Off-by-one bug in gain estimated by the charging model

[lucabaldini]

The following line in instrument/charging.py:

g = self.__gain_data[:,:,i-1]

introduces a subtle bug, as we later update the value of g in place before assigning it to the next time slice:

g -= r.T * delta_max / k_c * (1 - delta / delta_max) * dt
g += delta / tau_d * dt
self.__gain_data[:,:,i] = g

and this is actually updating the value of the gain either at time slice i (which is what we want) and at time slice i-1 (which we don't). As a result, at the end of the loop, the gain is off by one everywhere but in the very last bin.

Luckily, all we have to do is to replace the incriminated line with:

g = numpy.copy(self.__gain_data[:,:,i-1])

and we are good.

Original url: https://bitbucket.org/ixpesw/ixpeobssim/issues/404
Created on 2021-04-07T11:06:57.360903+00:00