[GreenKeys] TU replacement

[Jim Haynes]

Ancient times:  Armstrong and Schmitt independently proposed tuned
circuits for the mark and space frequencies, subtract the rectified
outputs of the two channels to determine mark or space predominance.
The wish was that noise would affect both channels equally and cancel
out, while signal would affect only one and prevail.

Classical period:  Sprague of Press Wireless wrote in November 1944
issue of Electronics magazine.  The FSK signal is treated as FM and
demodulated in a limiter-discriminator system.  Sprague says a little
about a bandpass filter ahead of the limiter, and says without a very
clear explanation that the discriminator must be linear over a frequency
range far in excess of the deviation band.  Then he goes on to advocate
AC coupling from the discriminator output into a threshold corrector
to remove the effects of frequency drift in the system.  And further
advocates introducing hysteresis using neon bulbs in the coupling
circuits.

It seems that many amateurs were aware of Sprague's article without
entirely understanding it.  Nearly all amateur TU designs of the limiter-
discriminator type suffered from several deficiences.  First they did
not have adequate filtering in front of the limiter, and simply relied
on receiver bandwidth.  Second the discriminators were not linear far
beyond the deviation width.  Instead they tended to be a pair of tuned
circuits centered on the mark and space frequencies.  This was a 
convenience in tuning and in setting the transmitter shift fairly 
accurately.  Third they did not employ much in the way of low-pass
filtering after the discriminator.  All this did not make its way into
the amateur literature until Don Wiggins wrote an article in RTTY in
November 1960.  The limiter improves SNR only when the desired signal
is stronger than any interfering signals.  Thus there needs to be
something like a twin-peak filter ahead of the limiter to try to suppress
interference that is not dead-on the mark and space frequencies.
Then the effect of interference on the output of the limiter is to
cause spikes in the instantaneous frequency.  These will average out
to zero only if the discriminator is widely linear.  And the post-
discriminator low pass filter is what makes them average out.

I haven't come across any corroborating information for Sprague's
recommendation of hysteresis in the signal path.  And there was at least
one amateur TU design that employed what might be called anticipation,
the opposite of hysteresis.

The immediate pre-modern TU architecture shifted to what is called
two-tone detection.  Instead of treating the signal as FM it is treated
as a diversity pair of on-off keyed signals, detected separately and
then subjected to some decision process.  For this purpose the mark and
space filters can be quite narrow.  This concept was the subject of
publication as early as 1947 and later in 1956 in papers published in
England.  I believe the first publication in the amateur literature was
in RTTY in 1963 by Frank Gaude.

With the two tones detected separately there are four possibilities for
the combination of detector outputs.  In two of these they agree: mark-on,
space-off and mark-off, space-on.  In those cases we know what we want to
call the output.  It is in the other cases mark-on, space-on and mark-off,
space-off. that we can do something interesting.  Mark-off, space-off
could mean the signals have faded out.  Mark-on, space-on is likely the
result of multipath propagation, or of interfering signals.  In the
ST-8000 the logic in the conflicting cases is to leave the output at
whatever it was before the conflict occurred.  Another possibility would
be just to assert mark, since that is more probable than space.  There
have been quite a variety of threshold-determining circuits applied to
the individual detectors or to their combined signals.

A good source of knowledge of all this stuff is on Chen's web site 
w7ay.net

I'm not much of a theoretician, so I don't know why the phased lock
loop performs poorly in the presence of interfering signals.  Perhaps
it is that the circuit is inherently a limiter, so you need sharp
filtering ahead of it for good FSK performance.

 	---

 	"Ya can argue all ya wanna, but it's dif'rent than it was."
 	"No it ain't! No it ain't!  But ya gotta know the territory."
 		Meredith Willson, The Music Man