RadioBanter - View Single Post

July 6th 05, 09:11 PM

wrote:
Michael Coslo wrote:

wrote:

Dee Flint wrote:

Please show me and everyone else how we can run more than 300
baud on HF
without exceeding reasonable band widths. There are a whole
lot of things,
not just video, that would be nice to do.

The answer (in large part) is to use different modulation and encoding
schemes. Such as QPSK instead of BFSK. Multiple carriers spaced just
far enough apart to avoid interference. Simultaneous AM and PM.

Of course such modes may not have the HF performance we're used to
from, say, PSK31 or SSB. You don't get something for nothing; Shannon's
Theorem shows that the increased data rate in a given bandwidth comes
at the price of needing more S/N to get acceptable results.

That old s/n bugaboo. PSK31 does very well at low power levels in great
part because the PA of present day transmitters work well at lower power
levels. Crank 'em up to full power, and things aren't always so pretty.
You can visually see the results on the waterfall, just as you can see
when the drive level is too high.

That's transmitter IMD - just one part of the problem.

Now is that an insurmountable problem? No. But it means that we would
probably have to design new transmitters that can put out a PSK-friendly
output at high power levels.

Which can be done and is being done.

Another method is of course to increase the ERP. So we can put up a
directional antenna.

And an amplifier.

Better be a skweeky clean one!

But now we're certainly a long way from simple, and many present day
rigs and users wouldn't be able to participate in the fun.

But all that is doable. After all, Morse Code has a big S/N advantage
over even SSB voice, yet hams manage to make SSB contacts on HF. Morse
Code simply does better under marginal conditions. Or perhaps we should
say that its margins are lower.

I wonder if any experiments have been performed by Amateurs along these
lines to find a practical limit to how many different phase angles can
be accommodated. Sounds like fun.

The problem is in two parts: First there's the accuracy of the
hardware, which is probably pretty good using modern parts and methods.

But second is the distortion of the RF path, which is not under our
control. If, say, the ionosphere causes the phase to wander a couple of
degrees each way, modes like BPSK and QPSK may still work, but "256PSK"
will be full of errors.

Note how there are times when PSK31, even if the signals are loud,
won't work in QPSK mode but will work in BPSK mode. That's not because
of narrower bandwidth or some hardware or software change. It's because
the path is introducing so much phase distortion (a form of noise) that
the distortion exceeds the QPSK demodulation criteria.

W0EX observed path-induced phase distortion that was so high that PSK31
wouldn't work even in BPSK mode, yet the PSK31 carrier could be heard
clearly and seen easily on the waterfall.

As K0HB points out, the 300 baud limit only applies in the CW/data
subbands. If you can stuff "TV" into a reasonable bandwidth, it can be
sent in the voice/image subbands.

As a simple example, imagine 25 PSK31 carriers spaced cheek-by-jowl in
a typical SSB bandwidth, running QPSK. If one gets you 100 baud, 25
will get you 2500 baud. Go to 8PSK and you get 5000 baud. Then add
compression on top of that...

PSK31 is 31 baud, so the numbers have to be shifted a bit.

Yes - I should have written "PSK31-like"

The baud
rate was chosen because it is around the level that a good typist can
type at. But it can be changed PSK100 baud easily, just sacrifice a bit
of bandwidth.

Exactly. The principle is what matters. The problem is that the
transmitter and receiver must be very linear to avoid IMD products
causing trouble.

How can we do it? Bandwidth is directly related to baud rate.

Only if "all else is equal". The trick is to make the tradeoff
somewhere else. The familiar "56K" modem trades off S/N rather than
bandwidth.

It's not a complex subject at all.

You've probably heard the old engineering adage:

"You can have it fast, good or cheap. Choose any two"

Same with light bulbs:

Bright, long lasting, or cheap.

I'd say "bright, long lasting, or efficient".

Maybe, but you ought to see people choke when I tell them what their
replacement bulb in a data projector is!

Bright, long lasting and .5 kilobucks.

All Shannon's Theorem does is equate fast to data rate, good to S/N,
and cheap to bandwidth.

There's also the factor of error rate. In the above simplified
discussion I assumed the same error rate for all cases. Obviously there
are some situations where a higher error rate is tolerable.
Error-correction can help, but error correction carries its own
overhead, slowing down data rate.

That it will. PSK *is* a delicate mode, and gets more delicate the more
phase shifts in use. Interestingly enough, it works pretty well at
frequencies where there is more bandwidth available

A dramatic example of the effect of errors can be seen on TV.
Conventional analog NTSC-type TV shows "errors" as "snow" and sometimes
even loss of sync. But you can still watch a "snowy" picture. Digital
TV methods often show errors as pixelation or complete loss - you see
*nothing*.

That brings up a useful analogy. When I got started in video, the
portable work was being done with the old 3/4 inch U-Matic tapes. As
time progressed, we shifted to formats like Betacam, S-VHS and the like.
Eventually some interesting formats such as Hi-8 came out. The Hi-8 had
a pretty decent video quality to it, and looked like it was going to
revolutionize things.

But there was a problem. I'm sure you are familiar with the way that
modern video lays the tracks down on the tape - the record heads are at
an angle, and there are at least two of them. This way the tape is
"striped", with the video laid on at an angle to provide more linear
space with which to write the image.

This was videotaping's bandwidth "cheat". And it works fairly well.

But the needed bandwidth didn't go away, and there was tremendous
demand to make the tape smaller and more slender.

So the fix was to lay the heads at an even greater angle, so as to
compensate for the smaller tape. What may have been a 70 degree angle
for the 3/4 inch tape might be 45 degrees for 1/2 inch tapes, and
perhaps 22 degrees for the 8mm tapes. (note: approximate angles)

A dropout - a fairly common thing - on U-Matic would just make a little
white dot on the screen. On the 1/2 inch tapes, an entire line might be
lost. On th 8mm tapes the entire signal might go away for a little bit.

This is because that missing oxide will be cutting across a lot more of
those stripes on the tape as it got smaller.

- Mike KB3EIA -