Need to count packets received per X engine per integration

[AaronParsons]

Currently, I think we just assume that we get all the packets. This means that if some packets are lost (which we have occasionally seen; we aren't sure about the root cause but it seem to be fixed by rebooting an X engine) we effectively have a flux scale change in the data that can be freq/time dependent. If we knew how many packets went into an integration, we would be able to back out what happened, track X engine health, and potentially restore a consistent flux scale by dividing out by the packet count.

[AaronParsons]

@jsdillon Perhaps you can post an image of this issue as evidenced in autocorrelations?

[jsdillon]

The impact is very hard to see in the autos, but it's much clearer in the mean-subtracted autocorrelations. See e.g. here: https://nbviewer.jupyter.org/github/HERA-Team/H1C_IDR3_Notebooks/blob/main/data_inspect/data_inspect_2458055.ipynb

In general, our RFI algorithms pick this up pretty well. Here's an example from the next day where those x-engines outright die toward the end of the night, but show periodic wonky behavior for the entire night.

https://nbviewer.jupyter.org/github/HERA-Team/H1C_IDR3_Notebooks/blob/main/rfi_inspect/rfi_inspect_2458056.ipynb

[david-macmahon]

Do the locations for the missing packets get filled with zeros or with leftover old data?