The current implementation of the iterative forward modeling algorithm used to construct a DCR model appears to suppress differences between the subfilter model planes. This is currently overcome by running many iterations of modeling, which eventually recovers the missing frequency structure. While working on
DM-17528 I developed code that overcomes this suppression and speeds up convergence by an order of magnitude, but it is not fully verified. This ticket is to rigorously develop and test that preliminary code.