[RTTY] Multipath, testing, soundcards

[Tom McDermott]

	A lot has been written about RTTY decoding lately. It's may be
worthwhile to clarify some of the points.

1. One suggestion has been made to record RTTY signals on various paths, and
use those as a standard of comparison for testing the different decoders.
This suggestion certainly can lead to insight in how to setup parameters and
compare various decoders. It's not so clear that it captures the different
effects repeatably. A lot of work was done in the 1970's and 1980's on this
subject. Turns out that the most significant distortion in RTTY paths can be
traced back to just a couple fundamental physical effects. The CCIR put
together a standard for this, and describes the different path conditions to
be used for testing. A standalone DSP based path simulator was developed by
Johan Forrer, KC7WW, and I think a PC-soundcard based equivalent was also
produced (by I can't find by whom...).

   Johan wrote quite a good paper on his test results. But his home page has
moved, and I can't find the new one quite yet. Maybe others can point to it.
These simulators allow setting repeatable test conditions, and even better
allow setting up various paths that define the 'box' over which the modem is
expected to work.

2. A suggestion was made about what text patterns to use. A repeating
pattern (specifically:  RYRYRYRY...) is the worst possible pattern to
utilize for such test. The pattern should have the most random-like pattern
of bits. This is because badly-designed modem filters can pass the repeating
pattern test with good results, yet yield poor copy with more random
characters. Modem filters (whether implemented in hardware or software,
linear phase or not) suffer the most with long pattern lengths (due to
filter ringing). Other problems (like AC coupling in the baseband) may pass
the repeating pattern test but can fail miserably in the real world.

3. Some descriptions of what multipath does to a signal have been made. The
most dominant effect is 'selective fading'. This is caused most simply by
two-ray multpath. At some frequency, the two propagation paths differ in
length by 180 degrees but are roughly equal in strength, causing
cancellation at that frequency. That frequency drifts with time, and
ultimately one of the two RTTY tones gets notched out. Sometimes the notch
spacing can be 170 Hz, such that both tones get notched out at the same time
(modems don't copy well through that!). Another problem is 'polar flutter',
which is caused by rapid changes in the polar ionosphere. This changes the
amplitude of the signal, essentially amplitude modulating it (broadening out
the mark and space tones, sometimes much more than the modulation of the
tones themselves). It is particularly hard on the phase of the received
signal. PSK is affected a lot, RTTY not as much. Since RTTY does not care
about the phase of the received signal, it more tolerant. If the polar
flutter is too severe, even RTTY gets killed.

   A very graphic demonstration of frequency-selective multipath can be
seen. There is a free program 'DIGTRX' that is used for digital SSTV on
14,233.  It uses 8 carriers in about 2 KHz of passband. The waterfall
display shows the complete passband, and the mulipath can easily be seen as
a moving notch. Sometimes the multipath is a deep narrow notch, sometimes a
broad diffuse notch. Sometimes increasing in frequency, sometimes
decreasing. The high-Q notch (sharp, narrow) is caused when two paths exist
that are significantly different in path length. The low-Q notch is caused
when two paths exist that are almost the same length. I see the multipath
most often in the evening just before 20 meters goes long, on ~1000 mile
paths.

	-- Tom, N5EG