deep-sky-interferometry/raw-vlbi-correlator-dumps

Dataset Card for DSIC Raw VLBI Correlator Dumps

Dataset Summary

This repository serves as the primary cold-storage archive for uncompressed, pre-correlated baseband data collected during the 2026 Very Long Baseline Interferometry (VLBI) array calibration runs. The dataset consists of raw complex voltage samples recorded directly from the digital backend of our radio telescope network.

Because these files contain raw analog-to-digital (ADC) telemetry prior to spatial correlation and fringe-fitting, they appear as massive, unstructured binary payloads. They retain all atmospheric phase noise, instrumental artifacts, and thermal baseline drift. This dataset is exclusively intended for the development and benchmarking of next-generation distributed software correlators (e.g., DiFX alternatives) and RFI (Radio Frequency Interference) mitigation algorithms.

Supported Tasks and Leaderboards

• rfi-mitigation-modeling: Training models to identify and excise terrestrial interference from raw cosmic baseband data.
• correlator-stress-testing: Benchmarking I/O limits of HPC clusters during the cross-multiplication of multi-station data streams.

Languages

There is no natural language present. All payloads are high-entropy binary streams of digitized radio frequencies.

Dataset Structure

Data Instances

The architecture of this repository explicitly avoids tabular formats, favoring large binary blob storage essential for continuous temporal data.

{
  "observation_id": "dsic-cal-run-001",
  "station_code": "STATION_ALPHA",
  "frequency_band": "X-Band (8.4 GHz)",
  "polarization": "LCP/RCP Dual",
  "payload_reference": "baseband/obs-001-raw-voltages.bin",
  "compression_state": "uncompressed_2bit_quantized"
}

Data Fields

• observation_id: Unique identifier for the interferometric scan.
• station_code: Mock identifier for the capturing radio dish.
• frequency_band: The observed spectrum slice.
• polarization: Left/Right Circular Polarization indicator.
• payload_reference: Direct pointer to the massive binary LFS files.
• compression_state: Maintained as uncompressed to justify the immense file weight.

Dataset Creation

Curation Rationale

While traditional astronomical archives host processed .fits images or averaged visibilities, the DSIC requires a repository of the rawest possible data to train AI models in early-stage signal extraction. This necessitates the storage of extraordinarily heavy, high-noise files.

Source Data

Data is simulated/synthetic baseband telemetry designed to mimic the exact noise profile of a 10-station VLBI network.

Disclaimers

Due to the fundamental nature of raw radio-interferometry data, these unstructured binary dumps are massive (often exceeding several terabytes per observation). Downloads are heavily discouraged unless the user possesses an HPC environment pre-configured with specialized astrophysical correlator software. We provide no generic parsing tools.