+
+Figure 5.5 High level overview of the LC3 encoder
+
+The first element of the encoder is the Low Delay Modified Discrete Cosine Transform
+module (LD-MDCT). LD-MDCT is a well-established method of performing time to
+frequency transformations in perceptual audio coding. It is low delay (hence the LD), but it
+still takes time to convert the audio input sample into spectral coefficients and group the
+corresponding energy values into bands. It’s where a fair proportion of the codec delay
+comes from.
+One of the modules fed from the LD-MDCT is the Bandwidth Detector, which detects
+incoming audio signals which have previously been sampled at different coding rates. It can
+detect the commonly used speech bandwidths in voice communication, i.e., NB (Narrow
+Band: 0-4 kHz), WB (Wide Band: 0-8 kHz), SSWB (Semi Super Wide Band: 0-12 kHz), SWB
+(Super Wide Band: 0-16 kHz) and FB (Full Band: 0-20 kHz). If it detects a mismatch, it
+signals to the Temporal Noise Shaper (TNS) and Noise Level modules to forestall and avoid
+any smearing of noise into any empty upper spectrum.
+The main path for the frequency components generated by the LD-MDCT is into the
+Spectral Noise Shaper (SNS) where they are quantised and processed. The job of the SNS is
+132
+
+Chapter 5 - LC3, latency and QoS
+to maximise the perceptual audio quality by shaping the quantisation noise so that the
+eventual, decoded output is perceived by the human ear as being as close as possible to the
+original signal.
+The remaining modules in the encoder are largely responsible for controlling artefacts. The
+most difficult sounds for a codec to handle are ones with a sharp attack, such as percussion
+instruments. Those transients are such a difficult thing for codecs to deal with that
+castanets, glockenspiel and triangles are key test sounds which are used for assessing a
+codec’s performance. Part of the problem with sharp attack transients is that overall, these
+sounds show a fairly flat spectrum. The Attack Detector signals their presence to the
+Spectral Noise Shaper, so that it can inform the Temporal Noise Shaping module (TNS) of
+their presence. The TNS then reduces and potentially eliminates the artefacts for signals
+which have severe transients.
+The next stage is to determine the number of bits required to encode the quantised
+spectrum, which is the job of the Spectral Quantiser. It can be considered as an intelligent
+form of automatic gain control. It also works out which coefficients can be quantised to
+zero, which the decoder can interpret as silence. This process risks introducing some coding
+artefacts, which are addressed by the Noise Level module, using a pseudo random noise
+generator to fill any gaps, ensuring that everything is set to the proper level for the decoder.
+It also uses the input from the bandwidth detector to ensure that the encoded signal is
+restricted to the active signal region. Once that is done, the spectral coefficients are entropy
+encoded and multiplexed into the bitstream.
+One other component of the resulting bitstream is a resampled input. Performed at a fixed
+rate of 12.8 kHz, this is passed through a Long Term Post Filter (LPTF). For low bit rates
+this reduces coding noise in any frames which contain pitched or tonal information. The
+Long Term Postfilter (LTPF) module perceptually shapes quantization noise by controlling a
+pitch-based postfilter on the decoder side.
+An encoded LC3 frame does not contain any timing information, such as time stamps or
+sequence numbers. It is up to the system using the LC3 to control the timing of packets,
+which we saw when we looked at the Core.
+
+133
+
+Section 5.4 - The LC3 codec
+
+5.4.2
+
+The LC3 decoder
+
+The decoder is shown in Figure 5.6 and essentially reverses the process.
+
+Figure 5.6 Overview of the LC3 decoder
+
+The bandwidth information is used to determine which coefficients are zero, with the Noise
+Filling model inserting information for those which are inband. The Temporal Noise Shaper
+and Spectral Shaper process these, before the Inverse LD-MDCT module transforms them
+back to the time domain. The Long Term Post Filter is then applied, using the transmitted
+pitch information to define the filter characteristic.
+Before the received packets for each frame are decoded, the Controller generates a Bad
+Frame Indication flag (BFI) if it detects any errors in the payload, along with a payload size
+parameter for each channel. If the BFI flag is set, the decoder will skip the packet and signal
+that a Packet Loss Concealment (PLC) algorithm should be run to replace missing data in
+the output audio stream. Any errors detected during the decode, will also trigger the PLC.
+
+5.4.3
+
+Choosing LC3 parameters
+
+From a Bluetooth LE Audio design viewpoint, the closest most developers will come to the
+LC3 specification is the parameters which they use to configure it in their applications. These
+are a subset of the features of Table 5.1 and are shown in Table 5.2.
+LC3 Parameter
+
+Values
+
+Sampling Rate
+Bits per sample
+Frame Size
+
+8kHz, 16kHz, 24kHz, 32kHz, 44.1kHz or 48kHz.
+16, 24 or 32.
+7.5ms or 10ms
+(The actual size for 44.1kHz sampling is generated
+automatically when 7.5 or 10ms is set.)
+20 to 400
+
+Bytes per frame (payloads per
+channel)
+Number of audio channels
+Table 5.2 LC3 configuration parameters
+
+134
+
+Typically 1 or 2, but limited only by the profile or
+implementation.
+
+Chapter 5 - LC3, latency and QoS
+The bits per sample at the encoder and decoder are local settings and may be different. In
+most cases, they are set to 16.
+For unicast streams, an Acceptor can expose which combinations of these values it supports,
+along with a preference for which configuration is used with a specific use case. An Initiator
+makes a selection from the supported values exposed by each Acceptor to configure each
+audio stream.
+To limit these to a sensible number of combinations, BAP defines sixteen configurations for
+the LC3 codec to help drive interoperability. These are reproduced below in Table 5.3 and
+cover sampling frequencies of 8 kHz to 48 kHz. These can be found in BAP Tables 3.5
+(Unicast Server), 3.11 (Unicast Client) and 3.12 (Broadcast Source) and 3.17 (Broadcast
+Sink).
+Codec
+Configuration
+Setting
+
+Supported
+Sampling
+Frequency
+
+Supported
+Frame
+Duration
+
+Supported
+Octets per
+Codec Frame
+
+Bitrate
+(kbps)
+
+8_1
+8
+7.5 ms
+26
+27.734
+8_2
+8
+10 ms
+30
+24
+16_1
+16
+7.5 ms
+30
+32
+16_2 1
+16
+10 ms
+40
+32
+24_1
+24
+7.5 ms
+45
+48
+2
+24_2
+24
+10 ms
+60
+48
+32_1
+32
+7.5 ms
+60
+64
+32_2
+32
+10 ms
+80
+64
+441_1
+44.1
+7.5 ms
+97
+95.06
+441_2
+44.1
+10 ms
+130
+95.55
+48_1
+48
+7.5 ms
+75
+80
+48_2
+48
+10 ms
+100
+80
+48_3
+48
+7.5 ms
+90
+96
+48_4
+48
+10 ms
+120
+96
+48_5
+48
+7.5 ms
+117
+124.8
+48_6
+48
+10 ms
+155
+124
+1 Mandated by BAP for Acceptors and Initiators acting as unicast Audio Sinks or Audio
+Sources, and Broadcast Sources.
+2 Mandated by BAP for Acceptors acting as unicast Audio Sinks or Broadcast Sinks.
+Table 5.3 BAP defined Codec Configuration Settings
+
+For Broadcast Streams where the Broadcast Source has no knowledge of the receiving
+devices, it has to make a unilateral decision on the LC3 parameters it will use. BAP currently
+mandates that every broadcast receiver must be capable of decoding 10ms LC3 frames
+encoded at 16kHz with a 40 byte SDU and 24kHz with a 60 byte SDU, so a Broadcast
+Source using these values knows that its audio streams can be decoded by every Bluetooth
+135
+
+Section 5.4 - The LC3 codec
+LE Audio device.
+Some top level audio profiles mandate support for receiving higher quality encoded LC3
+audio streams. TMAP mandates support for a range of codec configurations that employ
+48kHz sampling, 7.5ms and 10ms frames and a variety of bitrates. However, a Broadcast
+Source with no connection to all of the receiving devices cannot know whether or not such a
+stream can be decoded. We’ll look at the implications for broadcast design later in the
+chapter.
+
+5.4.4
+
+Packet Loss Concealment (PLC)
+
+An annoying feature of wireless transmission is that packets get lost. For Bluetooth, which
+shares the 2.4GHz spectrum with Wi-Fi, baby monitors and a host of other wireless
+products, that’s generally as a result of interference. The Bluetooth specification is one of
+the most robust radio standards, employing adaptive frequency hopping to try and avoid any
+interference, but there are occasions when data will be lost. If the audio source is a phone or
+a PC which is also using Wi-Fi or has other Bluetooth peripherals, there will also be
+occasions when they need priority, which results in a Bluetooth LE Audio transmission
+being missed. We’ll look at ways to mitigate these later on, but the reality is that occasionally
+a packet will be irretrievably lost.
+Unfortunately, losing a packet in an audio stream is very noticeable, and something that
+annoys users. To try and conceal it, a number of techniques have evolved. Inserting silence
+is generally annoying, unless it happens to be preceded by a silent or very quiet moment.
+Repeating the previous frame may work, but becomes noticeable if there are consecutive
+missed frames.
+To provide a better listening experience, the industry has developed a range of Packet Loss
+Concealment algorithms which attempt to conceal the missing audio by predicting what it
+was most likely to be. These work very well with voice and generally quite well with music,
+although if a segment with, or close to an attack transient is lost, that is difficult to conceal.
+The LC3 specification includes a Packet Loss Concealment algorithm which has been
+developed to match the LC3 codec. It is applied whenever the Bad Frame Indication flag
+signals a lost or corrupted frame, or when the decoder detects an internal bit error. It is
+recommended that it, or an alternative PLC algorithm, is always used.
+
+136
+
+Chapter 5 - LC3, latency and QoS
+
+5.5
+
+LC3 latency
+
+We looked at the basics of latency at the start of the chapter, but it’s important to
+understand it in more detail.
+
+Figure 5.7 The component parts of latency
+
+Figure 5.7 illustrates the main components of latency, showing how it is built up across the
+Initiator and Acceptor. Any frame-based codec starts by imposing a delay due to the
+sampling of the audio. For a 10ms frame length, (the standard frame length in LE Audio),
+that accounts for the first 10ms of latency. Once the incoming audio frame has been
+sampled, the encoding can start, which, for LC3 takes about 2.5ms, before it has a fully
+encoded SDU to pass forward for transmission.
+The diagram shows transmission starting immediately. If the receiver gets a valid packet on
+its first transmission, the Transport Delay can be less than a millisecond, but if one or more
+retransmissions have been scheduled, it will take a few more milliseconds before the last
+possible transmission would be received at the Synchronisation Reference Point. The
+earliest that can occur with Bluetooth LE Audio is around 14ms from the point where the
+first sample for that audio frame was taken, and is the fixed point in time at which every
+Acceptor can start to decode their received packets. The Synchronisation Reference Point is
+where the Presentation Delay starts. Within this period, the Acceptor needs to decode the
+LC3 packet, which takes a few milliseconds for the LC3 decoder, and apply the packet loss
+concealment, if that is required, which takes another few milliseconds. Although it’s not
+needed for most packets, the time to run the algorithm has to be allocated for the occasions
+where it is required. If any other audio processing needs to be done, such as algorithms for
+noise cancellation or speech enhancement, they need to be completed before the end of the
+137
+
+Section 5.6 - Quality of Service (QoS)
+Presentation Delay, which is where each Acceptor renders the reconstituted audio stream.
+Because the Basic Audio Profile requires that every Acceptor must support a value of 40ms
+for Presentation Delay , they need to support around 40ms of buffering to hold the decoded
+audio data before the rendering point. In practice, the value of Presentation Delay may be
+lower or greater – manufacturers of receiving devices may support a range from as low as
+5ms, up to several hundred milliseconds.
+If everything is optimised, the quickest this whole process can happen for a 10ms LC3
+packet is just over 20ms. Using a 7.5ms frame makes little difference, as the shorter frame
+needs a longer look-ahead delay, so the saving is only around 1 ms.
+A 20ms delay is equivalent to the time it takes sound to travel 7m. Our hearing has evolved
+to cope with this level of delay. If we hear an original sound and an echo 25 – 30ms later,
+the brain processes it without any difficulty. That means that we can use Bluetooth LE
+Audio Streams for earbuds and hearing aids which pick up the ambient sound as well as a
+Bluetooth stream without the wearer being distracted by any echo effects. However, the
+example above involved a lot of optimisations. It assumes that transmission can start as
+soon as the encoding is complete and that all retransmissions happen within a quarter of a
+frame, which is only valid for small packets and the lower sampling frequencies. In most
+real applications other factors come into play, which brings us to the Quality of Service or
+QoS.
+
+5.6
+
+Quality of Service (QoS)
+
+Quality of Service is a term applied to the received audio signal and encompasses latency, the
+perceived sound of the decoded audio and the incidence of any audio artefacts, such as pops,
+crackles and gaps. All of these features have trade-offs with each other. The latency
+example described above is a highly idealised one, where packets arrive when they’re
+expected. In practice, they don’t. The human body is very efficient at absorbing Bluetooth
+signals, so if someone is wearing earbuds, but has their phone in the back pocket of their
+jeans, the signal between the phone and the earbuds may be attenuated by up to 80dB. If
+you’re in a room, that may not matter, as the earbud will probably pick up reflected signals
+from the walls or ceiling. But if you’re outside, where you don’t have those reflecting
+surfaces, far more packets will be lost.
+In the previous chapter, we saw that the design of Isochronous Channels adds robustness by
+using retransmissions, pretransmissions, burst numbers, flush timeouts and frequency
+hopping. If we apply enough of these features, we have a very high confidence that almost
+every packet will get through, and the small number that don’t arrive intact can be filled in
+using PLC. However, as we apply these techniques, latency starts to increase, as does power
+consumption. Spreading retransmissions across more than a single Isochronous Interval
+adds an extra frame time to the latency for each additional Isochronous Interval. Putting
+more retransmissions within a single frame limits the number of different streams which can
+be accommodated and pushes up the power – both for the transmitter, and also for the
+receiver, which needs to stay active to look for consecutive transmission slots.
+138
+
+Chapter 5 - LC3, latency and QoS
+These robustness features are separate from the codec settings. But the codec settings also
+have an effect on robustness. Higher quality encoding, with 48kHz sampling, will produce
+larger packets, which will be more susceptible to interference. Similarly, if you encode
+multiple Audio Streams into a single packet, i.e., the channel allocation is greater than 1, that
+SDU will contain multiple codec frames, resulting in larger packets, with the same problem.
+Given the large number of possible codec and robustness configurations that are allowed,
+BAP has defined sets of standard combinations aimed at the two major use cases – Low
+Latency and High Reliability, which can be found in Tables 5.2 and 6.4 of BAP. They cover
+both 10ms and 7.5ms frame intervals.
+Low latency is interpreted as settings which will allow all of the retransmissions to fit within
+a single Isochronous Interval for sampling rates of 8kHz to 32kHz. The larger packets for
+48kHz extend into two Isochronous Intervals, going up to four Isochronous Intervals for
+44.1kHz. High reliability QoS configurations prioritise retransmission over latency, allowing
+retransmissions to be spread across six or more Isochronous Intervals for broadcast and ten
+or more for unicast.
+Table 5.4 shows the maximum number of Isochronous Intervals allowed for each sampling
+frequency, derived from these tables. The reason that the Low Latency 48 kHz sampled
+setting needs two Isochronous Intervals is to allow a sufficient number of retransmissions
+with the larger packets, as only a limited number fit into a single Isochronous Interval. For
+low sampling frequencies, the smaller packets mean that more retransmissions can be fitted
+into each Isochronous Interval. How many retransmissions are allocated is ultimately down
+to the scheduler in the Controller.
+Low Latency
+
+High Reliability
+Unicast
+
+Sampling Frequency
+
+7.5ms
+
+10ms
+
+7.5ms
+
+10ms
+
+Broadcast
+7.5ms
+
+10ms
+
+8 kHz
+1
+1
+10
+10
+6
+6
+16 kHz
+1
+1
+10
+10
+6
+6
+24 kHz
+1
+1
+10
+10
+6
+6
+32 kHz
+1
+1
+10
+10
+6
+6
+1
+44.1 kHz
+4
+4
+12
+9
+8
+6
+48 kHz (80 kbps)
+2
+2
+10
+10
+7
+7
+48 kHz (96 / 124 kbps) 2
+2
+2
+10
+10
+7
+7
+1 The 44.1 kHz figures differ because they are defined for framed PDUs. All other sampling
+rates are unframed.
+Table 5.4
+
+The maximum number of Isochronous Intervals allowed for BAP QoS settings
+
+In the latency example of Figure 5.7, we saw that using LC3 with a 10ms frame and a
+significant degree of optimisation gave an overall latency of just over 20ms. That used a
+139
+
+Section 5.6 - Quality of Service (QoS)
+Presentation Delay of just over 5ms. If low latency is important to an application,
+implementers need to be careful about their choice of QoS and codec configuration, as it can
+result in the latency increasing significantly. Table 5.5 shows the actual overall latency values
+which are likely to achieved. These have been calculated using a value of 12.5ms for the LC3
+to sampling and encode a 10ms frame.
+Low Latency
+
+High Reliability
+
+Unicast and Broadcast
+Sampling
+Frequency
+8 kHz
+16 kHz
+24 kHz
+32 kHz
+44.1 kHz
+48 kHz (80 kbps)
+48 kHz (96/124 kbps)
+
+Unicast
+
+Broadcast
+
+PD=10ms
+
+PD=20ms
+
+PD=40ms
+
+PD=40 ms
+
+PD=40 ms
+
+32.5 ms
+32.5 ms
+32.5 ms
+32.5 ms
+53.5 ms
+42.5 ms
+42.5 ms
+
+42.5 ms
+42.5 ms
+42.5 ms
+42.5 ms
+63.5 ms
+52.5 ms
+52.5 ms
+
+62.5 ms
+62.5 ms
+62.5 ms
+62.5 ms
+83.5 ms
+72.5 ms
+72.5 ms
+
+147.5 ms
+147.5 ms
+147.5 ms
+147.5 ms
+137.5 ms
+147.5 ms
+152.5 ms
+
+112.5 ms
+112.5 ms
+112.5 ms
+112.5 ms
+112.5 ms
+117.5 ms
+117.5 ms
+
+Table 5.5 Typical end-to-end latencies for BAP QoS settings
+
+The values in Table 5.5 are for single, mono audio streams. For stereo applications, the latency
+increases, as we will see. The message for implementers is to take care when choosing codec
+and QoS settings.
+The Quality of Service recommendations in Tables 5.2 and 6.4 of BAP, include suggestions
+for parameters to use in the HCI commands to set up unicast or broadcast streams. These
+are recommendations (remember that the Controller uses some of these as guidance, not as
+definitive values) covering the:
+•
+•
+•
+•
+•
+•
+
+SDU Interval (which is the same as the Frame Duration, other than for 44.1kHz)
+Framing requirements (all are unframed, except for 44.1kHz, which is framed)
+Maximum_SDU_Size (which is the same as the
+Supported_Octets_per_Codec_Frame)
+A recommended Retransmission Number (RTN) for Low Latency and High
+Reliability options
+Max_Transport_Latency for Low Latency and High Reliability options, and
+Presentation Delay, requiring a value of 40ms to be in the supported range.
+
+Using the values from these tables may not always produce the expected latencies. One of
+the reasons for that is the value for Presentation Delay. BAP requires all Audio Sinks to
+include a Presentation Delay of 40ms within their range of supported values, but it should
+not be taken as the default value – it is what it says in the note at the bottom of the table – a
+140
+
+Chapter 5 - LC3, latency and QoS
+value that must be supported by every Acceptor acting as an Audio Sink. It’s a compromise
+for interoperability to ensure that everything has time to decode the audio data, apply PLC
+and any additional processing. Most devices will be able to do better. Some higher layer
+profiles require tighter performance. TMAP and HAP both require Acceptors to support a
+Presentation Delay of 20ms for broadcast, but if an Initiator sets it to 40ms, that impacts
+latency.
+Recalling Figure 5.7, the Presentation Delay in that example is only around 5ms, leading to
+an overall latency below 25ms. Many hearing aids will be capable of supporting that, but it is
+highly optimised. In unicast, where the Initiator and Acceptor talk to each other, they can
+agree on a lower value for Presentation Delay when the Initiator is aware that it is delivering
+a low latency use case. However, a basic Broadcast Source has no way of knowing the
+capabilities of the Broadcasts Sinks around it, so will have to make a decision based on the
+use case and a knowledge of the capabilities of Broadcast Sinks which are on the market and
+which its designers expect will access it.
+A pair of Acceptors can act unilaterally if they can communicate with each other, by making
+a decision to render earlier, potentially on the basis of the Context Type for the stream.
+However, that is outside the scope of the Bluetooth LE Audio specifications.
+Returning to the Low Latency columns in Table 5.5, sampling frequencies above 32 kHz risk
+echo effects when they are used for ambient sound applications. None of the High
+Reliability settings are really suitable when reinforcing ambient sound, particularly for
+broadcast.
+Where an ambient audio stream cannot be heard, which is the case with most streaming
+music and telephony applications, latency becomes far less of a problem, unless there is an
+accompanying video stream, where it could lead to lip-synch problems. However, in these
+cases, the video application generally manages the audio stream, so can adjust the relative
+timing to prevent any issue.
+RTN – the Retransmission Number, benefits from some further explanation. The large
+values in some of the configurations suggest lots of retransmissions, but that is not
+necessarily the case. RTN is defined in the Core [Vol 4, Part E, Sect 7.8.97] as the number
+of times that a CIS Data PDU should be retransmitted from the Central to the Peripheral or
+Peripheral to Central before it is acknowledged or flushed. For Broadcast, it is simply the
+number of times the PDU should be retransmitted [Vol 4, Part E, Sect 7.8.103]. As we’ve
+already seen, the Host is not allowed to specify values for FT, PTO, NSE or BN, as it
+doesn’t know what other constraints the Controller might have. Hence RTN, is just a
+recommendation to help guide the Controller’s scheduling algorithm.
+Most of the time, audio data won’t be transmitted RTN times. RTN is better interpreted as
+the maximum number of retransmissions, not the average number. If an SDU has an FT
+greater than 1 and the SDU is transmitted the maximum number of (RTN + 1) times, the
+141
+
+Section 5.6 - Quality of Service (QoS)
+following SDU will only have 1 opportunity for transmission, with no retransmissions
+possible. RTN gives the opportunity to get more transmission slots to accommodate short
+bursts of interference, but setting it too high can penalise subsequent SDUs. The average
+number of transmission opportunities is always NSE/BN. Having pointed that out, the
+values given in Tables 5.3 and 6.4 of BAP have been well tested and should generally be
+followed. However, as Table 6.5 of BAP shows, the Controller may choose other values.
+Returning to the Isochronous Channel settings, it’s useful to look at what the Initiator’s Host
+asks for and what it actually gets. BAP gives an example of how a Controller might interpret
+the HCI parameters it receives based on different resource constraints. To understand that
+effect, we can look at the High Reliability setting for the 48_2_2 broadcast Audio Stream
+QoS configuration. That’s defined in Table 6-4 of BAP and is reproduced in Table 5.6
+below, where a maximum transport latency of 65ms is requested.
+SDU
+Interval (µs)
+
+Max SDU
+(octets)
+
+RTN
+
+Max Transport
+Latency (ms)
+
+Presentation
+Delay (µs)
+
+10,000
+
+100
+
+4
+
+65
+
+40,000
+
+48_2_2
+
+Table 5.6 Broadcast Audio Stream Configuration setting for 48_2_2 High Reliability audio data
+
+Table 6.5 of BAP provides recommended Link Layer parameters for a Controller to use
+when it receives an LE_Set_CIG_Parameters HCI command containing the values of Table
+5.6. These are shown in Table 5.7, which also shows the resulting transport latency and the
+percentage of available airtime which is used for each set of parameters.
+Option
+
+ISO_Interval
+(ms)
+
+BN
+
+NSE
+
+IRC
+
+PTO
+
+Num_BIS
+
+RTN
+
+Max_Transport_Latency
+Mono/Stereo
+
+Airtime Usage
+Mono/Stereo
+
+1
+2
+3
+
+30
+10
+20
+
+3
+1
+2
+
+9
+5
+8
+
+2
+1
+2
+
+1
+1
+1
+
+1 or 2
+1 or 2
+1 or 2
+
+2
+4
+3
+
+65 / 71 ms
+43 / 46 ms
+65 / 70 ms
+
+18 / 36%
+30 / 59%
+24 / 48%
+
+Table 5.7 Recommended BAP LL parameters for the 48_2_2 broadcast Audio Stream QoS configuration
+
+The three options in Table 5.7 are possible because the Max_Transport_Latency and RTN
+are only recommendations to guide the Controller when it works out its scheduling.
+Depending on what else it is doing, it will normally need to take other constraints into
+account. The three options in Table 5.7 are recommended Link Layer settings for the
+following three cases:
+•
+
+142
+
+Option 1, which has the lowest airtime, is optimised for coexistence with Wi-Fi and
+other Bluetooth devices operating at 7.5ms schedules. Using a larger Isochronous
+Interval, to carry multiple 10ms frames, provides the largest gaps for Wi-Fi
+operation. If the Broadcast Source is a phone or a PC, and is using Wi-Fi to obtain
+the music stream to send over Bluetooth LE Audio, that’s very important. This
+option uses the least airtime for the Bluetooth LE Audio transmissions, but has the
+
+Chapter 5 - LC3, latency and QoS
+
+•
+
+•
+
+highest latency.
+Option 2 is designed to provide the highest reliability and lowest latency,
+maximising the number of retransmissions in each frame. It uses the greatest
+amount of airtime, so is only really suitable for broadcast devices which have no
+other 2.4GHz radio operations.
+Option 3 aims for a balanced approach between reliability and coexistence. It
+would be a good choice for a Broadcaster which is using Wi-Fi to obtain the music
+stream, but which has no other coexistence issues to contend with.
+
+These are only three of many possible combinations which can be chosen by a chip’s
+scheduler. Over time, others may be added to the recommended list, as the industry
+acquires more experience with Bluetooth LE Audio.
+Before leaving this subject, Table 5.8 shows the overall latencies for stereo broadcast streams
+using each of these three options, with 20ms Presentation Delay mandated by TMAP and
+HAP, and the baseline 40ms of BAP. The Controller will inform the Host of which
+parameters it has chosen, but the Host has no control over what that choice will be. That
+will be down to the scheduling algorithm in the chip. Designers should be aware of this
+variation. If the latency for your application is critical, ask your silicon manufacturer for
+details of what choices they are making in their Controller scheduling.
+Overall Latency
+
+Option 1
+Option 2
+Option 3
+
+PD=20ms
+
+PD=40ms
+
+122.3 ms
+78.4 ms
+112.0 ms
+
+142.2 ms
+98.4 ms
+132.0 ms
+
+Table 5.8 Overall latency for a broadcast stereo stream using the BAP LL options for the 48_2_2 QoS configuration
+
+5.6.1
+
+Airtime and retransmissions
+
+We touched on airtime in Table 5.7. Although the LC3 codec is more efficient than
+previous Bluetooth codecs, as the bitrate is increased, the packets take up an increasing
+amount of the available airtime. The figures in Table 5.7 for Option 2 show that a stereo
+stream using these parameters accounts for almost 60% of the available airtime. That
+doesn’t include the airtime required for advertising, other active Bluetooth links or any other
+2.4GHz radio activity.
+
+143
+
+Section 5.6 - Quality of Service (QoS)
+
+Figure 5.8 Airtime usage for different unidirectional QoS configurations
+
+Figure 5.8 shows the effect of the higher bitrates in the QoS configurations (using a 10ms
+frame size), as well as the impact of more retransmissions specified by the High Reliability
+settings. For the 48kHz sampling configurations, the airtime is the same for Low Latency
+and High Reliability, as a minimum retransmission number of 4 is recommended for each
+because of the greater susceptibility of the larger packets to interference.
+Broadcast Sources have no idea of whether their retransmissions have been received, so have
+to transmit every packet. Unicast Initiators only need to retransmit packets if they fail to
+receive an acknowledgement. That means that in many cases a unicast Initiator will transmit
+far fewer packets, providing additional airtime for any other resource on the device which
+needs it. However, if the link budget for the connection is poor, as it may be for devices like
+earbuds and hearing aids, particularly when used outside, they may need to transmit the
+maximum number of retransmissions, so the airtime must be allocated for this eventuality.
+Airtime is a limited resource, and increasing the audio quality and reliability of a stream eats
+into it. One of the design requirements for Bluetooth LE Audio was the ability to support
+multiple audio streams, allowing a TV or cinema to transmit soundtracks in multiple
+languages. Looking at Figure 5.8, it’s clear that is not possible with any of the 48kHz
+sampling configurations, unless multiple radios are employed to transmit each different
+stereo language stream. In contrast, a Low Latency 24kHz or 32kHz configuration could
+easily cope with two different stereo streams. If product designers want to take advantage of
+Bluetooth LE Audio’s ability to transmit multiple streams, they need to consider the airtime
+implications. Which brings us to audio quality.
+
+144
+
+Chapter 5 - LC3, latency and QoS
+
+5.7
+
+Audio quality
+
+Since the early days of electronic audio reproduction, a small group of users have constantly
+pressed for higher quality. That led to technical advances in recording and reproduction,
+along with marketing of terms like Hi-Fi. As well as real technical advances, it saw the
+appearance of pseudo-scientific fashions such as oxygen free copper cables and gold-plated
+connectors to “ensure” that the analogue signals were not degraded. Digital technology
+didn’t lessen the enthusiasm for these, despite the fact that a gold-plated USB connector is
+an anachronism. Audio devotees kept clamouring for enhanced quality, with the digital age
+seeing calls for even higher sampling rates, lossless codecs and increased output levels. The
+fact that many people over 30 now have a level of hearing loss which means they are
+incapable of hearing any of these “improvements” doesn’t stop product marketing managers
+pushing for ever higher audio quality.
+In its early days, Bluetooth audio attracted a fair amount of negative coverage. Some was
+well-deserved, as some A2DP headsets had resource constrained codec implementations.
+The transducers used for rendering audio were also relatively primitive. Much has changed
+since then, with massive development in headphones, earbuds and speakers, to the point
+where few users have concerns over Bluetooth technology as an audio solution and it is
+routinely used in top-end audio equipment costing thousands of dollars.
+During the LC3 codec development, the Bluetooth SIG commissioned independent research
+on its audio quality compared with other codecs. The results confirm that it offers
+equivalent or better subjective performance to other codecs, across the entire spectrum from
+8kHz to 48kHz sampling. Examples of LC3 encoded audio streams are available to listen to
+at the Bluetooth SIG website37.
+The tests were performed by a bank of expert listeners, using high quality, wired headphones
+in audio listening booths. They were asked to rate each sample of sound against the
+reference recording, grading how close they felt it was to the original. The tests used the
+MUSHRA38 protocol, with a mixture of sounds from the EBU’s39 test recordings.
+It is always difficult to relate test results like these, which are done in perfect listening
+conditions, with the real-world experience, because they are subjective. However, I have
+tried to describe the general application for the different sampling frequencies in Table 5.9.
+
+https://www.bluetooth.com/blog/a-technical-overview-of-lc3/, which includes a link to an audio
+demo.
+38 Multiple Stimuli with Hidden Reference and Anchor methodology for testing perceived quality,
+defined in ITU BS.1543-3.
+39 European Broadcasting Union TECH 3253 - Sound Quality Assessment Material recordings for
+subjective tests
+37
+
+145
+
+Section 5.8 - Multi-channel LC3 audio
+Sampling Rate
+
+Description
+
+8 kHz
+16 kHz
+24 kHz
+
+Suitable for voice of telephony quality.
+Higher quality voice. Adequate for voice recognition applications.
+Adequate for music where there are imperfect listening conditions,
+such as background noise, or where a listener has any hearing
+impairment. Listeners are likely to detect a difference from higher
+sampling rates if they are concentrating on the audio stream in
+noise-free surroundings.
+Most users will not detect a difference from the original when
+listening with any background noise.
+Users cannot detect a difference to the original.
+
+32 kHz
+48 kHz
+
+Table 5.9 A subjective description of LC3 audio quality for different sampling rates.
+
+What is important is that audio application designers understand that there are compromises
+in designing a wireless audio experience and that some QoS choices may exclude the
+development of new use cases. Some sections of the audio industry will jump onto the
+higher quality that LC3 offers and promote the highest sampling rates, despite the fact that
+limitations in reproduction and the listening environment will probably mean that few, if any
+listeners will appreciate them. In addition, the resulting latency is unsuitable for live
+applications and some devices may not be able to decode them. Others will look at the new
+features and use cases that are enabled by Bluetooth LE Audio and choose 24kHz or
+32kHzas an acceptable compromise which allows new use cases to develop. Only time will
+determine what users value most. It may be more flexibility in their applications, or a desire
+to see a bigger “quality” number in the marketing literature.
+In the past, there have been two occasions when the audio industry took the decision to
+reduce audio quality. The first was the introduction of CDs. The second was the use of
+MP3, which enabled audio streaming. Each time, there was a balance between greater ease
+of use versus maintaining the current audio quality. On both occasions, consumers
+expressed a major preference for ease of use.
+
+5.8
+
+Multi-channel LC3 audio
+
+In Chapter 3, we introduced the concept of Audio Channels. It means that there are
+multiple different ways of transmitting the same audio data between devices. To understand
+how they work, it’s best to look at a couple of examples. In each of these, the following
+nomenclature is used to describe how many audio channels are multiplexed into a CIS or
+BIS. The number of arrowheads on each CIS or BIS corresponds to the number of audio
+channels that it is carrying.
+
+146
+
+Chapter 5 - LC3, latency and QoS
+
+Figure 5.9 Representation of the number of Audio Channels in a CIS or BIS
+
+Figure 5.10 shows the two possible options for transmitting a stereo stream between an
+Initiator and a single Acceptor. This is the simple unicast use case of streaming music from a
+phone to headphones or from a TV to a soundbar.
+
+Figure 5.10 Multiplexing options for a stereo stream from an Initiator to one Acceptor
+
+At the top of Figure 5.10, option (a) uses two separate CISes, one to carry the left audio
+channel and the other to carry the right channel. Below that, Option (b) sets the Channel
+Allocation to 2, so that both of the LC3 encoder outputs are concatenated into a single
+payload by the Controller (the Link Layer – LL), which is sent in a single CIS. At the
+Acceptor, the individual left and right payloads are separated in the Controller, decoded and
+rendered as individual channels after the Presentation Delay. (The Controller is shown as
+the LL block (Link Layer) in the diagrams. Where it is not multiplexing a stream, but simply
+transmitting or receiving a single channel codec frame, it is omitted to simplify the diagrams.)
+Sending the same unicast stereo input to a mono Acceptor is a little more interesting, as at
+some point between the input and output, the stereo signal needs to be downmixed to a single
+mono stream. There are three possible approaches to do that, which are shown in Figure 5.11.
+
+147
+
+Section 5.8 - Multi-channel LC3 audio
+
+Figure 5.11 Transmitting unicast stereo to a mono Acceptor
+
+For all three options, the Acceptor would set its Audio Sink Locations as Front Left plus Front
+Right, so its Supported Audio Location would be 0x0000000000000011. Note that there is
+no mono Audio Location, as mono is a property of a stream, not a physical Audio Location.
+In Option (a), the Initiator can deduce that the Acceptor will render a mono signal, as it has
+set the Channel Allocation to 1, which means it will only render the stream to one location. If
+it had wanted both the left and the right channel, it would have set Channel Allocation to 2.
+Setting both Audio Location bits, but stating that it only supports a single, non-multiplexed
+CIS, signifies that it requires the Initiator to downmix the input audio channel to mono before
+it is sent.
+In Option (b) and Option (c), the Channel Allocation and Audio Sink Location information
+the Initiator receives after the Codec Configuration procedure is identical to what it would
+have received if the Acceptor were a stereo device (see Figure 5.10). It means that the Initiator
+has no way of knowing whether it is a mono or a stereo device40. Therefore, it supplies it with
+stereo information, either as two separate CISes in Option (b), or a multiplexed stream on a
+single CIS in Option (c). In this case, the Acceptor is responsible for downmixing the resultant
+streams.
+BAP requires all devices which set multiple Bluetooth LE Audio Locations to be able to
+support at least the same number of streams, so in Option (a), the Acceptor would also need
+to be able to support the two CIS configuration of Option (b). If it requires the Initiator to
+
+The Initiator could look for an instance of the Volume Offset Control Service and infer from its
+presence that it’s a stereo device, but that may be trying to be too clever.
+40
+
+148
+
+Chapter 5 - LC3, latency and QoS
+perform the downmixing, it would specify the single Channel Allocation and Left plus Right
+Audio Locations in a preferred PAC record (which we’ll cover in Chapter 7), to signal its desire
+for a single mono stream.
+The popular use case of separate earbuds also has different possible configurations. Figure
+5.12 shows the two options for transmitting a stereo audio stream to a pair of Acceptors.
+
+Figure 5.12 Stream options for a pair of stereo earbuds
+
+Option (a) illustrates the configuration which will generally be used, where the earbuds set
+their Audio Location to just Left or Right (not Left and Right, as in the previous examples).
+The Initiator will send each earbud a single CIS corresponding to their Audio Location.
+In Option (b), both earbuds have set their Channel Allocation to 2, so they will receive a
+multiplexed stream. Each earbud then has to decode the stream they require, discarding the
+other. It is a less efficient option, but allows earbuds to swap streams if they detect that the
+user has turned around, so has application in 3D sound and virtual reality headsets, particularly
+if additional spatial audio content is available.
+The configurations shown above are only a small selection of the possible combinations and
+not every Initiator and Acceptor will support all of them. BAP lists 14 different combinations
+of stream configuration (which is not exhaustive), mandating the most common ones, and
+applying conditional and optional requirements on the others. Table 5.10 shows the
+requirements for connections between an Initiator and a single Acceptor. M means that they
+are mandatory and must be supported by Bluetooth LE Audio compliant devices. C is a
+conditional support, which is generally based on whether a device supports bidirectional
+streams. O means that support is optional. Full details of these, including what the conditional
+requirements are, can be found in Tables 4.2 and 4.24 of BAP.
+149
+
+Section 5.8 - Multi-channel LC3 audio
+Audio Configurations 6 to 9, and 11 involve two streams. In Table 5.10, they bear the suffix
+(i), indicating that they refer to the use case where a single Acceptor is involved in both streams
+and is therefore supporting two CISes.
+Audio
+Configuration
+
+Stream
+Direction
+(Initiator –
+Acceptor)
+
+1
+2
+3
+4
+5
+
+– – – – –>
+<–––––
+< – – – – –>
+– – – – –>>
+< – – – –>>
+– – – – –>
+– – – – –>
+– – – – –>
+<–––––
+– – – – –>
+<––––>
+<–––––
+<–––––
+<< – – – – –
+< – – – –>
+<––––>
+
+6 (i)
+7 (i)
+8 (i)
+9 (i)
+10
+11 (i)
+
+Unicast
+
+Broadcast
+
+Initiator
+
+Acceptor
+
+M
+M
+C
+O
+C
+
+M
+M
+C
+C
+C
+
+M
+
+O
+
+C
+
+C
+
+C
+
+C
+
+M
+
+C
+
+O
+
+C
+
+C
+
+C
+
+12
+13
+
+Not Applicable
+
+14
+Table 5.10 Stream configuration requirements for single Acceptors, denoted as (i)
+
+150
+
+Initiator
+
+Acceptor
+
+Not Applicable
+
+M
+
+M
+
+M
+
+C
+
+O
+
+C
+
+Chapter 5 - LC3, latency and QoS
+Table 5.11 shows the requirements where the Initiator establishes the streams with a pair of
+Acceptors, with one CIS connected to each. These are denoted by a suffix (ii).
+Audio
+Configuration
+
+6 (ii)
+7 (ii)
+8 (ii)
+9 (ii)
+11 (ii)
+
+Stream
+Direction
+(Initiator –
+Acceptor)
+– – – – –>
+– – – – –>
+– – – – –>
+<–––––
+– – – – –>
+<––––>
+<–––––
+<–––––
+< – – – –>
+<––––>
+
+Unicast
+
+Broadcast
+
+Initiator
+
+Acceptor
+
+M
+
+O
+
+C
+
+C
+
+C
+
+C
+
+M
+
+C
+
+C
+
+C
+
+Initiator
+
+Acceptor
+
+Not Applicable
+
+Table 5.11 Stream configuration requirements for sets of two Acceptors denoted as (ii)
+
+Many other combinations are possible, but a Bluetooth LE Audio device cannot automatically
+expect them to be supported on an Initiator or Acceptor. An Initiator can determine support
+for other configurations from the PAC records and Additional_ASE_Parameters values in the
+Codec_Specific_Configurations.
+
+5.9
+
+Additional codecs
+
+The LC3 is a very good codec, which can be used across a wide range of sampling rates for
+voice and music applications. It is mandatory for every Bluetooth LE Audio device to
+support it at the 16kHz and 24kHz sampling rates (Initiators only need to support 16kHz, as
+some public address systems may be voice only), but the majority of implementations are
+likely to support higher sampling frequencies.
+Despite that, there are specialised applications where other codecs may perform better. To
+accommodate this, the Bluetooth LE Audio specifications allow the use of additional codecs,
+or vendor specific codecs.
+Additional codecs used to be called optional codecs in A2DP. These are codecs which are
+designed by external specification bodies or companies, but the Bluetooth specifications
+include configuration information that allow qualified devices to recognise that they exist
+and how to set them up. This allows multiple manufacturers to add the capability to use
+them. Without that, a connection would default back to the mandatory Bluetooth codec.
+Normally, additional codecs are licensable from external standards organisations, so anyone
+can integrate them into their product. Currently, no additional codecs are defined for
+151
+
+Section 5.9 - Additional codecs
+Bluetooth LE Audio.
+Vendor specific codecs are ones which are licensed by manufacturers, who provide that
+Bluetooth configuration information to their licensees. Other Bluetooth products would not
+understand that information, so would ignore them and use the mandatory codec, or an
+additional codec (if they supported it). Vendor codecs are proprietary and outside the scope
+of the Bluetooth specifications. Where they are used, the Codec_ID is set to Vendor
+Specific.
+
+152
+
+Chapter 6 - CAP and CSIPS
+
+Chapter 6. CAP and CSIPS
+Within the Generic Audio Framework of the Bluetooth® LE Audio set of specifications, the
+four BAPS specifications (the Basic Audio Profile, the Audio Stream Control Service, the
+Broadcast Audio Scan Service and the Published Audio Capabilities Service) do most of the
+heavy lifting. If you implement these four specifications, you can build almost any unicast or
+broadcast application. However, they are designed to be as generic as possible, which means
+that there are multiple, different ways of putting the pieces together. CAP – the Common
+Audio Profile, defines a set of procedures to establish common ways to perform all of the
+everyday actions that are needed to set up Audio Streams between an Initiator and one or
+more Acceptors. For most developers, CAP provides the interface on which they build their
+applications.
+CAP ties in the content control, use case and rendering control concepts which we came across
+in Chapter 3. It defines when these need to be used during the stream configuration and
+establishment process. It is also key to preventing multi-profile issues when a Bluetooth LE
+Audio device transitions between different use cases within a connection, or makes
+connections with different devices.
+The other thing that CAP brings to the process is coordination. The biggest market for
+Bluetooth audio today is in earbuds – two separate devices which need to work as if they were
+one. The BAPS specifications deal with individual streams. CAP lays down rules for managing
+streams when an Initiator connects to multiple Acceptors, using the Coordinated Set
+Identification Profile and Service to bind them together.
+Finally, CAP defines the Commander role, which is what elevates broadcast from a simple
+telecoil replacement to a highly flexible audio distribution solution.
+In this chapter, I’ll provide an overview of the CAP procedures and how they’re used.
+Essentially, you can think of CAP as a recipe book. The BAPS specifications define all of the
+ingredients for Audio Streams and CAP tells you which ones to use for each procedure, and
+what order to use them in. Most of the detail will come out in the following chapters on
+setting up unicast and broadcast Audio Streams, where we’ll see how CAP augments and
+utilises the stream control and management procedures that are already in BAPS. But before
+we start on CAP, it’s important to understand CSIP and CSIS.
+
+6.1
+
+CSIPS – the Coordinated Set Identification Profile and Service
+
+Because A2DP was designed for streaming to a single device, every Bluetooth solution on the
+market today that sends audio to multiple devices uses some proprietary method of making
+them work together, whether they’re a pair of earbuds, hearing aids or speakers. Bluetooth
+LE Audio needed to define a standard method of doing this, so that any combination of
+products could be used together. The issue is not just making sure that they can have their
+volume controlled together, but also to make sure that if you change the device providing the
+153
+
+Section 6.1 - CSIPS – the Coordinated Set Identification Profile and Service
+audio stream, such as when a phone call interrupts you when you’re watching TV, then both
+left and right earbuds change their connection to your phone at the same time. That
+requirement led to the concept of Coordinated Sets.
+The Coordinated Set Identification Service is instantiated on devices which form a group
+(called a Coordinated Set), where the group members fulfil a specific use case, acting in a
+concerted manner. A typical example is a pair of earbuds. The specifications are not limited
+to audio devices – they could equally be used for sets of medical sensors, such as ECG patches,
+where data is being collected from separate sensors. Nor do all of the functions of those
+devices need to be exactly the same, but one feature should be. For example, in a pair of
+earbuds, the common feature is that they both would render audio streams, but only one need
+have a microphone.
+For a pair of earbuds, coordination performs four main functions:
+•
+•
+•
+
+•
+
+It identifies the members of the Coordinated Set, i.e., a left and right earbud
+It is used to apply volume controls and mute to both devices
+It is used to ensure they both receive audio streams from the same Audio Source
+(the streams themselves are generally different, being left and right, but that is
+irrelevant to CSIPS, and
+It allows a device to lock access to the earbuds, preventing other devices from
+accessing them.
+
+CSIP defines two roles:
+•
+•
+
+A device which is a member of a Coordinated Set is a Set Member, and
+A device which discovers and manages a Coordinated Set is a Set Coordinator.
+
+The Set Member role is essentially a passive role – it simply states that it is part of a set. The
+Set Coordinator not only finds the members, but then passes that knowledge on to other
+procedures to ensure that those procedures act on all members of the set.
+CSIS requires that every member of a Coordinated Set includes a Resolvable Set Identity (RSI)
+AD Type, which allows a Client device to recognize that it is a member of a Coordinated Set.
+This means that there must be two or more Acceptors that need to be found. The RSI is a
+random six-octet identifier, which changes over time. This reduces the risk of someone
+tracking your Bluetooth products.
+Once a Client device, which can be an Initiator or Commander, makes a connection to a device
+exposing the RSI AD Type, it pairs and bonds with the set member it has discovered and reads
+its Set Identity Resolving Key (SIRK) characteristic. The SIRK is a 128 bit random number
+which is common to all members of the Coordinated Set, which a Client can use to decode
+
+154
+
+Chapter 6 - CAP and CSIPS
+the Resolvable Set Identity. It does not change for the lifetime of the device41. The SIRK is
+normally programmed into all of the devices which comprise the Coordinated Set during
+manufacture. The Bluetooth LE Audio specifications do not specify a way to set it or change
+it, so if products need to be added to a Coordinated Set during their lifetime, such as when a
+lost or broken earbud is replaced, manufacturers will need to determine a method to
+accomplish this.
+The Client also needs to read the Coordinated Set Size characteristic, which tells it how many
+devices there are in that Coordinated Set. It then proceeds to find the other members of the
+set by using the Set Members Discovery Procedure defined in CSIP. This involves the Client
+device looking for other Acceptors exposing the RSI AD Type, connecting and pairing to
+them and reading their SIRK characteristic. If the SIRK has the same value as the first
+coordinated device, then it is a member of the same set. If it is different, the Client device
+should discard the pairing and look for the next device with an RSI AD Type and check its
+SIRK characteristic value. The Set Members Discovery process ends when all of the set
+members are found, the process reaches an implementation-specific timeout (usually around
+ten seconds), or it is terminated by the application. If the Initiator or Commander can’t find
+them all, they can proceed with those they have found, but should continue to try to find the
+missing members.
+If you are shipping Acceptors which are part of a Coordinated Set, then CAP demands that
+all four of the characteristics defined in CSIS are implemented in each device. These are
+shown in Table 6.1.
+Characteristic Name
+Set Identity Resolving Key (SIRK)
+Coordinated Set Size
+Set Member Lock
+Set Member Rank
+
+Mandatory Properties
+Read
+Read
+Read, Write, Notify
+Read
+
+Optional Properties
+Notify
+Notify
+None
+Notify
+
+Table 6.1 Coordinated Set characteristic properties for use with CAP
+
+Currently, no CAP procedure uses the Set Member Lock or Rank characteristics, although it
+mandates that they are implemented on the Acceptor.
+The reason for the Set Member Lock is that when a Client writes that characteristic, the Lock
+gives that Client exclusive access to features on that Acceptor. Which features the Lock
+controls is defined by the higher-layer application. Implementors should take care in using
+the Lock, as it can prevent other Commanders or Initiators accessing the Acceptors. When a
+
+41
+
+A firmware update is considered to be the start of a new life.
+
+155
+
+Section 6.2 - CAP – the Common Audio Profile
+Client no longer needs exclusive access, it should release the Lock.
+The Rank is a value that is normally assigned at manufacture to provide a unique number to
+each member of the Coordinated Set. It doesn’t imply any priority, but is a unique, positive
+integer within the Coordinated Set which is used to ensure that all Clients apply their
+operations to the members of the set in the same order. Rank is used in the Ordered Access
+procedure in CSIP. This states that when applying a Lock for any reason, Clients should start
+with the set member that it knows to have the lowest rank and then work up through the other
+members of the Coordinated Set in ascending numeric order of Rank. That prevents a race
+condition where two different Clients apply a Lock to different set members at the same time.
+The Rank values are normally set during manufacture.
+The Ordered Access procedure is also used by CAP as a precursor to running any CAP
+procedure. It checks whether any Acceptor is locked and only continues with the CAP
+procedure once it has determined that none of the set members have a Lock set. It then
+applies the CAP procedure to each Acceptor in order of increasing Rank.
+
+6.2
+
+CAP – the Common Audio Profile
+
+As I said above, CAP can be considered as the recipe book for all of Bluetooth LE Audio,
+bringing together all of the other specifications into five main sets of procedures which define
+the way that everything works. Top level profiles, like HAP, TMAP and PBP then refer to
+these CAP procedures, adding additional requirements for their specific use cases.
+CAP is where the Initiator, Acceptor and Commander roles that I’m using throughout this
+book are defined. Whilst only the Initiator and Acceptor can be involved in Audio Streams,
+CAP describes how all three of these roles can include a range of components from the other
+Generic Audio Framework specifications. These are listed in the matrix of Table 6.2, which
+shows Mandatory, Optional, Conditional and Excluded requirements. Conditional features
+are normally mandated for specific combinations of other features, and require that a role
+supports at least one of a number of optional components. Excluded combinations are not
+allowed. The full details of all of these combinations can be found in Table 3.1 of CAP.
+Boxes with a dash (-) are features which could be implemented in a device, but are outside the
+scope of the CAP procedures.
+Role
+Component
+BAP Broadcast Assistant
+BAP Scan Delegator
+VCP Volume Controller
+VCP Volume Renderer
+156
+
+Acceptor
+
+Initiator
+
+Commander
+
+X
+C
+X
+O
+
+O
+X
+X
+
+C
+C
+C
+X
+
+Chapter 6 - CAP and CSIPS
+Role
+Component
+
+Acceptor
+
+Initiator
+
+Commander
+
+X
+O
+X
+O
+X
+X
+C
+
+X
+O
+X
+O
+C
+X
+
+C
+X
+X
+X
+M
+X
+
+MICP Microphone Controller
+MICP Microphone Device
+CCP Call Control Server
+CCP Call Control Client
+MCP Media Control Server
+CSIP Set Coordinator
+CSIP Set Member
+
+“M” = Mandatory, “O” = Optional, “C” = Conditional, X = Excluded, “-” = Not
+relevant to this profile.
+Table 6.2 Component support requirements for CAP roles. See Table 3.1 of CAP for individual conditions
+
+6.2.1
+
+CAP procedures for unicast stream management
+
+The CAP procedures for managing streams can be grouped into three categories. The first is
+a set of three procedures for unicast streams, which are largely self-explanatory:
+•
+•
+•
+
+Unicast Audio Start procedure
+Unicast Audio Update procedure
+Unicast Audio Stop procedure
+
+These procedures involve both the Initiator and the Acceptor, as setting up each unicast
+stream uses command and response sequences.
+
+6.2.2
+
+CAP procedures for broadcast stream transmission
+
+The second set of three procedures is for broadcast streams:
+•
+•
+•
+
+Broadcast Audio Start procedure
+Broadcast Audio Update procedure
+Broadcast Audio Stop procedure
+
+The Broadcast Audio Start, Stop and Update procedures are only used by the Initiator, as that
+is set up unilaterally, with no knowledge of whether any Acceptors are present.
+
+157
+
+Section 6.2 - CAP – the Common Audio Profile
+
+6.2.3
+
+CAP procedures for broadcast stream reception
+
+To complement the broadcast procedures for the Initiator, the third set of stream management
+procedures is for Acceptors that want to acquire the broadcast Audio Streams, giving us two
+procedures:
+•
+•
+
+Broadcast Audio Reception Start procedure
+Broadcast Audio Reception Ending procedure
+
+There is no broadcast Audio Update procedure for the Acceptor, as an Acceptor is a passive
+entity in broadcast which only consumes the Audio Stream. It cannot do anything to update
+the content of the broadcast streams, hence there is no update procedure for broadcast audio
+reception.
+
+6.2.4
+
+CAP procedures for stream handover
+
+CAP includes two procedures specifically for a stream handover, where an Initiator wants to
+transition between transmitting unicast and broadcast Audio Streams (and vice versa). These
+are the:
+•
+•
+
+Unicast to Broadcast Audio Handover procedure, and the
+Broadcast to Unicast Audio Handover procedure
+
+These are very important, as they describe the audio sharing use case, where a user can
+transition from listening to a private stream from their phone or music source, to sharing it
+with friends by converting it to broadcast. Although the transmission topology changes
+between CIG to BIG, the original Initiator retains control of the Audio Stream and the user’s
+Acceptors.
+The reason for the importance of the handover procedures is that most Initiators and
+Acceptors do not have the resources to handle concurrent unicast and broadcast streams. For
+the High Reliability QoS settings, there simply isn’t enough airtime to do both. That means
+that an Initiator usually needs to stop its unicast stream before staring the broadcast stream,
+even for a 24kHz sampling rate. To ensure continuity for the primary user, i.e., the one who
+is sharing their audio with their friends, the Initiator has to apply the same Context Types to
+the new stream. In most cases, the primary user with the Audio Source will want to continue
+to control the stream, so although the Audio Stream is now broadcast, they will keep their
+ACL link up and associate the new broadcast Audio Stream with the same Content Control
+ID (CCID) they were using for the unicast stream.
+In practice, there are likely to be short gaps in transmission during this handover, which will
+be noticeable on the original Acceptors linked to the Initiator. Whilst implementations can
+try to minimise these, it may be simpler to conceal them by adding a simple announcement to
+the user to “please wait while we move to Audio Sharing”.
+158
+
+Chapter 6 - CAP and CSIPS
+
+6.2.5
+
+CAP procedures for capture and rendering control
+
+As well as the four sets of procedures for Audio Streams, CAP also ties together volume and
+rendering control with five Capture and Rendering Control procedures, which are also selfexplanatory:
+•
+•
+•
+•
+•
+
+6.2.6
+
+Change Volume procedure
+Change Volume Offset procedure
+Change Volume Mute State procedure
+Microphone Mute State procedure
+Change Microphone Gain Settings procedure.
+
+Other CAP procedures
+
+In addition to the Audio Stream, and Capture and Rendering Control procedures, CAP defines
+a set of procedures which are used with the Audio Stream procedures listed above. The first
+one, which we came across in the previous section, is the Coordinated Set Member Discovery
+procedure, which is performed by an Initiator before it sets up a unicast stream, and by an
+Initiator or Commander before they adjust the volume or capture settings on a Coordinated
+Set. Once the members of a Coordinated Set are known, CAP always applies the CSIP
+Ordered Access Procedure prior to any other CAP procedure. The second is the Distribute
+Broadcast_Code procedure, which defines how Broadcast_Codes are distributed for
+encrypted, private broadcast streams.
+
+6.2.7
+
+Connection establishment procedures
+
+With the exception of broadcast use cases, Bluetooth LE Audio uses the same peripheral
+connection procedures that are used for other LE applications, both for device discovery, LE
+ACL connection and bonding. Section 8 of BAP defines the procedures, referring to the
+relevant section in the Core specification and provides recommended values to use for Scan
+Interval, Scan Window and Connection Interval, for quick setup and reduced power situations.
+These are all standard connection procedures defined in the Generic Access Profile (GAP),
+so I won’t spend any more time on them here.
+Section 8 of CAP expands on the BAP requirements with considerations for multi-bond
+situations, where an Acceptor is bonded with multiple Initiators (for example, a phone and a
+TV). It also introduces two new modes which reflect the fact that both Initiators and
+Acceptors can make connection decisions, in order to support the concept of the “Sink led
+journey”. These are the “Immediate Need for Audio related Peripheral” mode and the “Ready
+for Audio related Peripheral” mode.
+
+159
+
+Section 6.2 - CAP – the Common Audio Profile
+6.2.7.1
+
+Immediate Need for Audio related Peripheral (INAP) mode
+
+The INAP mode is use when an Initiator needs to connect to an Acceptor, typically because
+of a user action, such as pressing a button, or an external action, such as an incoming phone
+call. As this is generally a high priority response, the Initiator should use the quick setup
+parameter values for connections from BAP Section 8. The Initiator should determine
+whether the Acceptor that responds is a member of a Coordinated Set, and if so, discover and
+connect to all of the other set members. If more than one Acceptor responds, the Initiator
+should present the user with the available options.
+6.2.7.2
+
+Ready for Audio related Peripheral (RAP) mode
+
+The opposite case to INAP is the RAP mode, where an Acceptor is looking for an Initiator.
+It signals this by transmitting Targeted Announcements containing a Context Type describing
+the use case that it wants supported. If an Initiator can support it, it should attempt to connect.
+When in the RAP mode, the Initiator should use the reduced power parameter values for
+connections, described in BAP Section 8. Once connected to the Acceptor which was sending
+Targeted Announcements, the Initiator should determine whether the Acceptor is a member
+of a Coordinated Set, and if so, discover and connect to all of the other set members.
+The range of Initiator responses in RAP or INAP mode to different forms of announcement
+is described in Table 8.4 of CAP.
+
+6.2.8
+
+Coping with missing set members
+
+There will be occasions when an Initiator has read the Coordinated Set Size characteristic of
+a Coordinated Set member, tried to locate the other members and discovered that some are
+missing. This could happen at the start of setting up an Audio Stream, or during the course
+of a session when the Initiator detects a link loss with one of the Acceptors. It may be because
+one or more of them are physically missing, out of range, or their battery has died. When this
+occurs, the Initiator should try to find the missing member, but should not stop with the
+connection process, nor terminate any existing streams. Instead, it should regularly try to find
+the missing set member(s) and add them back by re-establishing the connection once they
+have been discovered.
+An implementation may decide to adjust the content of a unicast Audio Stream if a member
+is lost. For example, if your phone had been streaming a stereo music stream to a Coordinated
+Set of left and right earbuds and the connection to one of them disappears, it may decide to
+replace the remaining stream with a down-mixed mono stream. However, this behaviour is
+not specified in Bluetooth LE Audio and is implementation specific, as is the retry cadence
+and any timeouts for attempting to find missing Coordinated Set members. This should not
+affect the Context Type, as the use case remains the same.
+-oOo-
+
+160
+
+Chapter 6 - CAP and CSIPS
+The Common Audio Profile is probably best summed up as a profile which says “this is how
+to connect” and “do this for all of the streams you’re dealing with”. As such, it pulls together
+all of the other Generic Audio Framework specifications, providing a common set of
+procedures for the top level profiles to use. As more specifications are developed within GAF,
+CAP will expand to include them.
+We’ll come across all of these CAP procedures in the next few chapters on setting up unicast
+and broadcast streams and Capture and Rendering Control. Most developers will find
+themselves working with these procedures, rather than the underlying BAP procedures, as the
+CAP procedures are the ones which are referenced by the top level profiles. However,
+understanding the BAP procedures and the parameters used for setting up and managing
+streams is a useful exercise to understand how everything fits together in the Bluetooth LE
+Audio world. That’s what we’ll do next.
+
+161
+
+Section 6.2 - CAP – the Common Audio Profile
+
+162
+
+Chapter 7 - Setting up Unicast Audio Streams
+
+Chapter 7. Setting up Unicast Audio Streams
+In this chapter we’ll look at how to configure and set up unicast Audio Streams. The four
+main specifications which are involved at this stage are:
+•
+•
+•
+•
+
+PACS – the Published Audio Capabilities Service,
+ASCS – the Audio Stream Configuration Service,
+BAP – the Basic Audio Profile, and
+CAP – the Common Audio Profile.
+
+CAP sits above the other three and defines procedures which use BAP, ASCS and PACS to
+configure and manage unicast Audio Streams, introducing coordination, Context Types and
+the Content Control ID to associate Audio Streams with use cases. Most of the time, CAP is
+just a set of rules which tells you the correct order to run BAP procedures and how to use
+Context Types and the Content Control ID with them. Top level Bluetooth® LE Audio
+profiles, such as HAP and TMAP, are built on top of CAP, mostly extending mandatory
+requirements. In this chapter I’ll concentrate on the underlying BAP procedures, as they do
+the heavy lifting, but will reference CAP and higher layer profiles where necessary.
+All of the Bluetooth LE Audio specifications are based on the GATT structure of Bluetooth
+LE, which has a Client-Server relationship. Higher layer specifications provide context and
+add control to unicast Audio Streams, but BAP, ASCS and PACS are the foundations they
+build on.
+
+7.1
+
+PACS – the Published Audio Capabilities Service
+
+PACS is the simplest of the specifications and is essentially a statement of what an Acceptor
+can do. Acceptors use PACS to exposes these capabilities in two characteristics – a Sink PAC
+characteristic if it supports the Audio Sink role and a Source PAC characteristic if it supports
+the Audio Source role. These contain Published Audio Compatibility records, called PAC
+records, which include information about the codec and the configurations which it supports,
+along with other optional features. Between them, they expose the full range of capabilities
+of a device.
+PACS can be thought of as a database of the audio capabilities of an Acceptor, which is
+populated at the point of manufacture and is static for the life of the product, although it may
+be updated through a firmware update. PACS specifies information which an Initiator obtains
+when it starts the process of configuring and establishing an audio stream. PACS can also be
+used by a Broadcast Assistant to allow it to filter the Broadcast Streams which it detects, so
+that a Broadcast Sink is only presented with compatible Audio Streams, but we’ll cover that in
+the broadcast chapter.
+
+163
+
+Section 7.1 - PACS – the Published Audio Capabilities Service
+An important concept about the information specified by PACS is that it is a statement of fact
+about the total capabilities of an Acceptor. PACS describes what an Acceptor is capable of
+doing. It is the first step in devices getting to know each other. After an Initiator has read
+these capabilities, an Acceptor will normally expose a more limited set of preferred capabilities
+during the process of establishing a stream, and the Initiator should use those values.
+However, an Acceptor should not reject any configuration an Initiator requests which is based
+on the PACS information it has exposed.
+The intent is that this should result in an efficient configuration process, especially when
+multiple Acceptors are involved, particularly if they come from different manufacturers and
+have differing capabilities. It’s an evolution from the classic Bluetooth audio profiles which
+allow Central and Peripheral devices to go through negotiation loops, where the two devices
+try to ascertain what the optimum settings are for an audio connection. In some
+implementations, that resulted in a deadlock, with a connection never being made, or a poor
+codec configuration which affected the quality of the audio. It’s an issue which resulted in the
+inclusion of the “S” and “D” coding settings as recommendations for CVSD and the “T”
+eSCO42 parameter sets for mSBC. The aim of PACS is to prevent those problems arising.
+
+7.1.1
+
+Sink PAC and Source PAC characteristics
+
+Both the Sink PAC and Source PAC characteristics contains an array of “i” PAC records,
+taking the form shown in Table 7.1.
+PAC Record
+
+Description
+
+Number_of_PAC_Records
+
+Number of PAC records [i] for this
+characteristic
+Octet 0: Codec (defined in the Bluetooth
+Assigned Numbers)
+LC3 = 0x06
+Vendor Specific = xnn
+Octets 1&2: Company ID, if vendor specific,
+otherwise 0x00
+Octets 3&4: Vendor specific Codec ID,
+otherwise 0x00
+Length of the codec specific capabilities for the
+codec in the ith PAC Record.
+
+Codec_ID [i]
+(5 octets)
+
+Codec_Specific_Capabilities_Length [i]
+(1 octet)
+
+The S “safe set” parameters for CVSD were introduced in the Codec Interoperability section (5.7)
+of HFP v1.5 in 2005, followed by the D and T parameters for the mSBC in version 1.6 in 2011, to
+help ensure compatibility when using these codecs.
+42
+
+164
+
+Chapter 7 - Setting up Unicast Audio Streams
+PAC Record
+
+Description
+
+Codec_Specific_Capabilities [i]
+(length varies)
+
+Identification of the capabilities of the codec
+implementation in the ith PAC Record,
+normally expressed as a bitfield.
+Length of the metadata associated with the ith
+PAC Record.
+0x00 if there is none.
+LTV formatted metadata for the ith PAC record
+
+Metadata length [i]
+(1 octet)
+Metadata [i]
+Table 7.1 Format of a PAC Record
+
+7.1.2
+
+Codec Specific Capabilities
+
+Every Bluetooth LE Audio specification includes at least one PAC record using the LC3
+codec, as this is mandated in BAP, which is the lowest layer of the Generic Audio Framework.
+The assigned number for LC3 is 0x06, so every Bluetooth LE Audio Acceptor will have at
+least one PAC record with the Codec_ID of 0x0000000006. Note that the Codec_ID is the
+Coding Format described in the Host Controller Interface Assigned Number list, and not the
+Audio Codec ID defined in the Audio/Video assigned numbers list.
+The Codec_Specific_Capabilities requirements for LC3 are defined in BAP Section 4.3.1 and
+consist of five LTV structures. Each LTV has a Type code, defined in the Generic Audio
+Assigned Numbers document, so that it can be identified by the device reading it.
+As three of these LTVs are bitfields, this means that there are typically multiple combinations
+described within the Codec_Specific_Capabilities structure.
+7.1.2.1
+
+The Supported_Sampling_Frequencies LTV
+
+The first of these five LTV structures is the Supported_Sampling_Frequencies (Type = 0x01)
+shown in Table 7.2, which is a bitfield of sampling frequencies covering the range from 8 kHz
+to 384 kHz. Support for each value is indicated by setting the corresponding bit to one. This
+LTV is mandatory. (Note that this structure is used for any codec. LC3 only specifies 8, 16,
+24, 32, 44.1 and 48 kHz sampling frequencies, shown as lightly shaded).
+15 14 13
+Bit
+RFU
+kHz
+
+12
+
+11
+
+10
+
+9
+
+8
+
+7
+
+6
+
+5
+
+4
+
+3
+
+2
+
+1
+
+0
+
+384
+
+192
+
+176.4
+
+96
+
+88.2
+
+48
+
+44.1
+
+32
+
+24
+
+22.05
+
+16
+
+11.025
+
+8
+
+Table 7.2 Supported Sampling Frequencies (0x01)
+
+7.1.2.2
+
+The Supported_Frame_Durations LTV
+
+The second codec capabilities LTV structure is the Supported_Frame_Durations (Type =
+0x02) of Table 7.3, which provides information on the two frame durations supported by the
+LC3 codec – 7.5ms and 10ms. When set to 1, Bits 0 and 1 indicate support for the two options
+of 7.5ms and 10ms frames for LC3. If both 7.5 ms and 10 ms are supported, bits 4 and 5 can
+165
+
+Section 7.1 - PACS – the Published Audio Capabilities Service
+be used to indicate whether one of them is preferred. Only one of bits 4 and 5 can be set.
+This LTV is also mandatory.
+Bit
+Frame
+Duration
+
+All other bits
+RFU
+
+5
+7.5ms
+preferred
+
+4
+10ms
+preferred
+
+3
+RFU
+
+2
+RFU
+
+1
+10ms
+supported
+
+0
+7.5ms
+supported
+
+Table 7.3 Supported Frame Durations (0x02)
+
+7.1.2.3
+
+The Supported_Audio_Channel_Counts LTV
+
+The third LTV structure is the Supported_Audio_Channel_Counts (Type = 0x03) shown in
+Table 7.4. This is another bitfield, which indicates the number of Audio Channels which can
+be included in a CIS or BIS. Channel Count is the number of multiplexed Audio Channels
+which can be carried in the same direction on an Isochronous Stream, which we covered in
+Section 5.8. Bits 0 to 7 indicate the number of channel counts that are supported, with a value
+of 1 representing a supported option. At least one bit must be set, otherwise it indicates that
+no audio channel can be set up. If multiplexing is not supported, then the Channel Count is
+1 and this LTV structure can be omitted.
+Bit
+15 14 13 12 11 10 9 8
+Channel Count
+RFU
+
+7
+8
+
+6
+7
+
+5
+6
+
+4
+5
+
+3
+4
+
+2
+3
+
+1
+2
+
+0
+1
+
+Table 7.4 Supported Audio Channel Counts (0x03)
+
+7.1.2.4
+
+The Supported_Octets_Per_Codec_Frame LTV
+
+The fourth LTV structure is the Supported_Octets_Per_Codec_Frame (Type = 0x04) shown
+in Table 7.5. This structure consists of two pairs of two octets, with the lower pair (Octets 0
+and 1) specifying the minimum number of octets per codec frame and the upper pair the
+maximum number of octets per codec frame. Both can be set to the same value where only a
+specific number of octets is supported. It defines the range of bitrates43 supported for the
+selected sampling rate (see Table 5.1). This LTV is also mandatory.
+Octet
+Value
+
+15 – 4
+RFU
+
+3
+2
+maximum number of octets per
+codec frame
+
+1
+0
+minimum number of octets per
+codec frame
+
+Table 7.5 Supported Octets per Codec Frame (0x04)
+
+7.1.2.5
+
+The Supported_Max_Codec_Frames_Per_SDU LTV
+
+The fifth and final LTV structure in the Supported_Codec_Specific_Capabilities is the
+Supported_Max_Codec_Frames_Per_SDU (Type = 0x05), which is a single octet stating the
+maximum number of codec frames which can be packed into a single SDU. It can be omitted
+
+43
+
+The bitrate is eight times the number of octets.
+
+166
+
+Chapter 7 - Setting up Unicast Audio Streams
+if there is only one frame per SDU.
+7.1.2.6
+
+The Preferred_Audio_Contexts LTV
+
+The Codec Specific Capabilities can be followed by metadata, which is formatted in an LTV
+structure and described in the Codec Specific Capabilities LTV structures section of the
+Generic Audio Assigned Numbers document. Currently, the only defined metadata that is
+relevant for the PAC record is the Preferred_Audio_Contexts LTV. This is a 2-Octet bitfield
+of Context Type values, using the standard values for Context Types from the Generic Audio
+Assigned Numbers document. If a bit is set to 0b1, it signifies that this Context Type is a
+preferred use case for the codec configuration in that PAC record. It would normally only be
+present if the Acceptor preferred specific PAC records to be used for specific use cases,
+providing more granularity than the Supported_Audio_Contexts characteristic (q.v.).
+
+7.1.3
+
+Minimum PACS capabilities for an Audio Sink
+
+The mandatory BAP LC3 requirements mean that an Acceptor which is acting as a unicast
+Audio Sink, i.e., receiving an Audio Stream, has to support reception of a minimum of one
+audio channel at 16 and 24kHz sampling frequencies with a 10ms frame duration. Similarly,
+every Acceptor which is acting as a unicast Audio Source, i.e., transmitting an Audio Stream,
+has to support encoding a minimum of one audio channel at a 16 kHz sampling frequency
+with a 10ms frame duration.
+The mandatory 16kHz sampling rate for both Audio Sink and Source requires support for 40
+octets per SDU, whilst the 24 kHz sampling rate for the Audio Sink requires 60 octets. For
+the following example, which illustrates PAC records, we’ll just look at the Sink PAC
+characteristic. The same principles apply to the Source PAC characteristic.
+The Sink PAC characteristic would normally expose these two mandatory codec settings in a
+single PAC record. Alternatively, they could be separated into two PAC records, each
+specifying a single set of capabilities. BAP only requires support for one audio channel (Audio
+Configuration 1 in Table 4.2 of BAP), but to illustrate how PAC records are built, let’s look at
+an example of an Acceptor which can support two Audio Channels (corresponding to Audio
+Configuration 4). If it exposed these as four separate PAC records, they would look like Table
+7.6.
+
+167
+
+Section 7.1 - PACS – the Published Audio Capabilities Service
+PAC Record
+Codec_ID (LC3)
+Supported_Codec_Specific_Capabilities_Length
+Supported_Codec_Specific_Capabilities
+Supported_Sampling_Frequencies
+Length
+Code
+Value
+Supported_Frame_Durations
+Length
+Code
+Value
+Supported Audio Channel Counts
+Length
+Code
+Value
+Supported Octets per Codec Frame
+Length
+Code
+Value
+Supported Max Codec Frames Per SDU
+Length
+Code
+Value
+
+0
+
+1
+
+2
+
+3
+
+0x0D
+0x0B
+
+0x0D
+0x0B
+
+0x0D
+0x0B
+
+0x0D
+0x0B
+
+(Example: 16 kHz = 0x04, 24 kHz = 0x10)
+
+0x02
+0x01
+0x04
+0x02
+0x02
+0x02
+
+0x02
+0x02
+0x01
+0x01
+0x04
+0x10
+(Example – 10ms only)
+0x02
+0x02
+0x02
+0x02
+0x02
+0x02
+
+0x02
+0x01
+0x10
+0x02
+0x02
+0x02
+
+(Example: 1 = 0x00; 2 = 0x01)
+
+0x02
+0x03
+0x00
+
+0x02
+0x03
+0x01
+
+0x02
+0x03
+0x00
+
+0x02
+0x03
+0x01
+
+(Example – 40 or 60 = 0x2828 or 0x3C3C
+
+0x03
+0x04
+0x2828
+
+0x03
+0x03
+0x03
+0x04
+0x04
+0x04
+0x2828 0x3C3C 0x3C3C
+(Optional. Default = 1)
+
+0x02
+0x05
+0x01
+
+0x02
+0x05
+0x01
+
+0x02
+0x05
+0x01
+
+0x02
+0x05
+0x01
+
+Table 7.6 Example of PAC records for mandatory 16 & 24 kHz sampling at 10ms, with 1 or 2 audio channels
+
+As an example of how they can be formed, Table 7.6 contains the Sink PAC record parameters
+for the mandated codec requirements for all supported Bluetooth LE Audio profiles i.e., the
+16_2_1, 16_2_2, 24_2_1 and 24_2_2 QoS configurations from Table 5.2 and Table 6.4 of
+BAP. To cover those:
+•
+•
+•
+•
+•
+
+168
+
+The Supported_Sampling_Frequencies characteristic has the values 0x04 for 16kHz
+and 0x10 for 24kHz.
+The Supported_Frame_Durations is 0x02 to signify it only supports 10ms frames.
+The Supported_Audio_Channel_Counts values are 0x00 and 0x01 to support one
+and two channels.
+The Supported_Octets_per_Codec_Frame has values of 0x2828 and 0x3C3C for 40
+and 60 octets, corresponding to the 16_2_n and 24_2_n QoS settings, and finally,
+The Supported_Max_Codec_Frames_Per_SDU indicates that there is only one
+codec frame per SDU with a value of 0x01. This could have been omitted, as it is
+the default value. In this Sink PAC Characteristic there is no metadata.
+
+Chapter 7 - Setting up Unicast Audio Streams
+This same information could be provided more compactly in the single PAC record of Table
+7.7:
+PAC Record
+
+Record 0 (L-T-V)
+
+Codec_ID (LC3)
+Codec_Specific_Capabilities_Length
+Codec_Specific_Capabilities
+Supported Sampling Frequencies
+Supported Frame Durations
+Supported Audio Channel Counts
+Supported Octets per Frame
+Supported Max Codec Frames Per SDU
+
+0x0D
+0x0B
+0x 02 01 14
+0x 02 02 02
+0x 02 03 03
+0x 03 04 3C28
+0x 02 05 01 (Could be omitted as default = 1)
+
+Table 7.7 The PAC record content of Table 7.6 as a single record
+
+These two representations are not completely equivalent. PACS states that an Acceptor must
+support all possible combinations of a parameter value with all other possible combinations
+of a parameter value stated in a PAC record. That means that where multiple values are
+included in a PAC record, as in the example of Table 7.7, then every possible combination is
+valid, as the Initiator can expand the PAC record into an array of all possible combinations.
+In theory, that means that an Acceptor exposing the single PAC record of Table 7.7 should,
+if requested, support a value of 40 octets for 24 kHz sampling and 60 octets for 16 kHz
+sampling, as well as any octet value between 40 and 60, although these are non-standard. For
+the sake of interoperability, there is an assumption that an Initiator should confine itself to
+using the specific configurations from Tables 5.2 and 6.4 of BAP, which are tried and tested,
+but an Initiator should choose another combination. That’s not good practice, but it is
+allowed. If not all of those combinations are valid in a device, then the PAC records should
+be expressed as individual PAC records. However, that may result in the number of records
+multiplying rapidly, which is not a good thing. The example above, which is the baseline
+support needed for BAP, results in four PAC records, which is manageable. Trying the do
+the same thing for TMAP would require at least ninety-six individual PAC records. As we will
+see shortly, the ASE configuration process can guide the preferred options for configuring an
+Isochronous Stream, so individual PAC records can be reserved for applications such as
+vendor specific codec capabilities. More details are provided in Section 3 of PACS.
+If an Acceptor supports more than one codec type, at least one Sink or Source PAC
+characteristic will be required for each, as the Codec_ID field of a PAC record is unique. It is
+up to the implementation to decide whether all PAC records for a codec are exposed in a
+single or multiple Sink PAC or Source PAC characteristics. If there is a large number of PAC
+records, an implementation may want to split them into separate PAC record characteristics
+to prevent the maximum ATT MTU size being exceeded when reading them.
+Sink PAC characteristics do not distinguish between unicast and broadcast streams, so the
+169
+
+Section 7.1 - PACS – the Published Audio Capabilities Service
+values apply to both. Source PAC characteristics are ignored for broadcast.
+
+7.1.4
+
+Audio Locations
+
+Both Sink PAC and Source PAC characteristics can have an associated Sink_Audio_Locations
+and Source_Audio_Locations characteristic respectively, although these are optional. The
+basic concept of Audio Locations was described in Chapter 3. The Sink_Audio_Locations
+and Source_Audio_Locations characteristics specify which audio locations are supported by
+the Acceptor for received and transmitted audio streams. Each characteristic has a four-octet
+field which is a bitfield indicating which rendering or capturing locations it supports.
+If they are present, at least one bit must always be set. If they’re not present, the device should
+accept any Audio Location proposed by an Initiator. In many cases, the values are set by the
+manufacturer and are read only; a right earbud will only ever be a right earbud, so has a single
+Sink and/or Source Audio Location of Front Right. A speaker is likely to have its
+Sink_Audio_Locations characteristic set to both Front Left and Front Right. When an
+Initiator wants to establish a stream, it will decide which Audio Stream to send to match the
+Audio Location. If it finds two speakers, it would normally send a left stereo stream to the
+one exposing the Front Left location and a right stereo stream to the one with the Front Right
+location. If they both say they can support both Front Left and Front Right (which most will
+do), the user would normally use an application to configure which is which. The Sink and
+Source Audio Locations characteristics may be writeable, which would allow a configuration
+utility to set speaker locations for use by other Initiators. Because these are all interoperable
+features in the Bluetooth LE Audio specification, it means that any compliant audio
+application should be able to perform this sort of configuration. Alternatively, a speaker
+manufacturer could provide a physical switch on its speakers to allow a user to specify whether
+a speaker receives a left or right stream.
+In Chapter 5, we looked at how Initiators and Acceptors would deal with downmixing stereo
+streams for mono reproduction, showing that this could be done either before encoding the
+Audio Stream, or after reception and decoding. An Acceptor can use its PAC records to
+indicate to an Initiator whether or not it is capable of downmixing. If it sets its Audio
+Locations for Front Left and Front Right, but only supports a single Bluetooth LE Audio
+Channel, then it implies that it is not capable of downmixing, but will only render the stream
+it has been supplied. If it shows support for two channels, the implication is that it can decode
+them into separate streams to be rendered. It is then down to the Acceptor implementation
+to render them as:
+•
+•
+•
+170
+
+separate left and right Audio Channels, which it would do if it were a sound-bar or
+stereo speaker,
+a single one of the two Audio Channels it decodes, which it might do if it were an
+earbud or hearing aid, or
+downmix them to a single mono Audio Channel, if it contained a single speaker.
+
+Chapter 7 - Setting up Unicast Audio Streams
+The latter two options would usually require some level of user configuration of the speaker,
+but that is down to implementation.
+The Source_Audio_Locations characteristic behaves in the same way, although in most cases
+it is likely to be limited to support of Front Left and Front Right. If the Acceptor supports a
+single Audio Channel, the Initiator would assume that it is a mono microphone. If it supports
+two Audio Channels, the Initiator would assume it is a stereo microphone.
+The use of the Sink and Source Audio Location characteristics, along with Audio Channel
+Counts provides a lot of flexibility for directing Audio Streams. Much of that is implied, so
+implementers need to think carefully about the combination of parameters which they use.
+Both the Sink and Source Audio Location characteristic values can change, including during a
+connection. However, should that happen during a connection, the audio stream does not
+need to be terminated. The new values would apply to the next stream establishment process.
+An application could also decide to change the streams, for instance swapping the left and
+right channel inputs between the left and right Audio Streams if a user wanted to mirror the
+stereo effect because of the direction they were facing. This is implementation specific and
+outside the specifications.
+
+7.1.5
+
+Supported Audio Contexts
+
+Every Acceptor needs to contain a Supported_Audio_Contexts characteristic, which lists the
+Context Types which it supports. This is generally a static list, which will only change as a
+result of a software update which changes the Acceptor’s functionality.
+The name of Supported Audio Contexts is a little misleading. What this characteristic does is
+list the Context Types (i.e., the use cases) for which the Acceptor can make itself unavailable,
+in order to prevent any Initiator trying to establish an Audio Stream for use cases that the
+Acceptor is not interested in. Support for each Context Type is optional, other than the
+«Unspecified» Context Type, which has to be supported in every Supported_Audio_Contexts
+characteristic [BAP 3.5.2.1]. If an Audio Context is not marked as supported, an Initiator can
+still attempt to establish a stream, by mapping that Context Type to the «Unspecified» Context
+Type for its Audio Stream. If the Acceptor currently has «Unspecified» set to available (see
+below), it has no reason to reject that request. As a result of receiving the Audio Stream, an
+Acceptor may decide to change its Available_Audio_Contexts setting for that Context Type
+value for future requests.
+Given this behaviour, an Acceptor should set every Context Type bit in the
+Supported_Audio_Contexts characteristic where it thinks it will ever have a reason to make
+that use case unavailable. (An Audio Sink is prohibited from being unavailable to everything,
+as the Assigned Numbers document prohibits an Available_Audio_Contexts value of 0x0000.)
+
+171
+
+Section 7.1 - PACS – the Published Audio Capabilities Service
+
+7.1.6
+
+Available Audio Contexts
+
+An Acceptor uses the Available_Audio_Contexts characteristic to inform an Initiator that it is
+not available for specific Context Types which it has claimed support for in the
+Supported_Audio_Contexts characteristic. To allow an Initiator to continue with the Audio
+Stream establishment process, the Context Type of the Audio Stream the Initiator wants to
+set up must be shown as available in the Available_Audio_Contexts characteristic. Unlike the
+Supported_Audio_Contexts values, which are static, an Acceptor may update its
+Available_Audio_Contexts values at any time, notifying connected Initiators when it does so.
+To illustrate this, Figure 7.1 shows a typical setting for the Supported_Audio_Contexts and
+Available_Audio_Contexts characteristics.
+
+Figure 7.1 An example of settings for Supported_Audio_Contexts and Available_Audio_Contexts
+
+In this example, the Acceptor supports «Unspecified», (which is mandatory), along with
+«Emergency Alarms», «Ringtone», «Conversational», «Instructional» and «Media». It shows all
+of these as available. If an Initiator wants to start a stream with any of these Context Types,
+the Acceptor should accept it.
+If the Initiator wanted to set up a stream to carry key-press sounds, which are covered by the
+«Sound Effects» Context Type it can. This is because the Available_Audio_Contexts bitmap
+shows «Unspecified» is available. As long as «Unspecified» is shown as available in the
+Acceptor’s, the Initiator is allowed to map an unavailable Context Type that it wants to use to
+«Unspecified» in the properties of its Streaming_Audio_Contexts metadata. In which case,
+the Acceptor must accept the Audio Stream.
+
+172
+
+Chapter 7 - Setting up Unicast Audio Streams
+
+Figure 7.2 An example of the Supported_Audio_Contexts and Available_Audio_Contexts with fewer bits set
+
+Figure 7.2 illustrates the value of setting Context Type bits in the Supported_Audio_Contexts
+characteristic. In this example, «Emergency Alarm» and «Ringtone» are supported, but are not
+currently marked as available in the Available_Audio_Contexts characteristic. This means that
+if an Initiator tries to establish a stream associated with these Context Types it will be rejected
+by the Acceptor as they are specifically set to unavailable. If these bits had not been set as
+Supported, the Initiator could have mapped them to «Unspecified». However, in this case that
+is not allowed, as they are set as Supported and not Available. However, the Initiator could
+still remap streams associated with «Alerts», «Notifications», or any of the other unsupported
+Audio Contexts to «Unspecified», which the Acceptor should accept.
+
+Figure 7.3 An example of settings for Supported_Audio_Contexts and Available_Audio_Contexts with
+«Unspecified» unavailable
+
+Finally, Figure 7.3 shows an example where «Unspecified», is indicated as unavailable
+(remember that «Unspecified» always has to be supported). Setting «Unspecified» to
+unavailable prevents the Initiator from mapping any other unavailable Context Type to
+«Unspecified», so in this case the Acceptor is only available for Audio Streams that are
+associated with «Instructional», «Media» or «Conversational».
+
+173
+
+Section 7.2 - ASCS – the Audio Stream Control Service
+Table 7.8 illustrates this behaviour in terms of how an Initiator interprets the settings.
+Effectively the Acceptor can set three options:
+•
+•
+•
+
+Allowing an Initiator to proceed with establishing an Audio Stream,
+Prohibiting it from starting the Audio Stream, or
+Saying it doesn’t care – have a go.
+
+The combination of Not Supported and Available is not allowed.
+Supported
+Audio
+Contexts
+
+Available
+Audio
+Contexts
+
+0b0
+0b0
+0b1
+
+0b0
+0b1
+0b0
+
+0b1
+
+0b1
+
+Interpretation for Initiator
+
+Audio Stream Context Type may be mapped to «Unspecified»
+This combination is not allowed
+An Initiator shall not establish an Audio Stream with this
+Context Type
+An Initiator may establish an Audio Stream with this Context
+Type
+
+Table 7.8 Result of Supported and Available Bluetooth® LE Audio Context Type settings
+
+Whilst the Supported_Audio_Contexts characteristic is valid for all Clients, an Acceptor may
+expose different values of the Available_Audio_Contexts characteristics to different Clients.
+This allows an Acceptor to make a decision on what different Initiators can do, such as
+controlling which devices can stream media to it or establish an Audio Stream for a phone call
+on a per Client basis. How this is managed is up to the implementation, and it is down to the
+Acceptor to manage the different namespaces required to do this. When implemented, it
+provides a way for Acceptors to set connection policies based on use cases, which was not
+possible with Bluetooth Classic Audio profiles.
+
+7.2
+
+ASCS – the Audio Stream Control Service
+
+PACS specifies the way that an Acceptor exposes its properties and what it is available to do
+at any point in time. For most devices, it will include a wide range of options of codec settings
+and contexts, whether it can act as a Source and/or Sink, and which Audio Locations it
+supports. That range needs to be narrowed down to individual sets of parameters to establish
+Audio Streams, typically because of current resource constraints, such as when an Acceptor is
+supporting multiple Audio Streams, which may need different parameters. This is where the
+Audio Stream Control Service comes into play, by defining Endpoints for each Audio Stream.
+The Audio Stream Control Service is implemented on an Acceptor and defines how an
+Initiator can configure and establish an Audio Stream between the two of them, using the
+procedures defined in BAP.
+
+174
+
+Chapter 7 - Setting up Unicast Audio Streams
+
+7.2.1
+
+Audio Stream Endpoints
+
+At the heart of the Audio Stream Control Service is the concept of Audio Stream Endpoints.
+An Audio Stream Endpoint or ASE is defined as the origin or destination of a unicast Audio
+Stream in an Acceptor. An Initiator does not contain any ASEs – it configures the ASEs on
+the Acceptors. If audio data flows into an ASE, it’s a Sink ASE. If it flows out of it, it’s a
+Source ASE, both of which are described by read-only characteristics. An Acceptor must
+expose at least one ASE for every Audio Stream it is capable of supporting.
+For each ASE, an Acceptor maintains an instance of a state machine for each connected
+Initiator. The state of an ASE is controlled by each Initiator, although in some cases an
+Acceptor can autonomously change the state of an ASE. Whereas a Sink and Source PAC
+provide an Initiator with device wide properties, a Sink and Source ASE represent the current
+state and properties of each connection. These can be read using the Sink or Source ASE
+characteristics, which will return the information shown in Table 7.9.
+Field
+
+Size
+(Octets)
+
+ASE_ID
+
+1
+
+ASE_State
+
+1
+
+State specific ASE Parameters
+
+varies
+
+Description
+A unique ID for each client namespace
+0x00 = Idle
+0x01 = Codec Configured
+0x02 = QoS configured
+0x03 = Enabling
+0x04 = Streaming
+0x05 = Disabling
+0x06 = Releasing
+Depends on the state (See ASCS
+Tables 4-2 to 4-5)
+
+Table 7.9 ASE Characteristic format
+
+Figure 7.4 shows the state machine for a Sink ASE. The state machine for a Source is slightly
+more complex, as it contains an additional Disabling state. We’ll start with the Sink ASE state
+machine, as the journey through to enabling an ASE is essentially the same up until the
+Streaming state, then look at the differences when we get to the Disabling state.
+
+175
+
+Section 7.2 - ASCS – the Audio Stream Control Service
+
+Figure 7.4 The state machine for a Sink ASE
+
+The ASCS defines how an Initiator moves each ASE on an Acceptor through these states, by
+writing a succession of commands to the ASE Control Point characteristic, using the opcodes
+listed in Table 7.10. They are defined in Section 4.2 of ASCS.
+Opcode Operation
+0x01
+0x02
+0x03
+
+Config Codec
+Config QoS
+Enable
+
+0x04
+
+Receiver Start
+Ready
+Disable
+Receiver Stop
+Ready
+(Source ASE only)
+Update Metadata
+Release
+
+0x05
+0x06
+
+0x07
+0x08
+
+Description
+Configures an ASE’s codec parameters.
+Configures the ASE’s preferred QoS parameters.
+Applies the CIS parameters and starts coupling an ASE
+to a corresponding CIS.
+Completes the CIS establishment and signals that an
+ASE is ready to receive or transmit audio data.
+Starts to decouple an ASE from its CIS.
+Signals that the Initiator is ready to stop receiving data
+and completes decoupling a Source ASE from its CIS.
+Updates metadata for an ASE.
+Returns an ASE to the Idle or Codec Configured state.
+
+Table 7.10 ASE Control Point operation opcodes
+
+Opcodes 0x01 (Config Codec) and 0x02 (Config QoS) can be used to transition the state of
+an ASE, or update the configuration of an ASE which is already in that state. Opcode 0x07
+updates the metadata, but doesn’t change the state. All of the other opcodes result in a
+transition to another state in the ASE state machine.
+
+176
+
+Chapter 7 - Setting up Unicast Audio Streams
+An Initiator can perform an ASE Control Point operation on a single ASE, or multiple ASEs
+on an Acceptor, as long as the ASEs are in the same state. If the CIG includes multiple CISes,
+either on one, or multiple Acceptors, the Initiator should ensure that it has collected the
+relevant information from every ASE on every Acceptor before moving the ASEs on each
+Acceptor to the Enabling state. That’s repeating what CAP says, which is to perform the BAP
+procedures on all members of a Coordinated Set in the correct order. (In most cases, all of
+the Acceptors within a CIG will be members of a Coordinated Set. However, ad-hoc sets are
+allowed in CAP, where the Acceptors can be allocated to work together without being
+members of a Coordinated Set, in which case the order of configuration is left to the
+implementation.)
+A feature of ASEs is that an Acceptor supports a separate instance of each ASE for every
+current connection to an Initiator. If there are multiple Initiators with active ACL connections
+to an Acceptor, the Acceptor will maintain an independent set of ASE values for each Initiator.
+It maintains a different ASE_ID namespace for each Initiator. Whilst the ASE_ID must be
+unique within each namespace, the same ASE_ID can exist in different namespaces. That
+means that each ASE_ID and its corresponding handle remains unique. Managing that is
+down to implementation. A change to an ASE in one namespace does not affect the state of
+the same ASE in a different namespace (although the change may result in an application
+moving a CIS from one ASE to another one as a result of that change, but that is
+implementation specific.)
+Where Acceptors have built up a connection history to multiple Initiators, they may decide to
+maintain sets of ASEs with cached configurations. This simplifies the process of transitioning
+to Audio Streams from a different Initiator. Although the specification allows an Acceptor to
+have concurrent Audio Streams enabled with multiple Initiators, in practice that requires a lot
+of resources and complex scheduling, so will probably be rare, at least in early implementations
+of Bluetooth LE Audio devices. In most cases, utilising cached configurations to allow simple
+transitions is likely to provide an easier solution.
+
+177
+
+Section 7.2 - ASCS – the Audio Stream Control Service
+
+Figure 7.5 An example of exposed ASE states for multiple Clients
+
+Figure 7.5 shows how these principles work. It shows how an Acceptor with three ASEs
+might expose its instantiated states for three different Clients. Currently, only Client A has
+established streams with the ASEs which have handles of 0x1234, 0x1236 and 0x1240. If
+Client A reads the three ASEs, it will learn that all three are in the streaming state.
+If Client B were to read the same handles, it would be informed that all three are Idle. Note
+that the ASE_IDs for Client B may be different, as the Acceptor allocates and maintains a
+different namespace for each Client.
+Finally, when Client C reads them, it will be told that they are in the Codec configured state.
+This will normally be because Client C has had a previous set of streams enabled, and when
+they were released, it asked that the Acceptor kept the ASEs in the Codec Configured state to
+speed up the connection next time Client C wanted to establish a stream. In this state, the
+codec configuration is exposed, so that Client C can see whether it needs to perform a
+reconfiguration before moving the ASEs to the QoS configured state. In other states, the
+Codec Configuration information is not exposed.
+An Acceptor must expose at least one ASE for every Audio Stream that it can support. If it
+supports multiple Bluetooth LE Audio Channels in either direction, it must support at least
+one ASE for each of those Audio Channels, regardless of whether it uses all of them. If it
+supports multiplexing, the number of ASEs in each direction must be equal or greater than
+the number of Audio Location bits set to 0b1 for that direction, divided by the highest number
+of Audio Channels set in the Audio_Channel_Counts LTV [BAP 3.5.3]. If an Acceptor can
+support simultaneous Audio Streams from two or more Clients at the same time, it needs to
+provide an ASE for each Audio Stream. Put more simply, the Acceptor has to support
+everything it claims can be thrown at it, with at least one ASE for each Audio Stream.
+178
+
+Chapter 7 - Setting up Unicast Audio Streams
+It might feel as though there is a conflict between the global scope of the PAC record and the
+more limited configurations of an ASE, but they have different purposes and scope. To
+understand the relationship between a PAC record and an ASE, it might be helpful to consider
+an analogy, so we’ll resort to food, again. If you think of a restaurant, the Published Audio
+Capabilities are like a full list of ingredients that are in the kitchen. The ASEs are the individual
+dishes, which only use a subset of those ingredients. The restaurant’s expectation is that
+people choose the dishes off the menu, because they’ve normally been developed to suit the
+use case, which, in the restaurant’s case, may be breakfast, lunch, afternoon tea or dinner.
+However, sometimes the customer might ask for something different. If a diner comes in and
+asks for a burger with chips, mash, toast, pasta and custard (which I hope is not an item on
+any menu), BAP takes the view that the customer is always right and allows it. To limit those
+occasions, it may be helpful to expose individual PAC records which correspond to the use
+cases supported by the top level profiles. In this restaurant analogy, that would be different
+menus for different times of day.
+Moving around the state machine involves an interplay of BAP issuing a command and ASCS
+responding. Rather than going into detail about ASCS in isolation, it’s more useful to look at
+the actual process of interactions in setting up a stream.
+
+7.3
+
+BAP – the Basic Audio Profile
+
+ASCS describes the states of an ASE, but not the procedures to use them, as it’s a service
+which resides on a Server. It is a statement of the current state of each Audio Stream Endpoint
+on an Acceptor. To perform the transitions that move us through the ASE state machine to
+configure an ASE and enable a CIS, we need to move to the Basic Audio Profile which defines
+the Client (Initiator) behaviour. BAP defines these procedures for both unicast and broadcast
+and CAP bundles them together. In this chapter we’ll confine ourselves to the unicast
+procedures. In the next chapter, we’ll do the same for broadcast.
+Setting up a Source ASE is essentially the same process, but with one extra state, which we’ll
+cover at the end.
+
+7.3.1
+
+Moving through the ASE state machine
+
+Moving through the ASE state machine uses a standard process. The Initiator sends a
+command to the Acceptor’s ASE Control Point characteristic, specifying the control operation
+to be performed (i.e., which state it wants the ASEs to transition to, or which state needs to
+be updated). The command specifies which ASE or ASEs the command is being applied to
+and the parameters for that particular state transition.
+The command takes the form shown in Table 7.11 and includes a set of operation specific
+parameters for each state. We’ll look at each set of operation specific parameters as we go
+through the state machine configuration process.
+
+179
+
+Section 7.3 - BAP – the Basic Audio Profile
+Field
+
+Size (Octets)
+
+Opcode
+
+1
+
+Number of ASEs
+
+1
+
+ASE_ID[i]
+Operation specific
+parameters
+
+1
+Varies
+
+Description
+Control Point Operation for the next
+state or update, as shown in Table 7.10).
+Total number of ASEs included in this
+operation
+The IDs of those ASEs
+A list of parameters for each of the [i]
+ASEs
+
+Table 7.11 Basic structure of ASE Control Point command operations for an Initiator
+
+At each stage, the Initiator is made aware of the capabilities of the Acceptor. As long as it
+does not request features outside those provided by the Acceptor, the Initiator can expect the
+Acceptor to honour all of its ASE Control Point command’s values. Once it has received each
+ASE Control Point command, the Acceptor responds with an ASE Control Point
+characteristic notification, indicating success or otherwise for each of the [i] number of
+ASE_IDs which were contained in the command, as shown in Table 7.12.
+Field
+
+Size (Octets)
+
+Number of ASEs
+ASE_ID[i]
+
+1
+1
+
+Response_Code[i]
+
+1
+
+Reason[i]
+
+1
+
+Description
+Total number of ASEs included in this operation
+The IDs of those ASEs
+0x00 if successful, otherwise an error response
+code from Table 5.1 of ASCS
+An extended reason code from Table 5.1 of ASCS.
+0x00 indicates success. Other codes provide
+reasons where the operation has failed.
+
+Table 7.12 ASE Control Point characteristic format as notified to its Client
+
+If any of them are rejected or have errors, there is a comprehensive set of error responses to
+inform the Initiator, so that it can try again. Note that all of these operations are for arrays of
+[i] ASEs. They can be sent as individual commands for each ASE, but it’s more efficient to
+include all of the ASE_IDs for each Acceptor in one operation.
+Assuming there are no issues, the Acceptor then proceeds to update the state of all of its Sink
+ASE and Source ASE characteristics with the values that the Initiator provided in its ASE
+Control Point command. If that results in a change to any ASE, the Acceptor will notify the
+new value for each ASE that has changed.
+As the purpose of the ASE Control Point command is to effect a change to an ASE, either by
+changing its state or updating a parameter value(s), the Initiator will be expecting these
+notifications. However, not all ASEs need to be transitioned or updated in each ASE Control
+Point operation. This means that different ASEs on an Acceptor, associated with the same
+180
+
+Chapter 7 - Setting up Unicast Audio Streams
+Initiator, could be in different states. If in doubt, the Initiator should read their ASE
+Characteristics. (This should not be the case if a CAP Audio Streaming procedure has been
+used, as that requires all Acceptors to be transitioned to the same state before proceeding with
+the next command. However, even here, some ASEs may not be in the Enabled or Streaming
+state if they were configured as “spare” ASEs for future use, which we’ll cover in Section 7.6.)
+A simplified sequence chart for this process is shown in Figure 7.6. The same process applies
+to both Sink ASEs and Source ASEs.
+
+Figure 7.6 Simplified sequence diagram for ASE Control Point operations
+
+The clever (or complicated, depending on your point of view) part of this process is that the
+parameters in the ASE Control Point command and Sink ASE and Source ASE characteristics
+are different for each state as you move through the stream establishment process, providing
+the information that the Initiator needs for its next operation. The different parameters in the
+first few operations allow the Initiator to understand all of the ASE’s capabilities and gather
+the information it needs to set up the CIG and its constituent CISes.
+All of the Initiator’s operations have the same basic format, which was shown in Table 7.9 but
+the State Specific ASE parameters change with each ASE_State. The individual sets of
+parameters are defined in Tables 4-2 to 4-5 of ASCS for the ASE characteristic notifications
+and Sections 5.2 – 5.6 for the ASE Control Point commands.
+The Initiator can configure multiple ASEs with its ASE Control Point operation command,
+but each configured ASE notifies its state independently. For example, if the Initiator were
+to enable four ASEs for a headset - separate Sink ASEs for left and right stereo and two Source
+ASEs for a microphone on each side, it would receive one notification of the updated ASE
+Control Point characteristic and four independent notifications from the two pairs of Sink and
+Source ASEs.
+Each step through the ASCS state machine is defined as a specific procedure in BAP. In the
+next section we’ll go through these to see how the process works.
+181
+
+Section 7.4 - Configuring an ASE and a CIG
+
+7.4
+
+Configuring an ASE and a CIG
+
+An Initiator that wants to make a connection, whether in response to a user action, an external
+event, or a request from an Acceptor, needs to start by determining the capabilities of all of
+the Acceptors involved in the use case. Assuming that it’s the first time the devices have
+connected, and nothing is cached, there are various options for the Initiator to determine the
+Acceptor’s capabilities, depending on whether the Initiator only requires the basic capabilities
+mandated by BAP, whether there are mandated features from another top level profile, which
+means it needs to follow CAP as well, or whether it wishes to use optional or proprietary
+features. These options are shown in Table 7.13. In subsequent connections, there may be
+cached information about the Acceptor’s capabilities that allow the Initiator to skip these steps.
+Initiator Action
+
+Result
+
+Discover Sink PAC
+
+Acceptor can receive unicast audio with BAP mandatory
+settings
+Acceptor can receive unicast audio with BAP mandatory
+settings
+Acceptor can receive unicast audio with BAP mandatory
+settings
+Acceptor can transmit unicast audio with BAP
+mandatory settings
+Acceptor can transmit unicast audio with BAP
+mandatory settings
+Acceptor can transmit unicast audio with BAP
+mandatory settings
+Acceptor supports additional mandatory requirements
+Discover the Acceptor’s capabilities for reception
+Discover the Acceptor’s capabilities for transmission
+
+Discover Sink Audio
+Locations
+Discover Sink ASE
+Discover Source PAC
+Discover Source Audio
+Locations
+Discover Source ASE
+Discover Profile or Service
+Read Sink PAC
+Read Source PAC
+
+Table 7.13 Options for an Initiator to determine an Acceptor’s capabilities
+
+Let’s assume that we’re dealing with either TMAP or HAP and want to connect to a left and
+right pair of Acceptors. Having discovered that HAP or TMAP is supported on at least one
+of the Acceptors, by discovering their service UUID, the Initiator knows the mandatory codec
+configurations and QoS settings that are supported, so it can jump straight into CAP and run
+the CAP Connection procedure for non-bonded devices [CAP 8.1.2] to make a connection.
+The Initiator would determine if the first earbud is a member of a Coordinated Set, and as it
+is in our example, would then run the CAP procedure preamble [CAP 7.4.2], bonding with
+both devices in that set. Having done that, it can start the main business of establishing a
+connection, using the CAP Unicast Audio Start procedure [CAP 7.3.1.2]. This tells the
+Initiator to step through the underlying BAP procedures.
+
+182
+
+Chapter 7 - Setting up Unicast Audio Streams
+First, the Initiator needs to discover all of the Sink and Source ASEs on each Acceptor and
+confirm that each Acceptor has at least one ASE supporting the correct direction for each
+Audio Stream it wants to set up. It should also check the PAC characteristics to make sure
+the Acceptor is capable of supporting the settings it wants to use. These procedures are
+described in BAP in the Audio role discovery procedure [BAP 5.1], the Audio capability
+discovery procedure [BAP 5.2] and the ASE_ID discovery procedure [BAP 5.3].
+Audio role discovery and ASE_ID discovery are both mandatory, as you can’t proceed unless
+you know there is a suitable ASE to connect to. Audio Contexts discovery [BAP 5.4] is
+optional, but is necessary if the Initiator wants to use optional settings.
+BAP has been designed to provide a fall-back level of interoperability that will allow every
+Initiator to be able to set up an audio stream with any Acceptor with a reasonable quality of
+audio, to ensure that they will always be able to connect. If an Initiator is just using mandatory
+settings for BAP or a profile which it has already discovered that the Acceptor supports, it
+may have enough information to skip this stage. Similarly, if the request for connection came
+from the Acceptor, either directly, or through an Announcement, that may have already
+provided this information.
+The next step is to configure the codec for each ASE. Before doing that, the Initiator should
+check that each ASE can support the intended use case by checking the Acceptor’s
+Supported_Audio_Contexts and Available_Audio_Contexts characteristics, using the
+Supported_Audio_Contexts procedure [BAP 5.4] and the Available_Audio_Contexts
+procedure [BAP 5.5]. These were described in Section 7.1.5.
+
+7.4.1
+
+The BAP Codec Configuration procedure
+
+Assuming that the requisite Sink and/or Source ASE requirements are met, the Initiator will
+start to set up the Isochronous Streams by configuring each ASE on each Acceptor. It does
+this by employing the BAP Codec Configuration Procedure [BAP 5.6.1], where it writes to the
+ASE Control Point characteristic with the Config_Codec opcode of 0x01, in order to move
+each ASE to the Codec Configured state. The Initiator has already determined which codec
+configurations are supported, so it now selects the configuration it wants for its current
+application and sends that, along with a target latency and PHY. The format of parameters
+for this operation are defined in Table 5.2 of ASCS and summarized in Table 7.14.
+
+183
+
+Section 7.4 - Configuring an ASE and a CIG
+Parameter
+
+Description
+
+Target Latency [i]
+
+0x01 = low latency 1
+0x02 = balanced latency and reliability
+0x03 = high reliability 1
+0x01 = 1M PHY
+0x02 = 2M PHY
+0x03 = Coded PHY
+Codec configuration (ID, length and configuration),
+as described in Section 7.1.1 - Sink PAC and Source
+PAC characteristics.
+
+Target PHY [i]
+
+Codec_ID [i]
+Codec_Specific_Configuration [i]
+1 These
+
+terms are generic and do not necessarily align with the configurations defined as Low Latency
+or High Reliability in BAP.
+Table 7.14 State Specific Parameters for the Config Codec operation (Table 5.2 of ASCS)
+
+There are a number of points to highlight here. The first is that these parameters do not need
+to be the same for each audio stream. With a bidirectional CIS, the parameters, even the PHY
+and codec, can be different for each direction. In most cases, they are not, but if you’re
+streaming music in one direction and using the return direction for voice control, they may
+well be. Currently, most profiles require that Bluetooth LE Audio is run using the LE 2M
+PHY.
+The second is that the first two of these parameters, which are prefixed with “Target”, are
+recommendations. The Acceptor will use these recommendations to select the QoS
+parameters it feels best satisfy the requirement and fit its current status, then return those
+choices in response to this command.
+In contrast, the Codec ID and Codec Specific Configuration parameters are a statement of
+fact and a directive to the ASE to use specific values. The Codec Configuration is generally
+based on the top level profile being used, which builds on the mandatory options specified in
+BAP. We looked at mandatory and optional QoS configurations defined by BAP and other
+profiles in Chapter 5 (Codec and QoS). In most cases, the Initiator will choose one of these
+predefined configurations, although it could also use settings for an additional codec which is
+defined in a top layer profile, along with its recommended QoS settings, or a manufacturer
+specific codec. It could also make up its own configuration, although that runs the risk of poor
+interoperability. The predefined configurations have been through a lot of testing to ensure
+they perform well and should always be the preferred option.
+Implementers should note that there are two very similar sounding Codec Specific LTV
+structures. The first, the Codec_Specific_Capabilities LTV structure is used in PAC records
+and normally indicates a range of capabilities. The second, the Codec_Specific_Configuration
+LTV structure is used in the Codec Configuration operation (described above) and is for a
+single specific configuration. The syntax of the individual parameters is subtly different.
+184
+
+Chapter 7 - Setting up Unicast Audio Streams
+Throughout this process there is a degree of the devices dancing around each other, making
+recommendations, so that the other can provide information on what best suits its current
+state of resources. This isn’t a true negotiation – it’s better described as an informed
+enumeration, allowing the Initiator to populate a command to send to its Controller to
+configure each CIS. Even then, some of the parameters in its HCI command are also
+recommendations, with the Controller having the final say over how the CIG is configured.
+A third and important point to note is that whilst the Acceptor notifies the range of codec
+parameters it can support through its PAC records, the Initiator writes specific values, which
+define how it will configure the encoded Audio Stream. Some of these are defined with
+different values in the Supported settings the Acceptor sends and the actual settings the
+Initiator writes. For example, as we saw above, the Supported_Sampling_Frequencies LTV is
+a bitfield, whereas the Sampling_Frequency LTV is a specific value mapped to a sampling
+frequency. So, if an ASE supports only 16kHz, the Acceptor will notify a value of 0x04. In
+response, the Initiator will configure the ASE codec to support 16kHz sampling by writing
+0x03. But, to return to the process…
+Having obtained the information about the ASEs, their capabilities and availability, the
+Initiator can issue a single Write without Response command with an opcode of 0x01 which
+includes an array of all of the ASE_IDs that it wants to configure on an Acceptor, which can
+cover both directions. The Acceptor will respond by updating and then notifying its ASE
+Control Point characteristic, including an appropriate response code (0x00 if it’s OK). The
+Acceptor then transitions all of the specified ASEs to the Codec Configured state. Once it
+has done that, the Acceptor will notify the new state of each ASE, including the following
+state-specific parameters (summarized in Table 7.15), which are described in Table 4.3 of
+ASCS.
+Field
+
+Description
+
+Framing
+Preferred PHY
+Preferred Retransmission
+Number (RTN)
+Maximum Transmit Latency
+Presentation Delay Min & Max
+Preferred Presentation Delay
+Min & Max
+Codec Configuration
+
+Support for framed or unframed ISOAL PDUs
+A bitfield of values. Normally includes 2Mbps
+How many times packets should be retransmitted
+The maximum allowable delay for Bluetooth transport
+Supported range of Presentation Delay
+Preferred range of Presentation Delay
+The codec configuration for this ASE
+
+Table 7.15 Notified parameters following a Config_Codec opcode (Table 4.3 of ASCS)
+
+That concludes the BAP Codec configuration procedure and moves us on the to the QoS
+configuration procedure [BAP 5.6.2].
+
+185
+
+Section 7.4 - Configuring an ASE and a CIG
+
+7.4.2
+
+The BAP QoS configuration procedure
+
+At the end of the Codec configuration procedure, the Initiator’s Host will compare the
+configuration parameters that each Acceptor notified at the end of the Codec configuration
+procedure, with the requirements of its current application, and decide what values it wants to
+use within the limits of what it has been told.
+Before issuing the Config QoS command, it is good practice for the Initiator to send these
+parameters to its Controller using the LE Set CIG Parameters HCI command. This will
+confirm that the selected parameter set can be supported. Remember that these are use case
+related recommendations which inform the scheduling algorithms in the Controller. The
+actual values the Controller chooses may differ slightly. The parameters used in this HCI
+command are shown in Table 7.16. Full details of the command are in the Core in Vol 4, Part
+E, Section 7.8.98.
+As we saw in Chapter 4, an application cannot force specific values for the Link Layer.
+Reiterating what is happening here, the Initiator and Acceptors have exchanged information
+which allowed the Initiator to determine what to put into its HCI command to instruct the
+Core to schedule CISes that best meet the application requirements. (Bluetooth
+implementations use the HCI commands for testing to check that they are compliant, but they
+don’t need to be exposed in production products. Silicon and stack vendors may provide
+alternative APIs to provide this functionality, but understanding the HCI commands helps to
+demonstrate how the process works.)
+Parameter
+
+Description
+
+CIG_ID
+SDU_Interval_C_To_P
+
+Assigned by the Initiator’s Host
+Time interval between SDUs generated from the
+Initiator’s Host
+Time interval between SDUs generated from the
+Acceptor’s Host
+Worst case Sleep Clock Accuracy of all of the
+Acceptors in the CIG.
+Preferred packing (sequential vs interleaved)
+Framed if set to 1, otherwise up to the Controller on a
+per-CIS basis.
+The maximum transport latency for each direction.
+
+SDU_Interval_P_To_C
+Worst_Case_SCA
+Packing
+Framing
+Max_Transport_Latency_C_To_P
+Max_Transport_Latency_P_To_C
+CIS_Count
+CIS_ID[i]
+Max_SDU_C_To_P[i]
+Max_SDU_P_To_C[i]
+186
+
+The number of CISes in this instance of this
+command
+Unique ID for each CIS in this CIG
+The maximum sizes of SDUs from the Initiator’s
+and Acceptor’s Hosts.
+
+Chapter 7 - Setting up Unicast Audio Streams
+Parameter
+
+Description
+
+PHY_C_To_P[i]
+PHY_P_To_C[i]
+RTN_C_To_P[i]
+RTN_P_To_C[i]
+
+Bitfield indicating the possible PHYs to use for each
+direction. If more than one bit is set, the Controller decides.
+The Preferred number of retransmissions in each direction.
+The Controller can ignore this.
+
+Table 7.16 LE Set CIG Parameters HCI parameters
+
+In Table 7.16, the parameters highlighted in italics are recommendations to the Controller,
+which it may choose to ignore.
+There are two values which are required in the LE Set CIG Parameters HCI command (Table
+7.16) which don’t come from the ASE configuration process. The first is the
+Worst_Case_SCA, which is the worst case sleep clock accuracy of all of the Acceptors. The
+Initiator needs to determine this by requesting the current value of each devices’ sleep clock
+accuracy before issuing the command, normally by using the LE_Request_Peer_SCA HCI
+command.
+The second is the Packing parameter, which determines whether the CIS events will be sent
+sequentially or be interleaved. If set to 0, they will be sequential; if set to 1 they will be
+interleaved.
+Normally the Framing parameter value is set to 0, so that the Controller can decide, but some
+profiles such as HAP, may mandate that in some cases it be set to 1, to force them to be
+framed. The recommended QoS settings for unicast Audio Streams in Table 5.2 of BAP state
+that they should all be unframed apart from those with a 44.1kHz sampling frequency, which
+should be framed.
+At this stage the Controller does not inform either the Initiator or Acceptor’s Host of the
+values it has chosen – that information will be conveyed when the CISes are established.
+Instead, the Initiator’s Host only gets confirmation that its requested configuration can be
+scheduled, along with Connection Handles for each of the CISes.
+Once the HCI_Command_Complete event has been received, confirming that the CISes can
+be scheduled, the Initiator should send the ASE Control Point operation command with the
+opcode set to 0x02 to each of its Acceptors in turn, with the operation specific parameters
+defined in ASCS Table 5.3 and shown below in Table 7.17. This will move the ASEs to the
+QoS configured state. The values that the Initiator uses in the Config QoS command replicate
+the parameter values that it used in its HCI LE_Set_CIG_Parameters command.
+
+187
+
+Section 7.4 - Configuring an ASE and a CIG
+Parameter
+
+Value
+
+Number of ASEs
+CIG_ID
+CIS_ID [i]
+SDU_Interval [i]
+Framing [i]
+PHY [i]
+Max_SDU [i]
+Retransmission_Number [i]
+Max_Transport_Latency [i]
+Presentation_Delay [i]
+
+i (must be at least 1)
+Values used in the HCI LE_Set_CIG_Parameters
+command.
+(Note: At this stage, these may not be the values which
+the Controller has actually scheduled.)
+
+Value which the Acceptor will use for rendering or capture.
+
+Table 7.17 CIS and Presentation Delay values used with 0x02 Config QoS opcode (ASCS Table 5.3)
+
+The Presentation Delay is the only parameter which is not included in the HCI LE Set CIG
+Parameters command; instead, it is sent directly to the Acceptor. In almost all applications,
+the value of Presentation Delay will be identical for all of the Sink ASEs. The reason the
+Presentation Delay value is the same for each Sink ASE is that this determines the point at
+which audio will be rendered. For earbuds and hearing aids, that would always be at the same
+time. In some situations, such as where there are multiple speakers in a conference room or
+auditorium, there might be a requirement to assign different values to cope with latency related
+to their relative position in the room, but that is not the norm.
+If there are one or more Source ASEs, they are likely to use a different value of Presentation
+Delay to the value for the Sink ASEs, not least because the overall LC3 encode time, which is
+part of the Presentation Delay, is significantly greater than the decode time (around 13ms, as
+opposed to 2ms). So, more time is required to encompass that encoding. However, all Source
+ASEs would normally use the same value as each other.
+The Presentation Delay values chosen by the Initiator for rendering should be within the
+preferred Presentation Delay ranges exposed by the Sink ASEs, and the capturing values
+within those exposed by the Source ASEs. The values should not extend beyond the common
+range within the Max and Min values that were exposed for each direction by the set of
+Acceptors and should ideally be within the range of preferred Max and Min values (see Table
+7.15). If the Context Type is «Live», it should work with the lowest common minimum value.
+As before, the Initiator sends these in a Write without Response command, after which it
+receives a notification of the updated ASE Control Point characteristic, confirming that each
+ASE has moved to the QoS Configured state, followed by notifications of the updated ASE
+characteristic for each ASE. The parameters in these notifications reflect the values set in the
+previous ASE Control Point operation:
+
+188
+
+Chapter 7 - Setting up Unicast Audio Streams
+Parameter
+
+Value
+
+CIG_ID
+CIS_ID
+SDU_Interval
+Framing
+PHY
+Max_SDU
+Retransmission_Number
+Max_Transport_Latency
+Presentation_Delay
+
+Values set by the Initiator in its ASE Control Point
+operation.
+(Note: At this stage, these may not be the values which
+the Controller has actually scheduled.)
+
+Value which the Acceptor will use for rendering or capture.
+
+Table 7.18 Parameters returned by an Acceptor after receiving a Config QoS Opcode (0x02)
+
+Repeating the obvious once again, the Initiator should step through each state for all of the
+ASEs it intends to use on all of the Acceptors before proceeding to the next state, i.e., it takes
+them all to the Codec configured state, then all to the QoS configured state, etc. That’s purely
+common sense, because the Initiator needs to obtain information from all of them to ensure
+that it sets configuration parameters for the ASEs that work across all of the Acceptors.
+However, just to be sure, CAP mandates it.
+If multiple Acceptors or ASEs don’t provide a consistent set of parameters, the Initiator can
+repeat each step of the Config and QoS configuration process for one or more of them,
+including going back from the QoS configured state to the Codec Configured state. However,
+the expectation is that this procedure would normally be a single step process, which is
+performed once only for each set of ASEs, as it’s more efficient if it keeps all of the state
+changes in sync. So, the process steps are:
+•
+•
+•
+
+The Initiator sends a command with target values and explicit values for multiple
+ASEs
+The Acceptor confirms the new state for all of those ASEs (which may include a
+failure code) by notifying its ASE Control Point Characteristic
+The Acceptor separately notifies its preferred values for each ASE using ASE
+characteristics. (This is the point where it states what has been configured and its
+preferences for the next configuration step.)
+
+At any point, the Initiator can check an individual ASE characteristic. The parameters will
+indicate the state of the ASE, along with values that are relevant to that state. If an Initiator
+reads an ASE_ID in the Idle state, the only information it will receive is the ASE_ID and an
+ASE_State value of 0, indicating the Idle state. The most common reason for reading an ASE
+characteristic is to confirm a previous ASE Control Point operation when an expected
+notification has not been received.
+
+189
+
+Section 7.4 - Configuring an ASE and a CIG
+When the Initiator’s Host sent the HCI LE Set CIG Parameters command to its Controller,
+it moved the CIG to its Configured state. The Initiator’s Host can still modify the properties
+of a CIS or add further CISes to the CIG at this point, by resending an HCI LE Set CIG
+Parameters command for specific CISes, using its array structure. If it does, it will need to
+update the relevant ASEs using the Config QoS command by writing to the ASE using the
+array capability of the ASE Control Point characteristic.
+Once all of the ASEs are in the QoS configured state and the CIG is configured, we can start
+to enable the Audio Streams.
+
+7.4.3
+
+Enabling an ASE and a CIG
+
+Having configured all of the ASEs, the Initiator will have all of the information it needs to
+instruct its Controller to enable the CIG and constituent CISes. The Acceptor’s Host knows
+the main parameters of the Isochronous Channels, although some values are still provisional,
+as the actual RTN value will depend on the scheduling. If the Controller is also having to
+support other wireless connections, it may decide to compromise to allocate airtime between
+the different and set a lower value than its Host requested. At the point that each CIS is
+established, the settings will be configured at the Link Layer level and notified to the respective
+Hosts of the Initiator and each Acceptor.
+The Initiator can now invoke the Enabling an ASE procedure defined in BAP 5.6, by writing
+to each Acceptor’s ASE Control Point characteristic with the opcode set to 0x03 (Enable),
+using the parameters shown in Table 7.19. From this point, the Initiator cannot change the
+CIG, CIS or ASE configurations, with the exception of the Audio Stream metadata.
+Parameter
+
+Description
+
+Number of ASEs
+ASE_ID [i]
+Metadata Length [i]
+Metadata [i]
+
+Total number of ASEs to be enabled (i)
+The specific ASEs which are being enabled
+Length of metadata for ASE [i]
+LTV formatted metadata. This is most commonly the
+Streaming_Audio_Contexts and CCID_List LTV
+structures, but can include other LTV structures.
+
+Table 7.19 State Specific Parameters for the Enabling operation – Opcode 0x03 (Table 5.4 of ASCS)
+
+CAP mandates that the metadata in the Enabling operation includes the
+Streaming_Audio_Contexts, with the correct values set for each unicast Audio Stream. If
+there is a content control procedure associated with the Audio Stream, then the Metadata in
+the Enabling command must also include the CCID_List LTV structure. [CAP 7.3.1.2.6]
+Each Acceptor will notify its ASE Control Point characteristic, move its ASEs to the Enabling
+state, then, in turn, notify the ASE characteristic for each of the ASEs specified in the ASE
+Control Point command, returning their CIG and CIS IDs, along with any metadata written
+190
+
+Chapter 7 - Setting up Unicast Audio Streams
+by the Initiator, as shown in Table 7.20.
+Parameter
+
+Value
+
+CIG_ID
+CIS_ID
+
+Values set by the Initiator in its previous Config QoS
+operation.
+
+Metadata Length
+Metadata
+
+(0 if there is no metadata for this ASE)
+LTV structures for the Sink or Source ASE
+
+Table 7.20 Additional parameters for the ASE characteristic when in the Enabling, Streaming and Disabling states
+
+The metadata values for an ASE can only be set or updated by the Initiator. An Acceptor
+cannot make changes to these, even if it is acting as the Audio Source. An Acceptor can
+update its Available_Audio_Contexts value at any point, but these do not appear in the ASE
+metadata. Nor do they result in the termination of a current stream. Any change in the
+Available_Audio_Contexts is only used for subsequent connection attempts.
+Now that the ASEs are in their Enabling state, it is time to enable the required CISes and
+couple them to the ASEs. Remember from Chapter 4, that not every configured CIS included
+in the CIG needs to be established, as an Initiator can configure CISes which it can swap in
+and out as its demands change. The Initiator enables the CISes it requires by invoking the
+Connected Isochronous Stream Central Establishment procedure from the Core [Vol 3, Part
+C, Section 9.3.13]. The Link Layer will generate a CIS_Request to the Acceptor, which
+establishes the CIS using the Isochronous Stream Peripheral Establishment procedure from
+the Core [Vol 3, Part C, Section 9.3.14]. At this point, the Initiator will start transmitting
+packets with zero length PDUs. If it is a bidirectional CIS, both Sink and Source ASEs need
+to be established.
+There is a condition on the behaviour of the ASE at this point which implementors need to
+be aware of. If the CIS existed before the enabling operation, the transition from the Enabling
+state to the Streaming state is not notified. Instead, the Acceptor autonomously transitions
+the ASE to its Streaming state, without the need for the Initiator to write the Streaming
+command. This is most likely to occur if the Acceptor is reusing an existing configuration and
+the original CIG had never been disabled.
+As we saw in the CIG state machine in Chapter 4, once the CIG has been enabled, the CIS
+configurations cannot be changed, so any change to the codec configuration, QoS parameters
+or Presentation Delay would require the CIG to be terminated and the configuration process
+restarted. An individual CIS can be disconnected or restored (using the LE Create CIS
+command), but only their metadata can be updated once they are in the Enabling or Streaming
+state.
+
+191
+
+Section 7.4 - Configuring an ASE and a CIG
+
+7.4.4
+
+Audio data paths
+
+The Isochronous Streams are now up and running, but we haven’t connected any audio data
+– the Isochronous Channels are just conveying empty packets. The next step, if it has not
+already been done, is to connect a data path for each ASE, which uses the Audio Data Path
+Setup Procedure (BAP 5.6.3.1), which in turn uses the LE Setup ISO Data Path HCI
+command.
+Bluetooth LE Audio provides a lot of flexibility for the audio data path and where the codec
+is located. The codec can reside in the Host or in the Controller, and the audio data can come
+through HCI, or more typically via PCM or a vendor proprietary route.
+
+Figure 7.7 Common Audio Data Path configurations
+
+Figure 7.7 shows three of the most common data path configurations, showing that the codec
+can be implemented in the Host or the Controller. Audio is typically routed from the codec
+via a PCM interface, but, if encoded in the Host, it can also be transported over HCI. To set
+up each data path, the Initiator and Acceptors will both use the LE Setup ISO Data Path HCI
+command to bind the codec configurations that were written during the Config Codec state
+to the Connection Handle of each CIS, specifying the direction depending on whether it is a
+Source or Sink ASE. The data path configuration may be different in the Initiator and
+Acceptor, as the codec location and data path implementation will normally depend on the
+specific chipset being used. As the data path setup is internal to each device, that’s an
+implementation choice. This level of detail is normally hidden from the application developer
+below a more general API, but it’s useful to know what happens at each stage of the
+configuration process.
+Up to this point, the process is common for both Sink and Source ASEs. Now the two
+processes diverge. They’re still on the same path within the State Machines, but there is a
+difference in the order of commands.
+192
+
+Chapter 7 - Setting up Unicast Audio Streams
+For a Sink ASE, once the Acceptor has successfully set up its data path, it may autonomously
+move the ASE to the Streaming state, then notify each Sink ASE characteristic to the Initiator
+with the ASE state code set to 0x05 (Streaming). As soon as it receives this notification point
+the Initiator can start streaming audio packets to that ASE Sink.
+
+Figure 7.8 Establishment of a Sink ASE, showing the ASE states
+
+Figure 7.8 shows a very simplified message sequence chart for establishing a Sink ASE,
+showing the overall sequence and the Acceptor’s autonomous transition to the Streaming
+state. More detailed MSCs are provided in BAP Section 5.6.3.
+For Source ASEs, the Acceptor needs to wait for the Initiator to confirm that it is ready to
+receive data from that ASE. Once it has received the ASE characteristic notification from the
+Acceptor, and as soon as it is ready to transmit audio data, the Initiator will write the ASE
+Control Point characteristic for that ASE with the opcode set to 0x04 (Receive Start Ready).
+(Note that Sink ASEs do not need to be included in the array of ASEs in this command, as
+they will independently move to the Streaming state.) At this point, the Acceptor will
+transition the Source ASE to the Streaming state, notify its ASE characteristic and commence
+audio streaming.
+A corresponding MSC for a Source ASE is shown in Figure 7.9. Here the Initiator needs to
+issue the Receiver Start Ready command to the Acceptor’s ASE Control Point characteristic
+to initiate the start of audio data packet transfer.
+193
+
+Section 7.4 - Configuring an ASE and a CIG
+
+Figure 7.9 Establishment of a Source ASE, showing the ASE states
+
+7.4.5
+
+Updating unicast metadata
+
+Whilst in the Enabling or Streaming states, an Initiator can update the metadata for any ASE
+by writing to the ASE Control Point with an opcode of 0x07 (Update Metadata) [BAP 5.6.4].
+This is a convenient way of reusing an ASE for a different audio application, without having
+to tear down and re-establish any Audio Streams.
+This is described in the CAP Unicast Audio Update procedure [CAP 7.3.1.3] and allows an
+ASE to be reused, keeping its current configuration, but changing the Context Type and
+CCID_List metadata. A typical application would be to use the same stream to move from a
+phone call to a music player on the same phone. There are inherent limitations, which is that
+the codec configuration cannot be changed. Another issue is that a phone call is normally
+bidirectional, whereas music streams are unidirectional. However, during the configuration
+process, enough ASEs could be configured to cope with this. At the point of transition of the
+use case, the Initiator could disable and enable ASEs until it has the number of configured
+ASEs which it needs, updating their metadata in the process. Note that in the case of a
+bidirectional CIS, an ASE might be disabled, but the CIS would remain enabled to support
+the other direction and its ASE. It is a good example of how the Bluetooth Classic Audio
+multi-profile issues have been designed out by adding flexibility for Audio Streams to be
+repurposed and reused.
+
+194
+
+Chapter 7 - Setting up Unicast Audio Streams
+
+7.4.6
+
+Ending a unicast stream
+
+The process of ending a unicast stream is where the Sink and Source ASE state machines
+diverge. We’ll cover the Source ASE in the next section. The difference is that a Source ASE
+has a Disabling state, whereas the Sink ASE does not.
+To stop streaming to a Sink ASE, the ASE needs to transition back to the QoS Configured
+state, using the CAP Unicast Audio Stop procedure [CAP 7.3.1.4]. This calls the BAP
+Disabling an ASE procedure [BAP 5.6.5], where the Initiator writes to an ASE with the 0x05
+(Disable) command. At this point, the Initiator stops transmitting audio data to the ASE. If
+the CIS is bidirectional and the Source ASE has not been disabled, the Initiator will continue
+to transmit null PDUs to allow the Acceptor to return its audio data.
+The associated CIS is not automatically disabled at this point. If the Initiator wishes to remove
+it, it should use the HCI_Disconnect command [Core Vol 4, Part E, 7.1.6]. As disabling the
+Sink ASE only moves it to the QoS configured state, the ASE can be re-established at a later
+point by writing the Enable opcode to the ASE Control Point characteristic. If the CIS has
+been disabled using the HCI Disconnect it can still be restored using the HCI LE Create CIS
+command.
+If there is no intention of reusing the ASE, it can be moved to the Releasing state by
+performing the BAP Releasing an ASE procedure [BAP 5.6.6]. Once the ASE is in the
+Releasing state, the Acceptor autonomously transitions it to the Idle state, or, if it wants to
+simplify the next ASE configuration cycle, it can cache the codec configuration by returning
+it to the Codec configured state.
+Once all of the ASE’s associated with a CIS have been disabled, the Initiator can disable any
+CISes which are still enabled and tear down the associated data paths. When all of the CISes
+in a CIG (across all Acceptors) have been disabled, the CIG should be moved to the Inactive
+state.
+
+195
+
+Section 7.4 - Configuring an ASE and a CIG
+
+7.4.7
+
+The Source ASE state machine
+
+A Source ASE behaves slightly differently, as we need a confirmation from the Initiator to
+ensure a clean transition. This introduces a Disabling state. It is only used for Source ASEs,
+and is shown in the Source ASE state machine of Figure 7.10
+
+Figure 7.10 State machine for a Source ASE
+
+For a Source ASE, the Acceptor needs confirmation from the Initiator that the Initiator is
+ready to stop receiving audio data. The Acceptor can autonomously stop streaming, but the
+Receiver Stop Ready command results in a clean termination of streaming. This is important
+if there is a potential requirement to reuse the ASE. Having notified the fact that it is in the
+Disabling state, the Acceptor should wait for a command from the Initiator to tell it to stop
+streaming and move to the QoS configured state. From there, the Source ASE can be moved
+to the Releasing State, but if it does, it will not be possible to reuse the CIS without taking the
+ASE through the state machine again.
+Once in the Releasing state, whether by a direct transition from the Streaming state for a Sink
+ASE, a direct transition from the Disabling state for a Source ASE, or being released from the
+QoS configured state for either, both Initiator and Acceptor should tear down their data paths
+and terminate the CIS. The Acceptor can choose to transition to either the Idle or Codec
+Configured states, both of which operations are performed autonomously. If the CIS is
+bidirectional, it should not be terminated until both Sink and Source ASEs are in the Releasing
+State. When the final CIS is terminated, the CIG moves from the Inactive CIG state to the
+No CIG state.
+
+196
+
+Chapter 7 - Setting up Unicast Audio Streams
+
+7.4.8
+
+Autonomous Operations on an ASE
+
+We’ve already seen that an Acceptor can autonomously transition a Sink ASE from the
+Enabling State to the Streaming state and from the Releasing state to the Idle state or the
+Codec Configured state. These are not the only occasions where an Acceptor can act
+autonomously. With the exception of the transitions shown in Table 7.21, all of the state
+machine transitions can be performed by either the Acceptor or Initiator. However, in most
+cases, moving around the state machine is performed as described above.
+ASE Type
+Sink and Source
+Sink and Source
+Sink and Source
+Source
+Sink and Source
+Sink and Source
+
+Current State
+
+Next State
+
+Initiating Device
+
+Codec Configured
+QoS Configured
+QoS Configured
+Disabling
+Releasing
+Releasing
+
+QoS Configured
+QoS Configured
+Enabling
+QoS Configured
+Codec Configured
+Idle
+
+Initiator
+Initiator
+Initiator
+Initiator
+Acceptor
+Acceptor
+
+Table 7.21 ASE transitions which are confined to Initiator or Acceptor actions
+
+7.4.9
+
+ACL link loss
+
+If the ACL link is lost between an Acceptor and an Initiator, all CISes to that Acceptor are
+disconnected. Where there are other Acceptors involved in the CIG, the Acceptor should
+move all of the ASEs associated with the lost ACL link to the QoS configured state, so that
+the CISes can be reenabled if the link returns. The Acceptor will notify this change of state,
+although it is questionable whether the Initiator will receive the notification.
+As Acceptors are not necessarily aware of the existence of any other Acceptor(s), they may
+not know whether the CIG is still active or streaming to other Acceptors. If they don’t receive
+a reconnection request after a link loss, they may employ an implementation specific timeout
+to return their ASEs to the Idle state. On a later reconnection of the ACL link, the Initiator
+should read the ASE characteristics of that Acceptor to check its state. If it was released from
+the QoS configured state because of a timeout, the Initiator will normally need to tear down
+and re-establish the entire CIG. Remember that there is an asymmetry between Initiator and
+Acceptor – the ASE state machine resides on the Acceptor, whereas the CIG state machine is
+on the Initiator. An Acceptor is never aware of the CIG state; it is what the Initiator, which
+has a global view of the ACL connection status, uses to drive the ASE states.
+
+197
+
+Section 7.5 - Handling missing Acceptors
+
+7.5
+
+Handling missing Acceptors
+
+Before starting to configure ASEs, CAP requires that an Initiator connects to all of the
+Acceptors in the target Coordinated Set. In the real world, there will be occasions when not
+all of them are present, either because one of them is turned off, missing, or out of range. In
+this case, the Initiator should go ahead with setting up the members of the Coordinated Set
+which it can find. How long it waits before this happens and whether it involves user feedback,
+as well as the choice of Audio Channels to send to the available Acceptor are all
+implementation specific. The Initiator should continue to search for the missing Acceptors,
+adding them in when it discovers them.
+The nature of a CIG is that once it is Active, additional CISes cannot be configured. Hence,
+if the Initiator only configures CISes for the Acceptors it can find, then, if missing ones turn
+up, and no CIS had been scheduled for them, it would have to tear down the CIG, reconfigure
+it and re-establish the CISes, which will disrupt the Audio Streams.
+To prevent that break in the audio, an Initiator can schedule the CIG with cached values for
+any missing ASEs. If it detects the presence of the missing Acceptor at a later point, it can
+configure their ASEs, then enable the associated CISes in the active CIG, (as they have already
+been scheduled), and start the transfer of audio data. CAP doesn’t describe this process; it’s
+an extension of functionality by using BAP procedures in addition to CAP ones, but it can
+result in a better user experience.
+
+7.6
+
+Preconfiguring CISes
+
+A similar case exists where an Initiator knows that an Acceptor is likely to participate in a
+number of different use cases which have different ASE configurations. A common example
+of this is streaming music (the analogue of A2DP), which just needs Sink ASEs to enable the
+CISes carrying audio from Initiator to Acceptor, but which may be interrupted by incoming
+phone calls, where a return audio stream from the microphone(s) is required. To make this
+transition faster, the Initiator can configure a set of ASEs which fulfil both use cases, i.e., two
+Sink ASEs and two Source ASE, so that when it wants to transition between the two
+applications, all it needs to do is enable or disable the Source ASEs, as opposed to tearing
+everything down and restarting.
+The downside to this is that the airtime for the return Isochronous Stream is always allocated,
+although it may never be used. A stereo 48_2_2 stream for High Reliability takes up around
+63% of the airtime. If it is scheduled to include a 32_2_2 return stream, that increases to 90%.
+By comparison, a bidirectional 32_2_2 stream takes only 41% of airtime. There is also a
+discrepancy in the latency requirements of the two applications. For music streaming, which
+generally uses the High Reliability QoS settings, the overall latency for a 48_2_2 stream, with
+a 32_2_2 return is just over 140ms. That’s a lot longer than the 56ms you would get with a
+bidirectional, Low Latency 32_2_1 stream. Table 7.22 illustrates the effect of different QoS
+configurations on airtime and latency for bidirectional streams.
+198
+
+Chapter 7 - Setting up Unicast Audio Streams
+QoS Configuration
+Outbound
+
+Inband
+
+48_2_1
+32_2_1
+24_2_1
+
+32_2_1
+32_2_1
+16_2_1
+
+Airtime
+
+Latency
+
+(Stereo)
+
+PD = 40ms
+
+90%
+41%
+31%
+
+141ms
+57ms
+56ms
+
+Table 7.22 Airtime and Latency for various bidirectional QoS configurations
+
+It means that there is a trade-off between airtime, QoS and call setup time, which designers
+need to be aware of. Making this decision will usually depend on other airtime resource
+demands in the Initiator. It illustrates the flexibility of Bluetooth LE Audio, but highlights the
+fact that implementors need to understand more of the overall system than they did with
+classic Bluetooth Audio profiles.
+
+7.7
+
+Who’s in charge?
+
+One of the questions for designers is deciding “who is in charge?” The development of
+Bluetooth LE Audio has seen some strong opinions; from phone manufacturers, who feel
+that the phone is responsible for everything, but also from speaker and hearable
+manufacturers who feel that more autonomy needs to be given to their products, leading to
+the concept of the Sink led journey. The specifications give room for both, but they need
+developers to be aware of some of the consequences.
+It’s useful to look at a simple example, which is a pair of earbuds, each of which has a
+microphone. A phone manufacturer might want to use both, as that would allow them to
+process both audio streams, which should result in a better voice signal. On the other hand,
+earbud manufacturers might prefer to use only one, conserving the battery life of the other
+earbud. They might even want to be able to swap which is being used during the course of
+a call, if they see the battery of the one with the active microphone starting to decline.
+The question is, how to enable these options? Assuming there is communication between
+the earbuds, one of them could consider not exposing its Source ASE, although that’s a
+brute force approach, especially as the phone could infer its presence by reading the PAC
+records. As we saw above, an Acceptor should expose an ASE for every Audio Stream it is
+able to support, even if it may not be able to support it at all points in time. A far better
+method is to indicate that the server is not available to transmit audio by using the
+Available_Audio_Contexts characteristic in PACS with the Available_Source_Contexts set
+to 0x0000 (which is one of the few occasions where you are allowed to set all Context Type
+bits to zero.)
+If the Initiator (phone) can see that each earbud has an available Source ASE, then it can
+effectively take charge. It may decide to select just one (not least because processing two
+incoming streams which are not time aligned is not trivial and increases its power
+consumption). It can use other information, such as checking the battery status of each
+199
+
+Section 7.7 - Who’s in charge?
+earbud, to guide its choice of microphone if it only wants to use one.
+If it’s more intelligent, it could look to see if phrases are regularly repeated during the call,
+infer that implies poor signal to noise, and add in the second microphone to try to improve
+the quality by gaining a few dBs of signal. Equally, a phone app could log the duration of
+calls with specific people, and from that history deduce the likely length of the call, the
+earbud battery life and use that to inform its decision of how many Audio Streams to set up
+and the QoS settings that would let the battery survive through the anticipated call duration.
+So much opportunity for differentiation! I suspect both of these approaches are unlikely in
+the short term, as the phone will continue to consider the earbuds as a resource to use as it
+likes.
+On the other hand, supporters of the Sink led journey would want more decision making
+ability in the earbuds. However, that implies an ability for them to communicate with each
+other, which is currently out of scope (other than the Preset local synchronisation in HAPS).
+If the pair decide that they only want to use the microphone on one earbud, they can decide
+between them which will be the one to take the battery hit, and the other can set its
+Available_Source_Contexts set to 0x0000. (They can do this while they’re still in the battery
+box, so this doesn’t have to use a separate, sub-GHz radio link.)
+This approach has the advantage that the phone knows that the Source ASE is there for
+both, so it can configure a stream in the CIS. It can’t enable it for the earbud showing
+Available_Source_Contexts as 0x0000, because the earbud will reject it at the config codec
+stage. However, the Initiator can schedule it, as it can put whatever it wants in the HCI LE
+Set CIG Parameters command. Although we’ve seen that the process normally uses
+information that the Initiator receives from an Acceptor, it can use cached or assumed
+values for missing Acceptors or ASEs. That means that if there is a need to swap
+microphones at some point, that can happen. Otherwise, the Initiator would need to tear
+down the CIG and rebuild it, resulting in a break in the call. (That shouldn’t terminate the
+call, as the loss of a stream doesn’t cause any change in the TBS state machine, but it’s not a
+good user experience).
+The point here is that the PACS and ASCS characteristics allow a surprising amount of
+ownership regarding how an Audio Stream is set up. An Initiator can act unilaterally, but
+that risks affecting how well devices work. For the best performance and user experience,
+designers need to understand the options offered and how to use them.
+--oOo-That concludes the methods for setting up unicast Audio Streams, so we can now turn our
+attention to broadcast.
+
+200
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+
+Chapter 8. Setting up and using Broadcast Audio Streams
+In this chapter we’ll look at how to configure broadcast Audio Streams. Broadcast is the major
+new feature of Bluetooth® LE Audio and has the potential to change the way that everybody
+uses audio. The exciting use cases it introduces include the ability to share audio, along with
+the ability to set up ad-hoc private connections. The latter is enabled by the Broadcast
+Assistant features which are defined in BASS and bring totally new control and user
+experiences.
+Once again, the main specification is BAP – the Basic Audio Profile, but we now need to add
+another one of the GAF specifications called BASS – the Broadcast Audio Scan Service. BASS
+defines how an additional device can be used to help find broadcast Audio Streams and
+instruct one or more Acceptors to synchronise with them.
+CAP once again comes into play. As well as providing procedures to set up and receive
+broadcast streams, it also defines the role of a Commander. A Commander can be a physical
+device, or an application on a phone or a TV. We’ll look at the Commander role in more
+detail once we’ve gone through the basics of sending and receiving broadcast Audio Streams.
+CAP’s main task is to lay down rules about associating Context Types and Content Control
+IDs and repeat the order in which things need to be done.
+We’ll start with the basics of setting up and receiving a broadcast Audio Stream, then look at
+what the Commander brings to the picture. In Chapter 12, we’ll delve further into the
+possibilities within broadcast audio, examining some of the new use cases that it can enable.
+Although the broadcast topology originated with a desire to improve the quality provided by
+inductive hearing loops, Bluetooth LE Audio provides far more versatility than an inductive
+loop. In many cases there will be an ACL connection between the Broadcast Source and the
+Broadcast Receiver in addition to the one between the Broadcast Source and a Commander.
+Devices can even support both broadcast and unicast at the same time, acting as a relay
+between broadcast and unicast connections. But before we get into those complications, we’ll
+start with the basics of broadcast by itself.
+The broadcast specifications fall into three separate functions:
+•
+
+•
+
+Transmitting a broadcast Audio Stream. At its simplest, a Broadcaster (which is an
+easier name to use for a Broadcast Source) acts independently – it generally has no
+idea whether any receivers are present and listening to it.
+Finding a broadcast Audio Stream. This will generally use a Broadcast Assistant,
+often referred to as a Commander, (although technically that’s just a role, of which
+the Broadcast Assistant is a sub-role). A Broadcast Assistant can find broadcast
+Audio Streams for a Broadcast Receiver. Broadcast Assistants elevate broadcast
+from being a simple telecoil replacement into a very powerful new topology, which
+201
+
+Section 8.1 - Setting up a Broadcast Source
+
+•
+
+8.1
+
+allows encrypted streams to be used for private audio, both at a personal and an
+infrastructure level. They also make it much easier to select amongst multiple
+broadcast Audio Streams. Broadcast Assistants can be designed as stand-alone
+devices, or can be collocated with any Broadcast Source.
+Receiving a Broadcast Audio Stream. A broadcast receiver, (which is essentially the
+same as a Broadcast Sink), can scan for the presence of a broadcast Audio Stream
+and synchronise to it. At this basic level, it also acts independently. A Broadcast
+Receiver can synchronise to encrypted or unencrypted broadcast Audio Streams,
+but needs to obtain the Broadcast_Code to decrypt an encrypted Audio Stream.
+This can be done out of band, or with the help of a Broadcast Assistant
+
+Setting up a Broadcast Source
+
+In Chapter 4 we covered the basics of Broadcast Isochronous Streams (BIS) and Broadcast
+Isochronous Groups (BIG). Unlike unicast streams, a Broadcast Source and Broadcast Sink
+operate independently. This makes the Broadcast Source very different from any other
+Bluetooth Central device, as it acts unilaterally. Unlike the unicast case, which we explored in
+the previous chapter, no commands, requests or notifications take place between devices.
+Instead, the Broadcast Source is driven entirely by its specific application.
+One result of this is that a Broadcast Source has a very simple state machine, shown in Figure
+8.1. As there are no interactions with any Acceptors, the procedures are very straightforward,
+consisting of commands from the Host to the Controller within the Broadcast Source.
+
+Figure 8.1 Broadcast Source State Machine and Configuration procedures
+
+To configure a Broadcast Source, the Host needs to provide the Controller with details of the
+BIG configuration. This is used by the Controller to schedule the BISes. It also needs to
+provide the information to populate the BASE, which describes the configuration of each BIS
+and its content. The configuration data is provided by the application running the Broadcast
+Source. In some applications, the metadata for inclusion in the BASE may be part of an
+application; in others it might be supplied externally, either from a control input, or extracted
+from an incoming audio stream, such as a TV’s electronic program guide data.
+202
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+BAP defines six procedures for transitions and configuration updates of a Broadcast stream:
+•
+•
+•
+•
+•
+•
+
+The Broadcast Audio Stream configuration procedure
+The Broadcast Audio Stream establishment procedure
+The Broadcast Audio Stream disable procedure
+The Broadcast Audio Stream Metadata update procedure
+The Broadcast Audio Stream release procedure, and the
+The Broadcast Audio Stream reconfiguration procedure
+
+Because there are no connections between devices, these procedures are mostly confined to
+HCI commands, sent when the Initiator is ready to start broadcasting. CAP bundles them
+together into just five procedures, combining configuration and establishment into its
+Broadcast Audio Start procedure:
+•
+•
+•
+•
+•
+
+8.2
+
+Broadcast Audio Start procedure
+Broadcast Audio Update procedure
+Broadcast Audio Stop procedure
+Broadcast Audio Reception Start procedure
+Broadcast Audio Reception Stop procedure
+
+Starting a broadcast Audio Stream
+
+As the Broadcast Source has no knowledge of what might be receiving its broadcast Audio
+Streams, none of the CAP preamble procedures concerning Coordinated Sets are relevant. All
+CAP requires is that the correct Context Types are included in the metadata. Content Control
+IDs are only required if there is an accompanying ACL connection carrying content control
+from the Broadcast Source. This would be required in applications like a personal TV, which
+uses broadcast to allow multiple family members to listen, but where their individual earbuds
+could be used to pause or change channel.
+Everything else we need is defined in BAP, where the procedures instruct the Controller to
+set up Extended Advertising and provide the data to populate the parameters which are
+exposed in the Extended Advertisements and the Periodic Advertising train.
+
+203
+
+Section 8.2 - Starting a broadcast Audio Stream
+
+Figure 8.2 Data required to set up a Broadcast Source
+
+Figure 8.2 takes the Extended Advertising diagram from Chapter 4 and highlights the specific
+elements of data required to set up a Broadcast Source. The first step in setting up a broadcast
+Audio Stream is to assemble all of the data needed for the Controller to populate the Broadcast
+Audio Announcement Service, BIGInfo and BASE, which is described in the BAP Audio
+Stream Establishment procedure [BAP 6.3].
+
+8.2.1
+
+Configuring the BASE
+
+The Application will determine how many BISes are going to be transmitted and their
+associated codec and QoS configurations. BAP recommends that at least one of the broadcast
+Audio Streams should be encoded with either the 16_2 or 24_2 settings of Table 3.2, i.e.,
+16kHz, 10ms SDU or 24kHz, 10ms SDU, to ensure that every Acceptor can decode it. Any
+other additional codec configurations will be determined by the application or a higher level
+profile. The Host should also obtain any metadata content which it requires to populate the
+BASE. Current metadata requirements are shown in Table 8.1.
+Profile
+
+Required LTV metadata
+structures
+
+BAP
+CAP
+CAP
+
+None
+Streaming_Audio_Contexts
+CCID_List
+
+HAP
+TMAP
+PBP
+
+None
+None
+None
+
+Comments
+
+Only if content control exists for the Audio
+Stream
+Inherited from PBP
+ProgramInfo is recommended to describe
+the Audio Stream
+
+Table 8.1 Metadata LTV requirements for BASE from different profiles
+
+204
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+The final piece of information needed to configure the stream is the value of Presentation
+Delay, which is generally determined by the QoS settings.
+The Controller can now put together its BASE structure, which describes exactly what streams
+it will be transmitting and their configuration.
+We went through the overview of the BASE in Chapter 4, looking at what it contains, which
+is repeated in Figure 8.3.
+
+Figure 8.3 Simplified BASE structure
+
+The entire BASE is an LTV structure which is presented as an AD Type, with the parameters
+shown in Table 8.2.
+Parameter
+Length
+
+Size
+(Octets)
+1
+
+Type
+1
+Level 1 - BIG Parameters (common to all BISes)
+Basic Audio Announcement Service UUID
+2
+Presentation Delay
+
+3
+
+Num_Subgroups
+
+1
+
+Description
+Total length of the
+structure.
+Service-Data UUID
+
+LTV
+
+0x1852 (from Bluetooth
+Assigned Numbers).
+Range from 0 to 16.7 sec in µs
+(0x000000 to 0xFFFF)
+The number of subgroups used
+to group BISes with common
+features in this BIG.
+205
+
+Section 8.2 - Starting a broadcast Audio Stream
+Parameter
+
+Size
+Description
+(Octets)
+
+Level 2 – BIS Subgroup Parameters (common parameters for subgroups of BISes)
+Num_BIS[i]
+
+1
+
+Codec_ID[i]
+
+5
+
+Codec_Specific_Configuration_Length[i]
+
+1
+
+Codec_Specific_Configuration[i]
+
+varies
+
+Metadata_Length[i]
+
+1
+
+Metadata[i]
+
+varies
+
+The number of BISes in this
+subgroup.
+Codec information for this
+subgroup. Typically, this will be
+LC3 (0x06).
+The length of the codec
+configuration data for the
+Codec_ID in this subgroup.
+The codec configuration data
+for the Codec_ID in this
+subgroup.
+Length of the metadata for this
+subgroup
+The metadata for this subgroup.
+CAP requires the
+Streaming_Audio_Contexts
+LTV Metadata for every
+subgroup, and also the
+CCID_List LTV structure if
+content control is applied.
+
+Level 3 – Specific BIS Parameters (if required, for individual BISes)
+BIS_index[i[k]]44
+
+1
+
+Codec_Specific_Configuration_Length[i[k]] 1
+
+Codec_Specific_Configuration[i[k]]
+
+varies
+
+Unique index for each BIS in
+the BIG.
+The length of the codec
+configuration data for a specific
+BIS [i[k]] in this subgroup.
+The codec configuration data
+for the specific BIS [i[k]] in this
+subgroup.
+
+Table 8.2 The parameters of the BASE.
+
+One point of possible confusion to be aware of is that BIS_index starts from 1, but CIS_IDs run
+from 0.
+44
+
+206
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+The BASE allows a lot of flexibility, the majority of which will not be needed in most everyday
+applications. The Level 1 parameters must exist and apply to every subgroup. Most BISes
+will probably only have a single subgroup, which will typically carry two or three BISes – one
+for left and one for right, plus an optional mono or joint stereo stream. The reason for
+defining more than one subgroup is for occasions when some of the BISes have different
+metadata, such as different languages. Differences like this can only be expressed in Level 2
+parameters. At Level 3, you can specify different codec configurations for BISes within a
+subgroup, such as having different quality levels, for example one stream at 48 kHz sampling
+and one at 24 kHz. But that is the only thing which changes at Level 3. Different languages,
+different parental ratings, etc., all need to have their own subgroup defined at Level 2.
+Where there is more than a single BIS, the Level 2 and Level 3 parameter sections should be
+repeated for each BIS in order of subgroup and BIS_index, as shown in Table 8.3.
+Level
+
+Parameter
+
+Value
+
+2
+
+Num_BIS [0]
+Codec_ID [0]
+Codec_Specific_Configuration LV [0]
+Metadata LV [0]
+BIS_index [0[0]]
+Codec_Specific_Configuration LV [0[0]]
+Num_BIS [0]
+Codec_ID [0]
+Codec_Specific_Configuration LV [0]
+Metadata LV [0]
+BIS_index [0[1]]
+Codec_Specific_Configuration LV [0[1]]
+Num_BIS [1]
+Codec_ID [1]
+Codec_Specific_Configuration LV [1]
+Metadata LV [1]
+BIS_index [1[0]]
+Codec_Specific_Configuration LV [1[0]]
+Num_BIS [1]
+Codec_ID [1]
+Codec_Specific_Configuration LV [1]
+Metadata LV [1]
+BIS_index [1[1]]
+Codec_Specific_Configuration LV [1[1]]
+
+Num_BIS = 2
+
+3
+2
+
+3
+2
+
+3
+2
+
+3
+
+0x01
+Num_BIS = 2
+
+0x02
+Num_BIS = 2
+
+0x03
+Num_BIS = 2
+
+0x04
+
+Table 8.3 Order of entries for multiple subgroups and BISes
+
+207
+
+Section 8.2 - Starting a broadcast Audio Stream
+
+8.2.2
+
+Creating a BIG
+
+As we saw with unicast, the BIG is created using an HCI command – in this case the LE
+Create BIG Parameters, with the parameters shown in Table 8.4.
+Parameter
+
+Description
+
+BIG_Handle
+Advertising_Handle
+Num_BIS
+SDU_Interval
+
+The BIG identifier, set by the Host.
+The handle of the associated periodic advertising train.
+The number of BISes in the BIG.
+The interval between the periodic SDUs containing encoded
+audio data in microseconds.
+The maximum size of each SDU, in octets.
+The maximum transport latency for each BIS, including pretransmissions.
+The requested number of retransmissions (which is a
+recommendation)
+A bitfield of acceptable PHY values. The Controller will
+choose from one of the supported options. A high-level
+profile may mandate only one value – normally 2Mbps.
+0 = 1Mbps, 1 = 2 Mbps, 2 = LE coded.
+The preferred method of arranging BIS Subevents. The
+Controller only uses this as a recommendation.
+0 = Sequential, 1 = Interleaved.
+If set to 1, the BIS data PDUs will be framed. If set to 1, the
+Controller can decide whether to use framed or unframed.
+Set to 1 if the audio stream needs to be encrypted and 0 for
+unencrypted.
+The code used to generate the encryption key for all BISes in
+the BIG. This should be set to zero for an unencrypted
+stream.
+
+Max_SDU
+Max_Transport_Latency
+RTN
+PHY
+
+Packing
+
+Framing
+Encryption
+Broadcast_Code
+
+Table 8.4 Parameters for the HCI LE Create BIG command
+
+An important difference between the LE Create BIG Parameters command and the equivalent
+LE Set CIG Parameters command, is that the same parameters are used across every BIS in
+the BIG. That limitation arises from the fact that the Isochronous Channel information is
+contained in the BIGInfo packet, which is not large enough to accommodate details of
+multiple, different BISes. The consequence is that every BIS is the same size, which must fit
+the largest SDU. Different BISes within the BIG may use different codec configurations and
+even different codecs, but the BIS structure must be configured to fit the largest possible
+packet across those different configurations. If different QoS settings are used for different
+BISes in the BIG, it means that there will be redundant space.
+
+208
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+Having created the BASE structure, the Host can ask the Controller to schedule the BIG and
+BISes, using the LE_Create_BIG command (defined in the Core Vol 4, Part E, Section
+7.8.103), with the parameters shown in Table 8.4.
+Once the command has been issued and the Controller has worked out its scheduling, it will
+respond to the Host with an LE_Create_BIG_Complete event [Core Vol 4, Part E, Sect
+7.7.65.27] containing the actual parameters which it has chosen to use, as shown in Table 8.5.
+Some of these are used by the Host application, others will reflect what is in the BIGInfo –
+the data structure which informs scanning devices of the configuration of the broadcast
+Isochronous Streams.
+Parameter
+
+Description
+
+Used in
+
+Subevent Code
+
+0x1B – Subevent code for the
+LE_Create_BIG_Complete event
+Set to 0 if the BIG could be successfully
+scheduled; otherwise set to 1, when an Error Code
+is available.
+The same BIG handle the Host sent in its
+LE_Create_BIG command
+The maximum time for transmission, in
+microseconds, of all PDUs of all BISes in a single
+BIG event, using the actual parameters included
+below.
+The actual transport latency for the BIG in
+microseconds. (This covers all BIG events used
+for an SDU.)
+The PHY that will be used for transmission.
+The number of Subevents for each BIS
+The Burst Number. (The number of new payloads
+in each BIS event.)
+The offset used for pre-transmissions.
+The Immediate Repetition Count. (The number of
+times a payload is transmitted in each BIS event.)
+The maximum size of the payload in octets.
+The value of the Isochronous Interval. (The time
+between consecutive BIG Anchor Points.)
+Expressed as a multiple of 1.25 msecs.
+The total number of BISes in the BIG
+A list of Connection Handles for all of the BISes
+in the BIG
+
+Host
+
+Status
+
+BIG_Handle
+BIG_Sync_Delay
+
+Transport_Latency_BIG
+
+PHY
+NSE
+BN
+PTO
+IRC
+Max_PDU
+ISO_Interval
+
+Num_BIS
+Connection Handle[i]
+
+Host
+
+Host
+Host
+
+Host
+
+BIGInfo
+BIGInfo
+BIGInfo
+BIGInfo
+BIGInfo
+BIGInfo
+BIGInfo
+
+BIGInfo
+Host
+
+Table 8.5 Parameters returned by an LE_Create_BIG_Complete HCI event
+
+209
+
+Section 8.2 - Starting a broadcast Audio Stream
+At this point, the Host has everything it needs to start the process, the first stage of which is
+to set up the Extended Advertising and Periodic Advertising trains. It needs to include the
+Broadcast Audio Announcements [BAP 3.7.2.1] and Broadcast_ID [BAP 3.7.2.1.1] in the
+AUX_ADV_IND Extended Announcements, and the Basic Audio Announcement, including
+the BASE [BAP 3.7.2.2] and BIGInfo
+The Host uses the HCI_LE_Set_Extended_Advertising_Data command [Core Vol 4, Part E,
+Sect 7.8.54] and enters the Broadcast mode [Core Vol 4, Part C, Sect 9.1.1] to start the
+Extended Advertising, then uses the HCI_LE_Set_Periodic_Advertising_Data command
+[Core Vol 4, Part E, Sect 7.8.62] to populate the BASE, before entering the Periodic
+Advertising Mode [Core Vol 4, Part C, Sect 9.5.2].
+Once the Extended Advertisements and the Periodic Advertising train are established, the
+broadcast Audio Source enters the Configured State. At this stage it is not yet transmitting
+any audio data.
+
+8.2.3
+
+Updating a broadcast Audio Stream
+
+At any point, a Broadcast Source can update the LTV structures in the BASE containing the
+metadata. This is normally done to reflect changes in the Audio Channel content, such as new
+ProgramInfo or Context Types. This is defined in the BAP Modifying Broadcast Sources
+procedure [BAP 6.5.5]. The Broadcaster does not need to stop the Audio Stream, but can
+make the change dynamically when in the Configured or Streaming States.
+There is no guarantee that an Acceptor which is receiving the Audio Stream will see such a
+change. Once an Acceptor has synchronised to a stream using the information in the BIGInfo,
+it has no further need to track the Periodic Advertisements or read the BASE, unless required
+to do so by another profile.
+
+8.2.4
+
+Establishing a broadcast Audio Stream
+
+Once the Extended and Periodic Advertising is running, the Broadcast Source needs to
+establish its Audio Streams and start transmitting. It does this in three steps, defined by the
+BAP Broadcast Audio Stream Establishment45 procedure [BAP 6.3.2].
+First, it enters the Broadcast Isochronous Broadcasting Mode [Core Vol 3, Part C, Sect 9.6.2],
+which prepares it to send PDUs in the BISes of its BIG. It then starts the Broadcast
+Isochronous Synchronizability Mode [Core Vol 3, Part C, Sect 9.6.3], which starts sending the
+BIGInfo in the ACAD fields of the Periodic Advertisement, alerting any devices that are
+scanning for broadcast Audio Stream to its presence.
+
+45
+
+This should not be confused with the BASE. It’s unfortunate it has the same initials.
+
+210
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+Finally, the Broadcast Source needs to set up the audio data path in the same way as for unicast,
+using the LE Setup ISO Data Path HCI command [Core Vol 4, Part E, Sect 7.8.109].
+At this point, the Broadcast Source is fully operational, transmitting its BIG and constituent
+BISes. If it has no connections, it will never know whether any device detects or synchronises
+to those broadcasts.
+
+8.2.5
+
+Stopping a broadcast Audio Stream
+
+To stop broadcasting, and return to the Configured state, an Initiator needs to use the
+Broadcast Isochronous Terminate procedure [Core Vol 3, Part C, Sect 9.6.5]. This stops the
+transmission of BIS PDUs and the BIGInfo. Synchronised Acceptors will receive a
+BIG_TERMINATE_IND PDU to alert them to the end of transmission. If they fail to
+receive this, they get no further information other than the fact that both the BIS and BIGInfo
+have disappeared.
+At the end of the Broadcast Isochronous Terminate procedure, the Broadcast Source will
+return to the Configured State. It can then be released, or re-enabled.
+
+8.2.6
+
+Releasing a broadcast Audio Stream
+
+To return a Broadcast Source to its Idle state, it needs to exit the Periodic Advertising mode,
+which returns it to the Idle state.
+
+8.3
+
+Receiving broadcast Audio Streams
+
+If a Broadcast Sink wants to receive a broadcast Audio Stream, it needs to scan to find it. The
+same procedures for finding it are used, whether this is being done by a Broadcast Sink itself,
+or a Broadcast Assistant. We’ll just look at the Broadcast Sink doing it in this section, then
+see how the task can be delegated when a Broadcast Assistant is available.
+Because Acceptors act as separate entities, CAP doesn’t really come into play for the reception
+of broadcast Audio Streams. CAP defines a Broadcast Audio Reception Start procedure [CAP
+7.3.1.8], but unless the Acceptors have a way of communicating with each other, they don’t
+know that the other is present. They know they’re a member of a Coordinated Set, but not
+whether the other set members are around. A Commander can do that coordination task,
+which we’ll come to later, or there could be an out of band mechanism, such as a sub-GHz
+radio which allows the Acceptors to talk to each other. But an Acceptor scanning on its own
+is essentially operating at the BAP level.
+
+8.3.1
+
+Audio Announcement discovery
+
+The first step in receiving a broadcast stream is to find it, which is performed using the
+Observation Procedure from the Core [Core Vol3, Part C, Sect 9.1.2]. This is a standard scan
+to find advertisements. Here, the ones of interest are Extended Advertisements which contain
+the Broadcast Audio Announcement Service UUID with a Broadcast_ID. The Extended
+211
+
+Section 8.3 - Receiving broadcast Audio Streams
+Advertisements may also include other Service UUIDs, such as those defined in the Public
+Broadcast Profile, which the scanner can use to filter the streams which it wants to investigate.
+Having found an appropriate Extended Advertisement, the scanner will look for the SyncInfo
+data and use this to synchronise to the Periodic Advertisements associated with the
+Broadcast_ID. Once synchronised, the scanner can find and parse the data in the BASE,
+which provides information about the set of BISes.
+At this point, the Acceptor is acting quite differently from previous Bluetooth peripherals.
+Instead of being told what to do by a Central device, it is collecting information about the
+available Broadcast Sources within range which have Audio Streams of interest and decides
+for itself which one to receive. This is entirely driven by the implementation. It may look for
+a known Broadcast_ID, which it has connected to before, or information in the BASE
+metadata. It could connect to the first stream it finds, and then step through any other streams
+which are available, or it could collect data from every Broadcast Source within range and
+present this to the user in some form to allow them to make a manual choice. All of that is
+implementation specific.
+The Acceptor can use additional information to inform its choice. It would not normally
+bother with streams if their Context Type did not match any of the Acceptor’s Available
+Context Types. Nor would it want to receive a stream for an Audio Location that it does not
+support. But the decision is the Acceptor’s. Without a Commander, or a proprietary link
+between two Acceptors, a user might have to choose a stream manually on both their left and
+right earbuds. It is unlikely that this will ever be the case, as it’s such a bad user experience;
+manufacturers will either provide a separate Commander (or a Commander application), or
+implement a proprietary link between left and right earbuds. However, designers should be
+aware of the need to allow pairs of devices to work together when they do not have a Bluetooth
+link between them. But that’s where the remote control / Commander comes in.
+
+8.3.2
+
+Synchronising to a broadcast Audio Stream
+
+Once it has decided on a Broadcast Source to receive, an Acceptor starts the Broadcast Audio
+Stream synchronization procedure [BAP 6.6]. This involves it reading the BASE information
+from the AUX_SYNC_IND (and any supplementary AUX_CHAIN IND) PDUs to
+determine the codec configuration and the index number of the BIS which corresponds to the
+Acceptor’s Audio Location. The BIS_Index for each BIS defines the order of the BISes within
+the BIG. Knowing this, the Acceptor can determine which BISes in the BIG it wants to
+receive.
+The Acceptor then retrieves the BIGInfo, which contains all of the information of the BIG
+structure, the hopping sequence and where the BIS Anchor Points are located. Once it has
+this, it can invoke the Broadcast Isochronous Synchronization Establishment procedure from
+the Core [Vol 3, Part C, Sect 9.6.3] to acquire the appropriate BIS. If it has not already done
+so, it needs to set up its Audio data path. Having received the incoming audio packets, it then
+212
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+applies the Presentation Delay value to determine the rendering point.
+
+8.3.3
+
+Stopping synchronisation to a broadcast Audio Stream
+
+If an Acceptor decides it wants to stop receiving a broadcast Audio Stream it acts
+autonomously to disconnect the Audio data path and stops synchronisation with the PA and
+the BIS.
+
+8.4
+
+The broadcast reception user experience
+
+Although an Acceptor is allowed to acquire a Broadcast Stream by itself, it’s not a very good
+user experience. At its most basic, it’s an exact analogue of the telecoil experience, where
+someone wearing a hearing aid presses its telecoil button to acquire a telecoil stream. That
+user experience works because telecoils are inductive loops and you are normally only within
+range of one loop, which means there’s no issue of whether or not you’re connecting to the
+right one. If there is only one Bluetooth LE Audio transmitter within range, that experience
+will be the same, but as soon as broadcast applications become popular, that’s no longer going
+to be true; in many cases you’re likely to be within range of many broadcast transmitters. It’s
+exactly the same situation as we have with Wi-Fi. If you scan for Wi-Fi access points with
+your phone or laptop, you’ll frequently find a dozen or more. The user experience for
+Bluetooth LE Audio is potentially a lot more difficult, because in most cases you’ll be wearing
+a pair of earbuds or hearing aids, which have a minimal user interface. There’s also the added
+complication of needing to start and stop streaming together, and always connecting both
+earbuds to the same Broadcast Source.
+So, although the explanations above are valid and indicate how to receive a broadcast stream,
+they’re largely academic. To make this work in the real world we need to involve another
+device which does a better job of the hard work of finding the broadcast streams and telling
+one or more Acceptors what to do with them. We also need a way to distribute the keys
+needed to decode encrypted Audio Streams. Which brings us to Commanders and Broadcast
+Assistants.
+
+8.5
+
+BASS – the Broadcast Audio Scan Service
+
+First, a few words about BASS - the Broadcast Audio Scan Service. BASS is a service which
+is instantiated in an Acceptor to expose the details of which broadcast Audio Streams it knows
+about, which one(s) it’s connected to, and whether it wants a Broadcast Assistant to help it
+manage that information and help it to make connections. BASS works with Broadcast
+Assistants (which are defined in BAP, and which we’ll get to in a minute) to manage broadcast
+connections that they select and manage. It’s a key part of expanding the broadcast ecosystem,
+using ACL connections to assist the distribution and control of broadcast information.
+The key elements of BASS are two characteristics, both of which are mandatory whenever it
+is used:
+213
+
+Section 8.6 - Commanders
+Characteristic
+
+Properties
+
+Quantity
+
+Broadcast Audio Scan Control Point
+Broadcast Receive State
+
+Write, Write w/o response
+Read, Notify
+
+Only one
+One or more
+
+Table 8.6 BASS characteristics
+
+The Broadcast Audio Scan Control Point characteristic allows one or more Broadcast
+Assistants to inform an Acceptor of whether they’re actively working on its behalf to look for
+Broadcast Sources. They can tell the Acceptor how to find a BIG, connect to a BIG,
+disconnect from a BIG and provide the Broadcast_Code to decrypt an encrypted Audio
+Stream. As we’ll see, there is one really important parameter within the Broadcast Audio Scan
+Control Point characteristic, which is the BIS_Sync. When a Broadcast Assistant sets this to
+0b1, it tells the Acceptor to start reception of a stream. When it sets it to 0b0, the Acceptor
+should stop receiving it.
+The “one or more Broadcast Assistants” statement on the first line of the previous paragraph
+is an important one. An Acceptor can use as many Broadcast Assistants as it likes. All that is
+required is that each has an ACL link to that Acceptor. Once they’re connected and have
+registered the fact that they’re helping the Acceptor to find a broadcast stream, the Broadcast
+Assistants operate on a first-come, first served basis, limited only by any Locks which are
+invoked by CSIP if they’re operating on a Coordinated Set of Acceptors. Each Broadcast
+Assistant needs to register with the Broadcast Receive States for notifications, which means
+that all of them will be updated whenever a change is made to the Broadcast State, either by a
+Broadcast Assistant writing to it, or as a result of a local action on the Acceptor.
+The Broadcast Receive State characteristic exposes what the Acceptor is doing – which BIG
+it’s connected to, which BISes within that BIG it’s currently receiving, whether it can decrypt
+them and the current metadata it has for each of them. An Acceptor has to have at least one
+Broadcast Receive State characteristic for each BIG it can simultaneously synchronise to. In
+many cases, that will be just one. Now on to the Commander.
+
+8.6
+
+Commanders
+
+Although most people see Broadcast as the big new feature of Bluetooth LE Audio, the most
+important innovation is probably the concept of the Commander, which provides remote
+control and management of broadcast Audio Streams. In a new world of multiple broadcast
+streams, Commanders provide a simple way for users to select what they hear. Although the
+concept of broadcast was inspired by the telecoil experience of picking up inductive loops in
+hearing aids, Bluetooth LE Audio broadcast applications go a long, long way beyond what
+telecoil can do.
+Looking back at the inductive loop paradigm, this was straightforward – you would only
+receive an audio stream if you were standing within the boundary of the loop. (There might
+be a problem if you were in the room directly above it, but that was an edge case.) With
+214
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+Bluetooth broadcasts, they can and will overlap and go through walls, so users need a simple
+way to select the broadcast they want to listen to. We’ve already seen how the metadata
+information in BASE and service information from profiles like the Public Broadcast Profile
+(PBP) help inform that choice. However, that is just for public broadcasts, which are open
+for anyone to hear. For personal broadcasts, the broadcast Audio Streams are encrypted,
+requiring methods to acquire the encryption keys. Once again, that is enabled by
+Commanders.
+
+Figure 8.4 The major elements of a Commander and broadcast Acceptor
+
+Figure 8.4 illustrates the major components of a Commander and the main broadcast-related
+parts of an Acceptor. Both are roles defined in CAP, but I’m using them more generally in
+this book. It’s useful to visualise a Commander as a device in its own right, like a TV remote
+control, because for a user, what it does is very similar. Commanders usually contain four
+main elements:
+•
+
+•
+•
+•
+
+A CSIP Set Coordinator, which ensures that any commands are sent to all of the
+members of a Coordinated Set, whether that’s information about a Broadcast
+Source, or volume control.
+A Broadcast Assistant, which is used to scan for Broadcast Sources and lets the user
+select which one they want to listen to,
+A VCP volume controller (which we’ll cover in Chapter 10) to control the volume
+of each member of the Coordinated Set, and
+The ability to obtain and distribute a Broadcast_Code to decrypt broadcast Audio
+Streams
+
+Without the flexibility that comes with Commanders, and the capabilities they provide,
+broadcast encryption becomes cumbersome, making use cases around private broadcast
+streams very difficult to implement. Private broadcasts bring a whole new dimension to
+Bluetooth LE Audio. Whether that’s for personal TV and music, access to a TV in a hotel
+room, shared audio in a conference room, or the new Bluetooth Audio Sharing experience, it
+215
+
+Section 8.6 - Commanders
+relies on a simple way for users to connect to the correct Broadcast Source and get access to
+the Broadcast_Code to decrypt the audio. All of this is managed by two new roles, which are
+defined in BAP. They are the Scan Delegator and the Broadcast Assistant. They work with
+the Broadcast Audio Scanning Service (BASS) to distribute the information gathering and
+control that provide the richness in broadcast applications.
+
+8.6.1
+
+Broadcast Assistants
+
+Broadcast Assistants allow other devices to perform the scanning job that we described in
+Section 8.3.1. The Broadcast Assistant Role is normally the main Role of a stand-alone
+Commander, scanning on behalf of a Broadcast Sink. There are two important reasons to
+offload this task. The first is that scanning is a fairly power-hungry operation, which is
+something that you don’t want a hearing aid or earbud to do, or at least not very often,
+particularly without knowing exactly what it’s looking for. It’s a job which is far better
+performed on a device like a phone, or a dedicated remote control, neither of which are being
+powered by tiny zinc-air batteries. The second reason is the user experience. Broadcasts
+should always contain readable metadata in their BASE structures, allowing a device with a
+display to show what each is. Displaying that information is something which is easy to
+implement on a phone or remote control, but largely impossible on an earbud. So, moving
+the scanning and selection task somewhere else not only saves battery life, it vastly improves
+the user experience.
+Broadcast Assistants have an ACL link with an Acceptor. If the Acceptors are running a CAP
+based profile, they may be a member of Coordinated Set, in which case each Commander that
+includes the Broadcast Assistant has to run the CAP preamble procedure to ensure that it
+operates on all of the set members. Multiple Commanders can be involved, so a user could
+have a Commander application in a smartwatch as well as their smartphone, and also a
+dedicated remote control as an additional Commander.
+
+8.6.2
+
+Scan Delegators
+
+Commanders work on behalf of a device with a Scan Delegator role. The role is defined in
+BAP and is normally implemented in a Broadcast Sink. It’s job is to find other devices taking
+the Broadcast Assistant Role, which can take over the job of scanning for Broadcasters. A
+Scan Delegator can also be implemented in a Commander, as multiple Commanders can be
+chained together, effectively relaying information from one to another. We’ll see the benefit
+of that in a later chapter.
+The Scan Delegator Role within a Broadcast Sink always contains an instance of BASS. Scan
+Delegators solicit for Broadcast Assistants which can scan on their behalf by sending
+solicitation requests using Extended Advertising PDUs which include a Service Data AD Type
+containing the BASS UUID.
+
+216
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+Any Broadcast Assistant within range can connect and let the Scan Delegator know that it has
+started scanning on its behalf, interacting through the BASS instance on the Broadcast Sink,
+where it writes to the Broadcast Audio Scan Control Point characteristic [BASS 3.1], to
+confirm that it is actively scanning (opcode = 0x01) or has stopped scanning (opcode = 0x00).
+It is up to the BASS instance to record this status for multiple Clients to determine whether
+any are actively scanning. This process of scanning on behalf of a Scan Delegator is called
+Remote Broadcast Scanning. Once it knows that a Broadcast Assistant is scanning on its
+behalf, a Broadcast Sink may decide to terminate its own scanning process to conserve power.
+The Broadcast Assistant can read the PAC records of the Broadcast Sink to discover its
+capabilities, and use that information to filter the Broadcast Sources it finds, so that it only
+presents the Scan Delegator with options which are valid for its Broadcast Sink. Most
+commonly, that would take into account the Broadcast Sink’s Audio Location, Supported and
+Available Context Types and the Codec Configurations it supports.
+Once the Broadcast Assistant has detected a suitable Broadcast Source, it can inform the Scan
+Delegator by writing to one of its Broadcast Receive State characteristics to start the process
+of the Broadcast Sink acquiring the broadcast Audio Stream. This could be an automatic
+action, or it could be user initiated. A user might want to automatically connect to a known
+Broadcast Source, such as when they walk into a place of worship or their office. Alternatively,
+they may ask their phone to scan and let them know what’s available using a scanning
+application, before selecting their preferred Broadcast Source. That choice is implementation
+specific.
+
+8.6.3
+
+Adding a Broadcast Source to BASS
+
+Once that choice has been made on a Commander, the Broadcast Assistant writes it to the
+first empty Broadcast Receive State characteristic on the Broadcast Sink, adding the selected
+Broadcast Source information [BAP 6.5.4]. Before it does this, it should have read the current
+status of the Broadcast Receive State characteristics. If the user has chosen a Broadcast Source
+which already exists in one of them, then it should modify the appropriate Broadcast Source
+value using the Modify Broadcast Source procedure [BAP 6.5.5]. It should also check that the
+Broadcast Sink is capable of decoding any proposed Audio Source.
+The parameters that it writes into the characteristic using the Add Source Operation opcode
+of 0x02 [BASS 3.1.1.4], fall into three broad categories, as shown in Table 8.7.
+
+217
+
+Section 8.6 - Commanders
+Parameter
+
+Size
+(octets)
+
+Description
+
+Source Information
+Advertiser_Address_Type
+Advertiser_Address
+
+1
+6
+
+Advertising_SID
+Broadcast_ID
+
+1
+3
+
+Public (0x00) / Random (0x01)
+Advertising address of the Broadcast
+Source
+ID of the advertising set
+Broadcast ID of the Broadcast Source
+
+Action
+PA_Sync
+
+1
+
+PA_Interval
+Num_Subgroups
+BIS_Sync[i]
+
+2
+1
+4
+
+0x00: Don’t synchronise
+0x01: Synchronise using PAST
+0x02: Synchronise without using PAST
+The SyncInfo field Interval
+Number of subgroups in the BIG
+BIS_Index values to connect to:
+0x00: Don’t synchronise to
+BIS_Index[i[k]]
+0x01: Synchronise to BIS_Index[i[k]]
+
+Configuration
+Metadata Length[i]
+Metadata [i]
+
+1
+Varies
+
+Metadata length for the [ith] subgroup in
+the BIG
+Metadata for the [ith] subgroup in the
+BIG
+
+Table 8.7 Format of the Add Source Operation
+
+The first set of parameters – the Source Information, informs the Broadcast Sink of the
+identity of the Broadcast Source and the BIG, so that it knows which device (or more correctly,
+device identity) this information refers to. The Advertising Set ID – the SID, which is
+contained in the primary advertisements and is defined when the Extended Advertising is set
+up [Vol 4, Part E, Section 7.8.53], generally stays static for the life of a device, and isn’t allowed
+to change between power cycles. Although it is only a single octet, it can help to identify the
+device. The Broadcast_ID is in the AUX_ADV_IND of the Extended Advertising and is
+static for the lifetime of a BIG (which may be shorter than the SID). Along with the
+advertising addresses, this is enough information for a Broadcast Sink to scan for the
+Broadcast Source.
+The next group of parameters tells the Broadcast Sink what it should do. This would normally
+be triggered by the user selecting a Broadcast Source on the user interface of their Commander.
+PA_Sync tells it to synchronise to the periodic advertising train, which lets it obtain the BASE
+218
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+and BIGInfo. Setting the PA_Sync to either 0x01 or 0x02 tells the Broadcast Sink to go and
+do that. The PA_Interval is the interval between successive AUX_ADV_IND packets and if
+known, can help the scanner. Num_Subgroups provides the number of subgroups in the
+BASE, followed by the most important parameter, which is the BIS_Sync.
+BIS_Sync is the instruction from the Commander to the Acceptor to tell it which specific
+Broadcast Stream or Streams to connect to. It’s a four octet wide bitmap of all of the BISes,
+with values set to 0x01 for any stream which the Acceptor should connect to. There’s a slight
+subtlety here, which is why it’s arrayed, rather than a single octet, which is that it is effectively
+masked for each subgroup. So, each BIS_Sync[i] is only valid for that subgroup – all other
+values are set to 0b0. The reason for that is that an Acceptor is never expected to receive
+BISes from different subgroups at the same time, as the whole point of a subgroup is to
+arrange associated streams together. If two subgroups carried different languages, say
+Japanese and German, the expectation is that you would only choose one.
+
+Figure 8.5 Illustration of mapping from BIS number to BIS_Index
+
+Figure 8.5 illustrates how BIS_Index values in subgroups relate to the ordering in a BIG, where
+BISes are arranged in numeric order. In this case, those don’t align with the BIS_Index values,
+hence the mapping of BIS_Index. In writing a BIS_Sync value, only one subgroup can be
+selected. So, if subgroup 0 were selected from Figure 8.5, the only allowable values for
+BIS_Sync[0] would be that one or both of bits 0 and 1 are set to 0b1.
+A Broadcast Assistant can set the value of BIS_Sync to 0xFFFFFFFF, meaning “no
+preference”, in which case the Broadcast Sink is free to make its own decision regarding which
+BISes to use.
+Finally, the metadata fields provide the Level 2 metadata for each subgroup – typically the
+Audio Contexts and the Audio Locations, allowing the Acceptor to decide whether they are
+appropriate for what it wants to do, i.e., do they match its current Available Audio Context
+Type settings. The data sent in the Add Source operation, which is written into each Broadcast
+219
+
+Section 8.6 - Commanders
+receive state can be modified at any time by the Modify Source operation, which will update
+the current information. Despite the specific connection information which is conveyed in
+these two operations, it is entirely up to the Acceptor whether it follows the request to
+synchronise with the PA and the BISes which are written to it. That’s up to the
+implementation.
+If the Acceptor decides to act on the instruction from the Commander, it will use the
+instructions to synchronise to the Periodic Advertising train, either by using PAST, or scanning
+with the Broadcast Source information it’s been given, retrieving the BASE and BIGInfo from
+the ACAD and Basic Audio Announcement in the PA. These give it all of the information it
+needs to synchronise to the BIG, after which it will use the BIS_Sync value to select which
+specific BIS or BISes it receives.
+PAST – the Periodic Advertising Sync Transfer procedure [Core, Vol 6, Part B, Section 9.5.4]
+is the most efficient method, as the Broadcast Assistant provides the Scan Delegator with the
+SyncInfo data allowing it to synchronise directly to the PA. This process is known as Scan
+Offloading.
+Disconnection can be performed autonomously by the Acceptor, or a Commander can use
+the Modify Source operation to set the appropriate BIS_Sync bit to 0b0. It may also tell the
+Acceptor to stop synchronising to the PA, but retain the rest of the Source information.
+Alternatively, it can delete the Source record, by performing the Remove Source operation on
+that Source_ID.
+8.6.3.1
+
+Set, Source and Broadcast IDs
+
+It’s worth a quick precis of the different IDs involved in these operations, as they can be
+confusing.
+
+Figure 8.6 Set and Broadcast IDs
+
+220
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+Referring to Figure 8.6, the Set and Broadcast IDs are properties of the Broadcast Source.
+Both are randomly generated numbers. The Advertising Set_ID is included in the primary
+advertisements and identifies a specific set of Extended and Periodic Advertising trains. The
+SID is normally static for the lifetime of the device, but must remain unchanged within every
+power cycle. SID can be useful when an Acceptor or Commander wants to reconnect regularly
+to a known source, whether that’s a personal device, or a fixed, infrastructure broadcast
+transmitter.
+The Broadcast_ID is specific to a BIG and is carried in the Broadcast Audio Announcement
+Service UUID. It remains static for the lifetime of the BIG.
+The Source_ID is an Acceptor generated number which is used to identify a specific set of
+broadcast device and BIG information. It is local to an Acceptor and used as a reference for
+a Broadcast Assistant. In the case of a Coordinated Set of Acceptors, such as a left and right
+earbud, the Source_IDs are not related and may be different, even if both are receiving the
+same BIS, as each Acceptor independently creates their own Source ID values.
+
+8.6.4
+
+The Broadcast Receive State characteristic
+
+On the Acceptor, the Broadcast Receive State characteristic is used to inform any Broadcast
+Assistant of its current status regarding broadcast Audio Streams. Its structure is shown in
+Table 8.8.
+Field
+
+Size
+Description
+(Octets)
+
+Source_ID
+Source Address Type
+
+1
+1
+
+Source_Address
+
+6
+
+Source_Adv_SID
+
+1
+
+Broadcast_ID
+
+3
+
+PA_Sync_State
+
+1
+
+Assigned by the Acceptor
+From the Add / Modify Source operation or an
+autonomous action
+From the Add / Modify Source operation or an
+autonomous action
+From the Add / Modify Source operation or an
+autonomous action
+From the Broadcast Audio Announcement Service
+UUID
+Current synchronisation status to the PA:
+0x00 – Not synchronised to the PA
+0x01 – SyncInfo request
+0x02 – Synchronised to the PA
+0x03 – Synchronisation failed
+0x04 – No PAST
+Other values – RFU
+
+221
+
+Section 8.7 - Broadcast_Codes
+Field
+
+Size
+Description
+(Octets)
+
+BIG_Encryption
+
+1
+
+Bad_Code
+
+Varies
+
+Num_Subgroups
+BIS_Sync_State[i]
+Metadata_Length[i]
+Metadata[i]
+
+1
+4
+1
+varies
+
+Encryption status:
+0x00 – Not encrypted
+0x01 – Broadcast_Code required
+0x02 – Decrypting
+0x03 – Bad_Code (wrong encryption key)
+If BIG_Encryption = 0x03 (wrong key), this contains
+the value of the current, incorrect key. Otherwise not
+present.
+Number of subgroups
+BIS synchronisation state for the ith subgroup
+Length of the metadata for the ith subgroup
+Metadata for the ith subgroup
+
+Table 8.8 Format of the Broadcast Receive State characteristic
+
+As with other characteristics which can be set by Client operations, as well as being changed
+by an autonomous Server action, the Broadcast Receive State characteristic can be used by a
+Client to check an operation and status, and also be notified by the Server to denote than an
+operation is complete, such as synchronising to a PA, or to request an action from a Client.
+When they’re being notified, some of these act as error reports, while others signal a request
+to a Broadcast Assistant. Amongst these, the most important are notifying a PA_Sync State
+of 0x01 to request SyncInfo and notifying a BIG_Encryption state of 0x02 to request a
+Broadcast_Code. Any autonomous change of BIS synchronisation or loss of BIS can be
+notified through the updated BIS_Sync_State.
+
+8.7
+
+Broadcast_Codes
+
+The ability to encrypt broadcast Audio Streams takes Bluetooth LE Audio into to a whole new
+area of audio applications. Encryption effectively provides a pseudo-geofencing capability,
+limiting broadcast coverage to those who have a decryption key. That can be used to limit
+access to overlapping broadcasts in a specific area, in the same way that Wi-Fi codes are used.
+For example, a hotel could allocate codes to prevent people listening to audio from an
+adjoining meeting room, or a TV on the floor above. By providing the means to send the
+encryption key over a Bluetooth link, or to allow a Broadcast Assistant to obtain it by using
+an out of band method, it becomes even more powerful. In Chapter 12, we’ll investigate the
+different ways that the Broadcast_Code can be distributed for different applications.
+
+222
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+Personal devices which implement audio sharing, such as TVs, tablets and laptops, can
+collocate a Broadcast Assistant with the Broadcast Source. The Assistant then has direct
+access to the Broadcast_Codes and can write them directly to the Broadcast Sink. It requires
+the Acceptors to be paired with the Broadcast Source, but this would be expected for personal
+devices.
+The life of a Broadcast_Code is implementation dependent. In some case it could be
+permanent – a Broadcast Source playing music in a coffee-shop would probably never change
+its code. A TV in a hotel room would maintain its Broadcast_Code for as long as the guest
+was staying in that room; a personal TV might renew it on each power cycle, so that friends
+who visited didn’t keep it indefinitely, whilst a mobile phone app which shared audio with
+your friends would probably renew it every session. These different lifetimes means that
+Broadcast Sources and Commanders need to support a variety of different methods to acquire
+Broadcast_Codes.
+
+8.8
+
+Receiving Broadcast Audio Streams (with a Commander)
+
+Having learnt about Commanders, we can revisit how devices are likely to receive Broadcast
+Streams in the real world, which will almost always involve a Commander, whether that’s a
+stand-alone device, a wearable, an app on a phone, or a Broadcast Assistant built into the
+Broadcast TV.
+
+8.8.1
+
+Solicitation Requests
+
+The first job is for the Scan Delegator to find itself some Broadcast Assistants, which it does
+by sending out Extended Advertisements which contain a Service AD Data Type containing
+a Broadcast Audio Scan Service UUID. These are called Solicitation Requests [BAP 6.5.2].
+
+Figure 8.7 The Scan Delegator solicitation process
+
+Figure 8.7 shows a Scan Delegator sending out Extended Advertisements containing the BASS
+Service UUID. To the left, three Broadcast Assistants are within range. Broadcast Assistant
+#1 is collocated with a Broadcast Sink and is actively scanning. Broadcast Assistants #2 and
+#3 are devices with just the Commander role, but only Broadcast Assistant #3 is actively
+223
+
+Section 8.8 - Receiving Broadcast Audio Streams (with a Commander)
+scanning. In this case, both Broadcast Assistant #1 and #3 should respond to the Solicitation
+requests.
+CAP, as always, tells each Broadcast Assistant to start off by connecting, establishing the
+members of a Coordinated Set, then lays out the BAP procedures to follow. At this point,
+each Broadcast Assistant should read each Broadcast Sink’s PAC records to determine their
+capabilities, mainly because there’s no point in them telling a Broadcast Sink about broadcast
+Audio Streams that the Broadcast Sink can’t accept. There may be many reasons for making
+that decision. It could be because the use case of the broadcast Audio Stream has a Context
+Type the Broadcast Sink can’t support; it might have a codec configuration it can’t decode,
+have an incompatible Channel Allocation or Audio Location, requires encryption, etc., etc.
+Only one Acceptor of a Coordinated Set needs to perform the Solicitation process. Once the
+Broadcast Assistant responds and determines that it is a member of a Coordinated Set, CAP
+requires it to find the other members, and from that point, read all of their PAC records, then
+interact with the instances of BASS on each of them.
+Figure 8.7 shows a simplified example of that process for Broadcast Assistant #3, which is
+responding to a Solicitation request by one member of a Coordinated Set. It connects to the
+Scan Delegator in Broadcast Sink #1, discovers that it is a member of a Coordinated Set of
+two Acceptors, then finds and connects to the second member.
+It discovers and reads the PAC records of each Acceptor and will set its broadcast filter policy
+so that it only reports broadcast Audio Streams which can be received by both Acceptors.
+Next, it will discover and read the BASS Broadcast Receive States and set up notifications for
+these characteristics, so that it will be aware of any changes. Reading the states tells it whether
+the Acceptors are currently synchronised to any Periodic Advertising train or are receiving any
+BISes. Once it has completed these processes, it is ready to start scanning on behalf of the
+Acceptors, and writes to their Broadcast Audio Scan Control Point characteristic to inform
+them of this.
+
+224
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+
+Figure 8.8 Discovery of Broadcast Sink Capabilities and State by a Broadcast Assistant
+
+Once that’s done and there is at least one active Broadcast Assistant scanning on behalf of the
+Scan Delegator, we move into BAP’s Broadcast Assistant procedures, starting with Remote
+Broadcast Scanning [BAP 6.5.3].
+
+8.8.2
+
+Remote Broadcast Scanning
+
+A Broadcast Assistant that has informed the Scan Delegator that it is scanning will start
+searching for broadcast transmitters. As it finds each Advertising Set, it will work its way
+through the Extended and Periodic Advertisements (AUX_ADV_IND and
+AUX_SYNC_IND), examining the contents and parsing the structures to determine whether
+the broadcast Audio Streams match the capabilities of the Acceptor. In most cases, it will
+build up a list of relevant, available broadcast Audio Streams and present them to the user. If
+the Commander has a suitable user interface, such as an app on a smartphone, this will
+probably be presented in the form of a sorted list, from which the user can select a broadcast
+Audio Stream to connect to.
+225
+
+Section 8.8 - Receiving Broadcast Audio Streams (with a Commander)
+
+Figure 8.9 Remote broadcast scanning
+
+Figure 8.9 follows on from Figure 8.8, where Broadcast Assistant #3 is scanning on behalf of
+the two Broadcast Sinks (shown here as a single entity). It has discovered Broadcast Source
+B and Broadcast Source C. The user makes a decision on the Commander to listen to
+Broadcast Source C and selects it, at which point Broadcast Assistant #3 writes the
+information about Broadcast Source C to the BASS instances in the Scan Delegators of the
+two Broadcast Sinks.
+Figure 8.9 also shows how a collocated Broadcast Assistant can behave. In this case, Broadcast
+Assistant #1 is collocated with Broadcast Source A. Broadcast Assistant #1 has never written
+to the Broadcast Scan Control Point to say it is scanning on behalf of the Scan Delegator, as
+its only purpose is to provide the Scan Delegator with information about its collocated
+Broadcast Source. As soon as it connects to the Scan Delegator it can write that information
+to a Broadcast Receive State. (A collocated Broadcast Assistant can also perform scanning, in
+which case it would have told the Scan Delegator that it is scanning on its behalf, but that is
+an implementation choice.)
+A practical point to bear in mind is that the receivers in Broadcast Assistants may have better
+sensitivity than the receivers in earbuds, as they are likely to have larger antennas. This means
+that they will probably detect Broadcast Sources which would be out of range of their
+Broadcast Sinks. Implementations may want to determine the RSSI of signals from Broadcast
+Sources and use that information to arrange the order of presentation of Broadcast Sources
+to the user.
+
+226
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+
+8.8.3
+
+Receiving a Broadcast Stream
+
+We are now at the point where the user has decided what they want to listen to, which takes
+us back to the operation of Adding a Broadcast Source, which we described in Section 8.6.3.
+In most cases, when the user selects a Broadcast Source on their Commander, the Broadcast
+Assistant will not just write the details of that source to the Broadcast Audio Scan Control
+Point characteristic, but will also set the PA_Sync and BIS_Sync parameters to instruct each
+Acceptor to start receiving the relevant streams.
+As the two Acceptors are acting independently, they may take different amounts of time to
+obtain the PA train before they can synchronise with the appropriate BISes, particularly if they
+have to scan rather than using PAST. That could result in a noticeable delay between the time
+that audio starts being rendered in the two earbuds. The Bluetooth specifications don’t
+contain any details on how to address this, but one method is for the Commander to do this
+in two stages, starting by setting the PA_Sync bits, waiting for notification that both Acceptors
+are synchronised and only at that point sending a Modify Source operation, instructing them
+to synchronise their BISes and receive and render audio.
+
+8.8.4
+
+Broadcast_Codes (revisited)
+
+In Chapter 12 we will see just how important Broadcast_Codes are for new audio applications.
+If we look back at Figure 8.9, and assume that the broadcast Audio Streams from both
+Broadcast Source A and Broadcast Source C are encrypted, there’s likely to be a slightly
+different way in which the two Broadcast_Codes would be obtained.
+Broadcast Source A’s Broadcast Assistant knows the Broadcast_Code for the streams – that’s
+why it’s collocated. So as soon as the user selects Broadcast Source A, the Broadcast_Code
+will be sent over in the Add Source Command.
+In the case of Broadcast Source C, the Broadcast Assistant may not know the code, particularly
+if it has not been paired with Broadcast Source C. In this case, one of the Scan Delegators
+can ask all of its Broadcast Assistants if they know the Broadcast_Code, by notifying the
+appropriate Receive State characteristic, with the BIG_Encryption field set to 0x01, indicating
+that it requires the Broadcast_Code. If the Broadcast Sink already has a Broadcast_Code for
+this stream, which no longer works (normally because it was obtained in an earlier session), it
+should set the BIG_Encryption field to 0x03 to indicate an incorrect Broadcast_Code, and
+include that value in the Bad_Code field, so that a Broadcast Assistant with the same bad code
+doesn’t waste time resending it. Any Broadcast Assistant can respond to this notification if
+it knows the code, by writing the Broadcast_Code value to the Scan Delegator’s Broadcast
+Audio Scan Control Point using the Set Broadcast Code operation (0x04) [BASS 3.1.1.6], along
+with the Source_ID.
+
+227
+
+Section 8.9 - Handovers between Broadcast and Unicast
+As we’ll see in Chapter 12, there is a range of out-of-band methods by which a Broadcast
+Assistant may obtain a Broadcast_Code, opening up some interesting new use cases. Some of
+these may directly trigger the broadcast Audio Stream acquisition process.
+
+8.8.5
+
+Ending reception of a broadcast Audio Stream
+
+Reception of a broadcast Audio Stream may be performed autonomously by an Acceptor, in
+which case it will notify this change in its Broadcast Receive State characteristic. If a
+Commander receives this notification and is aware that the Acceptor is a member of a
+Coordinated Set, it may decide to terminate the reception on the other Acceptor(s) by writing
+the appropriate value in their Broadcast Audio Scan Control Points. However, this is
+implementation specific.
+Alternatively, a user can use a Commander to stop reception by writing to the Broadcast Audio
+Scan Control Points of each Acceptor, with the BIS_Sync value set to zero for all BISes in all
+subgroups and the PA_Sync value set to 0x00. Once the Acceptor indicates that it is no longer
+synchronised to either a BIS or a PA by notifying its Broadcast Receive State characteristic
+with both PA_Sync_State and BIS_Sync_State[i] set to zero, the Broadcast Assistant should
+perform the Remove Source Operation (0x05) [BASS 3.1.1.7] to remove the associated
+Broadcast Receive State. Note that there is no state machine for the Broadcast Sink. It uses
+its Broadcast Receive State characteristic to synchronise or release broadcast Audio Streams.
+
+8.9
+
+Handovers between Broadcast and Unicast
+
+Two handover procedures are defined in CAP to cover the use cases where an Initiator wants
+to move between a unicast and a broadcast stream (or vice versa) to transport the same audio
+content. This not the same as the general case of changing from a unicast use case to broadcast
+use case, but is specific to an application like sharing personal audio with friends, where a user
+wants to convert from a single, personal stream to a broadcast one, or vice versa. The
+procedures in CAP allow an Acceptor to receive concurrent unicast and broadcast Audio
+Streams from the Initiator, with the intention that the handover could be seamless with no
+noticeable break in the stream. (In most cases that will not be the case, as an Acceptor will
+not have the resources to receive both types of streams at the same time, especially if they are
+using one of the higher sampling rate QoS configurations.) To provide a smooth listening
+experience, implementations should attempt to conceal any breaks in transmission for the
+original Acceptor, although that may be accomplished by an audible tone or message. (Friends
+joining the broadcast Audio Stream will not experience any break, as they would not have been
+receiving the original audio stream.)
+The procedures [CAP 7.3.1.10 and 7.3.1.11] also mandate that any Streaming Context Types
+and CCID_List used for the original stream are carried over to the reconfigured stream.
+Although the user of the original stream will have transitioned it from unicast to broadcast,
+they would still maintain the ACL connection with their Audio Source and expect any control
+functionality to continue, such as fast forward or volume control, hence the need to maintain
+228
+
+Chapter 8 - Setting up and using Broadcast Audio Streams
+the CCID association. Others who receive the broadcast will only have access to the audio.
+
+8.10
+
+Presentation Delay – setting values for broadcast
+
+With unicast, Acceptors let the Initiator know about their ability to support Presentation Delay
+by exposing their Maximum and Minimum Presentation Delay capabilities, as well as their
+preferred value range. In broadcast, there is no information transfer between Initiator and
+Acceptor. This means that the value which an Initiator sets may not be supported by all of
+the Acceptors which decide to receive it.
+BAP requires that all Broadcast Receivers must support a value of 40ms in their range, so this
+is probably a default value which many Initiators will use. However, it starts to introduce
+latency, which may be excessive for live audio.
+TMAP and HAP place tighter constraints on the value of Presentation Delay, mandating that
+an Acceptor must support 20ms. As most Acceptors are expected to be qualified to TMAP
+or HAP (most will probably support both), setting 20ms for a Broadcast Source seems a
+reasonable compromise. However, a BAP only compliant device may not support this.
+This raises the question of what an Acceptor should do if it does not support the value of
+Presentation Delay exposed by a Broadcast Source. The specification is silent on this, but the
+following guideline offer a pragmatic approach.
+•
+
+If the Presentation Delay is lower than an Acceptor can accommodate, it should use
+a value of 40ms (which all devices must support).
+
+The specification gives a Broadcast Source three octets to carry the Presentation Delay value.
+That gets over the limitation of 65ms, which would be the maximum if only two octets were
+used (the units are 1µsec), but it means that the value could be as high as 16.7 seconds. Most
+Acceptors will have limited memory for buffering the audio stream, so there may be occasions
+where an inappropriately high value of Presentation Delay is outside their capability. Whilst
+they could decide not to render the Audio Stream, the pragmatic approach is to revert to the
+BAP value of 40ms.
+In both of these cases, this will ensure that both left and right devices render at the same time.
+If they are able to communicate with each other, they may use a proprietary method to
+determine the most appropriate value, which will generally select the lowest common value
+they support.
+
+229
+
+Section 8.10 - Presentation Delay – setting values for broadcast
+
+230
+
+Chapter 9 - Telephony and Media Control
+
+Chapter 9. Telephony and Media Control
+After the complexity of setting up unicast and broadcast streams, the four specifications
+covering telephony and media control are refreshingly simple, albeit comprehensive. When
+Bluetooth technology was first being developed, most of our mobile products were fairly
+straightforward and just did one thing. Mobile phones made phone calls using the cellular
+networks and music players played music based on whatever physical media they supported –
+typically pre-recorded cassettes or CDs. Since then, both telephony and media (by which we
+mean music and any other audio which we can stop and start) have become a lot more
+complex.
+For telephony, we no longer constrain our phones to a cellular network. The move to Voice
+over IP (VoIP) has seen an explosion in internet-based telephony services, either as “Over the
+Top” (OTT) applications on our phones, or as programs on our laptops and PCs. The
+pandemic and the associated growth in home working have accelerated their use. Whilst we
+still make cellular calls, we also use Skype, Zoom, Teams and a growing host of other telephony
+services, sometimes using more than one at the same time.
+Media has also changed. As we saw in Chapter 1, the growth of Bluetooth Audio has paralleled
+the growth of music streaming services. We no longer own most of what we listen to, but
+borrow it on demand. That means that the traditional controls of Stop, Play, Fast Forward
+and Fast Reverse need to be supplemented with new controls which move beyond the physical
+device to the source of the audio.
+The control mechanisms of Bluetooth Classic Audio profiles have struggled to keep up with
+this evolution, particularly in telephony, where they are still based on the old “AT” command
+set, which was first designed for landline modems back in 1981, before being integrated into
+the early GSM standards in the 1990s and Bluetooth technology’s Headset and Hands-Free
+Profiles in the early 2000s. The development of Bluetooth LE Audio has given us the
+opportunity to go back to first principles for both telephony and media control, with universal
+state machines which are equally applicable to cellular and VoIP telephony, and all kinds of
+local and remote media sources.
+Content control is currently defined using two sets of profiles and services. In the case of
+telephony, they are:
+•
+•
+
+TBS – The Telephone Bearer Service, which defines the state of the device handling
+the call, and
+CCP – The Call Control Profile, which is the Client acting on TBS.
+
+231
+
+Section 9.1 - Terminology and Generic TBS and MCS features
+For Media, the respective pair of specifications are:
+•
+•
+
+9.1
+
+MCS – The Media Control Service, defining the state of the source of the media,
+and
+MCP – The Media Control Profile, which is the Client acting on MCP.
+
+Terminology and Generic TBS and MCS features
+
+Up until this point, I’ve mostly been using the Initiator and Acceptor terminology from CAP
+as the easiest way of describing what happens to an Audio Stream. Now that we’re looking at
+control, which is orthogonal to the audio streams, that’s no longer appropriate. As the basic
+architecture of Bluetooth LE Audio separates the audio data plane and the control plane it
+means that control features can be implemented on devices which don’t take part in Audio
+Streams. Because of that, we need to revert to Client-Server terminology in this chapter, where
+the Server is the instance of the Service – defining the state of the media player or the phone.
+For the Telephone Bearer and Media Control Services, the Server is located on the Initiator.
+The Client can be on the Acceptor, but is equally likely to be in a remote control, simply
+because that’s easier to use, particularly for small devices like earbuds and hearing aids. We’ve
+got used to the ease of use of remote controls – it’s why they were invented. It doesn’t need
+to look like a conventional remote control - it could be in the form of a smartwatch, another
+wearable device, or even buttons on a battery case. Moreover, there can be multiple Clients
+operating on the Server at the same time. You could accept a phone call on your earbud, but
+terminate it with your watch.
+As we’ll see, the control profiles can move us around the media and call states on an Initiator,
+reflecting the user’s desire to start or accept a phone call, or play or pause a piece of music.
+What they don’t do is start or stop any Audio Streams associated with those decisions. The
+link between control commands and the configuration and establishment of the Audio
+Streams which transport the audio data is entirely down to the implementation. It is up to
+applications on the Initiator to tie them together as it wishes.
+In both TBS and MCS, the service specification includes individual TBS and MCS instances,
+which can be instantiated for each application on the device. For example, if your phone
+supports Skype, WhatsApp and Zoom, as well as cellular calls, it can include a separate instance
+of TBS for each one of those applications. If it’s a media playback device, it can include an
+instance of MCS for Spotify, BBC Sounds and Netflix.
+However, if you’re controlling either telephony or music from a pair of earbuds, with a very
+limited user interface, you’re unlikely to want to, or even be able to discriminate between the
+different applications. To cope with this, the TBS and MCS specifications each contain a
+generic version of the service, called the Generic Telephone Bearer Service (GTBS) and
+Generic Media Control Service (GMCS) respectively. These behave in exactly the same way
+as TBS and MCS, but provide a single interface to the Server device. It is up to the
+implementation to map incoming commands from a Client to the appropriate application.
+232
+
+Chapter 9 - Telephony and Media Control
+
+Figure 9.1 Representation of Control Services and their generic counterparts
+
+Figure 9.1 is a simplistic representation of how this works. Every content control Server must
+include a single instance of the generic service – either GTBS or GMCS, and may have
+individual instances of TBS and MCS if it wants to expose control directly to an application.
+You can have as many instances of each service as you have telephony and media applications,
+with that mapping is down to the implementation.
+The respective profiles – CCP and MCP, define Client and Server roles, which are illustrated
+in Figure 9.2.
+
+Figure 9.2 Relationship of generic and specific control services
+
+Implementations can use the combination of these services to:
+•
+•
+•
+
+Treat each application as a unique media player or telephony service,
+Treat the device as a single telephony or media player where commands act on the
+whole device, or
+A combination, where a subset of commands may be directed to specific
+applications, being mapped according to their specific Content Control IDs
+(CCIDs).
+
+In all cases, the mapping is down to the product implementation.
+
+233
+
+Section 9.2 - Control topologies
+Both TBS and MCS include a wide range of features, which we’ll cover below. A lot of the
+time, Bluetooth LE Audio is concerned with earbuds, as that is by far the largest market by
+volume. The limited user interface (UI) of an earbud may make readers wonder why so many
+of these features are included and also why so few of them are mandated. That ignores the
+history of Bluetooth Audio, where, until recently, the major audio application was in carkits,
+where a very rich control and information interface is provided. Both TBS and MCS were
+designed to replicate many of the features currently available in carkits, so that Bluetooth LE
+Audio can be used to build the next generation of these products, as well as extending them
+to add new features, particularly for the control of media players.
+
+9.2
+
+Control topologies
+
+At the heart of both of the pairs of control specifications is the concept of state, defined in
+the Service specification. The state is held on the device where the audio originates – the
+media player for MCS, or the telephony device, acting as a gateway to the call bearer for TBS,
+but always the Initiator. Client devices can implement the corresponding profile, which writes
+to a control point on the Server, causing its state to transition. Once that transition has been
+made, which may also happen locally by the user pressing a button or touching a screen, the
+Server notifies its new state. It’s exactly the same principle we came across in the ASCS state
+machine - there is one device which has the state, and multiple Clients which can read and
+manipulate it. In the case of both MCP and CCP, the Client can be the device receiving the
+audio stream, or a device including the Commander role, such as a remote control or a
+smartwatch. However, in this case the Client in the Commander would be operating on the
+Initiator, not the Acceptor, as shown in Figure 9.3.
+
+.
+Figure 9.3 Topologies for content control
+
+234
+
+Chapter 9 - Telephony and Media Control
+In addition to the control point and state characteristics, each service contains a wide selection
+of characteristics which can be read or notified to determine further information about the
+application and external service supplying it.
+
+9.3
+
+TBS and CCP
+
+At the heart of the telephony control specifications is a generic call state machine, illustrated
+in Figure 9.4.
+
+Figure 9.4 The TBS state machine (asterisks denote remote operations)
+
+The heart of the state machine is the Active state, which signifies the presence of a call. In
+most cases, this state would be attained when a unicast Audio Stream had been established
+between the Initiator and Acceptor, normally consisting of a bidirectional stream. However,
+the state machine is equally valid if the call is being taken using a wired headset, or if the phone
+is being used as a speakerphone with no active Audio Streams. This emphasises the distinction
+between the control plane, which in this case is the TBS / CCP relationship, and the audio
+data plane, which is the BAP / ASCS relationship. It is up to the application to tie them
+together as it wishes. (Although they are an integral part of GAF, they are not restricted to
+Bluetooth LE Audio and could be used with other applications.)
+235
+
+Section 9.3 - TBS and CCP
+The transitions around the state machine can be prompted by external events, such as:
+•
+•
+•
+•
+
+An incoming call,
+A CCP Client writing to the TBS Call State characteristic,
+A local user action, such as answering a call on the phone, or
+An operation by the remote caller.
+
+The remote operations in Figure 9.4 (Remote Alert Start, Remote Answer, Remote Hold and
+Remote Retrieve are marked by an asterisk (*) at the end of each command). These transitions
+can only be made by the remote caller. The states are exposed by the Call State characteristic,
+which is notified to Clients whenever the state of the call changes.
+
+9.3.1
+
+The Call State characteristic
+
+The Call State characteristic is an array of all of the current calls on the device. As soon as any
+call enters a non-Idle state it is assigned a unique, single-octet, sequential index number by
+TBS, called the Call_Index, with a value between 1 and 255, which is notified in the Call State
+characteristic. (The Call Control state machine of TBS does not contain an Idle state, but one
+is shown in Figure 9.4 for clarity). The second octet of the Call State characteristic identifies
+the current state of each call, and the final octet holds Call Flags associated with that call.
+The format of the Call State characteristic is shown in Figure 9.5.
+Call_Index [i]
+(1 octet)
+
+State [i]
+(1 octet)
+
+Call_Flags [i]
+(1 octet)
+
+Figure 9.5 Call State characteristic
+
+Table 9.1 lists the state values which are returned in the Call State Characteristic.
+State
+
+Name
+
+Description
+
+0x00
+0x01
+
+Incoming
+Dialing
+
+0x02
+
+Alerting
+
+0x03
+0x04
+0x05
+
+Active
+Locally Held
+Remotely Held
+
+0x06
+
+Locally and
+Remotely Held
+RFU
+
+An incoming call (normally causing a ringtone).
+An outgoing call has started, but the remote party has
+not yet been notified (the traditional dialtone state).
+The remote party is being alerted (they have a
+ringtone).
+The call is established with an active conversation.
+The call is connected, but on hold locally (see below)
+The call is connected, but on hold remotely (see
+below)
+The call is connected, but on hold both locally and
+remotely (see below)
+Reserved for Future Use
+
+0x07-0xFF
+
+Table 9.1 States defined for the Call State characteristic
+
+236
+
+Chapter 9 - Telephony and Media Control
+9.3.1.1
+
+Local and Remote hold
+
+States 0x04, 0x05 and 0x06 refer to calls which have been placed on hold. These states are
+differentiated by who it was who put the call on hold. A value of 0x04 indicates that it has
+been put on hold locally, i.e., by the user with the phone or PC with this TBS instance. A
+value of 0x05 means that it has been put on hold by the remote party, and a value of 0x06,
+means that both ends on the call have put it on hold.
+A locally held call can be retrieved (brought back to the Active state) by writing to the control
+point. A remotely held call needs to be retrieved by the remote party. The only local action
+that can be applied to a remotely held call is to terminate it.
+9.3.1.2
+
+Call Flags
+
+The final octet of the Call State characteristic is a bitfield containing information on the call,
+with the meanings and corresponding values shown in Table 9.2.
+Bit
+
+Description
+
+Value
+
+0
+
+Call Direction
+
+1
+
+Information withheld by server
+
+2
+
+Information withheld by
+network
+RFU
+
+0 = incoming call
+1 = outgoing call
+0 = not withheld
+1 = withheld
+0 = provided by network
+1 = withheld by network
+Reserved for Future Use
+
+3-7
+
+Table 9.2 Call status bits for the Call State characteristic
+
+Bits 1 and 2 indicate cases where information, such as the URI (typically the caller ID), or
+Friendly Name may be deliberately withheld because of either a local or network policy. Either
+or both may be set, in which case the fields of the associated characteristics may be empty or
+null. Alternatively, the Server can fill them with appropriate text, such as “Withheld” or
+“Unknown Caller”. The choice of policy and text is implementation specific.
+9.3.1.3
+
+UCIs and URIs
+
+Two initialisms you will come across in telephony applications are URI and UCI, which stand
+for Uniform Resource Identifier and Uniform Caller Identifier. Both are designed to provide
+call information which covers a wide variety of telephony applications and bearers.
+The Uniform Caller Identifier is essentially the bearer; for example, “skype” or “wtsap”. The
+values for the UCIs are listed in the Bluetooth Uniform Caller Identifiers Assigned Numbers
+document and are generally abbreviated to no more than five characters. Hence WhatsApp is
+“wtsap”. Standard phone numbers use the UCI of “E.164” or “tel:”, signifying a standard
+dialer format.
+237
+
+Section 9.3 - TBS and CCP
+The Uniform Resource Identifier is a combination of the UCI and the caller ID, i.e., the caller’s
+number or username, depending on the application. It can be used for either an incoming
+call, where it is the Caller ID, or an outgoing call. An application can use the UCI portion of
+a URI for an outgoing call to select the appropriate telephony application to make the call.
+URIs are expressed as a UTF-8 string.
+
+9.3.2
+
+The TBS Call Control Point characteristic
+
+As we’ve seen with BAP and ASCS, a Client can move a Server around its state machine by
+writing to a control point characteristic. In this case, it’s the TBS Call Control Point
+characteristic. This requires a single opcode, followed by a parameter which depends on the
+specific opcode, as indicated in Figure 9.6.
+Opcode
+(1 octet)
+
+Parameter
+(varies)
+
+Figure 9.6 The TBS Call Control Point characteristic
+
+Table 9.3 shows the current opcodes and describes their purpose.
+Opcode
+
+Name
+
+Parameter
+
+Description
+
+0x00
+0x01
+
+Accept
+Terminate
+
+Call_Index
+Call_Index
+
+0x02
+
+Local Hold
+
+Call_Index
+
+0x03
+
+Local
+Retrieve
+Originate
+Join
+
+Call_Index
+
+Accepts the incoming call.
+Terminate the call associated with
+Call_Index, regardless of its state.
+Places an Active or Incoming call with
+Call_Index on Local Hold
+Moves a locally held call to the Active
+state.
+Starts a call to the URI.
+Joins the calls identified in the list of
+Call_Index values.
+Reserved for Future Use
+
+0x04
+0x05
+
+0x06 – RFU
+0xFF
+
+URI
+List of Call_index
+values
+
+Table 9.3 Call Control Point characteristic opcodes for moving around the TBS state machine
+
+The opcodes of Table 9.3 are requests to the Server, which it passes to its relevant telephony
+application. Some of these: Accept, Terminate and Originate, refer to actions which will be
+taken locally by the telephony application. Local Hold and Retrieve generally need to be
+passed back to the network, and Join is almost always a network function. Depending on the
+specific telephony bearer and application, not all of these functions may be available. They
+may also fail based on the current phone resources and the network connection. For example,
+some phone services do not allow a call to be dialled if a current call if active. Equally, if there
+is no cellular signal, an attempt to originate a call will fail.
+238
+
+Chapter 9 - Telephony and Media Control
+A TBS instance can indicate whether the Local Hold and Join functions are supported by
+using the Call Control Optional Opcodes characteristic. This is a bit field which a Call Control
+Client can read to determine whether it should issue these commands. The values in this field
+are normally static for at least the duration of a session, and generally for the lifetime of the
+device. If the operation is successful, the Call State characteristic is notified, informing the
+Client of the current State and Call_Index. In the case where the opcode was Originate (0x04),
+this is the first time the Client will be made aware of the Call_Index.
+When a Write to the TBS Call Control Point characteristic by a Client fails, the Call Control
+Point characteristic notifies the result of the opcode write using the following format:
+Requested Opcode
+(1 octet)
+
+Call_Index
+(1 octet)
+
+Result Code
+(1 octet)
+
+Figure 9.7 TBS Call Control Point characteristic notification format
+
+The Result_Code indicates Success, or provides the reason for the failure of the operation.
+These codes are listed in Table 3.11 of the TBS specification.
+
+9.3.3
+
+Incoming calls, inband and out of band ringtones
+
+In traditional telephony design, the first thing that a user knows about an incoming call is the
+phone ringing. As soon as you introduce a Bluetooth connection, this becomes more
+complicated. The phone can still audibly ring, or the ring may happen in your earbuds (if
+you’re wearing them). Or it can happen on both.
+There is a further complication. The ringtone that you hear in your earbuds may be identical
+to the sound on your phone, or it may be a locally generated ringing tone produced by the
+earbud, which will be different. If it’s the same as the sound of your phone, it requires the
+presence of an Audio Stream from the phone, which carries the same ringtone audio that is
+being played on your phone’s speaker. That’s called an inband ringtone. The problem with
+an inband ringtone is that you need to set up the Audio Stream to carry it, which may mean
+tearing down an existing Audio Stream to another device. Imagine the use case where you’re
+using your earbuds to listen to your TV and your phone rings. For an inband ringtone, your
+earbuds will probably need to terminate the Audio Stream with the TV and replace it with an
+Audio Stream from your phone. If you accept the call, that’s not a problem, but if you reject
+it, you will need to reconnect to the TV, which is a poor user experience.
+The alternative is for the phone to notify its Incoming Call characteristic to your earbuds. This
+contains the Call_Index and URI, i.e., the URI scheme and the Caller ID of the incoming call.
+The Call Control Client in your earbud can generate a local ringing tone alerting you to the
+call. If you accept the call, the Call Control Client will write an Accept opcode (0x00) to the
+phone’s Call Control Point characteristic. That instructs your earbud to terminate the Audio
+Stream with the TV and the phone (which is a different Initiator) will establish a bidirectional
+Audio Stream to carry the call.
+239
+
+Section 9.3 - TBS and CCP
+If you reject the call, your Audio Stream with the TV remains, as it has not been interrupted
+– your earbud will simply have mixed its locally generated ringtone with the TV audio. It’s a
+much cleaner user experience, but it does mean that your earbuds will make a different ringing
+sound to your phone.
+The use of out of band ringtones can be enhanced if the earbud can perform text-to-voice
+conversion, allowing it to speak the phone number contained in the Incoming Call
+characteristic as part of the call alert. If the phone supports the optional Friendly Name
+characteristic, which contains the text of the caller name from your contact list on the phone,
+this could also be spoken within the earbud.
+One final subtlety to call alerts is that the phone can be set into silent mode, where an incoming
+ringtone is not sent to the phone (or PC’s) speaker, but where the announcement is left to the
+Acceptor. A Call Control Client can determine this, along with whether the phone supports
+an inband ringtone, by reading the Status Flags characteristic, shown in Table 9.4.
+Bit
+
+Description
+
+0
+
+Inband Ringtone
+
+1
+
+Silent Mode
+
+2 - 15
+
+RFU
+
+Value
+0 = disabled
+1 = enabled
+0 = disabled
+1 = enabled
+Reserved for Future Use
+
+Table 9.4 Status flags characteristic for ringtone mode support
+
+If the Inband ringtone is disabled, the Call Control Client would normally announce an
+incoming call at the point when it is notified that the Call State characteristic has transitioned
+to the Incoming state. Note that this includes Call Control Clients which are not Acceptors.
+For instance, a smart watch that included a Call Control Client might vibrate or ring, although
+it could not accept an Audio Stream. This highlights the fact that the Call Control state
+machine has no immediate connection to an ASE state machine. One does not directly trigger
+the other. It is entirely up to the telephony application or operating system to link the call
+control states to Audio Stream management.
+If the Status Flags characteristic shows that Silent Mode is enabled, it means that the ringtone
+will not be played on the phone. Depending on the state of the Inband Ringtone (bit 0), it
+may be sent to the Acceptor using an Audio Stream, or as an out of band ringtone, or both.
+If both are enabled, it is up to the Acceptor to decide which to use.
+It is worth remembering that accepting or rejecting the call, along with all other state
+transitions, can be performed on the Call Control Server (i.e., the phone or PC) as well as on
+a Call Control Client. All connected Call Control Clients have equal access; any action that
+changes the state will generate a notification.
+240
+
+Chapter 9 - Telephony and Media Control
+
+9.3.4
+
+Terminating calls
+
+A call can be terminated by any Call Control Client, by a user action on the phone, or by the
+remote caller. It can also be lost for a variety of reasons, such as a fault or loss of signal on
+the telephone bearer network. Whenever a call is terminated, the Call Control Server will use
+the Termination Reason characteristic to notify the Call Control Clients of the reason for the
+termination. Like similar characteristics, it includes the Call_Index to identify the call and the
+termination reason, as shown in Figure 9.8.
+Call_Index
+(1 octet)
+
+Reason_Code
+(1 octet)
+
+Figure 9.8 Termination Reason characteristic
+
+If multiple calls are terminated, as might happen in the case of a network issue on a joined
+call, then a separate Termination Reason notification is sent for each Call_Index. The
+termination reasons are shown in
+Reason_Code Reason
+0x00
+0x01
+0x02
+0x03
+0x04
+0x05
+0x06
+0x07
+0x08
+0x09
+0x0A – 0xFF
+
+Incorrectly formed URI for originating call
+The call failed
+The remote party ended the call
+The server ended the call
+Line busy
+Network congestion
+A client ended the call
+No service
+No answer
+Unspecified
+Reserved for Future Use
+
+Table 9.5 Termination Code characteristic Reason_Code values
+
+9.3.5
+
+Other TBS characteristics
+
+As mentioned above, TBS contains a large number of other characteristics for the Call Control
+Server role. These characteristics provide more detailed information about the phone call.
+They can be read by more complex clients, such as carkits, which can use them to replicate
+the user experience of the phone by mirroring the level of information which is available from
+the original telephony application.
+Table 9.6 provides a list of the additional characteristics which have not been covered above.
+All of them are mandatory for a GTBS or TBS instance to support, with the exception of the
+Bearer Signal Strength characteristic and its dependent Bearer Signal Strength Reporting
+Interval characteristic. They are all described in the Telephone Bearer Service specification.
+241
+
+Section 9.4 - MCS and MCP
+Characteristic Name
+
+Description
+
+Bearer Provider Name
+Bearer Uniform Caller Identifier
+Bearer Technology
+Bearer Signal Strength (Optional)
+Bearer Signal Strength Reporting
+Interval
+Bearer URI Schemes Supported List
+Bearer List Current Calls
+Content Control ID (CCID)
+
+Telephony service. E.g., Vodafone, T-Mobile
+UCI, e.g., skype, wtsap, from Assigned Numbers
+As displayed on the phone. E.g., 2G, 3G, Wi-Fi
+Signal strength from 0 (no signal) to 100
+How often the signal strength is reported
+
+Incoming Call Target Bearer URI
+
+A list of supported URI schemes
+A list of all current calls and their state
+A CCID value which remains static until a service
+change
+The incoming URI of the call. E.g., your phone’s
+number.
+
+Table 9.6 Other characteristics specified in TBS
+
+There are corresponding procedures to read all of the TBS characteristics in the Call Control
+Profile. Although support for most of the characteristics in TBS are mandatory, the only Call
+Control procedure which is mandatory is the Read Call state procedure. This reflects the fact
+that for devices with a constrained user interface, such as hearing aids and earbuds, most
+control actions are likely to take place on the phone.
+
+9.4
+
+MCS and MCP
+
+Once you’ve understood telephony control, the media control specifications will look very
+familiar. The Media Control profile defines two roles – the Media Control Client and the
+Media Control Server. The latter resides on an Initiator, where the Media Control Service
+defines a state machine for media playback, (which is shown in Figure 9.9), along with a
+plethora of characteristics to notify what it’s doing.
+
+242
+
+Chapter 9 - Telephony and Media Control
+
+Figure 9.9 MCS state machine for media control
+
+For Media, the player normally resides in the Inactive state, moving to the Paused State when
+a Track is selected, from which it can be transitioned to Playing. From Playing, it can return
+to Paused by stopping it, or be moved to the Seeking state by issuing a Fast Forward or Fast
+Rewind command. There is no Stop state or command in the state machine. For digital
+media, there is no real distinction between the Stop and Pause command. That difference
+harks back to analogue devices where Stop turned off the motor of a record or cassette deck,
+whereas Pause left the motor running and lifted the stylus or playback head off the playback
+medium. When everything is in digital memory, the operations are the same.
+
+9.4.1
+
+Groups and Tracks
+
+The three main states – Paused, Playing and Seeking are only valid when there is a current
+track selected, which brings us to the concept of tracks and groups, which is a key part of
+MCS. Figure 9.10 illustrates the concept, showing a hierarchy of Groups, which contain
+Tracks. It is entirely up to an implementation to define how these are structured, but a Group
+is probably best visualised as a traditional album, where the tracks are individual songs.
+Groups can be nested to become parts of larger groups, so a Parent Group consisting of the
+entire works of Beethoven might contain subgroups of Orchestral music, Chamber music,
+Piano, Vocal works, etc., which could each contain further collections of subgroups. All that
+MCS requires is that there is a defined structure to create a playing order from the start of the
+first group to the end of the last group.
+243
+
+Section 9.4 - MCS and MCP
+
+Figure 9.10 MCS' arrangement of Groups and Tracks
+
+Groups contain tracks, which may also contain segments. Most of the MCP controls operate
+on tracks, which are what most people would regard as conventional tracks on a record or
+CD. Specifically, the operations of the state machine focus on the Current Track. The key
+elements of a Track are shown in Figure 9.11 , which are the Track Duration, Offsets from
+either end of the Track and the Playing Position.
+
+Figure 9.11 Key elements of a Track
+
+244
+
+Chapter 9 - Telephony and Media Control
+
+9.4.2
+
+Object Types and Search
+
+MCP and MCS include a lot of functionality to support complex user interfaces, such as the
+screen of a car’s AV system. Most of this is based on the existing Bluetooth LE Object
+Transfer Service (OTS) which allows extended information to be associated with Groups and
+Tracks. This includes extended names, icons and URLs which can be used to fetch more
+complex objects, such as album covers. OTS is also used for returning search results. MCS
+allows some very comprehensive search functions. You can search using combinations of
+Track Name, Artist Name, Album Name, Group Name, Earliest Year, Latest Year and Genre.
+For more information on these, look at the MSC specification. They’re unlikely to be used in
+products in the short term, so I’ll skip the detail and concentrate on the key aspects of media
+control.
+
+9.4.3
+
+Playing Tracks
+
+With the structure of Groups and Tracks explained, we can look at how media control works.
+Once a track is selected, which is generally a user action on the media player, the player will
+notify all of its connected Clients using the Track Changed characteristic. This characteristic
+has no value, i.e., it’s empty, but is a statement that the media player now has a current track,
+so a Client can start to control it. A Client can read the Track Title Characteristic and the
+Track Duration characteristic. It can also read the Track Position, which is returned with a
+10msec resolution from the start of the track. (The maximum track duration allowed is just
+over 16 months46.)
+A Client can now write to the Media Control Point characteristic with the values shown in
+Table 9.7.
+Opcode
+
+Name
+
+0x01
+0x02
+0x03
+0x04
+0x05
+
+Play
+Pause
+Fast Rewind
+Fast Forward
+Stop
+
+Table 9.7 Media Control Point characteristic opcodes for media control
+
+Once the Current Track starts playing, the Server should notify the Track Position
+characteristic, to inform Clients of the current track position. The frequency of notification
+
+This might seem ridiculous, but the longest single video, “Level of Concern” by the group “twenty
+one pilots” is 177 days, 16 hours, 10 minutes and 25 seconds long, and it is likely that someone will try
+to break this record.
+46
+
+245
+
+Section 9.4 - MCS and MCP
+is down to the implementation.
+Another set of opcodes are available for moving around the current track, which are shown
+in Table 9.8.
+Opcode
+0x10
+0x20
+0x21
+0x22
+0x23
+0x24
+
+Name
+
+Parameters
+
+Move Relative
+Previous Segment
+Next Segment
+First Segment
+Last Segment
+Goto Segment
+
+32 bit signed integer
+None
+None
+None
+None
+32 bit signed integer
+
+Table 9.8 Media Control Point characteristic opcodes for in-track movement
+
+The Media Control Service allows segments to be defined within a track. Each segment
+includes a name and an absolute offset from the start of the track. The commands in Table
+9.8 allow a move to a specific segment of the current track. None of these allow movement
+outside the current track. If the parameter value results in moving outside the current track,
+then the result will be limited to moving the position to the start of the track or the end of the
+track, but the track remains the current track.
+Issuing the Stop command will result in the current track becoming invalid, so before any of
+the commands in Table 9.7 can be used again, the user will need to select another track as
+current.
+Two further sets of opcodes allow a different track to be selected as the current track.
+Move within a Group
+
+Move to another Group
+
+Opcode
+
+Name
+
+Parameters
+
+Opcode
+
+Name
+
+Parameters
+
+0x30
+
+Previous
+Track
+Next Track
+First Track
+Last Track
+Goto Track
+
+None
+
+0x40
+
+None
+
+None
+None
+None
+32 bit signed
+INT
+
+0x41
+0x42
+0x43
+0x44
+
+Previous
+Group
+Next Group
+First Group
+Last Group
+Goto Group
+
+0x31
+0x32
+0x33
+0x34
+
+None
+None
+None
+32 bit signed
+INT
+
+Table 9.9 Media Control Point opcodes for moving around Tracks and Groups
+
+When another Group is selected, the first track of the current group resulting from the
+command becomes the current track.
+246
+
+Chapter 9 - Telephony and Media Control
+All segments, tracks and groups start numbering at zero. Where a Goto operation is used, a
+positive value “n” is equivalent to going to the first item and then executing the next item
+opcode n-1 times. A negative value is equivalent to going to the last item and executing the
+previous item opcode n-1 times.
+Moving to a new track makes it current and places it in the Paused state of the state machine
+of Figure 9.9.
+It is only mandatory to support a minimum of one of the opcodes in the Media Control Point
+characteristic. It is unlikely that any implementation would be that spartan, but Clients can
+check what the Server supports by reading the Media Control Point Opcodes Supported
+characteristic. This is a bitfield which indicates which opcodes the Server supports.
+As always with Control Point opcodes, a Client writes the opcode and receives a notification
+of the operation in the Media Control Point characteristic, with a Result_Code indicating
+Success, Opcode not supported, or Media Player Inactive. Once the Server has fulfilled the
+request it will also notify the appropriate characteristic for that operation if a change has
+occurred in its value.
+
+9.4.4
+
+Modifying playback
+
+The Media Control Service has a couple of neat features for modifying playback and Seeking.
+The first of these is the ability to request a change in playback speed, using the Playback Speed
+characteristic. This can be written by the Client to request that the current track is played
+faster or slower. The parameter for the characteristic is an 8-bit signed integer “p”, which
+ranges in value from -128 to 127, where the actual playback speed requested is:
+𝑝
+
+𝑠𝑝𝑒𝑒𝑑 = 264
+I.e., the speed is 2 to the power of (p/64).
+effectively ¼ to 4 times the standard speed.
+
+This gives a range of 0.25 to 3.957, which is
+
+Playback speed is a blind request; the Client does not know whether or not the value of “p” is
+supported. After the request, the Server notifies the value which has been set in the Playback
+Speed characteristic. If the request was outside the range that the Server supports, it should
+set it to the closest available speed. If the Media source is live or streamed, then the Playback
+Speed characteristic value may be fixed to a value of 1. It is entirely up to the implementation
+to decide how to adjust the speed and whether to maintain the pitch of the Audio Data.
+The other characteristic which can be varied is the Seeking Speed, by writing a signed integer
+value to the Seeking Speed characteristic. This is used as a multiple of the current Playback
+speed, with positive values signifying Fast Forward and negative values Fast Reverse. The
+Seeking Speed value is applied to the current Playback Speed, not the default Playback Speed,
+where the value of p would be 0. A value of 0 for the Seeking Speed indicates a Seeking Speed
+247
+
+Section 9.4 - MCS and MCP
+which is the same as a Playback Speed with a value of p = 0, regardless of the current Playback
+Speed.
+During Seeking, the Track Position characteristic should be notified to provide an indication
+of the current Track Position. The frequency of notification is determined by the
+implementation. Seeking will stop when the Track Position has reached the start or end of
+the current track.
+Whilst seeking, MCS states that no Audio Data should be played. However, the Context Type,
+which describes the use case (not the content of the Audio Stream) would not normally
+change.
+
+9.4.5
+
+Playing order
+
+The last feature to cover in MCS is the playing order. The Playing Order characteristic
+provides ten different ways to play tracks. Most of them cover the order of play when you’re
+playing an entire Group. The options are listed in Table 9.10.
+Value Name
+
+Description
+
+0x01
+0x02
+0x03
+0x04
+0x05
+0x06
+
+Single once
+Single repeat
+In order once
+In order repeat
+Oldest once
+Oldest repeat
+
+0x07
+
+Newest once
+
+0x08
+
+Newest repeat
+
+0x09
+0x0A
+
+Shuffle once
+Shuffle repeat
+
+Play the current track once
+Play the current track repeatedly
+Play all of the tracks in a group in track order
+Play all of the tracks in a group in track order, then repeat
+Play all of the tracks in a group once, in ascending order of age
+Play all of the tracks in a group in ascending order of age, then
+repeat
+Play all of the tracks in a group once, in descending order of
+age
+Play all of the tracks in a group in descending order of age, then
+repeat
+Play all of the tracks in a group once, in random order
+Play all of the tracks in a group once, in random order, then
+repeat
+
+Table 9.10 Playing Order characteristic values
+
+There are a couple of nuances in these. The first two values, which play the current track
+(0x01 and 0x02) set the current track to the next track when they have completed or been
+Stopped. For shuffle, the randomisation is left to the implementation. In most cases the
+implementation is not totally random, but weighted, as listeners do not expect the same tracks
+to be played close together at the start of a subsequent cycle. These decisions are left to the
+implementation. The Client simply sends the command. If the Server cannot support the
+requested playing order, for example when it doesn’t know the age of the Tracks, it should
+ignore the command.
+248
+
+Chapter 9 - Telephony and Media Control
+As with the Media Control Point characteristic, a Client can determine which Playing Order
+features are supported by reading the Playing Orders Supported characteristic. This is a 2octet bitfield with the bitfield meanings shown in Table 9.11
+Value
+
+Name
+
+0x0001
+0x0002
+0x0004
+0x0008
+0x0010
+0x0020
+0x0040
+0x0080
+0x0100
+0x0200
+
+Single once
+Single repeat
+In order once
+In order repeat
+Oldest once
+Oldest repeat
+Newest once
+Newest repeat
+Shuffle once
+Shuffle repeat
+
+Table 9.11 Meaning of bits in the Playing Orders Supported characteristic
+
+That’s it for telephony and media control Our next stop is volume, audio input and
+microphone.
+
+249
+
+Section 9.4 - MCS and MCP
+
+250
+
+Chapter 10 - Volume, Audio Input and Microphone Control
+
+Chapter 10. Volume, Audio Input and Microphone Control
+Anyone who has worked on audio specifications will probably tell you that a large part of their
+time was taken up with discussions about volume control; it’s a topic which generates even
+more debate than audio quality. The reason for that is twofold. The first is a never-ending,
+and generally academic discourse on the perception of volume and how to define the steps
+between minimum and maximum. The second is how to cope with multiple different ways of
+controlling the volume. The second problem didn’t exist when you had a single volume knob
+on your TV or amplifier. However, as soon as you have multiple remote controls, it became
+increasingly challenging for a user to work out how to set the volume.
+After the requisite hundreds of hours of debate, the Bluetooth® LE Audio working groups
+decided to more or less skip the first problem and concentrate on the second – how to
+coordinate multiple different points of control. So, this chapter is all about volume control,
+with only a passing reference to how volume is interpreted, because that’s left to the
+implementation. We’ll also cover microphone control, as it’s analogous, but involves the input
+of audio, rather than its output. All of these features are designed to be used on non-encoded
+audio signals. In the case of volume, that means after reception and decoding; for the
+Microphone control service and profile, it is before transmission.
+
+10.1
+
+Volume and input control
+
+After the previous chapters, this should be plain sailing. There are three services involved
+with volume and one profile:
+•
+•
+•
+
+•
+
+The Volume Control Service (VCS) which can be viewed as the main volume
+control knob
+The Volume Offset Control Service (VOCS), which can be considered as a
+“Balance” control
+The Audio Input Control Service (AICS), which allows individual gain control and
+mute on any number of audio inputs, which don’t need to be Bluetooth Audio
+Streams, and
+The Volume Control Profile (VCP), which lets multiple Clients control these
+Services.
+
+Although most of these have “volume” in their name, what they actually do is adjust the gain,
+i.e., the amplification of the audio signal.
+
+251
+
+Section 10.1 - Volume and input control
+
+Figure 10.1 Typical implementation of Volume and Audio Input services for a headphone
+
+Figure 10.1 shows a typical implementation for a pair of headphones, or banded hearing
+aids, which has three possible audio inputs – a Bluetooth Audio Stream, a Telecoil audio
+input and a microphone input. These inputs are mixed together, using the Audio Input
+Control Service to set their individual gains and to selectively mute and unmute them. The
+resulting stream has its gain controlled by the Volume Control Service setting. If the stream
+is a stereo one, then, once it is split into its left and right components, a separate instance of
+the Volume Offset Control Service can adjust the gain going to each speaker. (The drawing
+is a hardware representation. In practice the sum of VCS and VOCS gain settings are
+applied to each Audio Channel.) Used together differentially, these mimic the effect of a
+balance control. They can also be used individually to adjust the relative level of sound in
+each speaker to accommodate different levels of hearing loss in the left and right ear.
+
+10.1.1
+
+Coping with multiple volume controls
+
+From the outset, during Bluetooth LE Audio development, it was obvious that users would
+want to control the volume of their earbuds and headset from multiple different places. The
+assumption was that there would be local controls on the earbuds, volume controls on the
+audio sources (typically the phone), as well as separate volume controls in smart watches and
+other dedicated remote control devices.
+To make this level of distributed control work, you want to keep the primary volume control
+at the device which is rendering the audio. If you don’t, but let it be controlled at the audio
+source as well, you can end up with one Controller operating on the source level and one on
+the sink level. That leads to the situation where the source gain gets turned down, with the
+sink gain turned up to full to compensate. It means that the user can no longer increase the
+volume at their ears, and the audio quality is reduced, as the codec is sampling and encoding
+a signal with a very low amplitude. This decreases the all-important signal to noise ratio, as
+on decoding and amplification the noise gets amplified as well.
+252
+
+Chapter 10 - Volume, Audio Input and Microphone Control
+To address this, it’s important that the gain occurs at the end of the audio chain, but for that
+to work, all of the volume control Clients need to keep track of its value, so they always know
+where they are. It’s a bit like the old joke, where a driver stops to ask a passer-by if he could
+give him directions to get to Dublin, and the passer-by replies, “If I was going to Dublin, I
+wouldn’t start from here”. Fortunately, the GATT Client-Server model has notifications, so
+we can make use of these to ensure that every Volume Control Client knows the current gain
+level on the Server. But to be absolutely sure, the Volume Control Service includes an
+additional safeguard called the Change_Counter, which we’ll explore below.
+Looking back at Figure 10.1, it also explains the architecture, where the Volume Control state
+is as close as possible to the final point of rendering. This brings us on to the individual
+Volume Control Services. These are very thin documents, so we can skim through them fairly
+quickly.
+
+10.2
+
+Volume Control Service
+
+Like the control service we’ve seen before, all of the hard work is done by two characteristics:
+•
+•
+
+The Volume Control Point, which the Clients use to set the volume level, and
+The Volume State, which the Server uses to notify the current volume level after
+the Volume Control Point characteristic has been written.
+
+We should strictly be talking about gain, rather than volume, but as everyone is used to the
+colloquial use of volume, I’ll stick with it. A third characteristic – Volume Flags, is used to let
+a Client know about the history of the current volume setting.
+The Volume State characteristic has three fields, all one octet long, which are shown in Figure
+10.2. There is only one instance of the Volume State characteristic on an Acceptor.
+Field Name
+
+Values
+
+Volume_Setting
+Mute
+
+0 to 255. 0 = minimum, 255 = maximum
+0 = not muted
+1 = muted
+0 to 255
+
+Change_Counter
+
+Figure 10.2 The Volume State characteristic of VCS
+
+The Volume_Setting is defined as a value from 0 (minimum volume) to 255 (maximum), with
+no stipulation on how those figures are related to the output volume. It is left to the
+implementer to map these values to actual output volume, with the expectation that the user
+perception is approximately linear. I.e., if the Volume_Setting value is set to 127, then the
+resulting output should sound as if it is halfway to the maximum. This brings us back to the
+debate about how volume is perceived. Our hearing is generally logarithmic, which is why
+sound intensity is measured in decibels, but that can be affected by what it is we’re listening
+253
+
+Section 10.2 - Volume Control Service
+to. Leaving it to the manufacturer does mean that if speakers or earbuds from different
+manufacturers are used together, changes to their volume may not align with the listener’s
+expectation. One may appear louder than the other at certain settings. Solving that fell into
+the “too difficult” category, so it’s been left to implementation.
+The value of the Volume_Setting field is changed as a result of an operation on the Volume
+Control Point characteristic (q.v.) which modifies the current value, or a local operation on
+the user interface of the device. The value stored in Volume_Setting is not affected by any
+Mute operation.
+The Mute indicates whether the audio stream has been muted. This is independent of the
+Volume_Setting value, so that when a device is unmuted, the audio volume will resume at its
+previous level.
+Change_Counter is the feature mentioned above, which ensures that all Volume Control
+Clients are synchronised. When an Acceptor implementing VCS is powered on, the Server
+initialises Change_Counter with a random value between 0 and 255. Every subsequent
+operation on the Volume Control Point characteristic or the device’s local controls, whether
+to change volume or mute, which results in a change to the Volume_Setting or Mute field
+values, will result in the Change_Counter incrementing its value by 1. Once it reaches a value
+of 255, it will roll around to 0 and keep incrementing.
+The purpose of Change_Counter is to ensure that any Client attempting to change the current
+volume or mute settings is aware of what they are currently. This is checked by requiring every
+write procedure by a Client to include the current value of Change_Counter which they hold.
+Unless that matches the value in the Volume State Characteristic, it is assumed they are out of
+sync. The command is ignored, and no notification will be sent. In this case, the Client should
+recognise the lack of a notification, read the Volume State characteristic and try again.
+To change the volume, or mute a device, the Volume Control Client writes to the Volume
+Control Point characteristic of the Volume Control Service using one of the Volume Control
+Point procedures. This command contains an opcode, a Change_Counter value and, in the
+case of the Set Absolute Volume command, a Volume_Setting value. The six opcodes are
+listed in Table 10.1.
+Opcode
+
+Operation
+
+Operands
+
+0x00
+0x01
+0x02
+0x03
+0x04
+0x05
+
+Relative Volume Down
+Relative Volume Up
+Unmute / Relative Volume Down
+Unmute / Relative Volume Up
+Set Absolute Volume
+Unmute
+
+Change_Counter
+Change_Counter
+Change_Counter
+Change_Counter
+Change_Counter, Volume_Setting
+Change_Counter
+
+254
+
+Chapter 10 - Volume, Audio Input and Microphone Control
+Opcode
+
+Operation
+
+Operands
+
+0x06
+0x07-0xFF
+
+Mute
+Reserved for Future Use
+
+Change_Counter
+
+Table 10.1 Volume Control Point characteristic opcodes
+
+Most of these are self-explanatory. Relative Volume Down and Up change the volume setting,
+without affecting the Mute state, so can be used to change the volume level which will be
+applied when a device is eventually unmuted. Opcodes 0x05 and 0x06 unmute and mute
+without affecting the volume level, whilst opcodes 0x02 and 0x03 apply a relative volume step
+at the same time as muting or unmuting. Set Absolute Volume takes an additional parameter,
+which is the absolute volume level required.
+Although the Volume_Setting ranges from 0 to 255 for all devices, very few are likely to
+contain more than twenty-one47 discrete volume levels. This requires a Server to define a Step
+Size, which is applied whenever a volume setting command is issued. The Step Size is
+effectively 256 ÷ Number of steps. The following equations are used by the Server to calculate
+the new value written into the Volume_Setting value of the Volume State characteristic.
+Relative Volume Down: Volume_Setting = Max (Volume_Setting – Step Size, 0)
+Relative Volume Up: Volume_Setting = Min (Volume_Setting + Step Size, 255)
+
+10.2.1
+
+Persisting volume
+
+The final characteristic is the Volume Flags characteristic, which is a bitfield, of which only
+one bit is defined. This is Bit 0, which is the Volume_Setting_Persisted Field. If set to 0, it
+directs a Client to reset the volume by writing to the Volume Control Point characteristic with
+an opcode of 0x04. This sets an absolute volume, which may be a default value determined
+by an application or the Client device. It may also be the value remembered from the last time
+that application was used.
+If the value is 1, it indicates that the Volume Control Server still has the Volume_Setting from
+its last session, which should continue to be used for this session. The Volume Client should
+retrieve this value, either through a notification or by reading it, and use it as the initial value
+for this session. This can improve the user experience, by starting with the same volume
+settings which had previously been used.
+
+Assuming a 0 to 10 scale, with a resolution of 0.5. Unless you’re a Spinal Tap fan, in which case it’s
+twenty-three.
+47
+
+255
+
+Section 10.3 - Volume Offset Control Service
+
+10.3
+
+Volume Offset Control Service
+
+If all you have is a single, mono speaker, then the Volume Control Service is all you need. As
+soon as you are dealing with more than one audio stream, whether that’s going to a single
+Acceptor or multiple Acceptors, you need to include an instance of the Volume Offset Control
+Service for every rendered Audio Location. All of the characteristics are accessed using
+procedures in the Volume Control Profile.
+The Volume Offset Control Service includes four characteristics, shown in Table 10.2. All
+four of them are mandatory.
+Characteristic
+
+Mandatory Properties
+
+Optional Properties
+
+Volume Offset State
+Volume Offset Control Point
+Audio Location
+Audio Output Description
+
+Read, Notify
+Write
+Read, Notify
+Read, Notify
+
+None
+None
+Write without response
+Write without response
+
+Table 10.2 Volume Offset Control characteristics
+
+The Volume Offset State and Volume Offset Control work in the normal manner. The
+Volume Offset State includes two fields – Volume_Offset and Change_Counter
+Field
+
+Size
+
+Volume_Offset
+Change_Counter
+
+2 octets
+1 octet
+
+Table 10.3 The Volume Offset characteristic
+
+The Volume_Offset is two octets long, as it allows values from -255 to 255. The value of the
+Volume_Offset is added to the value of Volume_State to provide a total volume value for
+rendering.
+The Change_Counter is the same concept we saw in the Volume Control Service and is used
+to ensure that any Client issuing a Volume Offset Command is aware of the current status of
+the Volume Offset. Note that although it is the same concept and has the same name, it is a
+separate instance. Every instance of VOCS, as well as the single instance of VCS, maintain
+their own value of Change_Counter, which is updated when there is any change to their
+Volume_Offset or Volume_State field value.
+To effect a change, a Client writes to the Volume Offset Control Point characteristic with the
+following parameters:
+
+256
+
+Chapter 10 - Volume, Audio Input and Microphone Control
+Parameter
+Opcode
+Change_Counter
+Volume_Offset
+
+Size (octets)
+1
+1
+2
+
+Value
+0x01 = Set Volume Offset
+0 to 255
+-255 to 255
+
+Table 10.4 The Set Volume Offset parameters (opcode = 0x01)
+
+10.3.1
+
+Audio Location characteristics
+
+The Volume Offset Control service contains two characteristics relating to the Audio Location
+to which the offset is applied.
+The first of these is the Audio Location characteristic. This is a 4 octet bitmap which defines
+which Audio Location the offset should be applied to, using the format of Audio Locations
+from the Generic Audio assigned numbers. Normally, this will contain a single location, as a
+separate VOCS instance is applied to each Audio Location. This is normally a read-only
+characteristic, set at manufacture, but it can also be made writable.
+The Audio Output Description characteristic allows a text string to be assigned to each Audio
+Location to provide more information for an application, e.g., Bedroom Left Speaker. It may
+be assigned at manufacture, or made accessible for users to assign friendly names.
+
+10.4
+
+Audio Input Control Service
+
+The final part of the rendering trio of services is the Audio Input Control Service. This
+acknowledges that many products, such as hearing aids, earbuds and headphones, contain
+more than one source of audio streams and that more than one of them may be used at the
+same time. The most common example is the concurrent use of an external microphone along
+with a Bluetooth stream. For hearing aids, this has always been the way they work. More
+recently, with the appearance of transparency modes, it has become an increasingly popular
+feature in earbuds and headphones, so that wearers can be made aware of sounds from the
+world around them.
+An instance of the Audio Input Control Service is normally included for each separate audio
+path which is intended to be rendered on that device. If there is a single Bluetooth LE Audio
+path, it offers no advantages over VCS, so does not need to be included. Audio inputs which
+are not directly fed to the output, such as microphones which provide inputs for noise
+cancellation, would not have their own instance, as they would be used for processing another
+audio stream (which might have its own AICS instance).
+The Audio Input Control Service contains six characteristics, which are listed in Table 10.5.
+All of them are mandatory to support.
+
+257
+
+Section 10.4 - Audio Input Control Service
+Name
+
+Mandatory
+Properties
+
+Optional Properties
+
+Audio Input State
+Audio Input Control Point
+Audio Input Type
+Audio Input Status
+Audio Input Description
+Gain Setting Properties
+
+Read, Notify
+Write
+Read
+Read, Notify
+Read, Notify
+Read
+
+None
+None
+None
+None
+Write without response
+None
+
+Table 10.5 Audio Input Control Service characteristics
+
+Before jumping into these, we need to understand how AICS defines Gain. With volume,
+VCS and VOCS simply define a linear scale from 0 to 255, much like a knob on a Hi-Fi unit
+which goes from 0 to 10, and leaves it up to the manufacturer to convert that to what is usually
+a decibel based, internal mapping. With AICS, the assumption is made that gain will be decibel
+based, so the figures for Gain are in decibel-based units. However, to provide a bit more
+flexibility, the actual step in gain values can be defined in multiples of 0.1 dB. What those
+multiples are is a manufacturer specific choice, which is exposed in the Gain Setting Properties
+characteristic, along with minimum and maximum permitted values, as shown in Table 10.6.
+Name
+
+Size
+Description
+(octets)
+
+Gain_Setting_Units
+
+1
+
+Gain_Setting_Minimum
+Gain_Setting_Maximum
+
+1
+1
+
+The increment, in 0.1db steps, applied to all
+Gain_Setting values
+The minimum permitted value of Gain_Setting
+The maximum permitted value of Gain_Setting
+
+Table 10.6 Fields of the Gain Setting Properties characteristic (read only)
+
+In a further complexity, AICS allows an Audio Stream to have its Gain controlled
+automatically (traditionally known as Automatic Gain Control or AGC), or manually, which
+may either be by a Volume Control Client, or a local user control, with changed values exposed
+in the Audio Input State characteristic. When control is automatic, the Server does not
+support any changes made by the Client. The Server can expose whether or not the Client is
+allowed to change the mode from Automatic to Manual or vice versa through the Gain_Mode
+field of the Audio Input State characteristic, using the values shown in Table 10.7.
+
+258
+
+Chapter 10 - Volume, Audio Input and Microphone Control
+Gain_Mode
+
+Value
+
+Manual Only
+Automatic Only
+Manual
+
+0
+1
+2
+
+Automatic
+
+3
+
+Description
+A Client may not change these settings
+Manual, but a Client may change the Gain_Mode to
+Automatic
+Automatic, but a Client may change the Gain_Mode to
+Manual
+
+Table 10.7 Meaning of the Gain_Mode field values in the AICS Audio Input State characteristic
+
+Having sorted that out, we can move on to the Audio Input State and the Audio Input Control
+Point, which work much as we’d expect. The Audio Input State contains four fields, described
+in Table 10.8.
+Name
+Gain_Setting
+Mute
+Gain_Mode
+Change_Counter
+
+Size (octets)
+1
+1
+1
+1
+
+Table 10.8 Fields for the Audio Input State characteristic
+
+The Gain_Setting exposes the current value of Gain, in Gain_Setting_Units. Writing a value
+of 0 to Mute will not affect the current Gain_Setting value, so unmuting it by writing 1 to
+Mute will return the gain to the previous value in Gain_Setting. If the Server is in Automatic
+Gain Mode, it will ignore anything written to the Gain_Setting field.
+Mute says whether the Audio Stream is muted (value = 0) or unmuted (value = 1). AICS gives
+us an additional option, where the value of 2 indicates that a Mute function is disabled. The
+final audio stream can still be muted using the VCS Mute, but by exercising this option, the
+individual input stream cannot be muted separately.
+The Gain_Mode and the Change_Counter behave exactly as they do for VCS and VOCS.
+Again, it’s an independent instantiation for each AICS instance.
+Setting the AICS parameters is achieved by writing to the Audio Input State Control Point
+characteristic, with one of the five opcodes listed in Table 10.9
+
+259
+
+Section 10.4 - Audio Input Control Service
+Opcode Opcode Name
+
+Description
+
+0x01
+
+Set Gain Setting
+
+0x02
+0x03
+0x04
+
+Unmute
+Mute
+Set Manual Gain Mode
+
+0x05
+
+Set Automatic Gain
+Mode
+
+Sets the gain in increments of Gain_Setting_Units x
+0.1dB
+Unmutes (unless mute is disabled)
+Mutes (unless mute is disabled)
+Changes From Manual to Automatic Gain (if
+allowed)
+Changes From Automatic to Manual Gain (if
+allowed)
+
+Table 10.9 Opcodes for the Audio Input Control Point characteristic
+
+These all do pretty much what it says in the name, although as we’ve already discovered, there
+are quite a number of caveats where the Server can ignore them, especially if it is not prepared
+to give up control of the Gain and Mute functionality.
+The Set Gain Setting opcode (0x01) takes the form shown in Table 10.10.
+Parameter
+
+Size (Octets) Value
+
+Opcode
+Change_Counter
+Gain_Setting
+
+1
+1
+1
+
+0x01
+0 to 255
+-128 to 127
+
+Table 10.10 Format for the Audio Input Control Point characteristic Set Gain Setting procedure
+
+All of the other procedures use the same, shorter format for their parameters, shown in Table
+10.11.
+Parameter
+
+Size (Octets) Value
+
+Opcode
+
+1
+
+Change_Counter
+
+1
+
+0x02 = Unmute
+0x03 = Mute
+0x04 = Set Manual Gain Mode
+0x05 = Set Automatic Gain Mode
+0 to 255
+
+Table 10.11 Format for opcodes 0x02 - 0x05 for Audio Input Control Point characteristic
+
+As with VOCS, there are a few additional characteristics which can help with a user interface.
+The Audio Input Type is used to identify the Audio Input Stream with a read-only text
+description which can be used for a UI. Typical values (in English) are Microphone, HDMI,
+Bluetooth, etc. These are set by the manufacturer.
+
+260
+
+Chapter 10 - Volume, Audio Input and Microphone Control
+The Audio Input Status exposes the current status of each AICS Audio Stream. If the value
+is set to 0, the Audio Stream is inactive; if it is set to 1, the Audio Stream is active.
+The Audio Input Description allows for a more detailed description of an audio input, which
+is useful where there are multiple inputs of the same type, such as multiple Bluetooth or HDMI
+inputs. It may be set by the manufacturer or optionally made writable to allow a user to update
+it.
+
+10.5
+
+Putting the volume controls together
+
+The opening diagram showed how these three services work together. Figure 10.3 shows a
+typical implementation for a pair of stereo headphones, which support both a Bluetooth
+connection and a local microphone.
+The volume for each ear is set by combining the single value in VCS, with the Audio
+Location specific value from the VOCS instantiation for each ear. Either of the incoming
+audio streams can be individually muted or have their gain controlled (if the Server allows it),
+whilst VCS provides a global mute function.
+
+Figure 10.3 Typical implementation for a pair of headphones
+
+Figure 10.4 shows a similar approach for the left hearing aid of a stereo pair. Because
+only one channel is being rendered, there is only one instance of VOCS, serving the left hearing
+aid. The other hearing aid, in the right ear would be identical, but with a VOCS instance for
+the right Audio Location.
+
+261
+
+Section 10.6 - Microphone control
+
+Figure 10.4 Typical implementation for a left hearing aid from a stereo pair
+
+10.6
+
+Microphone control
+
+Although the examples above include microphones, they are all using the microphone as a
+local input, which is selected or mixed with other audio inputs to be rendered. However,
+microphones are also used as audio inputs which are captured and sent back to an Initiator.
+The Microphone Control Service and Profile (MICS and MICP) exist to provide device wide
+control over Microphones.
+The Microphone Control Service is probably the simplest service in all of the Bluetooth LE
+Audio specifications, comprising a single characteristic, which provides a device-wide mute
+for microphones. Its simplest usage is shown in Figure 10.5.
+
+Figure 10.5 Simple usage of the Microphone Control Service
+
+There is no corresponding Control Point characteristic. Instead, the Mute Characteristic can
+be written directly, as well as being read and notified. It has the same features as the Mute
+field in the AICS Audio State characteristic, namely:
+
+262
+
+Chapter 10 - Volume, Audio Input and Microphone Control
+Description
+
+Value
+
+Not Muted
+Muted
+Mute Disabled
+
+0x00
+0x01
+0x02
+
+Figure 10.6 MICS Mute characteristic values
+
+If control of microphone gain is required, MICS should be combined with an instance of AICS
+(Figure 10.7), although in this case MICS doesn’t provide much advantage over just using
+AICS. However, most Clients would expect to use MICS to perform a microphone mute,
+hence the reason to retain MICS.
+
+Figure 10.7 Combining the Microphone Service with an instance of the Audio Input Control Service
+
+The combination of AICS and MICS makes more sense when multiple microphones exist, as
+MICS gives the benefit of a device-wide mute for the microphones, which is its real purpose
+(Figure 10.8).
+
+Figure 10.8 Use of MICS and AICS for multiple microphones
+
+Finally, MICS can be used to provide a device-wide mute for multiple microphones as part of
+a volume control scheme, as shown in Figure 10.9. However, in most cases, multiple
+microphones in an Acceptor will be feeding directly into audio processing modules, so it’s
+unlikely that external control of individual microphones would be a requirement. What Figure
+10.9 does is demonstrate the flexibility of volume and microphone control services in
+Bluetooth LE Audio.
+
+263
+
+Section 10.7 - A codicil on terminology
+
+Figure 10.9 The combination of Microphone and Volume Control Services
+
+10.7
+
+A codicil on terminology
+
+Throughout this book I’ve been using the terms Initiator, Acceptor and Commander to
+describe the three main types of device in the Bluetooth LE Audio ecosystem. As I stated in
+Chapter 3, these terms are defined as roles in CAP, so purists would probably object to my
+conflating them with devices. I still feel that conflation leads to a clearer understanding of
+how everything fits together.
+Potential for confusion exists in the way a Commander (as a device) interacts with Initiators
+and Acceptors. As a role, a Commander can be collocated48 with an Initiator, but as a device,
+its interactions with an Initiator and each of its Acceptors can be confusing. Although all of
+these interactions are covered by CAP procedures, they are independent. Figure 10.10
+illustrates some of the profiles and services described in this chapter in a simple case of an
+Initiator, Acceptor and independent Commander.
+
+The spelling (or misspelling) of colocation and collocation can add further confusion. These are
+not alternative spellings, but totally different words. Colocation, with one “l” means “in the same
+place”, deriving from the Latin “locare”. In contrast, collocation, with two “l”s means “working
+together”, coming from the Latin “collegium”.
+48
+
+264
+
+Chapter 10 - Volume, Audio Input and Microphone Control
+
+Figure 10.10 An example of Profile and Server relationships for volume and control
+
+In this case, three profiles are implemented in the Commander. Two of them – Call Control
+and Media Control act on the complementary services in the Initiator, whereas Volume
+Control operates on the VCS, VOCS and AICS instances in the Acceptor. For the
+grammatically inclined, the profiles in the Commander and services in the Initiator are
+colocated; the services in the Acceptor are collocated.
+That brings us to the end of the GAF specifications. The only other specifications within
+Bluetooth LE Audio are the top level profiles, which we are about to come to.
+
+265
+
+Section 10.7 - A codicil on terminology
+
+266
+
+Chapter 11 - Top level Bluetooth® LE Audio profiles
+
+Chapter 11.
+
+Top level Bluetooth® LE Audio profiles
+
+Having covered everything in the Generic Audio Framework, we now come to the top level
+profiles of Bluetooth LE Audio. Although they’re still called profiles, in most cases they’re a
+lot simpler than the underlying profiles we’ve discussed, or the Bluetooth Classic Audio
+profiles. Instead of defining procedures, they generally confine themselves to configuration,
+specifying new roles which mandate a combination of optional features, and adding QoS
+requirements beyond those of BAP. In doing so they raise the bar by defining feature
+combinations to meet commonly experienced use cases.
+In this chapter we will look at what these profiles contain. As they rely on features which are
+already defined in the GAF specifications, they do not have complementary service
+specifications. HAP – the Hearing Access Profile49 is the exception, as the corresponding
+Hearing Access Service introduces a new Presets characteristic for Hearing Aids. TMAP –
+the Telephony Media and Audio Profile has a nominal TMAS service, which is included in the
+TMAP specification. The Public Broadcast Profile has no service. That’s an anomaly that is
+inherent with a broadcast application which does not expect a connection between Initiator
+and Acceptor. The lack of connection implies no possibility of a Client-Server relationship,
+hence no opportunity for a Service specification.
+A number of top level profiles are currently being developed in the Bluetooth working groups,
+but the first three scheduled for adoption are:
+•
+•
+
+•
+
+HAP and HAS – the Hearing Access Profile and Service, which define
+requirements for products which are intended for use in the hearing aid ecosystem.
+TMAP – the Telephony and Media Audio Profile. This aims to cover the main
+features of the classic HFP and A2DP profiles, adding in new capabilities arising
+from the Bluetooth LE Audio topologies.
+PBP – the Public Broadcast Profile, which is a foundation of the Bluetooth LE
+Audio Sharing use case (see Chapter 12). It identifies that a broadcast Audio
+Stream can be received by any Bluetooth LE Audio Acceptor.
+
+These are publicly available in draft form, but are still subject to change. The contents of this
+chapter reflect their state at the time of writing, which is documented in Section 13.2.2.
+
+The Hearing Access Profile and Service were previously called the Hearing Aid Profile and Service,
+but the names were changed as the term “Hearing Aid” has a specific, regulated medical meaning in
+some countries, so a declaration that a product complied with a Hearing Aid specification could cause
+confusion.
+49
+
+267
+
+Section 11.1 - HAPS the Hearing Access Profile and Service
+As the whole of the Bluetooth LE Audio development was started by the hearing aid industry,
+it seems only fair to start with HAP and HAS.
+
+11.1
+
+HAPS the Hearing Access Profile and Service
+
+The Hearing Access Profile and Service have been designed to meet the requirements of
+devices which are used in the Hearing Aid ecosystem, which covers hearing aids, products
+which supply Audio Streams to hearing aids and accessories which are used to control them.
+Hearing Aids are different to earbuds and other Acceptors because they are always on,
+capturing and processing ambient sound to assist their wearers. That means that all of the use
+cases driving the Hearing Access profile envisage Bluetooth LE Audio as an additional audio
+stream to the ambient one. This makes them unusual, as Bluetooth connectivity is not the de
+facto reason for buying them. They also differ in that everyone who wears them has hearing
+loss, so they have a different balance in requirements between audio quality (and hence QoS
+settings) and battery life. Taking your hearing aid out to recharge it during the day is a far
+more significant action than it is for an earbud wearer, as you may not be able to hear during
+that time. Pushing up audio quality increases the power consumption, so the Hearing Access
+profile imposes no additional QoS requirements over the settings mandated by BAP, using
+the 16_2 LC3 codec settings for voice (16kHz sampling rate, 7 kHz bandwidth) and the 24_2
+codec settings for music (24 kHz sampling rate, 11 kHz bandwidth). As many hearing aids do
+not occlude the ear, ambient sounds can be heard in addition to the Bluetooth stream. For
+this reason, hearing aids generally prefer to use the Low Latency QoS settings to avoid echo
+between the ambient and transmitted sound.
+The HAP specification describes the physical device configuration of products recognised as
+hearing aids. The profile defines four different configurations of Acceptor(s), all of which are
+included in the general term “hearing aid” through the document. These are:
+•
+•
+•
+
+•
+
+268
+
+A single hearing aid, which renders a single Bluetooth LE Audio Stream,
+A single hearing aid that receives separate left and right Audio Streams and
+combines the decoded audio into a single Audio Channel,
+A pair of hearing aids (also called a Binaural Hearing Aid Set) which are the
+members of a Coordinated Set, supporting individual left and right Audio Channels
+for each ear, and
+A hearing aid that uses a single Bluetooth link but provides separate left and right
+audio outputs to each ear. These are called Banded Hearing Aids, where there is a
+wired connection between the hearing aid device in each ear and the Bluetooth
+transceiver, which is typically in a neckband or an over-the-head band.
+
+Chapter 11 - Top level Bluetooth® LE Audio profiles
+The Hearing Access Profile defines four different roles for the hearing aid ecosystem:
+•
+•
+
+•
+•
+
+HA (Hearing Aid), which is any one of the four types of hearing aid described
+above.
+HAUC (Hearing Aid Unicast Client), which is an Initiator which can establish a
+unicast Audio Stream with a hearing aid. The unicast Audio Streams can be
+unidirectional or bidirectional.
+HABS (Hearing Aid Broadcast Sender), which is an Initiator transmitting broadcast
+Audio Streams, and
+HARC (Hearing Aid Remote Controller), which is a device that controls volume
+levels, mute states and hearing aid presets (which we’ll come to in a minute).
+
+The bulk of the Hearing Access Profile lists mandatory combinations of BAP and CAP
+features which are required for different product iterations. Whilst many are obvious, such as
+setting the CSIS Size characteristic to 2 for a pair of hearing aids, they ensure that any products
+claiming support for the profile will contain the same features and be interoperable. This is
+essentially the point of Bluetooth LE Audio top level profiles – ensuring that interoperability
+is assured for specific use cases by clearly specifying which underlying profiles, services and
+features must be present.
+There are limitations. Every hearing aid must be able to receive a Unicast or Broadcast Audio
+Stream that complies to the BAP mandated requirements, but does not need to receive one
+which has been encoded with other optional QoS settings. It must accept volume control
+commands from any device which supports the HARC role, which could be a stand-alone
+remote control or an application on a phone. For the first time, this means that hearing aid
+accessories will be globally interoperable with any make of hearing aid.
+All hearing aids need to support the «Ringtone», «Conversational» and «Media» Context Types,
+which mean that they will all support incoming phone calls. However, they do not need to
+support Call Control or return a voice stream when in a phone call. That’s a practical decision,
+as many hearing aids locate their microphones for best pickup of sound around the user, rather
+than the wearer’s voice, so it will often be better for a user to use the phone’s microphone
+when in a call. These are limitations which manufacturers will need to convey in their product
+marketing.
+HAP and HAS introduce a new concept, which is support for Presets. Presets are proprietary
+audio processing configurations which hearing aid manufacturers include to optimise the
+sound fed to the ear. Typically, they adjust the processing to cope with different environments,
+such as restaurants, shops, office, home, etc. HAP and HAS don’t attempt to standardise
+these settings, but provide a numbering scheme which manufacturers can map to their specific
+implementation. Users can then select a specific preset by its number, or cycle through them.
+It includes the ability to add Friendly Names, so that an application can display the current
+presets and other available presets. It also supports dynamic presets, where the availability of
+269
+
+Section 11.1 - HAPS the Hearing Access Profile and Service
+a preset may change depending on the status of the hearing aid. For example, a preset that
+was optimised for reception of a telecoil signal would not be available when there is no telecoil
+loop available. Presets are currently a feature which is specific to hearing aids. For more
+information on them, refer to the HAPS specifications.
+Within the preset feature of HAS, there is an interesting option, which is likely to expand into
+other profiles. This is the ability to use a non-Bluetooth radio to relay a command from one
+hearing aid to the other without the need for a Commander to institute a procedure to other
+set members based on a notification from one of them. This is the Synchronized Locally
+operation, which informs a Commander that any preset command sent to one hearing aid can
+be locally relayed to the other member of the set, which is described in Sections 3.2.2.8 –
+3.2.2.8 of HAS.
+The other enhancement in HAP is the option to include the Immediate Alert Service [HAP
+3.5.3]. This allows a device which does not support an Audio Stream to provide an alert to
+the Hearing Aid, which it can render as an alert to the user. No specific use cases are defined,
+but it is suggested that manufacturers might want to include this capability in products such
+as doorbells or microwave ovens to help inform hearing aid wearers of something they need
+to respond to.
+The aim of the HAP requirements is to support all of the use cases envisaged for hearing aids,
+which are illustrated in Figure 11.1.
+
+Figure 11.1 The four main use cases for HAP, showing the roles used in each
+
+As many of the hearing aid use cases will involve the user hearing the ambient sound as well
+as the received Bluetooth LE Audio Stream, it is expected that devices will favour the use of
+Low Latency QoS modes. To help ensure low latency, there is a requirement that all devices
+270
+
+Chapter 11 - Top level Bluetooth® LE Audio profiles
+in the HA role shall support a Presentation Delay value of 20ms for receive and transmit for
+Low Latency QoS modes.
+A final nuance in the Hearing Access Profile is a set of requirements in Section 4.1, which
+dictate Link Layer setting when the HAUC is using a 7.5ms Isochronous Interval and the
+hearing aid does not support the reception of 7.5ms LC3 codec frames. It is expected that
+this will be an edge case in situations where a hearing aid has a minimal LC3 implementation
+(as support for 7.5ms codec frames is not mandated) and where an Initiator is forced to use a
+7.5ms interval because of timing limitations of its other peripheral devices, which cannot
+accommodate the preferred 10ms interval. In this case, the scheduler should ensure the
+selection of the specified parameters for ISOAL to avoid segmentation of SDUs and a
+potential increase in the frame loss rate. It is anticipated that Initiators will move to the
+optimum 10ms transport interval as Bluetooth LE Audio becomes common, hence this is a
+short-term fix.
+All HAUCs, HAs and HABs must support the 2M PHY. Although its use is not mandated,
+it is highly recommended to conserve airtime, and hence battery life.
+
+11.2
+
+TMAP – The Telephony and Media Audio Profile
+
+Like HAP, the TMAP profile is predominantly a list of additional requirements beyond those
+mandated in BAP and CAP to emulate the use cases of the HFP and A2DP profiles. TMAP
+does not introduce any new behaviour or characteristics, so incorporates a minimal TMAS
+section within TMAP.
+Because it is bundling both telephony and media applications into one document, TMAP feels
+like a portmanteau profile, where implementers have a wide-ranging ability to pick and choose
+what they support. This can lead to some oddities, as many of its features are optional. In
+theory, depending on what you pick, it is possible to make a TMAP compliant device which
+supports telephone calls, but does not support music and vice versa. That’s a logical
+consequence of bundling so many different use cases together. It makes sense that a soundbar
+or speaker doesn’t support telephony, but that means that headphones don’t need to support
+telephony either. Instead, it is up to the implementer to choose the features and roles which
+are appropriate and let market Darwinism take care of any inauspicious choices.
+To try and prevent any surprises, the TMAS element of TMAP includes a TMAP Role
+characteristic [TMAP 4.7], which exposes the specific roles which a device supports. This
+allows Acceptors to determine the Roles that an Initiator supports, which is required for some
+use cases. Which brings us to the Roles. TMAP does not define any new Roles for a
+Commander, but defines three pairs of Roles for an Initiator and an Acceptor.
+
+271
+
+Section 11.2 - TMAP – The Telephony and Media Audio Profile
+
+11.2.1
+
+Telephony Roles – Call Gateway and Call Terminal
+
+For telephony, TMAP defines a Call Gateway (CG) and Call Terminal (CT) Role. The Call
+Gateway is the device which connects to the telephony network, such as a phone, tablet or
+PC, and is an Initiator. A Call Terminal is typically a headset, but can also be an extension
+speaker, or a microphone for a conference phone. Some common configurations are shown
+in Figure 11.2
+
+Figure 11.2 Typical configurations for the CG and CT roles
+
+Devices supporting the CG and CT Roles must support the higher codec settings of 32 kHz
+sampling at both 7.5ms and 10ms frame rates (32_1 and 32_2 from BAP), with the Low
+Latency settings. These correspond to the same audio quality as the Superwideband EVS
+speech codec that the 3GPP specifies for 5G phones, ensuring no loss of quality from local
+microphone to remote headset and vice versa.
+A CG must support the CCP Server role, but does not mandate any further features above
+those mandated in CCP.
+
+11.2.2
+
+Media Player Roles – Unicast Media Sender and Unicast Media
+Receiver
+
+Media Player use cases employ the Unicast Media Sender (UMS) and Unicast Media Receiver
+(UMR) Roles. The Unicast Media Sender is the Initiator, which is the audio source and the
+Unicast Media Receiver an Acceptor. Typical configurations are shown in Figure 11.3.
+
+Figure 11.3 Typical TMAP Unicast Media applications
+
+272
+
+Chapter 11 - Top level Bluetooth® LE Audio profiles
+For Unicast Media, TMAP pushes up the audio quality, requiring a Unicast Media Receiver to
+support all six of the 48kHz sampling codec configurations, form 48_1 to 48_6. Unicast Media
+Sources must support the 48_2 codec setting and at least one of the 48_4 or 48-6 settings. All
+TMAP roles must support the 2M PHY, which will almost certainly be necessary to find
+enough airtime for these configurations. Although support for these QoS configurations is
+required, it is up to the application to decide how it configures the ASEs. It can decide to use
+lower values. For UMS, the 32 kHz sampling rates remain optional, although it is unlikely that
+an Acceptor would not support them. A user is unlikely to hear the difference between 32
+and 48kHz, so the choice of setting is largely a marketing decision.
+TMAP does not mandate the support of any Context Type other than «Unspecified». If a
+Unicast Media Receiver wants to reject any attempt to establish a stream from a Call Gateway,
+it should support «Ringtone», but set it to not available.
+TMAP increases the requirement for media control by mandating support for Play and Pause
+opcodes. A UMS must also support the Media Control Point characteristic.
+
+11.2.3
+
+Broadcast Media Roles – Broadcast Media Sender and Broadcast
+Media Receiver
+
+The final two sets of Roles in TMAP are the Broadcast Media Sender and Broadcast Media
+Receiver roles. Unsurprisingly, these are designed for broadcast applications. Within TMAP
+they are generally envisaged to be personal applications, where higher audio quality is required,
+but may also be used in public broadcast applications, such as cinemas. Typical use cases are
+shown in Figure 11.4.
+
+Figure 11.4 Typical use cases for the Broadcast Media Sender and Receiver roles
+
+TMAP adds very few requirements on top of BAP and CAP for the broadcast Roles. It
+mandates support for higher quality QoS settings, requiring support for all of the Low Latency
+and High Reliability 48 kHz QoS modes defined in table 6.1 of BAP (48_1_1 to 48_6_1 and
+48_1_2 to 48_6_2) for a Broadcast Media Receiver and both 48_1 and 48_2 QoS
+configurations for a Broadcast Media Sender. It also mandates that a Broadcast Media sender
+273
+
+Section 11.3 - Public Broadcast Profile
+must support at least one of the 48_3 or 48_5 (7.5ms frame) codec configurations and one of
+the 48_4 or 48_6 (10ms frame) codec configurations.
+TMAP requires that Broadcast Media Receivers support a Presentation Delay value of 20ms
+within their range of Presentation Delays for both Low Latency and High Reliability QoS
+modes.
+TMAP also increases the mandatory support for broadcast Audio Configurations, requiring
+that a Broadcast Receiver can accept any of the broadcast Audio Configurations defined in
+Table 4.24 of BAP. These are shown in Table 11.1. The BMS requirements are unchanged
+from BAP.
+Audio
+Configuration
+
+Stream Direction
+(Initiator – Acceptor)
+
+BMS
+
+BMR
+
+12
+
+M
+
+M
+
+13
+
+M
+
+M
+
+14
+
+O
+
+M
+
+Table 11.1 Audio Configuration requirements for Broadcast Media Roles
+
+11.3
+
+Public Broadcast Profile
+
+The Public Broadcast Profile (PBP) is a simple, but very interesting profile. It is intended to
+support the Audio Sharing use case, providing a guarantee that a broadcast Audio Stream is
+configured such that any Acceptor can receive it. It contains a new UUID – the Public
+Broadcast Service UUID, which is added alongside a Basic Audio Announcement as an
+additional Service Type. This means that it appears in an Extended Advertisement, allowing
+a device to read it without having to synchronise to the Periodic Advertising train and then
+receive and parse the BASE. This reduces the amount of work for the scanner, reducing its
+power consumption. Essentially it acts as a filter which a Broadcast Sink or Broadcast
+Assistant can use to limit its scanning, reducing the power and time required to identify
+Broadcast Sources which they know the Broadcast Sink can decode.
+PBP defines three roles – the Public Broadcast Source (PBS), Public Broadcast Sink (PBK)
+and Public Broadcast Assistant (PBA). The only requirement on the PBK and PBA roles is
+that they are able to recognise and interpret the PBP UUID in an Extended Advertisement.
+A Public Broadcast Source may only include the PBP UUID in its Basic Audio Announcement
+if at least one of the BISes included in the associated BIG has been encoded using one of
+mandatory QoS settings defined in BAP for a Broadcast Sink, i.e., 16_2_1, 16_2_2, 24_2_1,
+274
+
+Chapter 11 - Top level Bluetooth® LE Audio profiles
+or 24_2_2. The BIG may contain BISes encoded at other QoS settings, but the PBP UUID
+can only be transmitted when a BIS with these mandatory settings is active.
+The reason for PBP is to alert Bluetooth LE Audio Sinks to a broadcast Audio Stream which
+can be universally received.
+That concludes our coverage of the Bluetooth LE Audio specifications. The final chapter will
+look at some of the new applications they enable.
+
+275
+
+Section 11.3 - Public Broadcast Profile
+
+276
+
+Chapter 12 - Bluetooth® LE Audio applications
+
+Chapter 12.
+
+Bluetooth® LE Audio applications
+
+The whole point of developing Bluetooth LE Audio was to support new audio applications,
+not just to produce a slightly lower power alternative to the existing Bluetooth Classic Audio
+profiles. Part of that was driven by the need to catch up with proprietary extensions,
+particularly for True Wireless Stereo. The bigger prize was to allow innovation in audio, to
+keep up the momentum which TWS and voice assistants had generated, making audio more
+universal and allowing greater flexibility in how we can listen to it.
+Broadcast is the child of the telecoil. The telecoil system has been around a long time. It
+performs well, but it is remarkably basic. It is mono, has a very limited audio bandwidth, and
+you can only hear it if you're located within the confines of the inductive loop. Whilst the aim
+of Bluetooth technology was to provide a more capable successor, it very quickly became
+obvious we could expand significantly on the user experience of telecoil.
+An initial concern with replicating the telecoil experience is that a Bluetooth transmission is
+not confined by its installation area. Being wireless, it penetrates walls, meaning that people
+in adjoining rooms and spaces can also hear any broadcast audio. In many situations, such as
+public halls and places of worship, that's not a significant problem, because the only source of
+broadcast audio will be the one that is relevant for that particular venue. However, in other
+situations, such as TVs in hotel rooms, or systems within conference centres with multiple
+meeting rooms, that becomes a major issue, as broadcasts will overlap, with the result that
+people will have difficulty in understanding which broadcast is the one they want to connect
+to, and potentially hearing things which they shouldn’t.
+That led to the implementation of encryption within a broadcast stream, so that only a user
+with the correct Broadcast_Code to decode that stream is able to listen to it. Although the
+broadcast streams overlap, the Broadcast_Code provides an access mechanism where only
+authorised listeners can decode a particular broadcast Audio Stream. That’s akin to how WiFi works today, where users are given an SSID name to identify a particular Wi-Fi network,
+and then need to key in an access code to be allowed to connect to it. Whilst people have
+become used to doing that with Wi-Fi, it’s a pretty basic and often frustrating user experience.
+Everyone wanted Bluetooth LE Audio to do it better.
+Tackling that needed a better mechanism for a user to access the code than a scrap of paper
+on a coffee shop table or a notice on the wall. It led to the development of the Broadcast
+Assistant and the Commander, providing ways for that information to be sent from a
+Broadcaster that is trusted or known to an individual listener. As the specifications developed,
+it quickly became obvious that this provided a very powerful mechanism for users, both to
+pick up broadcast codes, and gain access to individual broadcasts in a far more flexible manner
+than we have ever seen with previous Bluetooth connections.
+
+277
+
+Section 12.1 - Changing the way we acquire and consume audio
+
+12.1
+
+Changing the way we acquire and consume audio
+
+Changing the way we acquire and consume audio are important points for developers to
+understand. For most of the last two decades, the evolution of personal audio has been driven
+by mobile phones. Initially, we stored files that we’d acquired from music sharing networks
+on our phones. More recently, with the growth of audio streaming services, the phone became
+the router supplying music to our ears on demand. However, that experience largely depended
+on an interaction with the phone to select what we wanted to hear. With the features of
+Bluetooth LE Audio, and the predicted availability of multiple Bluetooth LE Audio sources,
+that is going to change. Those sources may be personal, in the case of our phones, laptops
+and TVs; public, such as a broadcast transmitter that's installed in a restaurant, pub, gym or
+office; or commercial, as in a cinema or when we’re listening to a live performance.
+Commanders mean that we will be able to do far more in terms of selection and control
+without touching our phones. It will be interesting to see what effect that has on the devices
+we carry with us and wear. If we can do more without touching our phones, their importance
+as the central device for much of our communication may wane. In turn, as voice becomes
+natural as a means of command, new applications may develop which use audio as their
+primary interface, with no need to interact with a screen. There is a long way to go, but
+smartphones were never going to be the final step in the evolution of our personal
+communications – some other product will emerge, as smartphones did themselves. The new
+capabilities that Bluetooth LE Audio brings to the equation, allowing greater ubiquity for voice
+and audio, may hasten that change.
+The ability to use multiple Commanders means that wearable devices gain increased utility.
+Whilst not all of that will relate to audio, it increases the reasons for purchasing and wearing
+them. That will help the wearables industry as a whole, which is still struggling to reach the
+volume or customer interest that they had hoped for.
+All of these changes will take time. Some will happen earlier, others later. Some may come in
+implementations which fail to convince users, but then get successfully reinvented a few years
+later. That’s been the case for most new personal technologies. What is clear, is that they will
+change the dynamics of the audio ecosystem, giving far more opportunity for audio
+manufacturers to innovate and chip away at the de facto status of smartphones. In this
+chapter, we’ll explore some of the different scenarios for Bluetooth LE Audio, look at the
+opportunities for new ways of presenting audio sharing services, determine what is needed to
+enable them, and assess what this may mean, both for phone design and for the design of
+other products.
+The Bluetooth LE Audio specifications will not, by themselves, transform the way we use
+audio. To be effective, many parts of the industry – hardware developers, apps developers,
+service providers and UX designers need to understand the possibilities and take on board
+how to make these new audio experiences compelling, letting users discover how audio can fit
+278
+
+Chapter 12 - Bluetooth® LE Audio applications
+in with their lives in new ways, and then build on that to deliver new ways of using audio.
+
+12.2
+
+Broadcast for all
+
+There is no question that both users and venues who currently use telecoil will welcome the
+advent of Bluetooth LE Audio as the next step in hearing reinforcement. The quality is
+considerably higher and installation costs should be significantly lower. For hearing aid users,
+it promises a much wider range of places where they can connect for hearing reinforcement.
+For building owners and architects, it should mean that hearing reinforcement becomes a
+standard feature of new buildings. (It’s a sad fact that the lack of telecoil in many new buildings
+has more to do with a lack of understanding from architects than the cost of the current
+technology.)
+This should all happen naturally, as broadcast products become available, helped by an
+ongoing promotion of Bluetooth LE Audio by the hearing aid industry. However, users will
+need new hearing aids that contain Bluetooth LE Audio to pick up these broadcasts. Currently
+only a small percentage of hearing aids contain any form of non-proprietary Bluetooth wireless
+technology, and it will take time for a critical mass of users to emerge. Compared to consumer
+volumes, the hearing aid market is relatively small – it will probably take at least five years to
+attain ten million users of new Bluetooth LE Audio hearing aids, not least because hearing
+aids need medical approval before they can be sold, which slows down time to market. This
+raises the interesting question of whether consumer products will drive the initial roll-out of
+audio broadcast infrastructure, which will benefit everyone, regardless of whether they have
+hearing loss.
+
+12.2.1
+
+Democratising sound reinforcement
+
+Today, providers of public broadcast/telecoil services are thinking predominantly about
+people wearing hearing aids when they design their products. These services most commonly
+replicate and reinforce an audio announcement with low latency. Very few people are thinking
+about whether this will be relevant to people without hearing loss, and how they can be
+extended. The likelihood is that many users of earbuds and headphones would find them
+beneficial. If that is going to happen, then most people, certainly in the short term, are likely
+to find and select them using apps on their smartphones. So, the first step will be for the
+phone operating systems to expose the information about these broadcasts, allowing users to
+select and listen to them. In other words, we need to see app interfaces which look something
+like those in Figure 12.1, mimicking what is already done today for discovering and pairing to
+Wi-Fi and Bluetooth devices.
+
+279
+
+Section 12.2 - Broadcast for all
+
+Figure 12.1 Simple user interface to find Bluetooth® LE Audio Broadcasts
+
+Although this is a format that most users will be familiar with, it’s still more difficult than it
+needs to be. The example of Figure 12.1 illustrates the fact that one of the first applications
+is likely to be in public transport. Many bus stops have announcement boards, but unlike train
+platforms, don’t make public audio announcements because they are intrusive on a street.
+National guidelines in many countries already require telecoil to be incorporated in new public
+infrastructure, and are likely to include Bluetooth LE Audio in their future recommendations.
+As these appear, good transport apps should start to include support for them, so that travel
+apps automatically detect the presence of an appropriate audio stream and ask the user if they
+wish to listen to it, saving them the inconvenience of going into their Bluetooth settings to
+search for it.
+There are several tools within the Bluetooth LE Audio specifications which help the
+development of a good user experience for integrated applications. Looking back at the
+chapters on broadcast, there are a number of pieces of information which should be included
+in the broadcast advertisements, to help a scanning device filter and sort different broadcasts.
+These are:
+•
+
+•
+
+•
+
+280
+
+The Local Name AD Type, which provides the device name. Normally this would
+be the primary item of information displayed in a list of available broadcast
+transmitters.
+The Program_Info LTV. An LTV structure which includes information about the
+content which is being broadcast. It’s similar to the top level of information that is
+included in a TV’s Electronic Program Guide, e.g., the TV channel.
+The BASE, which may contain further information about the content, including the
+name of the content, e.g., the specific program name for a TV broadcast, as well as
+the language of transmission.
+
+Chapter 12 - Bluetooth® LE Audio applications
+With this information, applications which are scanning on behalf of your earbuds or hearing
+aids can obtain a lot of information to direct them if they want to start embedding broadcast
+audio into their user experience. Equally, the providers of broadcast audio information need
+to think carefully about how to use it and how best to support applications. The primary use
+of most public broadcast will remain sound reinforcement, where the broadcast audio is
+designed to be low latency, so that it can be received and rendered alongside ambient audio.
+It’s now possible to include multiple languages for those announcements, but they should be
+scheduled so they don’t conflict with an ambient announcement in another language,
+otherwise they will be more difficult to understand. It’s a small, but obvious nuance, but one
+of many that developers will need to learn.
+Broadcast stream providers also need to think about the information they use to tag the
+streams. As you move from a bus stop onto a bus, your application should track the loss of
+the bus stop as a Broadcast Source and switch to the Broadcast Source on the bus you’re
+travelling on. In somewhere like London, with a high density of buses, it’s quite possible that
+you may be next to another bus on the same route, but travelling in the opposite direction, so
+an application should contain enough basic intelligence to detect that it has connected to the
+right transmitter. As long as the Local Device names and ProgramInfo values are sensibly
+ascribed, it should be straightforward to determine that. However, this needs service
+providers, application developers and equipment manufacturers to work together to give their
+users the best experience.
+Most public broadcast streams will be mono voice streams, which can be adequately encoded
+using the 16_2_2 QoS settings, using very little airtime. The messages are likely to be
+infrequent – in the example of a bus, either the announcement of the arrival of a bus, or, when
+you’re on it, an announcement of the next stop. If your phone has sufficient resources, a
+travel app should be able to monitor the appropriate BIS, mixing it in to any other Audio
+Stream at the point where it detects the BIS contains an audio message and ignoring it when
+there are only null packets.
+
+Figure 12.2 Mixing broadcast announcements into an audio stream
+
+281
+
+Section 12.2 - Broadcast for all
+Figure 12.2 shows the basic setup, where a phone is running both a music streaming app and
+a travel app, which is monitoring the local broadcast transmitters for relevant audio
+announcements. Figure 12.3 shows how the phone would combine audio from the different
+sources into a single stream which is encoded and sent in a single CIS. It makes the point that
+an Initiator is free to mix whatever audio channels it wants to place in a stream to send to the
+earbuds.
+
+Figure 12.3 Mixing application specific audio streams into a CIS
+
+Given that people are starting to leave their earbuds in for longer, using the transparency mode
+to hold conservations, this is a useful way of receiving relevant announcements, particularly if
+they are “silent” foreign language announcements.
+
+12.2.2
+
+Audio augmented reality – the “whisper in your ear”
+
+The example above of mixing broadcast travel announcements into other information
+generated by a travel app is the entry point into using broadcast audio as an element of
+augmented reality. The initial promises of augmented reality and virtual reality have proven
+to be rather disappointing, with few successful applications outside specialised business ones
+and products for stay-at-home gamers. Audio may offer a more acceptable entry step,
+particularly for everyday augmented reality.
+The advantage audio brings is that it is far less obtrusive. It’s the little whisper in the ear that
+helps you do whatever else you’re doing, with no need to purchase or wear what are often
+inconvenient wearable tech products. We are already seeing multi-axis sensors incorporated
+into earbuds which can detect your head position, and hence know where you’re looking.
+Combined with information from broadcast transmitters and spatially aware applications,
+these allow some innovative new audio applications where sound can start to merge into
+your everyday experience, without the need for any special AR or VR hardware. Directions
+can tell you which way to turn and where to look. A lot of the underlying technology for
+these already exists. It is just waiting for the potential of Bluetooth LE Audio to make it
+possible.
+
+12.2.3
+
+Bringing silence back to coffee shops.
+
+The examples above show how existing telecoil applications can be expanded to a wider
+audience and integrated into today’s smartphone apps. There is likely to be a parallel evolution
+of broadcast infrastructure in new areas. One that is getting a lot of attention is employing
+broadcast to provide background music in cafes and restaurants.
+282
+
+Chapter 12 - Bluetooth® LE Audio applications
+At some point in the past, somebody though it was a good idea to equip most cafes, bars and
+restaurants with background music. The design chic of the past decade has largely been to
+remove any furnishings that act as noise absorbers, making conversation increasingly difficult
+and raising the background noise level. That generates a positive feedback loop where
+customers start talking more loudly, so the staff turn the music up, resulting in customers
+having to raise their voices and turning what should be an enjoyable experience into a far less
+pleasant one. Restaurant reviews now routinely include a measure of noise to show whether
+it’s possible to hold a conservation. Despite negative consumer feedback, the music remains.
+It would be nice if Bluetooth LE Audio could effect a change.
+Parts of the hospitality industry believe that there is an obvious place for Bluetooth LE Audio
+to replace loudspeakers, providing music for those who want to listen and reducing the noise
+levels for those who want to talk.
+The addition of hardware can be extremely simple. All it needs is a Bluetooth LE Audio
+broadcasting module fitted to the output of an existing audio system. Chip vendors are already
+developing reference designs for units like this, and within twelve months of the specifications
+being adopted, devices like this should be readily available for a relatively low price. Product
+designers should make sure that they can be easily configured by the venue owners, so that
+they are easy for customers to discover.
+12.2.3.1
+
+Making audio universally available
+
+This is where the Public Broadcast Profile becomes important. A commercial Bluetooth LE
+Audio transmitter should allow the selection of any of the QoS settings which are defined in
+the BAP specification [BAP Table 6.4]. However, not all Acceptors will be able to decode all
+of them. Resource constrained devices, such as hearing aids, which want to achieve the
+greatest possible battery life, will probably not support decoding for sampling rates above 24
+kHz. So, if a broadcast transmitter only supports a higher rate, wearers will be unable to hear
+the broadcast audio. There are some simple solutions to this, but they require broadcast device
+manufacturers to implement them.
+The simplest approach is to confine transmission of music in a public space to the 24_2_1
+QoS setting. If there is any background noise, i.e., it is less that a perfect listening environment,
+most users are likely to find this satisfactory.
+
+283
+
+Section 12.2 - Broadcast for all
+Sampling Rate
+24 kHz (24_2_1)
+32 kHz (32_2_1)
+48 kHz (48_2_1)
+
+Mono
+airtime
+
+Stereo Airtime
+
+Universally receivable?
+
+13%
+15%
+30%
+
+26%
+30%
+60%
+
+Yes
+No
+No
+
+Note: These figures are based on the Low Latency settings. The 48kHz sampling figures have a higher airtime
+requirement, as the configuration requires a RTN setting of 4, rather than the value of 2 which is used for the
+24_2_1 and 32_2_1 settings.
+
+Table 12.1 Airtime requirements for different broadcast stream QoS settings
+
+Table 12.1 provides the data for this discussion. Given current marketing trends in the audio
+industry, it’s likely that many manufacturers will want to be able to promote their products as
+“48kHz audio quality”, which poses a dilemma. If these transmitters support a stereo 48kHz
+stream, yet still want to provide universal support, they need to add at least a 24kHz mono
+stream. If that were included within the same BIG, it would take the airtime to 90%, because
+although it uses less airtime, every BIS in a BIG is allocated the same timing parameters. The
+alternative is to transmit a second, separate BIG for the 24kHz mono stream where the
+combined BIGs would account for around 72% of airtime, but would require more advertising
+resource. If a broadcast transmitter uses Wi-Fi to obtain its audio stream, that is probably not
+going to work. A more practical solution, which addresses the quality and universal access
+requirements would be to transmit 32kHz stereo and 24kHz mono Audio Streams in one BIG,
+which only consumes around 45% of the total airtime. Transmitting both a stereo 24kHz pair
+and a 32kHz pair of audio streams in the same BIG takes the same airtime as a single 48kHz
+stereo stream.
+In the examples above, I have chosen the Low Latency QoS settings, despite that fact that
+they include fewer retransmissions for the 24kHz and 32kHz sampling rates. In this particular
+application, latency isn’t that important, as there are usually no visual cues, or simultaneous
+ambient sound (although there could be). However, in this application, the broadcast
+transmitter is likely to be placed high on a wall or attached to the ceiling to get the best
+coverage, so there’s not much in the way to absorb the transmissions. That should mean that
+the Low Latency settings are adequate. If the café owner located the broadcast transmitter in
+a cupboard or under the serving counter, that would be different. This is something that
+equipment designers need to think about in their physical design and installation instructions,
+ensuring that broadcast transmitters for this application can easily be wall mounted and that
+any leads attached to a unit are long enough to do that.
+Equipment manufacturers need to understand these constraints and design products with the
+flexibility to support multiple BIGs and allow the installer and venue owner to position and
+configure them appropriately. We should not expect café owners to understand airtime or
+QoS parameters – manufacturers should provide them with solutions which are easy to install
+and use. That includes the ability to customise all of the AD Type names and LTV strings
+with values which are appropriate for each installation.
+284
+
+Chapter 12 - Bluetooth® LE Audio applications
+A basic broadcast transmitter which just accepts an audio input, either from a 3.5mm jack,
+RCA plug or a USB input is about as simple as you can get, but will be adequate for very many
+premises and applications. Equipment vendors will almost certainly provide more complex
+broadcast units for larger establishments, as well as managed services to configure them,
+supply the audio stream and integrate them into other audio services. Even with basic
+Bluetooth LE Audio broadcast, there are plenty of opportunities for differentiation.
+
+12.3
+
+TVs and broadcast
+
+From an early point in the Bluetooth LE Audio development, the TV industry became very
+excited about the possibility of connecting TVs to earbuds, headphones and hearing aids.
+Although TVs can use unicast, there is an obvious limitation to its scalability, as it requires
+separate CISes for every listener. Hence, the consensus is that most TV usage will use
+broadcast, adding encryption to ensure that only authorised users can decode the audio
+content. In this section, we’ll look at the requirements for the three most common application
+areas.
+
+12.3.1
+
+Public TV
+
+Today, a large number of publicly installed TVs are silent. It’s not because they have no audio
+output, but because they are installed in areas where ambient audio would be annoying or
+conflict with other audio. Common examples are airports, where operators don’t want users
+to be distracted from being able to hear flight and security announcements; gyms, where
+multiple TVs display different channels, but are silent, as multiple conflicting audio from them
+would be cacophonous; pubs and bars, where the sound from even one TV (and there are
+often multiple TVs with different channels), would disrupt normal conversation, and outdoor
+installations, where the sound would not be heard or be intrusive
+These installations are all akin to the simple Broadcaster that we looked at above. They don’t
+need encryption; they just need a broadcast transmitter. That could be a similar plug-in device
+attached to their TV’s audio outputs, or a built-in Bluetooth LE Audio broadcast transmitter.
+It needs to allow a device location to be entered so that a user can match an audio broadcast
+with the TV, or it can take the Wi-Fi route of sticking its access information on a notice on
+the wall.
+As we’ve seen above, many of these devices, including those integrated into TVs, will probably
+be managed, not least to allow specific messages to be mixed into the audio stream. For
+example, in an airport, flight announcements would probably be mixed into an audio stream,
+so that someone listening to a TV would never miss them. This is where we will probably see
+the start of multiple language streams. If multiple language soundtracks are available for a TV
+broadcast, they can be sent in separate BISes on one BIG. Using a 24kHz mono coding, it
+should be possible to provide six different language streams from one broadcast transmitter.
+Each input stream would mix the flight announcements in real-time with the audio input for
+that program, with any unannounced languages being mixed at a subsequent point where they
+285
+
+Section 12.3 - TVs and broadcast
+don’t conflict with any ambient announcements. If a transmitter needed to support more
+channels, it could add a second Bluetooth transmitter. Each language is identified by a
+Language LTV value in the BASE, indicating the language of each BIS. A user can set their
+scanning application to select specific languages, so that these are presented preferentially.
+Alternatively, it can show all of the language variants which are available and let the user
+choose the one they want to hear. This requires the phone’s APIs to expose this information
+and application developers to understand how to use it.
+
+12.3.2
+
+Personal – at home
+
+Personal TV has different priorities. Many TVs already support Bluetooth Classic Audio
+profiles, but that is generally limited to connecting to a soundbar, a single set of earbuds or a
+headphone. The reality is that in most homes, multiple members of the family or friends will
+be listening. Although TV manufacturers could do a like-for-like transition by moving the
+Bluetooth LE Audio unicast, that will quickly hit an airtime limit of a few users. It makes far
+more sense to move to broadcast.
+Unlike the public TV cases described above, most users would not want their neighbours to
+hear what they are listening to, so for personal TVs (and other audio sources in the home), the
+expectation is that broadcast Audio Streams would always be encrypted. That means that
+there needs to be a simple mechanism for the user to obtain the Broadcast_Code to decrypt
+the Audio Streams.
+This is where the Broadcast Assistant comes in. Personal audio sources, which expect to be
+heard regularly by multiple users, would expect each user to pair and bond with them. That
+provides a long-term trusted connection. When a user comes within range of the TV, they
+would ask their earbuds to connect and the Broadcast Assistant in the TV would inform them
+of the broadcast stream location using PAST, followed by the current Broadcast_Code. This
+is an important point - once a user has paired, they no longer need to use their phone to
+connect to a broadcast. When the user comes within range of the TV, the basic LE connection
+can be automatic. A pair of earbuds can be informed of that connection with an internally
+generated audio message, alerting the user to the presence of audio source. If they want to
+connect, they press a button or perform whatever gesture the earbuds require, allowing them
+to receive the information they need from the TV to start audio streaming, with no further
+action from the user. For users walking between rooms with different audio sources, it’s an
+elegant way for them to connect to the nearest source. Notifications are provided to all
+Broadcast Assistants informing them that the earbuds are dropping synchronisation or
+synchronising to a new source. It means that every connected device which is involved keeps
+track of the current status, making it possible to engineer a very elegant solution.
+
+286
+
+Chapter 12 - Bluetooth® LE Audio applications
+If friends come around and want to connect with their earbuds or hearing aids, the TV owner
+would put their TV into a Bluetooth Audio Sharing mode, (which would ideally be a button
+on their TV remote control, not a menu item buried many layers down on the TV). That
+would activate the advertising train associated with the broadcast stream, which would provide
+a method for the friend to obtain the Broadcast_Code.
+
+Figure 12.4 An example of how to add a friend to a TV broadcast
+
+The method of obtaining the Broadcast_Code is not described in the Bluetooth LE Audio
+specifications, as it is out of scope. It could be discovered by displaying a number on the TV
+screen for a user to enter into a phone app, displaying a QR code, or the use of NFC or
+another proximity technology. The market will probably agree on a standard option in the
+near future.
+It is expected that Broadcast_Code values would be refreshed at the end of each session, to
+ensure ongoing privacy. In practice that is likely to happen automatically at each power-down.
+Bonded users would be unaffected, as they would automatically be issued with the new value
+each time they connect.
+
+12.3.3
+
+Hotels
+
+Anyone who has ever been disturbed during their stay in a hotel by the TV in the neighbouring
+room being far too loud will welcome the advent of Bluetooth LE Audio in TVs, allowing the
+occupants to use them silently. It’s an ideal application, but carries the same problem of
+Bluetooth transmissions penetrating through walls, floors and ceilings, exposing what you are
+watching to all of your neighbours.
+The same approach can be used as in personal TVs, where a user is given the Broadcast_Code
+whenever they want to connect. Equally, the hotel could rely on a user having a generalpurpose broadcast scanning app which lets them enter a pin-code from the TV, or scan a QR
+code, with their phone acting as a Commander for their earbuds. Although these work, they
+are cumbersome and don’t give a particularly smooth user experience. A more effective option
+287
+
+Section 12.4 - Phones and broadcast
+would be to incorporate the Broadcast_Code provision into the hotel app, so that as soon as
+someone checks in, their phone would be provided with the correct information to access the
+TV from that app, with their personal Broadcast_Code static for the duration of their stay.
+That’s a fairly simple extension of the TV management systems which are already integrated
+in most hotel TVs.
+
+12.4
+
+Phones and broadcast
+
+In the early days of Bluetooth LE Audio development, broadcast was mainly seen as an
+infrastructure application, broadcasting to large numbers of people. As the potential of the
+technology became better understood, there was growing excitement about what it offered for
+smartphones. What excited people most was the ability to share music on their phone with
+their friends.
+The use case is a simple one. It has been around since we first started storing music in personal
+devices. Back in the days of Sony’s original Walkman, you’d see pairs of people sharing their
+earbuds to listen together to the music. As we’ve progressed from portable music players to
+streaming music on our smartphones, the technology for sharing has not advanced. In
+contrast, it’s become more difficult with the removal of the 3.5mm audio jack from
+smartphones, as the simple sharing of wired earbuds is no longer possible with many handsets.
+Bluetooth LE Audio broadcast solves that problem. If you’re listening to your favourite music
+and your friends come along, you can ask them to join, transition from a unicast stream (or
+from an A2DP stream – we don’t know what underlying technology will be used) to a
+broadcast stream. That sets your phone up as a Broadcast Source which your friends can find.
+Unless you want to be a public Broadcaster, your application will almost certainly want to
+encrypt your audio, so your phone application needs to distribute the Broadcast_Code. This
+can be accomplished through a short-term pairing (without a requirement to bond). That has
+the advantage that your friend’s phone will notify yours once it acquires the broadcast stream.
+Your phone can then stop transmitting the advertising set, so that nobody else is aware of
+your broadcasts. If other friends want to join in, you just repeat the process. At any point in
+the process, your friends can leave. Once they have all left, your phone will probably revert
+back to unicast, as the acknowledged packets in CISes means it is more power efficient for an
+Initiator – a broadcast transmitter has to transmit every possible retransmission Subevent.
+Equally, any of your friends can take the role of Broadcaster to share their favourite music.
+It’s a process which can continue for as long as any of them want. As applications develop,
+we may find ways to make swapping the Broadcaster within a group even more seamless.
+In some cases, the stream doesn’t need to be encrypted. Anyone wanting to set up an ad-hoc
+silent disco just needs to start broadcasting.
+Phones can even take advantage of Bluetooth LE Audio without being phones. Many people
+with hearing loss like to use table microphones to help pick up voices in a meeting or around
+288
+
+Chapter 12 - Bluetooth® LE Audio applications
+the dinner table. Once your phone has Bluetooth LE Audio, an application could set it up as
+a private Broadcaster, broadcasting its microphone pickup to everyone else around the table.
+It uses no cellular functionality, but acts as a local community microphone.
+
+12.5
+
+Audio Sharing
+
+All of these use cases have the same thing in common – they’re evolving the telecoil
+experience, which was a hearing aid only experience, to everybody who owns a Bluetooth LE
+hearable, whether that’s a hearing aid, headphone, set of earbuds, or even a portable speaker.
+That’s a new concept for the vast majority of people. The Bluetooth SIG, working with
+hearing loss user groups, as well as hearing aid and consumer hearables manufacturers, is
+developing guidelines for manufacturers and venues to promote the universal nature of
+broadcast under the name “Audio Sharing” to help encourage its uptake and ensure
+interoperability for all users.
+The interoperability issue is an important one for broadcasts as there are two conflicting
+categories of devices:
+•
+
+•
+
+Hearing aids and resource-constrained hearables, generally supporting the Hearing
+Access profile with the ability to support only the 16kHz and 24kHz sampling rates
+for LC3, and
+Consumer devices which may want to use the highest LC3 sampling rates of
+48kHz.
+
+To cope with this discrepancy, and ensure the optimum experience for all users, the Audio
+Sharing guidance separates broadcast transmitters into two categories – Public and Personal.
+•
+
+•
+
+Public broadcast transmitters are ones which everyone should be able to access.
+These must transmit a stream at the lower sampling rates of either 16kHz or 24kHz,
+supporting the Public Broadcast Profile to inform users where and when those
+audio streams are available. A user detecting the Public Broadcast Service UUID
+will know that they can synchronise to that stream. If they have the resources,
+public devices can also transmit higher quality streams at the same time.
+Personal broadcast transmitters are devices like TVs, phones, PCs and tablets. The
+Audio Sharing guidelines recognise that the user may normally prefer to receive a
+higher quality stream, which would not be universally interoperable. However,
+when a user selects an Audio Sharing compliant broadcast mode on their device,
+such as when a friend asks to access your TV or listen to your music stream, it must
+be possible to configure it to a universal setting, which, if selected would broadcast
+a 16kHz or 24kHz sampled, PBP compliant Audio Stream, so that any device could
+pick it up. It is up to the user to ask the other stream recipients which option they
+need. Again, if the device is capable, it can simultaneously transmit streams with
+higher sampling rates.
+289
+
+Section 12.6 - Personal communication
+The Bluetooth SIG plans to promote these guidelines in a similar manner to the way that the
+Wi-Fi Alliance has for publicly accessible Wi-Fi access points, to help encourage
+interoperability and confidence in the new Audio Sharing ecosystem.
+
+12.6
+
+Personal communication
+
+Whilst all of the broadcast applications described above can be accessed by individuals, most
+are likely to be used by groups of people, and the interface designs for them should be designed
+with that in mind. However, there are applications which will be predominantly designed for
+individual conversations. Once again, these mimic many of the telecoil applications, where a
+small inductive loop is used to provide an audio feed for a single person. These are typically
+found at ticket desks, taxis, banks, supermarket checkouts and hotel receptions. These use a
+static microphone to pick up the hearing aid wearer’s voice, and a telecoil loop to return an
+audio stream from the other person to their hearing aid. What is important in this application
+is that the broadcast audio stream is private and never mixed up with another one nearby.
+While the conversation can always be heard by someone standing nearby, you do not want it
+to be picked up by someone several metres away. Nor do you want the wrong stream to be
+picked up. So, once again, encryption and authentication are necessary.
+The techniques to find the Broadcast Source and to obtain the Broadcast_Code are the same
+as the ones described above, but these particular use cases are generating interest in developing
+an industry-wide method to connect.
+
+12.6.1
+
+Tap 2 Hear – making it simple
+
+Although nothing has yet been specified, industry interest is focused on the use of NFC to
+provide a simple “Tap 2 Hear” interface, with the user tapping their phone, or any Commander
+device, onto a touchpad. That contact will transfer the information that the hearing aids or
+earbuds need to find the correct Broadcast Source and the appropriate Broadcast_Code. As
+soon as that is done, the conversation can begin, with the Broadcast_Code and BIG details
+remaining static for the duration of the session. It provides an attractive and simple user
+experience, and is applicable for all types of personal connection, whether at a supermarket
+checkout, accessing the right conference meeting room, or even connecting to the broadcast
+stream in a theatre or cinema. It is equally applicable to Audio Sharing, where friends would
+just need to tap your phone to share music, or to gain access to a Private TV.
+
+12.6.2
+
+Wearables take control
+
+Although the Bluetooth Classic Audio profiles allow for remote control on separate devices,
+there has been very little use of these capabilities. These have mainly been limited to carkit
+functionality, with an occasional foray into wearable devices. With Bluetooth LE Audio, that
+is likely to change. There are a number of reasons for that.
+One of the main drivers for remote controls is the continuing growth of earbuds. As the
+hearing aid industry discovered several decades ago, adding user controls to something as small
+290
+
+Chapter 12 - Bluetooth® LE Audio applications
+as an earbud or hearing aid is not easy, either for the manufacturers, or the customer. As a
+result, customers have been encouraged to use phone apps to control their audio. However,
+constantly taking your phone out and opening the appropriate app is far from being the best
+user experience. The control features within the Bluetooth LE Audio specifications make it
+much easier to incorporate remote controls into other devices and to have as many of them
+as a user wants. These are low power products which can easily run off coin cells, so can be
+low cost, and will be interoperable, allowing any Bluetooth product to integrate the features.
+As a result, we expect to see a trend for wearable devices to implement this functionality,
+whether that’s wristband, smart watch, glasses or clothing. It may be a feature which increases
+the usefulness of wearables for many users, bringing life back to what has become a relatively
+moribund product sector.
+
+12.6.3
+
+Battery boxes become more important than phones
+
+It’s worth adding a few words on the humble battery charger box for earbuds. It’s a necessary
+accessory which was developed alongside earbuds to give them the semblance of a useable
+battery life. The first generation of earbuds mostly struggled to run for an hour before they
+needed recharging. The battery box was a neat idea, both to hold them safely, but also to
+recharge them constantly, giving users the perception that they would run for a day. Since
+those early products, the battery life of earbuds has improved significantly, with models
+claiming up to 10 hours of continuous music (albeit with processing features like Automatic
+Noise Cancelling turned off). Bluetooth LE Audio will increase the basic battery life, although
+manufacturers will probably take the opportunity to add more features which suck up power.
+Regardless of that, battery boxes are here to stay, not least because they provide a very
+important function alongside their charging capability, which is a container to stop you losing
+your earbuds.
+Many battery boxes already contain a Bluetooth chip to help with pairing, and to provide an
+audio stream in situations where one doesn’t exist, such as on an aircraft. Here, the battery
+box can plug into the 3.5mm jack provided by the aircraft’s personal entertainment system,
+transmitting sound to your earbuds. Bluetooth LE Audio’s lower power and lower latency
+make it the ideal choice to replace Bluetooth Classic Audio in these applications. However,
+the battery box is also ideal for many remote control functions. Unlike earbuds and hearing
+aids, it is big enough for buttons, so is ideal as an easily accessible volume controller. It can
+also act as a Broadcast Assistant, scanning for available Broadcasts. In general, it will provide
+a much faster and easier interface than getting your phone out and opening an app, because
+the Bluetooth LE Audio control functionality is always there as buttons.
+In designing Bluetooth LE Audio Controllers, it is a useful exercise to think about how you
+would implement them on a battery box. It’s such an easy thing for users to interact with, but
+because of its inherent simplicity, it’s often overlooked. It won’t necessarily be the best device
+to implement them on, but its accessibility and small size, fitting into pockets that won’t hold
+a phone, make it an attractive alternative.
+291
+
+Section 12.7 - Market development and notes for developers
+As designers find ways to turn small devices like this into compelling user interfaces, we will
+probably find that we spend less time interacting with our phone. As earbuds get better at
+mixing Bluetooth audio streams with ambient sound and conversations (which audio
+processing can enhance), we will probably spend more time wearing our earbuds, talking to all
+sorts of things around us and interacting with audio. Phones are not going to go away any
+time soon, although we have already passed peak smartphone, but Bluetooth LE Audio may
+drive the applications that lead us to whatever comes next.
+
+12.7
+
+Market development and notes for developers
+
+There’s a simplistic view that broadcast infrastructure in buildings, theatres, hotels and cafes
+will happen because the cost of a Bluetooth LE Audio broadcast devices will be cheap, so
+people will buy them from eBay and Amazon, plug them in to their existing audio system and
+put up a Bluetooth Audio Sharing sign. That will certainly happen. It’s what happened in the
+early days of Wi-Fi, when venue owners did exactly that. With Bluetooth LE Audio, it should
+be easier, as you don’t need to install an internet connection, you just need a cable to attach it
+to your existing audio system.
+However, with Wi-Fi, many venue owners discovered that it was easier to work with a service
+provider, who could install it, manage it and provide support for the venue and the customers.
+The same model is likely to appear with Bluetooth LE Audio. I suspect that we will see WiFi Access Point providers selling access points which include Bluetooth LE Audio, so that a
+single device can manage both. There are similar opportunities for the companies currently
+making and installing telecoil loops, where they will see their business open up to a far wider
+range of customers. It will need apps developers to be aware of how broadcast works to
+provide the user interfaces.
+On which front, it is important to point out that these applications will only happen when
+silicon, operating systems and application developers understand the full potential of
+Bluetooth LE Audio and include support for the features which enable these different use
+cases. In the early days, some of these may not be possible, as developers naturally concentrate
+on releasing what they consider to be the most obvious applications. Over time, as the market
+learns, and we get a critical mass of products, the more complex use cases will emerge,
+hopefully with simple and intuitive user interfaces.
+
+12.7.1
+
+Combining Bluetooth Classic Audio with Bluetooth LE Audio
+
+Bluetooth Classic audio implementations will not disappear in the short term. The majority
+of audio products in the market today use pre-5.2 chipsets which aren’t upgradeable, and many
+products, from TVs to cars, have a life of ten years or more. For at least the next five years,
+phones and most earbuds will support both Bluetooth Classic Audio and Bluetooth LE Audio.
+Where both support the same use case, such as with HFP and A2DP, it will be up to the
+implementation to decide which to use. In phones, that will largely be down to the individual
+manufacturer and the silicon implementation. Products with a wider variety of stacks and
+292
+
+Chapter 12 - Bluetooth® LE Audio applications
+open-source applications may be more diverse.
+The important point for developers is to make sure it works for the user. Whilst CAP includes
+handover procedures for moving from unicast to broadcast, these only apply where both are
+Bluetooth LE Audio. In the real world, it’s equally possible that the transition may be from
+A2DP to LE broadcast and then back to LE unicast or classic Bluetooth. For product
+developers, the rule must be to anticipate and allow any combination. The Bluetooth SIG
+runs regular unplugfests where developers can test their products with those from other
+manufacturers. As products start to incorporate more and more of the new features of
+Bluetooth LE Audio, these will be invaluable to ensure an interoperable ecosystem and a
+degree of uniformity in user experience.
+
+12.7.2
+
+Concentrate on understanding broadcast
+
+One of the learnings from developing the specification is how difficult the concepts of
+broadcast are for anyone who is used to the peer-to-peer model of HFP and A2DP. Whilst
+unicast includes a lot of new concepts (see Chapter 3 to review them), the acknowledged
+packet model in a piconet is relatively familiar. Broadcast, with a transmitter that is unaware
+of whether anyone is listening to it, and receivers which act totally independently and have no
+state machine, have been surprisingly challenging for many to take in. Adding BASS, alongside
+the Broadcast Assistant and the delegation of scanning, does seem to be a challenge to that
+orthodox thinking.
+The examples in this chapter try to illustrate the flexibility which is allowed. Developers need
+to understand the limitations that are imposed by broadcast – that each BIG has a fixed
+structure for all BISes, which may mean there are times when multiple BIGs are more efficient.
+Scanning and filtering broadcasts is all important for a good user experience, so the relevant
+fields need to be supported, and where appropriate, configurable by users. They must
+understand the Basic Audio Announcement, the use of Targeted and General Announcements
+by Initiators and Acceptors and the meaning of the BASE structure. The interaction between
+Broadcast Assistants and Scan Delegators and the use of the characteristics in BASS are
+fundamental to the sharing and authentication processes described above.
+Finally, anyone designing with Bluetooth LE Audio should take time to understand and
+implement the control features and appreciate the fact that they can be distributed between
+multiple different devices. These are all simple Client-Server relationships, but the contents
+of each of the services are very comprehensive and make the difference between a rich or a
+basic user experience.
+Although it is a basic LE feature, developers should also make sure they understand the role
+of notifications. In Bluetooth LE Audio, notifications are the means whereby Servers ensure
+that everything within the topology is up to date. Knowing when to expect them, and what
+to do if they don’t arrive when expected (which is normally to go and read the characteristic)
+is vital to provide a robust ecosystem, where a user can pick and choose multiple products
+293
+
+Section 12.7 - Market development and notes for developers
+from different manufacturers and have confidence that they will all work with each other.
+
+12.7.3
+
+Balancing quality and application
+
+Developers should not forget the difference between Optional and Mandatory in the
+Bluetooth LE Audio specifications. In many of them, very little is mandated. Many features
+are mandated to be supported, such as the mandated codec settings in BAP and TMAP, but
+that doesn’t mean that an application needs to use them. The mandate is that if an application
+chooses to use them, they must be supported. If they’re only optional and an application on
+an Initiator wants to use them, the Acceptor can reject that request. It means that there is a
+lot of flexibility in what a product can do.
+Having said, that, going off-piste may impinge on interoperability. As an example, the QoS
+settings in BAP have been thoroughly tested and developers can be sure that every silicon
+supplier will have made sure their implementation is interoperable, so they should be used.
+However, there will be times when physics gets in the way, typically if you want to transmit
+multiple Audio Streams, where you will begin to run out of airtime. We know from a history
+of audio development that consumers don’t necessarily want the highest quality – they want
+ease of use. CDs and streaming services became successful because they were easy to use, and
+the audio quality was adequate. The companies that invented them understood that by being
+brave and taking a step back from the audio quality treadmill, they could provide an experience
+that allowed them to win customers’ hearts. Bluetooth LE Audio has potential for new,
+compelling services, which should be designed with that in mind. The audio quality is there
+when it is needed, but so is a lot else.
+
+12.7.4
+
+Reinventing audio
+
+In the last few decades, the way we consume audio has become concentrated into the hands
+of phone manufacturers and streaming services. What actually renders it, notwithstanding the
+inordinate success of Apple’s Airpods and their competitors, is largely the tail of the dog.
+Hearables have added neat features, but they haven’t really played any major part in the
+development of the audio chain – they remain a passive peripheral to the phone. Bluetooth
+LE Audio’s new topologies and distributed control features provide the potential for a new
+generation of innovation, where devices we have yet to imagine become an important part of
+our lives. At that point the hearables’ tail could start wagging the dog. The aim of everyone
+involved in the Bluetooth LE Audio development has been to provide the tools for a new
+generation of innovation. I hope this book has inspired you to think outside the box and
+consider how that can be achieved.
+
+294
+
+Chapter 13 - Glossary and concordances
+
+Chapter 13.
+
+Glossary and concordances
+
+In this final chapter, I’ve listed all of the specifications which comprise Bluetooth® LE
+Audio, the acronyms used in this book, along with a list of all of the Bluetooth LE Audio
+procedures and sub-procedures, along with where they are defined. I can only hope to
+provide an overview of the specifications within this book. These cross-references should
+help you navigate your way through the specifications.
+
+13.1
+
+Abbreviations and initialisms
+
+The following abbreviations and initialisms are used in this book and throughout the
+Bluetooth LE Audio specifications.
+Acronym /
+Initialism
+
+Meaning
+
+3GPP
+A2DP
+ACAD
+ACL
+ACL
+AD
+AdvA
+AICS
+ASCS
+ASE
+ATT
+AVDTP
+AVRCP
+BAP
+BAPS
+BASE
+BASS
+BFI
+BIG
+BIS
+BMR
+BMS
+BN
+BR/EDR
+BW
+CAP
+CAS
+
+Third Generation Partnership Project
+Advanced Audio Distribution Profile
+Additional Controller Advertising Data
+Asynchronous Connection-oriented link
+Asynchronous Connectionless
+Advertising Data
+Advertiser Address
+Audio Input Control Service
+Audio Stream Control Service
+Audio Stream Endpoint
+Attribute Protocol
+Audio Video Distribution Transport Protocol
+Audio Video Remote Control Profile
+Basic Audio Profile
+The set of BAP, ASCS, BASS and PACS
+Broadcast Audio Source Endpoint
+Broadcast Audio Scan Service
+Bad Frame Indication
+Broadcast Isochronous Group
+Broadcast Isochronous Stream
+Broadcast Media Receiver
+Broadcast Media Sender
+Burst Number
+Basic Rate/Enhanced Data Rate
+Bandwidth
+Common Audio Profile
+Common Audio Service
+295
+
+Section 13.1 - Abbreviations and initialisms
+Acronym /
+Initialism
+
+Meaning
+
+CCID
+CCP
+CG
+CIE
+CIG
+CIS
+CMAC
+CRC
+CSIP
+CSIPS
+CSIS
+CSS
+CT
+CTKD
+CVSD
+DCT
+DECT
+DSP
+DUN
+EA
+EATT
+EIR
+FB
+FT
+GAF
+GAP
+GATT
+GC
+GMCS
+GPS
+GSS
+GTBS
+HA
+HAP
+HARC
+HAS
+HAUC
+HCI
+
+Content Control Identifier
+Call Control Profile
+Call Gateway
+Close Isochronous Event
+Connected Isochronous Group
+Connected Isochronous Stream
+Cipher-based Message Authentication Code
+Cyclic Redundancy Check
+Coordinated Set Identification Profile
+The set of CSIP and CSIS
+Coordinated Set Identification Service
+Core Specification Supplement
+Call Terminal
+Cross-Transport Key Derivation
+Continuous Variable Slope Decode
+Discrete Cosine Transform
+Digital Enhanced Cordless Telecommunications
+Digital Signal Processor
+Dial-up Networking
+Extended Advertisement
+Enhanced ATT
+Extended Inquiry Response
+Full Band (20 kHz audio bandwidth)
+Flush Timeout
+Generic Audio Framework
+Generic Access Profile
+Generic Attribute Profile
+Group Count
+Generic Media Control Service
+Global Positioning System
+GATT Specification Supplement
+Generic Telephone Bearer Service
+Hearing Aid (as in the role described in the Hearing Access Profile)
+Hearing Access Profile
+Hearing Aid Remote Controller
+Hearing Access Service
+Hearing Aid Unicast Client
+Host Controller Interface
+
+296
+
+Chapter 13 - Glossary and concordances
+Acronym /
+Initialism
+
+Meaning
+
+HDMI
+HFP
+HQA
+IA
+IAC
+IAS
+INAP
+IRC
+IRK
+ksps
+L2CAP
+LC3
+LD-MDCT
+LE
+LL
+LSB
+LSO
+LTK
+LTPF
+LTV
+MAC
+MCP
+MCS
+MDCT
+MEMS
+MIC
+MICP
+MICS
+MSB
+mSBC
+MSO
+MTU
+NB
+NESN
+NPI
+NSE
+OOB
+OTP
+
+High-Definition Media Interface
+Hands-Free Profile
+High Quality Audio
+Identity Address
+Immediate Alert Client
+Immediate Alert Service
+Immediate Need for Audio related Peripheral
+Immediate Repeat Count
+Identity Resolving Key
+kilo samples per second
+Logical Link Control and Adaption Protocol
+Low Complexity Communication Codec
+Low Delay Modified Discrete Cosine Transform
+Low Energy (as in Bluetooth Low Energy)
+Link Layer
+Least Significant Bit
+Least Significant Octet
+Long Term Key
+Long Term Postfilter
+Length | Type | Value
+Message Authentication Code
+Media Control Profile
+Media Control Service
+Modified Discrete Cosine Transform
+Microelectromechanical System
+Message Integrity Check
+Microphone Control Profile
+Microphone Control Service
+Most Significant Bit
+Modified SBC (codec)
+Most Significant Octet
+Maximum Transmission Unit
+Narrow Band (4 kHz audio bandwidth)
+Next Expected Sequence Number
+Null Payload Indicator
+Number of Subevents
+Out of Band
+Object Transfer Profile
+297
+
+Section 13.1 - Abbreviations and initialisms
+Acronym /
+Initialism
+
+Meaning
+
+OTS
+PA
+PAC
+PACS
+PAST
+PBA
+PBAS
+PBK
+PBP
+PBS
+PCM
+PDU
+PHY
+PLC
+PSM
+PSM
+PTO
+QoS
+RAP
+RFU
+RSI
+RTN
+SBC
+SDP
+SDU
+SIRK
+SM
+SN
+SNS
+SSWB
+SWB
+TBS
+TMAP
+TMAS
+TNS
+UCI
+UI
+uint48
+
+Object Transfer Service
+Periodic Advertisement
+Published Audio Capabilities
+Published Audio Capabilities Service
+Periodic Advertising Synchronization Transfer
+Public Broadcast Assistant
+Public Broadcast Audio Stream Announcement
+Public Broadcast Sink
+Public Broadcast Profile
+Public Broadcast Source
+Pulse Code Modulation
+Protocol Data Unit
+Physical Layer
+Packet Loss Concealment
+Protocol Service Multiplexer
+Protocol/Service Multiplexer
+Pre-Transmission Offset
+Quality of Service
+Ready for Audio related Peripheral
+Reserved for Future Use
+Resolvable Set Identifier
+Retransmission Number
+low-complexity Sub Band Codec
+Service Discovery Protocol
+Service Data Unit
+Set Identity Resolving Key
+Security Manager
+Sequence Number
+Spectral Noise Shaping
+Semi Super Wide Band (12 kHz audio bandwidth)
+Super Wide Band (16 kHz audio bandwidth)
+Telephone Bearer Service
+Telephony and Media Audio Profile
+Telephony and Media Audio Service
+Temporal Noise Shaping
+Uniform Caller Identifier
+User Interface
+unsigned 48-bit integer
+
+298
+
+Chapter 13 - Glossary and concordances
+Acronym /
+Initialism
+
+Meaning
+
+UMR
+UMS
+URI
+URL
+UTF
+UUID
+UX
+VCP
+VCS
+VOCS
+VoIP
+WB
+
+Unicast Media Receiver
+Unicast Media Sender
+Uniform Resource Identifier
+Uniform Resource Locator
+Unicode Transformation Format
+Universally Unique Identifier
+User Experience
+Volume Control Profile
+Volume Control Service
+Volume Offset Control Service
+Voice over IP
+Wide Band (8 kHz audio bandwidth)
+
+Table 13.1 Acronyms and initialisms
+
+13.2
+
+Bluetooth LE Audio specifications
+
+The following specifications are part of the new specifications developed for Bluetooth LE
+Audio. They build on new features which are present in the Core version 5.2 and above.
+
+13.2.1
+
+Adopted Specifications
+
+These specifications have been adopted and implementations can be qualified to them
+Specification
+
+Full Name
+
+Family
+
+AICS
+ASCS
+BAP
+BASS
+CCP
+CSIP
+CSIS
+GMCS
+
+Audio Input Control Service
+Audio Stream Control Service
+Basic Audio Profile
+Broadcast Audio Scan Service
+Call Control Profile
+Coordinated Set Identification Profile
+Coordinated Set Identification Service
+Generic Media Control Service (part of
+MCS)
+Generic Telephone Bearer Service (part of
+TBS)
+Low Complexity Communication Codec
+Media Control Profile
+Media Control Service
+Microphone Control Profile
+Microphone Control Service
+
+Generic Audio Framework
+Generic Audio Framework
+Generic Audio Framework
+Generic Audio Framework
+Generic Audio Framework
+Generic Audio Framework
+Generic Audio Framework
+Generic Audio Framework
+
+GTBS
+LC3
+MCP
+MCS
+MICP
+MICS
+
+Generic Audio Framework
+Codec
+Generic Audio Framework
+Generic Audio Framework
+Generic Audio Framework
+Generic Audio Framework
+299
+
+Section 13.3 - Procedures in Bluetooth LE Audio
+Specification
+
+Full Name
+
+Family
+
+PACS
+TBS
+
+Published Audio Capabilities Service
+Telephone Bearer Service
+
+Generic Audio Framework
+Generic Audio Framework
+
+Table 13.2 Adopted Bluetooth® LE Audio Specifications
+
+13.2.2
+
+Draft specifications
+
+These specification are undergoing interoperability testing. Draft versions have been
+publicly released to help implementers understand the contents, but are subject to change.
+The text in this edition is based on the version shown in Table 13.3.
+Specification Version
+
+Full Name
+
+Family
+
+CAP
+CAS
+HAP
+HAS
+PBP
+TMAP
+TMAS
+
+Common Audio Profile
+Common Audio Service
+Hearing Access Profile
+Hearing Access Service
+Public Broadcast Profile
+Telephony and Media Audio Profile
+Telephony and Media Audio Service
+(part of TMAP)
+
+GAF
+GAF
+Top level profile
+Top level profile
+Top level profile
+Top level profile
+GAF
+
+VS_r06
+VS_r07
+v09
+v09
+VS_r00
+VSr02
+
+Table 13.3 Draft Bluetooth® LE Audio Specifications
+
+13.3
+
+Procedures in Bluetooth LE Audio
+
+The following procedures are defined in the Bluetooth LE Audio specifications. Note that
+there are some cases where procedures have the same name. However, these are different
+procedures which are context sensitive.
+Procedure or sub-procedure name
+
+Specification
+
+Section
+
+Answer Incoming Call
+ASE Control operations
+ASE_ID discovery
+Audio capability discovery
+Audio data path removal
+Audio data path setup
+Audio role discovery
+Available Audio Contexts discovery
+Broadcast Audio Stream configuration
+Broadcast Audio Stream disable
+Broadcast Audio Stream establishment
+Broadcast Audio Stream Metadata update
+Broadcast Audio Stream reconfiguration
+
+CCP
+BAP
+BAP
+BAP
+BAP
+BAP
+BAP
+BAP
+BAP
+BAP
+BAP
+BAP
+BAP
+
+4.4.13.1
+5.6
+5.3
+5.2
+5.6.6.1
+5.6.3.1
+5.1
+5.4
+6.3
+6.3.3
+6.3.2
+6.3.3
+6.3.1
+
+300
+
+Chapter 13 - Glossary and concordances
+Procedure or sub-procedure name
+
+Specification
+
+Section
+
+Broadcast Audio Stream release
+Broadcast Audio Stream state management
+Call Control Point Procedures
+CIS loss
+Codec configuration
+Configure Audio Input Description Notifications
+Configure Audio Input State Notifications
+Configure Audio Input Status Notifications
+Configure Audio Location Notifications
+Configure Audio Output Description Notifications
+Configure Mute Notifications
+Configure Volume Flags Notifications
+Configure Volume Offset State Notifications
+Configure Volume State Notifications
+Coordinated Set Discovery procedure
+Disabling an ASE
+Enabling an ASE
+Fast Forward Fast Rewind
+Join Calls
+Lock Release procedure
+Lock Request procedure
+Move Call To Local Hold
+Move Locally And Remotely Held Call To Remotely
+Held Call
+Move Locally Held Call To Active Call
+Move to First Group
+Move to First Segment
+Move to First Track
+Move to Group Number
+Move to Last Group
+Move to Last Segment
+Move to Last Track
+Move to Next Group
+Move to Next Segment
+Move to Next Track
+Move to Previous Group
+Move to Previous Segment
+Move to Previous Track
+Move to Segment Number
+
+BAP
+BAP
+CCP
+BAP
+BAP
+VCP
+VCP
+VCP
+VCP
+VCP
+MICP
+VCP
+VCP
+VCP
+CSIP
+BAP
+BAP
+MCP
+CCP
+CSIP
+CSIP
+CCP
+CCP
+
+6.3.5
+6.2
+4.4.13
+5.6.8
+5.6.1
+4.4.3.8
+4.4.3.1
+4.4.3.5
+4.4.2.3
+4.4.2.7
+4.4.1
+4.4.1.3
+4.4.2.1
+4.4.4.1
+4.6.1
+5.6.5
+5.6.3
+4.5.25
+4.4.13.7
+4.6.4
+4.6.3
+4.4.13.3
+4.4.13.5
+
+CCP
+MCP
+MCP
+MCP
+MCP
+MCP
+MCP
+MCP
+MCP
+MCP
+MCP
+MCP
+MCP
+MCP
+MCP
+
+4.4.13.4
+4.5.39
+4.5.29
+4.5.34
+4.5.41
+4.5.40
+4.5.30
+4.5.35
+4.5.38
+4.5.28
+4.5.33
+4.5.37
+4.5.27
+4.5.32
+4.5.31
+301
+
+Section 13.3 - Procedures in Bluetooth LE Audio
+Procedure or sub-procedure name
+
+Specification
+
+Section
+
+Move to Track Number
+Mute
+Mute
+Ordered Access procedure
+Originate Call
+Pause Current Track
+Play Current Track
+QoS configuration
+Read Audio Input Description
+Read Audio Input State
+Read Audio Input Status
+Read Audio Input Type
+Read Audio Location
+Read Audio Output Description
+Read Bearer List Current Calls
+Read Bearer Provider Name
+Read Bearer Signal Strength
+Read Bearer Signal Strength Reporting Interval
+Read Bearer Technology
+Read Bearer UCI
+Read Bearer URI Schemes Supported List
+Read Call Control Point Optional Opcodes
+Read Call Friendly Name
+Read Call State
+Read Content Control ID
+Read Content Control ID
+Read Current Track Object Information
+Read Current Track Segments Object Information
+Read Gain Setting Properties
+Read Incoming Call
+Read Incoming Call Target Bearer URI
+Read Media Control Point Opcodes Supported
+Read Media Information
+Read Media Player Icon Object Information
+Read Media State
+Read Mute
+Read Next Track Object Information
+Read Parent Group Object Information
+Read Playback Speed
+
+MCP
+VCP
+VCP
+CSIP
+CCP
+MCP
+MCP
+BAP
+VCP
+VCP
+VCP
+VCP
+VCP
+VCP
+CCP
+CCP
+CCP
+CCP
+CCP
+CCP
+CCP
+CCP
+CCP
+CCP
+MCP
+CCP
+MCP
+MCP
+VCP
+CCP
+CCP
+MCP
+MCP
+MCP
+MCP
+MICP
+MCP
+MCP
+MCP
+
+4.5.36
+4.4.1.6.7
+4.4.3.7.3
+4.6.5
+4.4.13.6
+4.5.24
+4.5.23
+5.6.2
+4.4.3.9
+4.4.3.2
+4.4.3.6
+4.4.3.4
+4.4.2.4
+4.4.2.8
+4.4.8
+4.4.1
+4.4.5
+4.4.6
+4.4.3
+4.4.2
+4.4.4
+4.4.14
+4.4.16
+4.4.12
+4.5.44
+4.4.9
+4.5.12
+4.5.11
+4.4.3.3
+4.4.15
+4.4.10
+4.5.42
+4.5.1
+4.5.2
+4.5.22
+4.4.2
+4.5.14
+4.5.18
+4.5.8
+
+302
+
+Chapter 13 - Glossary and concordances
+Procedure or sub-procedure name
+
+Specification
+
+Section
+
+Read Playing Order
+Read Playing Order Supported
+Read Seeking Speed
+Read Status Flags
+Read Track Duration
+Read Track Position
+Read Track Title
+Read Volume Flags
+Read Volume Offset State
+Read Volume State
+Receiver Start Ready
+Receiver Stop Ready
+Relative Volume Down
+Relative Volume Up
+Released ASEs or LE ACL link loss
+Releasing an ASE
+Search
+Set Absolute Track Position
+Set Absolute Volume
+Set Audio Input Description
+Set Audio Location
+Set Audio Output Description
+Set Automatic Gain Mode
+Set Bearer Signal Strength Reporting Interval
+Set Current Group Object ID
+Set Current Track Object ID
+Set Gain Setting
+Set Initial Volume
+Set Manual Gain Mode
+Set Members Discovery procedure
+Set Mute
+Set Next Track Object ID
+Set Playback Speed
+Set Playing Order
+Set Relative Track Position
+Set Volume Offset
+Stop Current Track
+Supported Audio Contexts discovery
+Terminate Call
+
+MCP
+MCP
+MCP
+CCP
+MCP
+MCP
+MCP
+VCP
+VCP
+VCP
+BAP
+BAP
+VCP
+VCP
+BAP
+BAP
+MCP
+MCP
+VCP
+VCP
+VCP
+VCP
+VCP
+CCP
+MCP
+MCP
+VCP
+VCP
+VCP
+CSIP
+MICP
+MCP
+MCP
+MCP
+MCP
+VCP
+MCP
+BAP
+CCP
+
+4.5.19
+4.5.21
+4.5.10
+4.4.11
+4.5.4
+4.5.5
+4.5.3
+4.4.1.4
+4.4.2.2
+4.4.1.1
+5.6.3.2
+5.6.5.1
+4.4.1.6.1
+4.4.1.6.2
+5.6.7
+5.6.6
+4.5.43
+4.5.6
+4.4.1.6.5
+4.4.3.10
+4.4.2.5
+4.4.2.9
+4.4.3.7.5
+4.4.7
+4.5.17
+4.5.13
+4.4.3.7.1
+4.4.1.5
+4.4.3.7.4
+4.6.2
+4.4.3
+4.5.15
+4.5.9
+4.5.20
+4.5.7
+4.4.2.6.1
+4.5.26
+5.4
+4.4.13.2
+303
+
+Section 13.4 - Bluetooth LE Audio characteristics
+Procedure or sub-procedure name
+
+Specification
+
+Section
+
+Track Discovery - Discover by Current Group Object
+ID
+Unmute
+Unmute
+Unmute/Relative Volume Down
+Unmute/Relative Volume Up
+Updating Metadata
+
+MCP
+
+4.5.16
+
+VCP
+VCP
+VCP
+VCP
+BAP
+
+4.4.1.6.6
+4.4.3.7.2
+4.4.1.6.3
+4.4.1.6.4
+5.6.4
+
+Table 13.4 Procedures and sub-procedures defined in Bluetooth® LE Audio specifications
+
+13.4
+
+Bluetooth LE Audio characteristics
+
+The following characteristics are defined the Bluetooth LE Audio service specifications.
+The reference is to the table in the specification in which the characteristic is defined.
+Characteristic
+
+Specification
+
+Table
+
+ASE Control Point
+Audio Input Control Point
+Audio Input Description
+Audio Input State
+Audio Input Status
+Audio Input Type
+Audio Location
+Audio Output Description
+Available Audio Contexts
+Bearer List Current Calls
+Bearer Provider Name
+Bearer Signal Strength
+Bearer Signal Strength Reporting Interval
+Bearer Technology
+Bearer Uniform Caller Identifier (UCI)
+Bearer URI Schemes Supported List
+Broadcast Audio Scan Control Point
+Broadcast Receive State
+Call Control Point
+Call Control Point Optional Opcodes
+Call Friendly Name
+Call State
+Content Control ID
+Content Control ID (CCID)
+Coordinated Set Size
+
+ASCS
+AICS
+AICS
+AICS
+AICS
+AICS
+VOCS
+VOCS
+PACS
+TBS
+TBS
+TBS
+TBS
+TBS
+TBS
+TBS
+BASS
+BASS
+TBS
+TBS
+TBS
+TBS
+MCS
+TBS
+CSIS
+
+4.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.5
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+5.1
+
+304
+
+Chapter 13 - Glossary and concordances
+Characteristic
+
+Specification
+
+Table
+
+Current Group Object ID
+Current Track Object ID
+Current Track Segments Object ID
+Gain Setting Properties
+Incoming Call
+Incoming Call Target Bearer URI
+Media Control Point
+Media Control Point Opcodes Supported
+Media Player Icon Object ID
+Media Player Icon URL
+Media Player Name
+Media State
+Mute
+Next Track Object ID
+Parent Group Object ID
+Playback Speed
+Playing Order
+Playing Orders Supported
+Search Control Point
+Search Results Object ID
+Seeking Speed
+Set Identity Resolving Key
+Set Member Lock
+Set Member Rank
+Sink ASE
+Sink Audio Locations
+Sink PAC
+Source ASE
+Source Audio Locations
+Source PAC
+Status Flags
+Supported Audio Contexts
+Termination Reason
+TMAP Role
+Track Changed
+Track Duration
+Track Position
+Track Title
+Volume Control Point
+
+MCS
+MCS
+MCS
+AICS
+TBS
+TBS
+MCS
+MCS
+MCS
+MCS
+MCS
+MCS
+MICS
+MCS
+MCS
+MCS
+MCS
+MCS
+MCS
+MCS
+MCS
+CSIS
+CSIS
+CSIS
+ASCS
+PACS
+PACS
+ASCS
+PACS
+PACS
+TBS
+PACS
+TBS
+TMAP
+MCS
+MCS
+MCS
+MCS
+VCS
+
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+5.1
+5.1
+5.1
+4.1
+3.2
+3.1
+4.1
+3.4
+3.3
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+3.1
+305
+
+Section 13.5 - Bluetooth LE Audio terms
+Characteristic
+
+Specification
+
+Table
+
+Volume Flags
+Volume Offset Control Point
+Volume Offset State
+Volume State
+
+VCS
+VOCS
+VOCS
+VCS
+
+3.1
+3.1
+3.1
+3.1
+
+Table 13.5 Characteristics defined in the Bluetooth® LE Audio specifications
+
+13.5
+
+Bluetooth LE Audio terms
+
+The following specific terms are defined and used throughout the Bluetooth LE Audio
+specifications. They are generally capitalised when used with their defined meanings. Where
+abbreviations, acronyms or initialism are used, these are indicated after the term.
+Phrase
+
+Specification Section
+
+Additional Controller Advertising Data
+(ACAD)
+Application Profile
+ASE identifier (ASE_ID)
+ASE state machine
+Audio Channel
+Audio Configuration
+Audio Location
+
+Core
+
+Audio Sink
+Audio Source
+Audio Stream Endpoint (ASE)
+Broadcast Audio Source Endpoint
+(BASE)
+Broadcast Audio Stream
+Broadcast Isochronous Group (BIG)
+
+BAP
+BAP
+ASCS
+BAP
+
+Broadcast Isochronous Stream (BIS)
+
+Core
+
+Broadcast Sink
+Broadcast Source
+Broadcast_ID
+BIG_Sync_Delay
+
+BAP
+BAP
+BAP
+Core
+
+Call Control Client
+Call Control Server
+Call Gateway (CG)
+
+CCP
+CCP
+TMAP
+
+306
+
+Core
+ASCS
+ASCS
+BAP
+BAP
+BAP
+
+BAP
+Core
+
+Volume 6, Part B, Section
+2.3.4.8
+Volume 1, Part A, Section 6.3
+Section 4.1
+Section 3
+Section 1.6
+Section 4.4
+Section 1.6 (and GA Assigned
+Numbers)
+Section 1.6 and 3.3
+Section 1.6 and 3.3
+Section 4.1
+Section 3.7.2.2
+Section 1.6
+Volume 6, Part B, Section
+4.4.6.2
+Volume 6, Part B, Section
+4.4.6.1
+Section 1.6
+Section 1.6
+Section 3.7.2.1
+Volume 6, Part B, Section
+4.4.6.4
+Section 2
+Section 2
+Section 2.2
+
+Chapter 13 - Glossary and concordances
+Phrase
+
+Specification Section
+
+Call Terminal (CT)
+Caller ID
+CIG Identifier
+
+TMAP
+TBS
+Core
+
+CIG_Sync_Delay
+
+Core
+
+CIS Identifier
+
+Core
+
+Connected Isochronous Group (CIG)
+
+Core
+
+Connected Isochronous Stream (CIS)
+
+Core
+
+Context Type
+
+PACS
+
+Coordinated Set
+Enhanced ATT (EATT) bearer
+
+CSIP
+Core
+
+Extended advertising (EA)
+
+Core
+
+Generic Access Profile (GAP)
+Inband Ringtone
+Link Layer (LL)
+Local Retrieve
+Low Energy asynchronous connection
+(LE ACL)
+Media Control Client
+Media Control Server
+PA_Interval
+
+Core
+TBS
+Core
+TBS
+Core
+
+Packet Loss Concealment (PLC)
+Periodic Advertising Synchronization
+Transfer (PAST)
+Periodic Advertising Train (PA)
+
+LC3
+Core
+
+Presentation Delay
+Published Audio Capabilities(PAC)
+record
+Remote Broadcast Scanning
+Remote Hold
+
+BAP
+PACS
+
+Section 2.2
+Section 1.9
+Volume 6, Part B, Section
+4.5.14
+Volume 6, Part B, Section
+4.5.4.1.1
+Volume 6, Part B, Section
+4.5.13.1
+Volume 6, Part B, Section
+4.5.14
+Volume 6, Part B, Section
+4.5.13
+Section 1.9 (and GA Assigned
+Numbers)
+Section 2.1
+Volume 3, Part F, Section
+3.2.1
+Volume 6, Part B, Section
+2.3.1
+Volume 3, Part C
+Section 1.9
+Volume 6, Part B
+Section 1.9
+Volume 1, Part A, Section
+3.5.4.6
+Section 2.1
+Section 2.1
+Volume 6, Part B, Section
+2.3.4.6
+Appendix B
+Volume 3, Part C, Section
+9.5.4
+Volume 6, Part B, Section
+4.4.5.1
+Section 7
+Section 2.2
+
+BAP
+TBS
+
+Section 6.5
+Section 1.9
+
+MCP
+MCP
+Core
+
+Core
+
+307
+
+Section 13.5 - Bluetooth LE Audio terms
+Phrase
+
+Specification Section
+
+Scan Offloading
+Service Data AD data type
+Service UUID
+Set Coordinator
+Set Members
+Silent Mode
+Sink ASE
+Source ASE
+SyncInfo
+
+BAP
+CSS
+CSS
+CSIP
+CSIP
+TBS
+ASCS
+ASCS
+Core
+
+Unenhanced ATT bearer
+
+Core
+
+Unicast Audio Stream
+
+BAP
+®
+
+Table 13.6 Defined terms in the Bluetooth LE Audio specifications
+
+308
+
+Section 6.5
+Section 1.11
+Section 1.1
+Section 2
+Section 2
+Section 1.9
+Section 4
+Section 4
+Volume 6, Part B, Section
+2.3.4.6
+Volume 3, Part A, Section
+10.2
+Section 1.6
+
+Index
+ACAD .............................................. 113, 210
+Acceptor ............................. 53, 78, 156, 164
+ACL link loss ........................................... 197
+Additional Controller Advertising Data
+............................................................... 113
+ADV_EXT_IND ................................... 112
+Advertising Set_ID ................................. 221
+AICS................................................... 43, 251
+AirPods ....................................................... 17
+Airtime ......................................... 55, 91, 284
+Anchor Point .............................. 80, 90, 110
+Announcements ........................................ 71
+ASCS .......................................... 41, 163, 174
+ASE Control Point ........................ 180, 190
+ASE state machine .................................. 179
+ASE_ID ........................................... 175, 190
+ASHA.......................................................... 20
+Audio Announcement............................ 211
+Audio Channel........................................... 54
+Audio Configuration ..................... 150, 274
+Audio Data Path ..................................... 192
+Audio Input Control Service.......... 43, 251
+Audio Location................ 61, 170, 219, 257
+Audio quality..................................... 84, 145
+Audio Sharing ................................. 117, 289
+Audio Sink......................................... 43, 167
+Audio Stream Configuration Service ... 163
+Audio Stream Control Service ....... 41, 174
+Audio Stream Endpoint ......................... 175
+Audio_Channel_Counts ........................ 178
+Audio_Channel_Location ....................... 63
+Automatic Gain Control ........................ 258
+Autonomous Operation......................... 197
+AUX_ADV_IND .......................... 112, 210
+AUX_CHAIN_IND PDU ................... 113
+AUX_SYNC_IND ........................ 112, 212
+Auxiliary Pointer ..................................... 111
+Availability.................................................. 60
+Available Audio Contexts ............... 60, 172
+Available Audio Contexts characteristic 59
+Available_Source_Contexts .................. 200
+BAP ............................................ 41, 163, 179
+BAP Broadcast Audio Stream
+Establishment ..................................... 210
+BASE ......................115, 118, 204, 210, 280
+
+Basic Audio Announcement ................... 72
+Basic Audio Announcement Service ... 112
+Basic Audio Profile .................. 41, 163, 179
+BASS ................................. 41, 119, 201, 213
+Bidirectional CIS ....................................... 92
+BIG..................................................... 55, 100
+BIG_Offset .............................................. 115
+BIG_Sync_Delay .................................... 123
+BIGInfo .......................... 102, 113, 210, 212
+BIS ............................................................... 76
+BIS Spacing .............................................. 114
+Bitrate........................................................ 144
+BN ............................................................... 85
+Bragi ......................................................17, 27
+Broadcast Assistant................ 211, 215, 217
+Broadcast Audio Announcement ........... 72
+Broadcast Audio Announcement Service
+............................................................... 211
+Broadcast Audio Reception Ending
+procedure ............................................. 158
+Broadcast Audio Reception Start
+procedure .................................... 158, 203
+Broadcast Audio Reception Stop
+procedure ............................................. 203
+Broadcast Audio Scan Control Point .. 214
+Broadcast Audio Scan Service 41, 201, 213
+Broadcast Audio Source Endpoint ...... 115
+Broadcast Audio Start procedure 157, 203
+Broadcast Audio Stop procedure 157, 203
+Broadcast Audio Stream configuration
+procedure ............................................. 203
+Broadcast Audio Stream disable
+procedure ............................................. 203
+Broadcast Audio Stream establishment
+procedure ............................................. 203
+Broadcast Audio Stream Metadata update
+procedure ............................................. 203
+Broadcast Audio Stream reconfiguration
+procedure ............................................. 203
+Broadcast Audio Stream release
+procedure ............................................. 203
+Broadcast Audio Streams ...................... 201
+Broadcast Audio Update procedure ... 157,
+203
+Broadcast Isochronous Group ...... 55, 100
+309
+
+Broadcast Isochronous Stream .........76, 98
+Broadcast Isochronous Terminate
+procedure ............................................. 211
+Broadcast Receive State ......................... 214
+Broadcast Receive State characteristic . 221
+Broadcast Receiver.................................. 201
+Broadcast Source ..................................... 201
+Broadcast Source State Machine........... 202
+Broadcast to Unicast Audio Handover
+procedure ............................................. 158
+Broadcast_Code .. 119, 208, 215, 222, 223,
+227, 290
+Broadcast_ID........................................... 221
+Burst Number ............................... 63, 85, 87
+Call Control ID ......................................... 63
+Call Control Profile.......................... 45, 231
+Call Gateway ............................................ 272
+Call State characteristic........................... 236
+Call Terminal ........................................... 272
+CAP ................................... 47, 153, 156, 163
+CCID.................................................. 63, 158
+CCP .................................................... 45, 231
+Change Microphone Gain Settings
+procedure ............................................. 159
+Change Volume Mute State procedure 159
+Change Volume Offset procedure ....... 159
+Change Volume procedure.................... 159
+Change_Counter ..................................... 254
+Channel Allocation ................................... 61
+CIG.............................................................. 55
+CIG reference point ................................. 94
+CIG state machine .................................... 95
+CIG synchronisation point ...................... 94
+CIG_Sync_Delay .................................... 123
+CIS .................................................. 55, 76, 80
+Close Isochronous Event ........................ 83
+Codec Configuration procedure ........... 183
+Codec Configuration Settings ............... 135
+Codec Configured ................................... 176
+Codec Configured state.......................... 178
+Codec Specific Capabilities ...................... 63
+Codec Specific Configuration ............... 184
+Codec_Frame_Blocks_Per_SDU ........... 63
+Codec_ID ................................................. 164
+Codec_Specific_Capabilities ........ 165, 184
+Codec_Specific_Configuration ............. 206
+Commander ... 73, 118, 156, 201, 214, 223,
+265
+Common Audio Profile . 47, 153, 156, 163
+310
+
+Connected Isochronous Group.............. 55
+Connected Isochronous Stream 55, 76, 80
+Content Control ID ....................... 158, 233
+Context Types............................................ 56
+Control Subevent ............................. 99, 108
+Coordinated Set ........................ 64, 154, 215
+Coordinated Set Identification Profile.. 46,
+64, 153
+Coordinated Set Identification Service .. 46
+Coordination Control ............................... 46
+CSIP ......................................................46, 64
+CSIS.............................................. 46, 64, 154
+CSSN ........................................................... 99
+CSTF ........................................................... 99
+Disabling state ......................................... 196
+EHIMA......................................................... 9
+Encryption ............................................... 208
+Ending a unicast stream ......................... 195
+Extended Advertising .................... 110, 210
+Extended Attribute Protocol .................. 40
+Flush Point ................................................. 85
+Flush Timeout .................................. 85, 107
+Frame Length .......................................... 137
+Frame Size ................................................ 134
+Framing...................................... 89, 140, 185
+Frequency hopping ................................... 83
+FT ................................................................ 85
+General Announcement........................... 71
+Generic Audio Framework...................... 41
+Generic Media Control Service...... 46, 232
+Generic Telephone Bearer Service 46, 232
+GMCS ................................................ 46, 232
+Group Count ........................................... 102
+Groups ...................................................... 243
+GTBS ................................................. 46, 232
+Handover.................................................. 228
+HAP .......................................................... 267
+HAS ........................................................... 267
+Hearing Access Profile ........................... 267
+Hearing Access Service .......................... 268
+Immediate Need for Audio related
+Peripheral ...................................... 71, 160
+Immediate Repetition Count................. 103
+INAP .................................................. 71, 160
+Inband ringtone ....................................... 239
+Incoming call .................................. 236, 239
+Initiator ........................................ 53, 78, 156
+IRC ............................................................ 103
+ISO PDU.................................................... 81
+
+ISO_Interval ................................... 114, 209
+ISOAL ............................................... 89, 122
+Isochronous Adaptation Layer ............. 122
+Isochronous Interval ................................ 80
+Isochronous Payloads .............................. 81
+Isochronous Stream............................54, 75
+Latency............................................. 126, 137
+LC3 ..................................................... 47, 125
+LE Create CIS command ........................ 97
+LE Remove CIG command .................... 97
+LE Set CIG Parameters ......................... 186
+LE_Set_Extended_Advertising_Data . 210
+LE_Set_Periodic_Advertising_Data ... 210
+Link Layer ID ......................................89, 99
+Local Name AD Type ............................ 280
+Locally Held ............................................. 236
+Lock ............................................................ 46
+Low Complexity Communications Codec
+................................................................. 47
+Low Latency ............................................ 139
+Made for iPhone ....................................... 20
+Max_Transport_Latency .............. 140, 208
+Maximum Transmit Latency ................. 185
+Maximum_SDU_Size............................. 140
+MCP ........................................... 45, 232, 242
+MCS............................................ 45, 232, 242
+MCS state machine ................................. 243
+Media Control Client .............................. 242
+Media Control Profile ..................... 45, 232
+Media Control Service ............. 45, 232, 242
+MEMS microphone .................................. 25
+Metadata ........ 120, 165, 190, 206, 219, 222
+MICP........................................................... 44
+Microphone control................................ 262
+Microphone Control Profile ................... 44
+Microphone Mute State procedure ...... 159
+MICS ........................................................... 44
+Missing Acceptors................................... 198
+Missing set members .............................. 160
+Modify Broadcast Source procedure ... 217
+MP3 ........................................................... 126
+mSBC ........................................................ 128
+Multi-channel ........................................... 146
+Multi-profile ............................................... 51
+Mute .......................................................... 255
+Near Field Magnetic Induction.........22, 65
+NFMI .......................................................... 22
+NSE ............................................................. 85
+Number of Subevents .............................. 85
+
+Object Transfer Service .................. 45, 245
+Out of band ringtones ............................ 239
+PAC record .............................................. 163
+Packet Loss Concealment............... 67, 136
+Packing............................................. 187, 208
+PACS .................................................. 41, 163
+PAST ........................................ 118, 121, 220
+PBP................................................... 267, 274
+Periodic Advertising ...................... 113, 210
+Periodic Advertising Sync Transfer ..... 220
+Periodic Advertising Synchronisation.. 118
+Playback speed......................................... 247
+Playing order ............................................ 248
+Playing Tracks.......................................... 245
+PLC .................................................... 67, 136
+Preferred Audio Contexts............... 59, 167
+Preferred Retransmission Number ...... 185
+Presentation Delay . 65, 109, 137, 140, 185,
+229
+Presets ....................................................... 269
+Pre-Transmission Offset......... 63, 103, 107
+Program_Info .......................................... 280
+PTO .......................................................... 103
+Public Broadcast Profile ...... 113, 215, 267,
+274
+Published Audio Capabilities Service .. 163
+Published Audio CapabilitiesService...... 41
+QoS ........................................................... 138
+QoS configuration procedure ............... 186
+QoS Configured state ............................. 195
+Quality of Service .................................... 138
+Rank ............................................................ 46
+RAP .................................................... 71, 160
+Ready for Audio related Peripheral71, 160
+Receiver Start Ready ............................... 176
+Receiver Stop Ready ...................... 176, 196
+Releasing state.......................................... 196
+Remote Broadcast Scanning.................. 225
+Remote Control......................................... 73
+Remotely Held ......................................... 236
+Resolvable Set Identity ........................... 154
+Retransmission ........................................ 137
+Retransmission Number ............... 140, 141
+Retransmissions....................................... 139
+Ringtone ..................................................... 58
+Robustness ........................................ 84, 102
+RTN ......................................... 140, 141, 208
+Sampling Frequency ............................... 139
+Sampling Rate .......................................... 131
+311
+
+SBC............................................................ 128
+Scan Delegator................................ 216, 223
+SDU Interval................................... 140, 188
+Set Coordinator .............................. 154, 215
+Set Identity Resolving Key .................... 154
+Set Member .............................................. 154
+Set Member Lock .................................... 155
+Set Member Rank.................................... 155
+Silent Mode .............................................. 240
+Sink ASE ......................................... 175, 193
+Sink ASE state machine ......................... 175
+Sink Audio Location ............................... 171
+Sink led journey ......................................... 51
+Solicitation Requests ............................... 223
+Source ASE ..................................... 175, 193
+Source ASE state machine..................... 196
+Source Audio Locations ......................... 171
+Streaming Audio Contexts....................... 59
+Sub_Interval ............................................. 114
+Sub_Interval spacing ................................ 83
+Subevent ..................................................... 82
+Supported Audio Context Types ............ 60
+Supported Audio Contexts ............. 59, 171
+Supported_Audio_Channel_Counts.... 166
+Supported_Codec_Specific_Capabilities
+............................................................... 165
+Supported_Frame_Durations ............... 165
+Supported_Max_Codec_Frames_Per_SD
+U ............................................................ 166
+Supported_Octets_Per_Codec_Frame166
+Supported_Sampling_Frequencies ....... 165
+Synchronisation Point .............................. 67
+Synchronisation Reference ........... 123, 137
+SyncInfo ................................................... 112
+
+312
+
+Target Latency ......................................... 184
+Target PHY .............................................. 184
+Targeted Announcement ......................... 71
+TBS ............................................................ 231
+TBS Call Control Point characteristic.. 239
+TBS state machine .................................. 235
+Telephone Bearer Service ...................... 231
+Telephony and Media Audio Profile... 267,
+271
+Terminating calls ..................................... 241
+TMAP .............................................. 267, 271
+Top level profiles ...................................... 47
+Track Position ......................................... 245
+Tracks........................................................ 243
+Transport Delay ............................. 123, 127
+True Wireless ............................................. 21
+Unicast Audio Start procedure ............. 157
+Unicast Audio Stop procedure ............. 157
+Unicast Audio Update procedure ......... 157
+Unicast Media Receiver .......................... 272
+Unicast Media Sender ............................. 272
+Unicast to Broadcast Audio Handover
+procedure ............................................. 158
+Updating unicast metadata .................... 194
+URI ............................................................ 237
+VCP .................................................... 43, 251
+VCS .................................................... 43, 251
+VOCS ................................................. 43, 251
+Volume Control Profile .................. 43, 251
+Volume Control Service.................. 43, 251
+Volume Controller .................................. 215
+Volume Offset Control Service ..... 43, 251
+Volume State characteristic ................... 253
+
+313
+
+ABOUT THE AUTHOR
+
+Nick Hunn is the author of the Fundamentals of Short-Range Wireless – the first book to
+cover Bluetooth Classic, Wi-Fi, Zigbee and Bluetooth low energy. He developed some of the
+very first Bluetooth products to come to market and has started a number of successful
+technology companies. Nick has been involved in Bluetooth since its inception and has helped
+author most of the major Bluetooth requirements documents, including those for Bluetooth
+Low Energy and LE Audio. Since 2013 he has chaired the Bluetooth Hearing Aid working
+group, which developed the concept of Bluetooth LE Audio and he has participated in writing
+all of the Bluetooth LE Audio specifications, including the LC3 codec.
+Nick regularly talks and reports on the audio industry. In 2014, he coined the word
+“Hearables” to describe the new generation of personal audio products. He has produced two
+major reports on the market for hearable devices, which correctly predicted the growth and
+outstanding success of personal Bluetooth audio. These, and many other articles on the
+market and accompanying technology can be accessed on his blog at www.nickhunn.com.
+Nick’s aim with this book is to demystify the complexity of the specifications and help readers
+understand the potential they bring to the future market for hearables.
+
+314
+
+315
+
+
\ No newline at end of file