Procedure for cluster-based gain analysis while the calib db is down or pedestals are bad

First one needs to make a fake pedestal from a fixed-gain file.  This can be done for e.g. run 666 using

python CalcNoiseAndMean.py -r 666 --skipNevents 1000

This creates

CalcNoiseAndMean_mean_r666_step0.npy

which has the shape of an epixHR frame and can be subtracted as a pedestal.  The point of skipping N events is to get to a more T-settled period so adjust 1000 as needed.

One can look at test pixels vs timestamp and as a 1d histogram using e.g.

python EventScanParallelSlice.py -r 777 --fakePedestal CalcNoiseAndMean_mean_r666_step0.npy --special regionCommonMode

and

python EventScanParallelSlice.py -r 777 -f ../scan/EventScanParallel_c0_r777_n1.h5 -L cm

One can run the clustering by doing

python SimpleClustersParallelSlice.py -r 777 --fakePedestal CalcNoiseAndMean_mean_r666_step0.npy --special regionCommonMode,FH

and

python SimpleClustersParallelSlice.py -r 777 -f ../lowFlux/SimpleClustersParallel_c0_r777_n1.h5 -L cm

Or do the usual equivalent with sbatch if you don't want to wait forever.

This makes per-pixel gain histograms with fits and plots all the slice pixel gains (peaks).  It also makes a histogram of all slice pixels energy across the run before clustering.