Eternal, Top posting 'cuz U R.



No, I think I had it in the first place, or at least, was close. It's
just difficult to express in few words. Lets see if I have it...



Although you don't say it, I thought that this "pattern" of phase shifts
is not simply alternating phase shifts at regular intervals, but a definite
pattern with some of the times between shifts being, lets say, twice as
long, or 20ms. This pattern is what is unique and what your (so called "non
computer") detector will be looking for to generate a "beep" (what you
symbolize with the *) to be listened to, right?



I got this from your:

... I found no less than eight
orthogonal logic conditions by which I could identify the
presence of this pattern in at most 10 time units...



Assuming you have a basic understanding of PSK31, what I think you are
proposing can be described in terms of PSK31. What if we simply use ONLY
ONE letter of PSK-31. However, we send it over and over. This gives your
detector something to recognize to ake the audio * beep for the listener.
If is doesn't see that specific _pattern_, it won't make the * beep.

Then just key the whole thing on & off just like ordinary CW.



I don't have the code here, so I'll make something up and say we use a
pattern of phase shifts like this:



_ _ _ __ Phase "A"

| | | | | | | |
| |_| |__| |__| |Â*_ Phase "B"



Looking just at the Phase A part for simplicity, we have "short, short,
pause, short, pause, long". If we did this correctly, we will not falsely
decode this if we slip just one time slot in a longer string of these
patterns. We must be in "bit sync". That is, the receiver system (not the
ear-listening-to-the-audio-out-of-the-receiver), must "know" where we are in
this pattern at all times.



This gives us 11 times or 110 ms. [[we need to set the timing so you can
get one or more of the patterns into one Morse dot, but I'll ignore that for
this discussion]]





Let's say this is a Morse DOT. String three of these together to get a
dash - which will be 330 ms long. Sounds like pretty slow code, but in
principle, this is what I thought you meant.



In reality, if this was a truly synchronous system, the pattern could
actually be a truly pseudo-random pattern which doesn't repeat in such a
short time, but may be seconds long. If we provide some "key-down" time at
the start for the receiver to find its place in the pattern, and achieve
"bit sync" this would also "work".



To firm up understanding, I compare this to the following "tone
detector" system. Assume we have, for the moment, returned to ordinary CW.
Tune your receiver for an 800 HZ tone and build an 800 Hz tone-detector.
Use the output of this tone-detector to Key an audio oscillator to produce a
"* beep" which you can listen to and copy the CW with.

What I have for your Phase shift system, simply replaces the CW with
this PSK pattern and the tone-detector with a "phase shift pattern
detector".



This is certainly "a way", but the problem of the modulation sidebands
is still there. The ear will hear a T6 Morse signal, but your demodulator
system still needs considerable computing (or calculating) power if using
"circuitry" of some kind.

In my previous post, I mentioned the calculations for FM not PM, but the
result has to be similar. The filtering used to define your phase shift
transition times will determine the modulation index and therefore the
bandwidth. 100 Hz modulation is 100Hz modulation and the sidebands need to
be there at the demodulator to demodulate the waveform and the receiver will
still require a 200Hz bandwidth. It will also have to be something rivaling
a computer to sync up and detect the pattern so it can properly "know" when
it is present and when it is absent and make the beep for the "enhanced
system" CW operator.

So it seems to me that you have a 200HZ CW system and that can't be
better than the standard 10HZ or whatever it is---not to mention the ability
of the ear-brain to pull CW out of the noise. The *ORIGINAL* DSP, no?.



Make sense?

73,
--
Steve N, K,9;d, c. i My email has no u's.



P.S. U got a real name?




"The Eternal Squire" wrote in message
...
Steve,

Evidently I've confused a few people as well as yourself.

1) The frequency shift keying contains no intelligence, only clocking
along a time axis.

2) The phase shift keying contains no intelligence, only clocking
along the phase axis.

3) The audio tones being shifted contain no intelligence, only
clocking along the amplitude axis.

The pattern below describes the shifting between tones A and B

Phi-0 Phi1-1
| |
_ _ v_ _v Tone A ^
| | | | | | | Frequency clocking.
|_| |_ _| |_| Tone B v

Note the phase directions at Phi-0 and Phi-1, each is the reverse
of the other.

Now, suppose the time unit is 10 millisecond. If I've done my job right
and can correctly detect the pattern to exist at each time
unit that the pattern travels through a special detector, and assuming
that our logic high from the detector runs a beeper, the human ear
should hear a solid beep for the 80 millisecond that the pattern
travels through the detector.

* * * * * * * *

Let's emit this pattern for just a couple time units longer:

_ _ _ _ _ _
| | | | | | | Frequency clocking.
|_| |_ _| |_| Tone B v

so that we can hear a solid beep for 100 millisecond

* * * * * * * * * *

Now you can humor me and imagine a contracted time scale, where
a plus represents ten stars, this is also 100 millisecond.

+

and if we continue the pattern for 2000 millisecond, we can hear
a solid beep for that period of time.

+ + + + + + + + + + + + + + + + + + + +

So we now have the equivalent of an unbroken carrier. Suppose
we turn off our pattern emitter, our detector fails to lock
onto the carrier, and so we hear silence for those periods.

+ + + + + + + + + + + + + +

And so we get the Morse letter K !!!


So, the pattern itself would be heard as morse code which
sounds like a buzzsaw on an ordinary receiver, at the
usual distance range.

If we place a special detector after the audio in an
ordinary receiver, the morse would be heard as a pure
tone from a greater distance without benefit of computer.

If we place a computer after the audio in an ordinary
receiver, then the morse would be heard at a greater
distance than can normally be detected by a computer.

Make more sense?










Steve Nosko wrote:
Bottom POST:

"The Eternal Squire" wrote in message
...

I don't understand what threshold effect is.

I'm sure everyone saw the FSK/PSK alternative to the analog FM scheme,
what would be the effect of thresholding would be on either alternative?

Giving the system more thought from input by other people, I am
probably leaning toward a repeated FSK/PSK pattern.

I worked out a pattern that shifts both in frequency and phase
at regular intervals:

_ _ _ _ _ _ _ _
| | | | | | | | | | | |
|_| |_ _| |_| |_| |_ _| |_|

where each underscore character is one time unit, and each
vertical bar is one frequency unit. Ignore the space in between
the underscore characters.

I drew up some timing diagrams, and I found no less than eight
orthogonal logic conditions by which I could identify the
presence of this pattern in at most 10 time units.

It would seem to me that if the incoming signal satisfied even three
of these measures for any given time unit, I could conclude that that
signal matched the pattern and emit a logic high. That may defeat many
problems introduced by a high noise floor.

That logic high could drive a tone generator to headphones. If each
time unit were 10 milliseconds, then a dit would be easily audible
if the signal satisfied the pattern at 5 out of every 10 time units.
More thoughts anyone? The Eternal Squire


"Squire",

1st, Threshold is best understood at an intuitive level with the
explanation
of the noise in my previous post. Because the noise is not correlated
to
produce an output, but the modulation is, the noise does not contribute
as
much as the desired sidebands, so as soon as you have enough sideband
power
at the receiver that isn't corrupted by the noise, you get signal out.
The
less the noise is correlated (the wider the bandwidth) the quicker this
happens as the signal increases out of the noise.

On to your new system:
What you now describe looks almost like PSK31, but somehow coded in
Morse
rather than the one used for PSK31. If the "warbles" are Morse based,
then
the CW operator should be able to decode it. It won't be as easy as CW,
though. The warbles are pretty subtle.

I see two things in your description.
1 - Is that it is still no better than CW bandwidth wise for the ear and
you'll need a very narrow Rx to reap the benefit. In addition, the
warbles
are much more subtle than on-off keying -- for the ear.
2 You talk about "orthogonal logic conditions", which sounds just like
the
code used for PSK31, no?? It has codes for each character which are
unique
and not confused with each other --a.k.a. "orthogonal". So, on the
detection side you still need the hardware (read computer sound card) to
decode it for your "full" system.
Sounds like you re-invented PSK-31 (which can sort of be decoded by
ear). Except, I think there may be a conflict trying to find the PSK
patterns which fit within the Morse characters and remain orthogonal.
Hmmmm How about only "processing" this with the sound card to do
the
narrow band filtering then send it out as audio tones for the final
decoding
by ear for the last error correction step.