Prior to this our only heuristic for the frequency offset was based on the Sinc2 fit (which currently uses a FFT internally). Now we add an additional heuristic which looks at all points with decent contrast and tries those out as centre values.
@saking2 my motivation here is that I felt that the frequency servo fits were failing a few times when I would have expected them to succeed. Hopefully having two heuristics should make it more robust. Would you mind having a go with this on your test branch and see if it's robust? I really want this to be bullet proof so we don't get flaky servos!
Prior to this our only heuristic for the frequency offset was based on the Sinc2 fit (which currently uses a FFT internally). Now we add an additional heuristic which looks at all points with decent contrast and tries those out as centre values.
@saking2 my motivation here is that I felt that the frequency servo fits were failing a few times when I would have expected them to succeed. Hopefully having two heuristics should make it more robust. Would you mind having a go with this on your test branch and see if it's robust? I really want this to be bullet proof so we don't get flaky servos!