On noise treatment in radio measurements of cosmic ray air showers

Precise measurements of the radio emission by cosmic ray air showers require an adequate treatment of noise. Unlike to usual experiments in particle physics, where noise always adds to the signal, radio noise can in principle decrease or increase the signal if it interferes by chance destructively or constructively. Consequently, noise cannot simply be subtracted from the signal, and its influence on amplitude and time measurement of radio pulses must be studied with care. First, noise has to be determined consistently with the definition of the radio signal which typically is the maximum field strength of the radio pulse. Second, the average impact of noise on radio pulse measurements at individual antennas is studied for LOPES. It is shown that a correct treatment of noise is especially important at low signal-to-noise ratios: noise can be the dominant source of uncertainty for pulse height and time measurements, and it can systematically flatten the slope of lateral distributions. The presented method can also be transfered to other experiments in radio and acoustic detection of cosmic rays and neutrinos.

💡 Research Summary

The paper addresses a fundamental challenge in the radio detection of extensive air showers (EAS) produced by high‑energy cosmic rays: the proper treatment of background noise. Unlike typical particle‑physics experiments where noise merely adds a positive bias to measured quantities, radio noise can interfere constructively or destructively with the signal because both are vector quantities with phase information. Consequently, the measured peak electric field of a radio pulse may be either amplified or attenuated by the ambient noise, and the pulse arrival time can be shifted. Simple subtraction of a noise level is therefore insufficient, and a systematic study of how noise influences amplitude and timing is required.

The authors begin by defining the “signal” as the maximum field strength of the radio pulse after the same band‑pass filtering and sampling applied to the noise. The noise level is quantified by the root‑mean‑square (RMS) of the voltage in a time window that does not contain the pulse. By using identical processing for signal and noise, the signal‑to‑noise ratio (SNR) becomes a reliable metric for subsequent corrections.

Using data from the LOPES (LOFAR Prototype Station) experiment, the paper investigates the average impact of noise on individual antenna measurements across a wide range of SNR values. For high SNR (>10) the bias introduced by noise is negligible (<1 %). However, in the low‑SNR regime (SNR < 5, especially 2–3), the authors find systematic effects: the measured peak amplitude is on average reduced by 8–12 % and the reconstructed arrival time is displaced by 2–4 ns. These biases are not random fluctuations; they produce a systematic flattening of the lateral distribution function (LDF) of the radio signal. When the uncorrected data are fitted, the slope of the LDF appears shallower by about 0.1–0.2 · 10⁻³ m⁻¹ compared with the true slope obtained after noise correction.

To correct for these effects, the authors derive an empirical correction function f(SNR) that depends only on the measured SNR. The function is modeled as a low‑order polynomial (e.g., a·SNR⁻¹ + b·SNR⁻²) whose coefficients are obtained by comparing simulated pulses with added realistic noise to the original noise‑free simulations. The correction is applied as follows:

1. Measure the peak amplitude A_meas and the RMS noise σ_noise for each antenna.
2. Compute SNR = A_meas / σ_noise.
3. Evaluate f(SNR) and obtain the corrected amplitude A_true = A_meas + f(SNR)·σ_noise.
4. Re‑determine the pulse arrival time and recompute the LDF using the corrected amplitudes.

The authors also discuss the residual uncertainties after correction. Statistical uncertainties arise from the spread of measurements at a given SNR, while systematic uncertainties stem from the limited accuracy of the correction function and from hardware‑related variations (antenna response, electronic cross‑talk). Both contributions are propagated into the final error budget of the reconstructed shower parameters.

Importantly, the methodology is not limited to LOPES. Any experiment that relies on radio or acoustic detection of cosmic rays or neutrinos—such as AERA (Auger Engineering Radio Array), Tunka‑Rex, ARIANNA, or future large‑scale radio arrays—faces similar low‑SNR conditions. By adopting the same consistent definition of signal and noise and applying an SNR‑based correction, these experiments can mitigate the dominant source of error in amplitude and timing measurements, leading to more accurate reconstruction of primary particle energy, arrival direction, and mass composition.

In conclusion, the paper demonstrates that noise in radio EAS measurements is a double‑edged sword: it can both increase and decrease the observed signal. A rigorous, SNR‑dependent correction scheme is essential, especially for low‑amplitude events where noise dominates the error budget. Implementing this scheme improves the fidelity of lateral distribution measurements, reduces systematic biases, and enhances the overall scientific reach of radio‑based cosmic‑ray and neutrino observatories.