Separation of gain fluctuations and continuum signals in total power spectrometers with application to COMAP

We describe a time-domain technique for separating $1/f$ gain fluctuations and continuum signal for a total power spectrometer, such as the CO Mapping Array Project (COMAP) Pathfinder instrument. The $1/f$ gain fluctuations of such a system are expected to be common-mode across frequency channels. If the instrument's system temperature is not constant across channels, a continuum signal will exhibit a frequency dependence different from that of common-mode gain fluctuations. Our technique leverages this difference to fit a three-parameter frequency model to each time sample in the time-domain data, separating gain and continuum. We show that this technique can be applied to the COMAP Pathfinder instrument, which exhibits a series of temporally stable resonant noise spikes that effectively act as calibrators, breaking the gain degeneracy with continuum signals. Using both simulations and observations of Jupiter, we explore the effect of a $1/f$ prior for the gain model. We show that the model is capable of cleanly separating Jupiter, a bright continuum source, from the gain fluctuations in the scan. The technique has two applications to COMAP. For the COMAP observations performing line intensity mapping (LIM), the technique better suppresses atmospheric fluctuations and foregrounds than the COMAP LIM pipeline. For the Galactic COMAP observations, which map Galactic continuum signals, the technique can suppress $1/f$ gain fluctuations while retaining all continuum signals. This is demonstrated by the latest COMAP observations of $λ$-Orionis, where our method produces far cleaner maps than a destriper alone, typically reducing the noise power by a factor of 7 on beam scales and up to 15 on larger scales.