[GreenKeys] on-off radio teletype (not FSK)

[Jim Haynes]

In the days of 850 Hz shift for hams, it was popular in TU designs to
be able to switch off one or the other tone and copy single tone.  The
reason was the tendency of CW operators to sit on one of the tone 
frequencies to try to disrupt the operation.

One thing I have learned over the years is that very few if any amateur
TU designs were done correctly.  The original paper we all read was by
Robert Sprague, "Frequency-Shift Radiotelegraph and Teletype System"
in November 1944 issue of Electronics.  In that article he talks about
a TU design which uses AC coupling and and a threshold restorer so that
the near-DC offset in the discriminator output caused by frequency drift
is removed.  He further states that the discriminator needs to be linear
far beyond the width of the signal, without explaining why.

Just about all the ham designs, and some commercial ones, violated that
condition of the discriminator.  A commercial example is the Bell Labs
designed AN/FGC-1, which uses bandpass filters for mark and space rather
than a linear discrminator.  I'm sure the reason for this was to tolerate
frequency drift in the signal by having the pseudo-discriminator outputs
be fairly unchanging as the signal drifted, until it went too far.

The ham designs almost all put the discriminator peaks right at the mark
and space frequencies.  For one thing, this made a tuning indicator
convenient, because you could set the transmitter frequencies right at
the peaks and know your shift was correct.

Don Wiggins W4EHU an electrical engineering professor wrote some articles
in RTTY beginning in 1960 where he explained the reason for making the
discriminator much wider than the signal peaks.  When a strong signal and
a weaker signal are put through a limiter, what comes out has the average
frequency of the stronger signal.  This is called limiter capture.  But
the weaker signal causes spikes to occur in the instantaneous frequency,
which result in spikes in the discriminator output.  If the discriminator
is widely linear these spikes will average out to zero and disappear
when the discriminator output is passed through a low-pass filter.  But
if the discriminator is nonlinear the spikes do not average out to zero
and and the discriminator output is distorted.

Then he goes on to stress that the other important thing is to limit
bandwidth ahead of the limiter, so that the desired signal is as much
as possible made stronger than the interfering signals.  This calls for
a twin-peak filter ahead of the discriminator to enhance the desired
signal as much as possible.  By the W4EHU published these observations
the ham community was swinging toward limiterless two-tone TU designs,
so I don't suppose much came of his work.  If you consider something like
the ST-6 and other TU designs of that period it seems that in two-tone
limiterless mode they work pretty well; but with the limiter switched on
they are all wrong.