Thu Dec 6 19:16:16 CET 2007

too much carrier drift

so, to get a bit of full-circle understanding: why not mix a signal to
DC and filter its absolute value? looks like the thing i did wrong was
not the mixer, but the place where the smoothing is going on.

  * mix + filter: isolates a frequency region
  * full-wave rectify + filter.

2 filter operations are essential here, so it's probably easier to do
only one, and instead of full-wave use the amplitude/power of a

But but but... maybe the filter is actually too sharp? I measured the
carrier at 1/7Hz, expecting it at 1/8Hz.. i'm missing a parameter:
bandwith and time decay are related, but increasing the sample rate..

look: this is just a shifted one-pole filter: it's the equivalent of
passing the difference signal 1/7-1/8 to the lowpass filter. that
probably won't survive.. so i'm stuck with the same problem: the
carrier shift is much more than the bandwidth of the signal!

it's about 80Hz at 600Hz, while the signalling frequency is around
9Hz. this means i have to do something about it... it's either going
to be manual tuning, or adaptive tuning.. synchronous demod is
starting to look like the only solution. or i should just use the 2
filter approach of above:

  * wide filter to eliminate noise: it should be wide enough to
    capture the carrier tolerance.

  * narrow filter to perform demodulation after full-wave rectify.

it's starting to look like synchronous is going to be a lot less
hassle.. again.. what do i need? an AGC to normalize the Q output such
that i can use it as feedback to phase offset.

go over this again.. something's wrong.