I am suggesting here that parameters be specified on a baseband pair basis. This forcloses, for instance, different spectral resolutions in the two basebands. One can think of cases in which that is useful, but they do not seem to me to justify the complexity in the real-time system. The suffix X indicates which of the four baseband pairs is indicated. I have no particular opinion about whether this should be A, B, C, D, or 1, 2, 3, 4 or [1], [2], [3], [4] or whatever.
decimationX = <1 | 2 | 4 | 8 | 16 | 32> ! Sets the total bandwidth analysed crosscorrelationX = <YES | NO> ! NO is the multiple single dish mode polarizationX = <YES | NO> ! four vs two products per BB pair (even if, due to switches, some are silly) integrationX = <integration time in seconds> ! the u,v FITS for the VLBA correlator requires the same integration time ! in all basebands, which I here propose to relax. ! having a baseline specific integration time is a big complication in ! the real-time system (and elsewhere also), which I do not believe is ! justified by the reduction in output rate one would realize nmbrlagsX = <16 | 32 | 64 | ... | 2048> ! one may want to process fewer than the full number of lags the hardware ! provides, to exchange spectral resolution for data volume or data rate. archiveX = <list of channel numbers to be passed to archiving system> ! We probably need a mechanism to discard the ends of the spectra ! to get data output rates and volumes down. This is the most ! general way of describing this, and pretty easy to implement. ! However, I can imagine the programmer of a 'filler' program looking ! like the famous Muench painting when asked to implement this in all ! its generality. In practice, one would presumably have a small number ! of spectral windows. But having the full generality in the real-time ! system does not seem harmful to me.