Use the total-power calibrated data, plus the information from the calibration module to arrive at the data in the units in which they will be imaged (is that always
, antenna temperature (K) corrected for opacity and efficiency?).
Glen's example script does this, which is a modification from what GBTIDL currently does (the model for tau_zenith, n_atmos, and eta_a is more sophisticated plus there is an attempt to estimate and remove the differential effect of the atmospheric correction to Tsys.
data = Tsys_ref * [exp(tau_zenith * n_atmos)/eta_a] * (sig_tp - ref_tp) / ref_tp - (tsysAir - tsysAir_ref)
- sig_tp is the total-power calibrated data for this integration (it is the SIG=T data for the frequency-switched case, otherwise all data to be imaged).
- ref_tp is the reference spectrum (SIG=F data for the frequency-swtiched case otherwise it's the reference spectrum returned from the calibration module, which was stored there in the previous step, in the case where this is an interpolated quantity, the calibration module is responsible for that interpolation).
- zenith_tau is the zenith opacity from the calibration module
- n_atmos = number of atmospheres. At high elevations this is simply 1/sin(elevation). This value is supplied by the calibration module.
- eta_a = an estimate of the point source efficiency returned by the calibration module.
- Tsys_ref is the system temperature associated with the ref_tp data.
- tsysAir is an estimate of the effective atmospheric contribution to Tsys at the elevation at which the data was taken
- tsysAir_ref is the effective atmospheric contribution to Tsys at the elevation of the reference spectra (this may not generalize well to complicated schemes to arrive at the reference spectra).
In the vector calibration case, Tsys_ref is a vector quantity. Which of the other quantities should be treated as vector function of frequency (vary sufficiently across the bandpass)? Is it ever desirable to smooth ref_tp to improve the signal to noise in data?