astro-legacy-archive/eht-m87-2017

EHT M87 2017 calibrated visibilities

These are the Event Horizon Telescope Collaboration's low- and high-band EHT-HOPS network-calibrated Stokes I visibility products for M87 on days 095, 096, 100, and 101 of the April 2017 campaign. They are the calibrated-data release 2019-D01-01, derived from Science Release 1 and used as the primary dataset for the First M87 EHT Results. Each configuration is named verbatim from its source filename stem and has one row per UVFITS random group in source order.

Target	M87, J2000, RA 187.7059307575226 deg, Dec 12.39112323919932 deg
Products	Four observing days, low and high band
Averaging	10 s and all 32 intermediate frequencies
Contents	One frequency channel, four FITS polarization slots, complex data and weights
Random groups	51,119 across eight configurations
Parquet downloads	164,360–427,670 bytes each

Reconstruction demonstration

The image above was reconstructed from SR1_M87_2017_095_lo_hops_netcal_StokesI and SR1_M87_2017_095_hi_hops_netcal_StokesI using ehtim 1.3.2. The ingestion path was load_dataset → embedded visibility metadata → UVFITS reconstruction byte-size-matched to the released files → eht-imaging. The reconstruction used the EHT Collaboration's published 2019-D01-02 eht-imaging RML pipeline and defaults from First M87 Event Horizon Telescope Results. IV. Imaging the Central Supermassive Black Hole, and recovered a compact flux of 0.5961 Jy.

How to use

python -m venv .venv && .venv/bin/pip install datasets huggingface_hub pyarrow

from datasets import load_dataset
dataset = load_dataset("astro-legacy-archive/eht-m87-2017", "SR1_M87_2017_095_lo_hops_netcal_StokesI", split="train[:1]", columns=["BASELINE", "DATE"])
row = dataset[0]
print(row["BASELINE"], row["DATE"])

262.0 [2457848.5, 0.03200231492519379]

import json
from huggingface_hub import hf_hub_download
import pyarrow.parquet as pq
path = hf_hub_download("astro-legacy-archive/eht-m87-2017", "SR1_M87_2017_095_lo_hops_netcal_StokesI.parquet", repo_type="dataset")
print(json.loads(pq.read_schema(path).metadata[b"visibility"])["schema"])

astro-legacy-archive/visibility

The embedded visibility document records the AIPS baseline-number and DATE formulas. BASELINE resolves as antenna 1 // 256 and antenna 2 % 256; the two DATE elements sum to the Julian Date. DATA retains the declared DEC → RA → IF → FREQ → STOKES → COMPLEX shape, and each COMPLEX triple is real part, imaginary part, and source weight; non-positive weight denotes a flagged slot.

Dataset structure

column	type	meaning
UU---SIN, VV---SIN, WW---SIN	float32	Raw source projected-baseline coordinates; resolve physical light-travel seconds as stored * pscal + pzero using the matching visibility.random_groups.parameters record
BASELINE	float32	Source AIPS baseline number
DATE	list[float32, 2]	Both source parameters named DATE, in PTYPE order
INTTIM	float32	Integration duration in seconds
TAU1, TAU2	float32	Source atmospheric-opacity correction quantities for the first and second stations
DATA	six nested fixed-size lists, lengths 1 × 1 × 1 × 1 × 4 × 3	Source DEC → RA → IF → FREQ → STOKES → COMPLEX array; real and imaginary are in source BUNIT JY, followed by the source weight

The source PTYPE order is UU---SIN, VV---SIN, WW---SIN, BASELINE, DATE, DATE, INTTIM, TAU1, TAU2. The two identically named source parameters are held in one length-2 DATE list because the Parquet dataset reader requires unique field names; both source values are stored unchanged. Their sum is the visibility Julian Date.

For these files, antenna numbers are floor(BASELINE / 256) and BASELINE modulo 256. The 21 baselines use antennas 1–7. The preserved AIPS AN table has eight records; NOSTA 8, SR, participates in no baseline. The source-provided lo- and hi-band centre frequencies are respectively 227070703125 Hz and 229070703125 Hz. The source GROUPS axes are COMPLEX, STOKES, FREQ, IF, RA, and DEC. Arrow nesting reads all six in reverse, outermost-first order. A one-based axis element j has value CRVAL + (j - CRPIX) * CDELT. For STOKES this gives the four FITS codes -1, -2, -3, and -4. FREQ additionally uses the AIPS FQ IF offset preserved in the metadata.

config suffix	random groups	timestamps	baselines	flagged slots	Parquet bytes
095_lo_hops_netcal_StokesI	6,453	701	21	3,530	326,701
095_hi_hops_netcal_StokesI	6,458	701	21	3,540	320,310
096_lo_hops_netcal_StokesI	8,645	900	21	4,262	418,384
096_hi_hops_netcal_StokesI	9,081	900	21	4,428	427,670
100_lo_hops_netcal_StokesI	2,367	186	21	1,106	164,360
100_hi_hops_netcal_StokesI	2,610	186	21	1,200	170,831
101_lo_hops_netcal_StokesI	7,447	781	21	3,142	365,059
101_hi_hops_netcal_StokesI	8,058	801	21	3,414	381,972

The source represents an estimated Stokes I product in all four polarization slots. RR and LL both carry an estimated Stokes I value; RL and LR and their errors were deliberately set to zero by the publisher. The slots are not four independent polarization measurements. RR and LL measurement errors are each sqrt(2) times the statistical error for Stokes I. A non-positive source weight marks a flagged value. Infinite weights are source values, not nulls.

Provenance

The eight UVFITS files total 4,495,680 bytes and are pinned to release commit 27ae76aacb9067a83388395c3826012b97aac371.

source file	bytes	SHA-256
SR1_M87_2017_095_lo_hops_netcal_StokesI.uvfits	567,360	97492ee1146b05e12b1783e6c9a8253a0f50d789835286ffcc3ed5d248d51a8e
SR1_M87_2017_095_hi_hops_netcal_StokesI.uvfits	567,360	a11a874bdbf205089192b5270abc5894ad85e4ab54f0f677e3280c5639ea3e0d
SR1_M87_2017_096_lo_hops_netcal_StokesI.uvfits	751,680	f3394b2c272ba92980c21e3522987344f9175a78c333648a36458c917216c365
SR1_M87_2017_096_hi_hops_netcal_StokesI.uvfits	786,240	5bc42247ccd01b58460d95e8c17f91e337c79ffe99a0718aacf0a91c77ad24ec
SR1_M87_2017_100_lo_hops_netcal_StokesI.uvfits	224,640	ee90ca5aec0bd8ed27f5ea13c941ae3f24c188d5c8a9aa87ab4b5606f57334be
SR1_M87_2017_100_hi_hops_netcal_StokesI.uvfits	244,800	76104537313cd6879c497901f313bad667f4608c48df961e67ac5615d84b2c00
SR1_M87_2017_101_lo_hops_netcal_StokesI.uvfits	650,880	697af2bb3bbf732115108ffefacd3e59e307f38fe685c3c4579146b0bd661298
SR1_M87_2017_101_hi_hops_netcal_StokesI.uvfits	702,720	618c3019f60e88268980267a9db68f638b379a85842d83c06003d28550d191f5

Rebuild with python tools/make_eht_m87_visibility.py.

Each file carries a versioned visibility metadata document containing field and axis definitions, the ordered random-group parameter definitions, complete FITS headers, AIPS AN and FQ tables, source provenance, credit, and regeneration command. PyArrow schema inspection exposes field descriptions; load_dataset retains types but strips field metadata. The loaded table contains the raw, unscaled random parameters, both source DATE occurrences in one length-2 field, and DATA with every declared source axis at its declared length. Reconstruction restores the source parameter definitions and array shape, every random parameter occurrence, the full group array, and each auxiliary bintable with identical dtype, shape, order, and values. FITS container serialisation and padding are outside that comparison.

Licensing and credit

The pinned release states:

The data files in this data set are licensed under the ODC-PDDL license https://www.opendatacommons.org/licenses/pddl/1.0/index.html.

The release publishes this credit line and no separately worded requested acknowledgement:

Authors: The Event Horizon Telescope Collaboration et al.

Citations

The release names First M87 EHT Results III: Data Processing and Calibration as its primary reference. It also cites the complete First M87 EHT Results series: I, II, IV, V, and VI.

The release also cites the reduction and calibration papers Blackburn et al. (2019), arXiv:1903.08832 and Janssen et al. (2019), arXiv:1902.01749. The same pinned release reference list includes Issaoun et al. (2017), EHT Memo 2017-CE-02, Janssen et al. (2019), EHT Memo 2019-CE-01, and Wielgus et al. (2019), EHT Memo 2019-CE-02.

Errata and requests

To report an erratum, propose a correction, or ask a question about this dataset, open a Discussion or Pull Request in its Community tab; posting requires a free Hugging Face account. Dataset requests and archive-wide questions may be posted in the README space's Community tab or the Community tab of any dataset. Confirmed corrections are made in the source conversion tools and published by rebuilding the dataset.