Yandasoft Regularised Bandpass Solver Investigations

In some realistic scenarios, the calibration solution smoothing may oversmooth the data, see SPO-470 for more details. This problem can be addressed by using better designed cost function in the calibration inverse problem.

Over PI5 and PI6 a few investigations have been flagged that may improve solver solution quality (for more details, see SP-733 and SP-927). For example, there was some indication that fourth-order smoothing constraints may be useful in some cases (see YAN-189), and it appears that the number of solver minor cycles may be related to the scale (number of channels) over which smoothing occurs.

Another possibility would be introducing the "minimum level of smoothness" parameter, to better control how we smooth the solution. This should be more robust, than varying the number of minor cycles, as it will be possible (1) to directly impose the minimum value of the smoothing term norm (instead of controlling the number of minor cycles), and (2) the full solution convergences can be assured with larger number of minor cycles.

These will be investigated in this feature.