astro-legacy-archive/vsa-released-products

VSA February 2004 released products

This dataset contains LAMBDA's February 2004 VSA spectrum, x factors, 16 window functions, and 16×16 Hessian matrix. Source identifiers, numeric values, shapes, and row order are retained. Unlabelled source columns remain Column 1, Column 2, and so on.

Use

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

from datasets import load_dataset

spectrum = load_dataset("astro-legacy-archive/vsa-released-products", "VSAF", split="train[:5]")
print(spectrum.column_names, spectrum.num_rows)

Output:

['Column 1', 'Column 2', 'Column 3', 'Column 4'] 5

import json
import pyarrow.parquet as pq
from huggingface_hub import hf_hub_download

path = hf_hub_download("astro-legacy-archive/vsa-released-products", "VSAF_mat.parquet", repo_type="dataset")
print(json.loads(pq.read_schema(path).metadata[b"dense_matrix"])["array"]["shape"])

Output:

[16, 16]

Structure and decode guidance

There are 19 configurations:

• VSAF has 16 rows. Column 1 contains the source keys data1 through data16 verbatim. Column 2 is the CMB band-power T0^2 l(l+1)C_l/(2pi); Column 3 and Column 4 are the magnitudes of its asymmetric lower and upper 68-percent likelihood errors. The three numeric columns are in uK^2. The source does not label its columns, so their numbered names remain unchanged.
• VSAF1 through VSAF16 are the 16 unequal-length window-function tables. In each, Column 1 is multipole l and Column 2 is the corresponding band-window value. The files state no units.
• VSAF_x has 16 rows and one source column, Column 1: the offset-lognormal x_i in source band order and in uK^2.
• VSAF_mat is a dense matrix with 16 Parquet rows. Its sole column, VSAF_mat, contains the rows of the source 16×16 Hessian M_ij in log-offset variables. No unit or axis names are documented.

Every configuration embeds its complete source-backed per-column descriptions, units, source URL, source SHA-256, and requested credit in Parquet schema metadata. The table configurations use the statistical_product metadata key; VSAF_mat uses dense_matrix. Read the latter with the second snippet.

LAMBDA labels VSAF.tar “Hessian Matrix”, but that tar contains the 16 window functions. The actual 256-value Hessian is VSAF_mat.txt, represented here as the VSAF_mat dense-matrix configuration. This release has calib_uncertainty = 0.03. VSAF_x supplies the 16 offset-lognormal x_i values, in band order and in the same uK^2 unit as the observed powers. The release defines O_i=log(O_i0+x_i) and chi^2=-2 log_like=sum_ij (O_i-log(B_i+x_i)) M_ij (O_j-log(B_j+x_j)); VSAF_mat is the Hessian M_ij used by that expression.

Sources

The product page and download page serve the following pinned sources:

Source	Bytes	SHA-256
VSAF.tar	153,600	19e49068d6d5163d62fe29e836587f59e11142a0e90848e209b40c820324031d
VSA-0402.tar	163,840	0742da8f65fc17cf87f187ff296743ac5a623f5575455a93ecd2bac31318e726
VSAF.txt	858	a989677da188d29ce27c60af59c85480a43b915e4c756ac355a7aa50ae5536ff
VSAF_x.txt	224	933672ba63fe642231325cda5913770a2979330f1dd17773b630a65104f2ea0c
VSAF_mat.txt	1,767	e796df7ce51f08562d5f127a60c35c900ff92f432943bf024e37919d64ac92c4
dickinson-2004.pdf	630,318	4f7bf176bb2138d1a3ef8cb51a4fefa23ec15116452798c11f50cb4975caac19
vsa_hessian_m_info.html	9,979	57f7f775db7849ec518ca3c1ba3bc7992d525e57af141487768dc099fdb1f7b8

The individual window sources are:

Source	Bytes	SHA-256
VSAF1.txt	4,256	45e36c5e3be839994668499fcd99efff58d2fce76d9980952b14d406a4f21d97
VSAF2.txt	4,395	a37e98462fd58173d49739599970574192ffb50514430bfd6a632ff9a98a873d
VSAF3.txt	8,435	471554218362451ab6631c354f555df9cba273ffff0af878957b9b0649b167df
VSAF4.txt	8,367	b28dbc653acf5eeddbdfd631ca3d6867b4f0d2fe4b5ac6390a98461f7edc7d7c
VSAF5.txt	8,871	42c81a1eaaf87cf8c032f85e460cf89094eae85603b08a5bd6b5324f0ae72d0a
VSAF6.txt	8,103	3fdf7e0ddbd4c27745b024c10b5de999e795ab50bbe6bfd0a50f0b9e633b484f
VSAF7.txt	8,869	a237ff2b9fce78743777bf90ec7ee09dd96ec20fd8fe15e755de1a3efe94764b
VSAF8.txt	8,522	d67419be73f1448452424adee65f7741529ffa1774a02ef9b5c157c01fa4a134
VSAF9.txt	8,365	c7db7daa19c05f7166ff4837635a885f8145560c6c425e11839a9e07fc79b262
VSAF10.txt	8,314	d4cd934e8c2f7667515be46ba1124801048e78229c33f2640bb82d4c45b0a669
VSAF11.txt	9,067	ad074feace47e4dafa5120e3f0344901d906b316be87d343fde56e7b9b8cc2a7
VSAF12.txt	9,154	316111402639da920b5657eda340d01f94b0557144549d3e267d6f6268654cc7
VSAF13.txt	9,109	b1fb08837a404d0f9bc4061d3512afb47476d5cec81b70eeb75a2bf3ffaf4c4c
VSAF14.txt	10,070	fb39b0bef4e691a9d0bc5aa56a90c7368853ec08d199deb22c8f1378c233e89f
VSAF15.txt	10,165	51ad293430d5c1b32bc2504384a0937cb0283956188af18f939049be6a4cb589
VSAF16.txt	12,474	bf413945373d273a0ea9a9e7af6bcf97b57359a9918f7f9aba86e631850f1efc

VSAF.tar duplicates the 16 individually served window files under member names VSAF1 through VSAF16. VSA-0402.tar duplicates all 19 logical products under the source member names VSAF.dataset, VSAF_x.dat, VSAF_mat.dat, and VSAF1 through VSAF16; the member bytes equal the corresponding individual files. All identifiers and filenames remain verbatim. No released sky-map file is served on the current product table. This is the February 2004 3%-calibration release, not the older 7%-calibration products described by archived collaboration pages.

Licence, credit, and citation

No named or formal data licence was found on the VSA product or download pages. HEASARC states verbatim: “HEASARC materials are all available freely for your use.”

Requested LAMBDA acknowledgement, verbatim: “We acknowledge the use of the Legacy Archive for Microwave Background Data Analysis (LAMBDA), part of the High Energy Astrophysics Science Archive Center (HEASARC). HEASARC/LAMBDA is a service of the Astrophysics Science Division at the NASA Goddard Space Flight Center.”

The product overview identifies C. Dickinson et al. (2004), “High sensitivity measurements of the CMB power spectrum with the extended Very Small Array,” MNRAS 353, 732, DOI 10.1111/j.1365-2966.2004.08206.x, and R. Rebolo et al. (2004), “Cosmological parameter estimation using Very Small Array data out to l=1500,” MNRAS 353, 747. No additional product-specific acknowledgement or citation boilerplate is published.