astro-legacy-archive/cobe-firas-c-vector
COBE/FIRAS C-Vector LAMBDA says the FIRAS C-Vector values “represent the variance (random noise) derived from multiple observations within the same pixel.” It serves one length-182 C_VECTOR for each of six spectral modes. The source FITS unit is MJy/sr. The distinct Supplement definition and resulting conflict are recorded below rather than resolved here. configuration NUM_FREQ zero suffix coverage source resolution FIRAS_C_VECTOR_HIF2 55 127 600–1350 GHz 24.6 GHz… See the full description on the dataset page: https://huggingface.co/datasets/astro-legacy-archive/cobe-firas-c-vector.
COBE/FIRAS C-Vector
LAMBDA says the FIRAS C-Vector values “represent the variance (random noise) derived from multiple observations within the same pixel.” It serves one length-182 C_VECTOR for each of six spectral modes. The source FITS unit is MJy/sr. The distinct Supplement definition and resulting conflict are recorded below rather than resolved here.
How to use
python -m venv .venv && .venv/bin/pip install datasets huggingface_hub pyarrowData
from datasets import load_dataset
dataset = load_dataset("astro-legacy-archive/cobe-firas-c-vector", "FIRAS_C_VECTOR_HIF2", split="train")
row = dataset.select([0]).select_columns(["C_VECTOR"])[0]
print(len(row["C_VECTOR"]), row["C_VECTOR"][:3])182 [3.862412214279175, 3.3703856468200684, 2.550980806350708]Metadata
from huggingface_hub import hf_hub_download
import json
import pyarrow.parquet as pq
path = hf_hub_download(repo_id="astro-legacy-archive/cobe-firas-c-vector", repo_type="dataset", filename="FIRAS_C_VECTOR_HIF2.parquet")
print(json.loads(pq.read_metadata(path).metadata[b"dense_matrix"])["source"]["keywords"])Dataset structure
Each configuration has one row and one fixed-size-list column:
The source headers give the mode in CHANSCAN and the spectral coordinate in NU_ZERO, DELTA_NU, and NUM_FREQ. The source defines NUM_FREQ as the “Number of frequency points with good data”: respectively 55, 55, 60, 170, 182, and 43 in the six files. The archive retains every stored element, including the exact zero suffixes in the table above. Consumers should read the source-verbatim structured keyword records rather than infer a boundary from the fixed storage width. Those records also expose PRODUCT, MODLLABL, MISS_SEG, DATE-BEG, DATE-END, DATE, and GALATEXC, with the FITS comments. Any immediately following COMMENT cards are retained separately under following_comment_cards, in their complete source order and spelling. The delivered GALATEXC values are 20 for HIF2/HIF3/HIF4/HIGH and 15 for HRES/LOWF; its following cards state that input below that absolute Galactic latitude was excluded in producing the dataset.
Scientific context
LAMBDA's product description uses the attributed “variance” wording above. Explanatory Supplement §7.1.2 instead defines the C-Vector as the square root of the covariance diagonal and says it includes random detector error plus remaining systematic variance. That amplitude definition agrees with the linear FITS unit, but conflicts with LAMBDA's variance wording. Both source statements are retained without choosing one or changing the source unit.
For every mode <MODE>, the exact companion is FIRAS_DESTRIPED_SKY_SPECTRA_<MODE>.FITS, mirrored in `firas-spectral-maps` as configuration FIRAS_DESTRIPED_SKY_SPECTRA_<MODE>. Each source table names its two relevant fields exactly WEIGHT and SPECTRUM; this applies to HIF2, HIF3, HIF4, HIGH, HRES, and LOWF. Explanatory Supplement §6.2 says the observation weights are dimensionless, a common variance supplies the scale, the corresponding sigma is called the C-Vector, and relative weights for the individual channel combinations are derived from their C-Vectors. No formula, weight, or frequency axis is derived or embedded here.
HIF2, HIF3, and HIF4 partition the HIGH band. HIGH, HRES, and LOWF use different spectral ranges or resolutions. LAMBDA publishes 24.6 GHz resolution for LOWF, HIGH, HIF2, HIF3, and HIF4 and 6.9 GHz for HRES. These are instrumental spectral resolutions, not a generated axis and not the FITS DELTA_NU interval. Retain the exact source mode association when combining a C-Vector with another FIRAS product.
Provenance
The six files are listed on the LAMBDA download page. Each source FITS file is 11,520 bytes and carries one C_VECTOR[182] row. Its individual SHA-256, source URL, complete primary and table headers, field unit, and structured regeneration record are embedded in the corresponding Parquet artifact.
LAMBDA records 2003-03-01 as the delivery date for all six legacy C-Vector files. Their filenames have no _v2 suffix. LAMBDA names Dale Fixsen and Nils Odegard only for the separate 2006 HEALPix reprocessing, not for these legacy C-Vector files.
The FIRAS Explanatory Supplement Appendix G gives the record definitions. The source array reconstructs with identical float32 bits and element order.
Licensing and credit
HEASARC states, verbatim: “HEASARC materials are all available freely for your use.” Its requested acknowledgement is: “This research has made use of data and/or software provided by the High Energy Astrophysics Science Archive Research Center (HEASARC), which is a service of the Astrophysics Science Division at NASA/GSFC.” LAMBDA's COBE credit is, verbatim: “The COBE datasets were developed by the NASA Goddard Space Flight Center under the guidance of the COBE Science Working Group.” No explicit product licence was found or published. The HEASARC statement and URL are retained as archive-use rights evidence, not presented as a source-issued licence named heasarc-free-use.
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 Community tab of any dataset. Confirmed corrections are made in the source conversion tools and published by rebuilding the dataset.