[hpsdr] frequency-specific frequency domain AGC

[Roger Rehr W3SZ]

Thanks again to all for the great suggestions and comments on the
frequency-specific frequency domain AGC!

I just have a little bit of follow-up.

I had previously indicated that on-the-air tests with the FACG with JT65
and FT8 were very good and I was quite pleased.

I was concerned about what the results would be with MSK144 though,
because that uses OQPSK (as noted below, however, this concern is purely
theoretical; there is no need for my technique to be used when doing
MSK144).

I don't have good 6M receive at home so I haven't been able to test how
the FAGC algorithm does with real-world MSK144 signals, and likely won't
get to my remote site any time soon to do that testing.  But this
morning I had a few minutes so I just quickly set up my Odyssey (Hermes
look-alike) to transmit MSK144 and one of my Hermes to receive the
MSK144, first with "normal" time-domain AGC and then with the FAGC.  In
both cases the MSK 144 was copied on every sequence so under ideal
conditions there was no discernible difference in decoding between the
two receiver AGC settings.

So at least my algorithm doesn't completely destroy the MSK 144 phase
information and totally prevent decoding :)

More tests with MSK144 when I eventually get time to do real on the air
tests with weaker signals, in the presence of significant channel noise,
with short bursts, etc will be important and may well show degraded
decoding even though things look good "in the lab".

But the results with MSK144 are really only of theoretical interest,
anyway.  The channel conditions with MSK 144 are completely different
than they are with JT65, FT8, etc.

With JT65, FT8, JT9, etc. the channel conditions are: many stations with
differing signal amplitude occupying the same frequency space at one
point in time, with multiple "conversations" occurring in the same
frequency space.

With MSK144, only one station can occupy the channel at any one time.  A
stronger signal will displace a weaker one.  So the problem that brought
about my trying FAGC is not present when doing MSK144, and there is not
a need to use FAGC when doing MSK144.  One can just use regular
time-domain AGC, or no AGC with the audio gain control backed off if
necessary when doing MSK144.

Warren, I've started reading about WOLA, but as I thought about using it
to mitigate the effects on the frequency-domain signal of making serial
time-dependent changes in the frequency-specific gain values my brain
began shouting to me that its decoding functions needed immediate
enhancement. So this will be slow going while I work on that by digging
out the DSP books that I bought long ago to read "someday".  Up until
now I have only ever learned enough about DSP and FFTs for [1] what I
needed to know when I was doing basic NMR research many years ago and
[2] what I needed to know to do the simple things that I've done in the
SDR world, so I've just scratched the surface of this domain.  It looks
like "someday" has arrived and that I will need to be digging into DSP
theory much more deeply if I am going to first make use of WOLA for this
purpose.  I am starting down that path.  This will be better for my
brain than doing Sudoku puzzles :)

Thanks Again and Have a Great Weekend all, and

73,

Roger