A first analysis of the data you gave me yesterday may be found here.  This will have to be redone, after our timing/gate scan, but this was a good exercise.

I did not yet do the linear interpolation to put all peaks at the same ADC value, but that will be the easy part.  I'll do it today.

You can see that the fitting is reasonable (not perfect) and I'd say the failure rate is acceptable (those a human has to deal with).

The final page shows the distribution of MPV for the lowest (Vbias=57) and highest (60) biases, for both QT32B (top) and QT32C (bottom).  You can see that there is always finite overlap, meaning that if we demand a fixed MPV, then some channels will require bias voltages *outside* of the range we've sampled.  Which is not a problem per se, but probably we should include higher and lower Vbias in our scan.  It would be nicer to interpolate than to extrapolate.

A final note: 744 tiles is a lot more than 93 tiles....