Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
The number of columns (2055) exceeds the maximum supported number of columns (1000). This is a current limitation of the datasets viewer. You can reduce the number of columns if you want the viewer to work.
Error code:   TooManyColumnsError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

21cm Hydrogen Line Sky Survey

Continuous 1-second-cadence power spectra of the 21cm neutral hydrogen (HI) line at 1420.405 MHz, recorded 24/7 from a fixed omnidirectional observer on the US East Coast.

What this is

A radio telescope pointed at the whole sky, recording one spectrum per second, indefinitely. The Earth's rotation scans the beam across the galactic plane daily, producing a natural drift scan. Every row is a self-timestamped power spectrum spanning 2 MHz centered on the HI rest frequency.

Hardware

Component Detail
Antenna Omnidirectional L-band feed probe (wide-beam, zenith-biased)
LNA Nooelec SAWbird+ H1 — SAW bandpass + LNA, centered 1420 MHz
SDR Nooelec SMArTee v5 (RTL2832U + R820T), TCXO 0.5 ppm
Compute Raspberry Pi 5, Debian, headless
Location US East Coast, 41.0°N 73.0°W (quantized)

The SAWbird+ H1 provides a ~68 MHz usable passband (1374–1442 MHz) with peak response near 1399 MHz. The observation window sits on the upper slope of this filter. Bias-tee powered from the SMArTee v5.

Schema

Each row is one integration period from rtl_power.

Column Type Unit Description
date string — UTC date (YYYY-MM-DD)
time string — UTC time (HH:MM:SS)
freq_start int Hz Lower band edge (1419500000)
freq_end int Hz Upper band edge (1421500000)
bin_width float Hz Frequency bin spacing (~976.56)
num_samples int — Samples averaged in this integration
power_0..power_2048 float dB Power spectral density per frequency bin

Frequency axis: 2049 bins from 1419.500 MHz to 1421.500 MHz at ~976.56 Hz spacing. To reconstruct: freq_mhz[i] = freq_start/1e6 + i * bin_width/1e6.

Power values: Log-scale (dB) relative to the SDR's internal reference. Not flux-calibrated. Relative measurements (e.g., ON-OFF, time differencing) are valid.

What's in the signal

The 21cm line from neutral hydrogen in the Milky Way. The galactic plane contains vast clouds of HI gas at different distances and velocities. Galactic rotation Doppler-shifts each cloud's emission to a slightly different frequency, spreading the line over 400 kHz (100 km/s). The main features:

  • Local/Orion arm — nearest gas, dominant contribution, near the rest frequency
  • Perseus arm — redshifted (receding), lower frequency
  • Outer arm — further redshifted
  • Scutum-Centaurus, Sagittarius-Carina — inner galaxy, complex velocities

The signal strength and Doppler structure change throughout the day as Earth's rotation sweeps different galactic longitudes through the beam.

Calibration notes

This is uncalibrated data. The power values include the receiver bandpass shape (SAWbird filter rolloff), gain drift (thermal), and system noise. To extract the HI signal:

  1. Earth-rotation OFF: Hours when the galactic plane is below the horizon serve as a natural bandpass reference. Subtract from on-plane hours.
  2. Polynomial baseline: Fit and subtract the bandpass shape from each spectrum, excluding the HI frequency range.
  3. Matched filter: Cross-correlate against a galactic rotation model for optimal SNR.

Validated ON-OFF calibration recipe

The recommended OFF reference is UTC 05:00-08:00 (LST ~10.6-13.6h), when the North Galactic Pole is near zenith and HI column density is minimal. Subtract the mean OFF spectrum from each ON hour to remove the bandpass and system response.

LST approximation for this site (73W): LST = (UTC_hour + 5.63) mod 24.

Expected results: ON-OFF differencing reveals a clear diurnal cycle in HI power. The signal is near zero during OFF hours (Galactic pole overhead), rises to ~0.5-1.0 dB as the Galactic plane enters the beam, and peaks above +1 dB when the Cygnus-Cassiopeia-Perseus region transits (UTC ~17-22, LST ~22-4h). The omnidirectional beam produces a broad velocity profile spanning roughly -150 to +150 km/s rather than discrete peaks at individual spiral arm velocities. A slight asymmetry favoring the negative-velocity (approaching) side is consistent with Perseus arm gas along these sightlines.

Velocity axis: Standard radio Doppler convention, positive = receding.

Known artifacts:

  • SAWbird bandpass slope across the 2 MHz window (higher power at lower frequencies)
  • Thermal gain drift (~1-3 dB over hours, correlated with ambient temperature); broadband power shows a secular +3 dB trend over 48 days (gain drift or seasonal RFI floor)
  • Occasional RFI spikes (narrowband, identifiable by persistence and bandwidth)
  • No RFI channels detected above 5-MAD threshold across 2049 channels

Usage

from datasets import load_dataset

ds = load_dataset("phanerozoic/hi-21cm-survey", split="train")

Plot a single spectrum

import numpy as np

row = ds[0]
freqs = np.array([1419.5e6 + i * row["bin_width"] for i in range(2049)]) / 1e6
powers = [row[f"power_{i}"] for i in range(2049)]

import matplotlib.pyplot as plt
plt.plot(freqs, powers)
plt.xlabel("Frequency (MHz)")
plt.ylabel("Power (dB)")
plt.axvline(1420.405, color="r", linestyle="--", alpha=0.5, label="HI rest freq")
plt.legend()
plt.show()

Daily average

df = ds.to_pandas()
df["datetime"] = pd.to_datetime(df["date"] + " " + df["time"])
df["hour"] = df["datetime"].dt.hour

power_cols = [f"power_{i}" for i in range(2049)]
hourly = df.groupby("hour")[power_cols].mean()

Research applications

  1. Galactic structure — Doppler tomography of spiral arms via velocity-resolved HI emission
  2. Drift-scan mapping — Earth rotation provides full RA coverage; stack days to build sky maps
  3. RFI characterization — long-term L-band interference environment near urban areas
  4. Receiver stability — thermal drift, gain variation, bandpass evolution over months
  5. Scintillation — interstellar scintillation of compact HI features at 1-second cadence
  6. Educational — accessible radio astronomy dataset for students (no telescope required to analyze)

Collection method

Automated daemon on Raspberry Pi 5:

  • rtl_power runs continuously, 1-second integrations
  • Python watchdog handles crash recovery, daily file rotation, health logging
  • Daily upload to HuggingFace via huggingface_hub
  • Receiver: 49.6 dB gain, 2 MHz bandwidth, 2048-point FFT

Citation

@dataset{hi-21cm-survey,
  title={21cm Hydrogen Line Sky Survey},
  author={Norton, Charles C.},
  year={2026},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/phanerozoic/hi-21cm-survey}
}
Downloads last month
1,546