FM stereo decoder is not working well

[kpreid]

[kpreid]

Improved in commit e62d1ada13a0b8af4e13b416932b6e57ec4f05fa.

For some reason I haven't yet tracked down, the difference signal is weak, which is why the output was basically mono. Now I've amplified it by the empirical right amount, but this also adds noise.

[kpreid]

The stereo decoding was happening after FM deemphasis, which accounts for the weakness of the difference signal. That's fixed in f8d959c1f761be3197f124421e50e00affb7fd78.

However, there is still an extreme sensitivity to the phase of the pilot signal. Unfortunately, GR does not make it easy to keep phase in sync along multiple signal paths, and my attempt to do so systematically seems to be sensitive to flowgraph reconnection (as receiver validity changes) even though it shouldn't be if my understanding of GR is right. (The VOR receiver has the same issue.)

We could calibrate out the phase error using a known test signal but this would be awkward as long as reconnection changes the delay (we would have to interrupt actual operation to recalibrate).