[RTTY] Article by Brian Beezley, K6STI

[Bill Turner]

Regarding the 23 Hz RTTY debate, here is an interesting article by
Brian, K6STI about RTTY shifts.  Brian is the author of the program
RiTTY by K6STI, recognized by many as the best of all the soundcard
programs for decoding RTTY under poor conditions.

_________________________________________________________

(Posted to the reflector with permission of the author.)


	                  The Case for Wide-Shift RTTY

	                    By Brian Beezley, K6STI



	        Amateur radioteletype began in the late 1940s with
	surplus commercial and military equipment.  Hams transmitted
	teletype signals over radio using frequency-shift keying.  In
	FSK two tones, called mark and space, alternately represent the
	digital 1s and 0s.  The tones were transmitted as amplitude-
	modulated audio at VHF or as two carrier frequencies at HF.
	Typically the tones were 850 Hz apart, a standard tone spacing
	for telephone equipment since the 1930s.  Early RTTY equipment
	used wide channel filters or FM discriminators to demodulate
	FSK, so the exact frequency shift wasn't critical.

	        It soon became apparent that the simple demodulators of
	the day were sensitive to HF-signal fading.  Unless you used
	high-gain limiter and decision circuits, the demodulated output
	would stick in mark or space during a deep fade and cause text
	to be missed or garbled.  Even high-gain circuitry couldn't help
	when the mark or space channel took a dive while the other
	channel remained steady.  This selective fading effect fooled
	the slicer circuit that determined whether a pulse was a mark or
	space.  If the space signal faded below the residual mark-
	channel noise, the slicer would mistakenly declare the pulse to
	be a mark.


	Narrow Shift

	        Plagued with selective fading, especially on the lower
	frequencies, hams began to experiment with narrow frequency
	shift in the 1950s.  Eventually everyone standardized on 170-Hz
	shift.  Narrow shift reduces selective fading because the closer
	in frequency you transmit two signals, the more correlated they
	remain at the receiver.  Selective fading occurs when a signal
	takes multiple paths through the ionosphere.  The multipath
	signals sum at the receiver.  If two signal components are close
	in amplitude but differ by about 180 degrees in phase, the
	composite-signal strength will drop dramatically.  The phase of
	multipath components depends on path length and signal
	frequency.  Mark and space signals tend to fade together more
	often with narrow shift because the tone frequencies are more
	alike.  When selective fading occurs, a signal slicer will make
	the correct decision more often with narrow shift than with
	wide.  This can noticeably improve print.


	Automatic Threshold Correction

	        In addition to narrow frequency shift, experimenters
	developed another technique to combat selective fading.
	Advanced RTTY demodulators began to employ automatic threshold
	correction.  ATC monitors the signal envelopes in the mark and
	space channels.  When they become unequal, the slicer threshold
	that distinguishes mark from space is altered.  The demodulator

	combines the mark and space signals into a composite signal by
	subtracting the space-channel waveform from that in the mark
	channel.  Without ATC, the slicer declares mark when the
	waveform is positive and space when negative.  ATC modifies the
	zero threshold.  For example, if the space-channel envelope
	momentarily fades by half, the ATC circuit adds half this amount
	to the threshold, moving it into the mark region.  This new
	decision point provides the most reliable discrimination between
	mark and space during the momentary fade.

	        A good ATC circuit is tricky to design.  The threshold
	must not be allowed to change so quickly that individual data
	pulses affect it, nor so slowly that rapid fading can't be
	accurately tracked.  AM-to-PM conversion in a poor ATC circuit
	easily can turn benign channel-amplitude differences into
	serious data-sampling timing errors.  Finally, simple ATC
	circuits may set the threshold to half of the mark amplitude
	during a pure-mark idle.  This makes the demodulator 6 dB more
	sensitive to false start pulses caused by noise.  False starts
	can cause spurious characters and loss of synchronization.  (All
	of these problems are easily overcome with software ATC where
	noncausal and nonlinear functions can be readily implemented.)
	Despite its limitations, the introduction of analog ATC in RTTY
	demodulators advanced the fight against selective fading,
	particularly when combined with narrow frequency shift.


	Diversity Reception

	        With selective fading held at bay, hams found themselves
	limited mainly by poor signal-to-noise ratio during frequency-
	insensitive fades.  When the mark and space channels fade
	together, the composite signal may momentarily drop below the
	noise.  This can cause a few garbled characters.  Worse, a fade
	may cause the demodulator or teleprinter to lose sync.
	Resynchronization sometimes may take a dozen characters or more
	after the signal emerges from the noise.  A quick, deep fade
	thus can have side effects that last much longer.

	        Diversity reception came to the rescue.  This technique
	uses two antennas, two receivers, and a combiner circuit.  The
	antennas are arranged so that the signal fades independently in
	each receiver.  You can separate the antennas by some distance
	to achieve space diversity.  (Often this can be accomplished
	with an antenna separation of less than one wavelength.)  Or,
	you can orient the antennas at right angles for polarization
	diversity.  When the incoming sky wave arrives crosspolarized to
	one antenna and yields little signal, it will match polarization
	with the other antenna and maximize its output.  For both space
	and polarization diversity, the combiner circuit selects the
	receiver output with highest S/N.  (Advanced diversity systems
	can coherently combine the two channels to provide better S/N
	than either channel alone.  These systems essentially are
	adaptive beamformers.)  It's much less likely that a signal will
	simultaneously fade in both channels of a well-designed
	diversity receiving system than in either channel alone.


	        Although effective against ionospheric fading, the extra
	hardware required for diversity reception limited its use
	primarily to commercial circuits.  (There's a chance that
	amateur RTTY diversity reception may finally become popular due
	to the recent availability of transceivers with two independent
	receiver channels and the advent of software modems.  It's easy
	to process and combine a second RTTY channel in software,
	particularly when both analog signals can be digitized by a
	single stereo sound card.  Even a simple random-wire receiving
	antenna can fill in many signal fades that occur on a high-gain
	directional antenna.)


	Mark/Space Diversity

	        While tuning for RTTY signals outside the ham bands, I
	often wondered why the vast majority I found used wide shift.  I
	might come across one or two signals at 170 Hz or 425 Hz, but
	most used 850-Hz shift--some even wider.  Watching FFT spectra
	and demodulated waveforms of wide-shift signals, I often was
	struck by how frequently selective fading was clearly evident.
	My FSK demodulator[1] implements robust ATC so selective fading
	seldom results in bad print, but I had no idea whether the
	commercial systems used something similar.  I kept wondering why
	the commercials didn't run narrow shift like amateur RTTY
	stations.  Surely the system designers were aware of the
	selective fading advantages discovered so long ago.

	        Then one day it dawned on me that the commercial RTTY
	stations must be implementing a form of frequency diversity by
	using a combination of wide shift with ATC.  With wide shift the
	mark and space channels are far enough apart that fading usually
	occurs independently in each channel.  When the mark signal
	disappears, ATC automatically reverts to space-only copy by
	shifting the threshold over to that side of the demodulated
	waveform.  (Without ATC, you've stepped back 50 years to fight
	the selective fading battle all over again.)  The key idea is
	that ATC lets you use wide-shift selective fading to combat
	narrow-shift frequency-insensitive fades.  It was suddenly clear
	that the combination of wide shift with ATC offered superior
	immunity to fading of all types.  It provides automatic, built-
	in frequency diversity!

	        Absent fading, the bit-error rates for well-designed
	narrow- and wide-shift FSK systems are essentially identical
	because it's possible to optimally filter the mark and space
	tones independently.  Therefore, wide shift with ATC should
	provide better results than narrow shift because the combination
	decorrelates channel fading to achieve frequency-diversity
	reception without incurring a signal-to-noise ratio penalty.

	        Mark/space diversity won't work as well as a two-
	receiver, two-antenna diversity system.  When you demodulate FSK
	using just one tone, you may lose up to 6 dB in signal-to-noise
	ratio (the same noise power competes with half as much signal
	amplitude, or one-quarter as much signal power).  Since an
	adaptive-combiner, dual-receiver diversity system can provide as
	much as 2-dB processing gain during a single-tone fade, it

	should have up to 8 dB S/N advantage over simple mark/space
	diversity.  Moreover, it's much less likely that all four
	channels in a full-diversity system will simultaneously fade
	than it is for both channels to die in mark/space diversity.
	But all you need for mark/space diversity is wide shift and good
	ATC.


	When and Where

	        I think a wholesale change by amateur RTTY stations back
	to 850-Hz shift would be a disaster.  The bands are just too
	crowded these days.  In addition, simple FSK demodulators may
	suffer significant S/N degradation when set for wide shift (they
	may respond to noise in the region between tones).  But if
	you're operating RTTY on an uncrowded band and encounter severe
	fading, try switching to wide shift.[2]  If your demodulator has
	good filters and effective ATC, you're likely to get much better
	copy.




	[1]  I use RITTY 1.0 DSP software that runs in my PC and uses a
	Sound Blaster card for analog I/O.  RITTY has a limiterless
	front-end, optimal matched filters, ATC, numerical flywheel, FFT
	tuning indicator, and other features especially designed for
	weak-signal recovery.

	[2]  FCC rules permit a frequency shift of up to 1 kHz.