I like Darrel's suggestion of using the loss of quantization efficiency as a measure of the increase in variance across the spectrum. The resulting change in bandpass shape is probably more important than the sensitivity loss. As expected, the 4 level digitizer (DG) case (Correlator TDM mode) is a lot more sensitive to DG input power level than is the 8 level case (Correlator FDM mode). Here are the numbers for the 4 level case (from TMS(Ed2) equation 8.52 and Figure 8.5 (curve 2)).
E(dB) QuantEffic 2 .963 1 .989 .5 .997 0 1.0 -.5 .996 -1 .985 -2 .936
where E is the DG input power compared to optimum and QuantEffic is the quantization efficiency compared to optimum. So with a tolerance on DG power level of +- 1 dB the potential changes in bandpass shape are not negligible compared to the desired spectral fidelity.
So we are going to have to carefully develop a calibration strategy related to the allowed change in DG input power level. Certainly I think it will be necessary to do an amplitude and and phase bandpass calibration on either side of an IF or baseband attenuator change because I would not expect these attenuators to be spectrally flat to the desired level.
Peter-- JeffMangum - 2009-06-04