[GreenKeys] FSK converter madness

[Jim Haynes]

On Fri, 5 Jul 2024, Harold Hallikainen wrote:

> I'd like to do more with 850 Hz shift since, I believe, it will get
> through selective fades better. Some quick experiments are at
> https://w6iwi.org/rtty/DspTU2/#:~:text=of%20153%20mA.-,RYcount,-4/17/24 .
> There is some "impaired audio" there that people can experiment with.
>
See


Apr 1996 - QST (Pg. 78)
The Case for Wide-Shift RTTY
(Technical Correspondence)

Author: Beezley, Brian, K6STI
Article: QST Archive [PDF]
Report a problem with this entry

Apr 1996 - QST (Pg. 78)
The Case for Wide-Shift RTTY
(Technical Correspondence)

Author: Beezley, Brian, K6STI
Article: QST Archive [PDF]
Keywords: RTTY
Report a problem with this entry

However wide shift caused some problems for hams.  Such as CW ops 
complaining about using so much bandwidth to get only 60 wpm.  Irv
Hoff reported that narrow shift seemed to greatly cut down on the
amount of possibly delibate QRM to RTTY contacts.  And perhaps one
of the reasons ARRL/QST practically ignored RTTY for quite a while
And then in contests there were so many stations trying to operate
that the "excessive" bandwidth seriously crimped things.

It has long been my belief, without any documentary evidence, that
wide shift was originally chosen to improve operation in the years
when drifty equipment was serious.  In a ham qso we could keep a hand
on the tuning knob while watching the tuning indicator, but in
commercial/govenment use they would like to have a circuit operate
for hours hands-off.  Well one bit of evidence is the AN/FGC-1 FSK
demodulator designed by Bell Labs.  The discriminator in that set is
not linear, but rather uses a pair of bandpass filters that are
fairly flat over a range of frequencies.

Contrariwise, the article on RTTY in Electronics magazine, November 1944,
Robert M. Sprague was the original source of RTTY knowledge for many
of us hams.  He advocated a linear discriminator, linear far beyond
the mark and space frequencies.  I believe practially every ham designed
demodulator was defective in having the peaks of the discriminator at
the mark and space frequencies.  This was convenient as a tuning
indicator, but didn't take full advantage of discriminator capture by
the strongest signal in the passband.  Sprague didn't explain the reason
for wide linearity.  That remained for Prof. Don Wiggins W4EHU in some
articles in RTTY circa 1960.  If two signals of unequal strength are
presented to a limiter and linear discriminator the result is suppression
of the weaker signal - what is called limiter capture in FM operation.
The discriminator output contains spikes in instantaneous frequency which
average out in the post-discriminator low pass filter.  But if the
discriminator is nonlinear they don't average out to zero.  So an optimum
design besides the wide discriminator linearity would have filtering
ahead of the limiter to try to weaken undesired signals, those removed
from the mark and space frequencies.

So an experiment for the DSP guys is to build a 
filter-limiter-discriminator demodulator with a widely linear 
discriminator and experiment with filtering ahead of the limiter,
such as a twin peak filter to weaken signals not near the mark and
space frequencies.

Sprague also advocated hysteresis in the mark/space decision threshold
circuit following the discriminator.  That is, it takes more voltage
to flip the demodulator output from one state to the other than if
there is a single threshold halfway between.  Experiment with whether
this is so, and if so how much hysteresis is best.  I recall one
amateur design that used anticipation rather than hysteresis.  If it
looked like the signal was changing from one state to the other then
speed the change rather than waiting for it to be manifest.

There is a paper (which at the moment I can't find) on the optimum 
bandwidth of a post-discriminator low pass filter.  The conclusion
was that the supposedly optimum filter bandwidth was a little too
narrow for normal operating conditions.  It reduced the error rate
on really bad signals, but they were too bad to be of any use.

Now there were a bunch of papers by Elie Baghdady of MIT on the
use of positive feedback around a limiter.  I guess the idea was
that if a limiter improved the SNR then an additional limiter would
improve it more and positive feedback was a way to achieve that.
I believe his final conclusion was that the positive feedback should
be just sufficient to cause the limiter to oscillate in the no-signal
condition.  Then the slightest signal would override the oscillation.
For RTTY use an additional feature is that the oscillating limiter
would act as a mark hold if the oscillation takes place near the
mark frequency.

So there's a whole program of research on the limiter-discriminator
style of FSK demodulator to keep you busy.  It seems just possible that
the limiter discriminator circuit might outperform the two-tone
designs in the special case of amateur contests where there is a lot
of interference.  A lot of theory is concerned with receiving the
weakest of signals, which is not the problem in an RTTY contest.

Jim W6JVE