This MR improves the PID performance for upgrade data by better tuning the pixel background estimates. This is done by adding a new scaling parameter to the background estimation algorithm, that allows the values computed by said algorithm to be scaled, in principle independently for each RICH, although at the moment I use 0.75 for each.
Longer term the algorithm needs to be rethought for the upgrade. It was originally designed for HPDs, which have O(800) active pixels in each PD. The algorithm considers each PD independently and computes a background term for all pixels in a PD, based on a comparison of the observed to expected signals. With the upgrade, and MaPMTs, we have an order of magnitude more independent PDs with an order less (64) pixels each, and perhaps a new algorithm is required for this scenario. This MR improves things in the meantime.
The attached PDFs show examples of the improvement I see.
Scale 1.0 means the same as in current master. A scale of 0.75 for each seems the best overall improvement.
Note this MR relies on LHCb!1288 (merged). It also contains the updates in !1020 (merged) so that should be merged first.
Made WIP, as this will change the references, so I leave it to @cattanem to test and apply whenever is convenient.
FYI @seaso @fkeizer - The update here and in LHCb!1288 (merged) is also now in the
RichTTFUStudies branches, so just pull those in to get the update for the Upgrade^2 studies. Also remember to update your local Gaudi to the current master, as well.