Impact of simulated 1/f noise for HI intensity mapping experiments

Abstract

Cosmology has entered an era where the experimental limitations are not due to instrumental sensitivity but instead due to inherent systematic uncertainties in the instrumentation and data analysis methods. The field of HI intensity mapping (IM) is still maturing, however early attempts are already systematics limited. One such systematic limitation is 1/f noise, which largely originates within the instrumentation and manifests as multiplicative gain fluctuations. To date there has been little discussion about the possible impact of 1/f noise on upcoming single-dish HI IM experiments such as BINGO, FAST or SKA. Presented in this work are Monte-Carlo end-to-end simulations of a 30 day HI IM survey using the SKA-MID array covering a bandwidth of 950 and 1410 MHz. These simulations extend 1/f noise models to include not just temporal fluctuations but also correlated gain fluctuations across the receiver bandpass. The power spectral density of the spectral gain fluctuations are modelled as a power-law, and characterised by a parameter $\beta$. It is found that the degree of 1/f noise frequency correlation will be critical to the success of HI IM experiments. Small values of $\beta$ ($\beta$ < 0.25) or high correlation is preferred as this is more easily removed using current component separation techniques. The spectral index of temporal fluctuations ($\alpha$) is also found to have a large impact on signal-to-noise. Telescope slew speed has a smaller impact, and a scan speed of 1 deg s$^{-1}$ should be sufficient for a HI IM survey with the SKA.