[CQ-Contest] Fwd: Why?

[Hank Greeb]

Andy:

My only comment:  Excellent analysis and hypotheses.

How can we entice a sufficient number of contest ops "pack the bands" in 
a contest using BPSK (31 or 63) to determine if the bandwidth advantages 
for PSK outweigh the disadvantages concerning multi-path distortion, etc.?

72/73 de n8xx Hg

-------- Forwarded with permission --------
Subject: 	[CQ-Contest] Why? (off list)
Date: 	Wed, 7 Jan 2015 03:55:54 +0000 (UTC)
From: 	aflowers at frontiernet.net <aflowers at frontiernet.net>
Reply-To: 	aflowers at frontiernet.net <andyk0sm at yahoo.com>
To: 	n8xx at arrl.org <n8xx at arrl.org>



Hank,

I saw your post to the CQ-Contest reflector asking about why RTTY and 
PSK-31.  I'm not subscribed to the CQ-Contest reflector, but I do scroll 
through the chatter from time to time. Your question actually touches on 
some really interesting technical subjects that don't often get 
discussed, so I thought I would address them as best as I could.  You 
have my permission to forward this to the CQ-Contest reflector if you 
wish.  If you do, please do so in whole.

Your question about why not use a more "spectrally efficient" modulation 
method than 45.45baud 170 Hz-shift Baudot for contesting, like one of 
the PSK modes.  I think a more efficient contest mode is a great idea 
(and I'll get to the inefficiencies in RTTY in a moment), but finding a 
substitute is not as simple as "spectral efficiency" from some 
equation.  Most importantly, one has to be careful when you talk about 
"spectrum efficiency" without taking into account the nature of the 
channel.  Contesting puts a premium on maintaining a station that is 
able to communicate under less-than ideal channel conditions.  In NAQP 
you usually have a bunch of multipliers on the backscatter path on 10 
and 15 meters (a real challenge here in the Northeast).  In DX contests 
you have to work that EX8 over the pole when the K index is 5.  Unlike 
the cable company or the SW broadcaster, you don't get to choose the 
optimum channel; amateur radio (and radiosport in particular) puts a 
premium on those unusual, weird, and even bizzare propagation 
phenomena.  This has interested me since I was a child and I think 
always will.

So, without getting into theorems and equations here are two major 
impediments to copying a digital code through an HF channel:

1) Noise. There's a minimum signal-to-noise ratio that will work for any 
given scheme.  All things equal, more noise in the channel means more 
bit errors. That means worse copy.  Noise might not be time or amplitude 
constant (e.g., QRN, QRM).
2) Intersymbol interference (ISI) due to multipath and other ionospheric 
effects.  You can think of the ionosphere as a filter with randomly 
changing group delay and amplitude variations.  We call this "QSB", 
"selective fading", "auroral flutter", "watery signals" or whatever.

In grossly oversimplified terms, you can think of a binary FSK 
demodulator as looking at two frequencies every 22ms (for 45.45 baud 
RTTY) and asking "is there more energy in the 'mark' frequency or the 
'space' frequency?"  It outputs whichever bucket wins at that moment.

If you have a varying multipath time delay and amplitude variations 
between the mark and space channels the bits to smear into each other by 
the time it arrives at the receiver.  When this happens the demodulator 
sees both mark and space energy at the same time.  From the point of 
view of the receiver at any one instant, the ratio of energy between the 
correct bit and the incorrect bit is less than it would be in a channel 
without the time delay.  The demodulator is more likely to make a bit 
error, and this means that the SNR is effectively a little worse than it 
would be in a channel without the time distortion.  For disturbed paths 
you can usually push a binary FSK signal through with loud signals.  
There is a benefit to QRO and loud signals.

Incidentally, the 170 Hz shift is enough to ensure that the channels 
have essentially uncorrelated ionospheric effects, so you can play 
tricks at the demodulator to track the independent channels in an 
attempt to put humpty-dumpty back together again.  An extreme case is 
what serial-tone HF modems do with very fast sample rates--they actually 
try to reconstruct the transfer function of the ionosphere and track it 
over time and put the bits back in the right order...of course, this 
requires quite a bit of bandwidth, but it may be worth it in some cases 
depending on how fast you want transfer data and the state of the 
channel.  The most extreme case I can think of is a STANAG variant out 
there that uses about 3 KHz of spectrum for what amounts to 75-baud 
effective throughput.  More occupied bandwidth can be better if that's 
what the channel demands and you don't have to worry about your channel 
neighbors.  A good contest mode requires us to balance both of those.

...so there's a cautionary tale about "spectral efficiency".  It really 
depends on what you are trying to do, what the channel is really like, 
and who your neighbors are.

Anyway, unlike FSK, a binary-PSK signal compares the *phase* of a 
carrier from one bit to the next.  In the case of PSK31, it looks at the 
carrier at 31ms intervals and asks, "Is the phase of the carrier the 
same or 180 degrees off from the last sample?"  The requirement here is 
very different from the FSK case--the ionosphere must be phase stable 
from bit to bit at a particular frequency.  If the phase stability isn't 
there (and it often isn't on those disturbed paths contesters are 
interested in) the phase of the carrier is jumping around more rapidly 
than you are going to get bit errors.  Perhaps a little more surprising, 
louder signals are not going to help in this situation--*if the phase 
stability isn't there in the channel, 500KW isn't going to allow 
communication from point A to point B.*  That might be a bit frustrating 
in a DX contest!

The converse is also true: if the necessary SNR is there and the channel 
is phase-stable, you don't need more than a QRP transmitter most of the 
time.  That's one of the great things about PSK.

Anyway, if you want to blue-sky a potentially workable idea, how about a 
binary FSK source-coded scheme for callsigns and exchanges that uses a 
22 ms bit period (same as 45.45 baud RTTY so everyone can use legacy 
hardware, and modifications to existing software would be 
minimal).  JT65 uses less than 28 bits to encode a callsign.  That's the 
equivalent of 5-1/2 symbols in 5-bit baudot.  For comparison, a nice 
short callsign like "K0SM" requires 6 symbols in Baudot and 9L/K0SM 
needs 12, or 60 bits!.  In the same 60 bits with a more efficient code I 
could send both calls and the exchange, and possibly a good deal of 
error correction!  Run rates could skyrocket for those who like that 
kind of thing.

I will say however, that I have had a good deal of fun building a Baudot 
FSK demodulator for HF--the code is terrible in a lot of ways, but those 
make for really interesting optimization problems!

Anyway, I hope that's useful to get you thinking :-)

Andy K0SM/2