[Laser] Re: Pulse Density Modulation - PPM to PSK31

[TWOSIG at aol.com]

Tim

Thank you for the response on PDM.   I was aware  that it is used for high 
efficiency (class D, and now I think class E )  amplifiers, but did not know 
that the pulse rate is about 2 MHz.

It  is interesting to contrast PWM and PDM.  One reference I found suggests  
that PWM is a special case of PDM.  I see the logic, but I prefer to think  of 
them separately.  Lets say that you want to pulse modulate a signal that  has 
frequency component of 20KHz.  Sampling theory says you have to  "sample" at 
least 40KHz, and realisticly for PWM you should sample faster than  60KHz, or 
a sample period of 16.67 microseconds.  With a 2MHz PDM, that  sample period 
would have 33 pulses which can be on or off.  The result  would be 33 discrete 
levels of intensity within the sample period.  If you  were to cluster all of 
the "on" pulses to be adjacent in time, leaving the "off"  pulses to fill the 
rest of the sample period, then I can see the argument that  PWM can be 
expressed as a special case of PDM.  The more common form of PDM  would disperse the 
on and off pulses, so I would expect that filtering the  output would be 
easier.  The high frequency component of the output would  be a sub-multiple of 
the clock frequency, and vary depending on the input.   ( I will have to think 
on that a while yet. )

This suggests that  PDM has a relationship between sample time, or said 
differently the high  frequency component of the input, and the resolution of the 
output.  Lower  frequency input allows the same clock frequency to use more 
time slots for each  sample, hence greater resolution.  (  I think that can be 
used for  multi-frequency light comm.  More later. )

I am not sure about  the limits to the pulse rate and repetition rate on a 
Xenon strobe tube.   If the rep rate is not fast enough for PSK31, then I would 
suggest that Pulse  Position Modulation be considered:   Assume that it takes 
at least 20  milliseconds to reset for the next pulse.  I would suggest that 
you  allocate 25 milliseconds for the overhead to set up for the pulse.   
Assuming that you want to send ten characters per second, or 100 milliseconds  per 
character.  You have 75 milliseconds to choose when you fire the next  pulse.  
If you want to model the ASCII 7-bit code, you allocate a time slot  75/128 
milliseconds for each character.  If you restrict the character set,  you can 
do better than the 586 microsecond in this example.  The pulse  leading edge or 
trailing edge will need to be detected with enough accuracy to  be able to 
identify the character.  This system would have a very high peak  to average 
power figure.

If you have a light source that can be  fired much more rapidly, say a pulsed 
laser, then PSK31 should be  possible.  Suppose that you can excite the laser 
with an LED or other  source that is fast enough to follow the amplified 
audio from a sound  card.  Even if the laser only sends a short pulse, each of 
those pulses  will fall in the time window of the positive (or negative ) peak of 
the audio  signal, which will preserve the phase information needed for  
decoding.

A more efficient way to use pulses, would be to fire one  pulse per bit, 
timed so that you can tell which phase is being sent ( two  possibilities with 
BPSK, four with QPSK ). The transmitter and receiver system  need to use the same 
frequency (tone) for PSK.   On the receive side,  you could send a tone burst 
triggered by the pulse detection.  Assuming  that BPSK is used, the tone 
burst would need to be about 30 milliseconds  long.  ( One way would be to run a 
clock through a counter chip with a  reset line.  If you toggle the reset line 
when the pulse is detected, the  output of the counter is at a known phase.  
If a pulse is missed, the  counter output will simply continue with the last 
phase synchronization it  received.

On the transmit side, it would be easier if the computer  output included 
either the varicode bit stream, or a square wave of the output  tone without the 
filtering that forces the phase changes to take place at near  zero amplitude. 
 However, a one-shot with triggering about eighty percent  of the tone peak 
amplitude should give a reliable phase sync.  It should be  obvious, that PSK31 
tone frequency should be very low.  A 250 Hz tone has a  two millisecond 
difference in the phases of a zero and a one.  An 800 Hz  tone has only 625 
microseconds.  Tighter tolerance is needed for the pulse  detector for higher tones.

In essence, this system uses a pulse  position modulated signal to 
communicate the phase information used for  PSK31.  The receiver could decode the pulse 
information into a bit stream,  instead of creating a local PSK31 signal.  I 
do not suggest this is an  eligant solution, just one that has a fair chance of 
working if  built.

James
N5GUI


#################################################################

Previous  message:
 
 
hi James,
My emails to the mailing list are getting rejected for
some  reason - it says i have an attachment attached,
but i dont... anyway i  figured i'd send this directly
to you and maybe you can forward it to the  group,
thanks.
 
PDM looks like a perfect match for xenon strobe tubes
and pulsed lasers  where you have no choice but one
pulse width - or at least not easily pulse  width
modulated. And it would use a very simple 'detector'
also, a low  pass filter or a receiver that has a
bandwidth lower then the basic pulse  rate at least. On
the other hand, PDM looks more complicated to  generate
than PWM, but there are chips out there that can do it
i  assume.
 
One 'all digital' (class D) Hi-Fi audio power amp i
saw online recently  had a basic PDM pulse rate of 2
Mhz for a bandwidth of 20Hz to 20Khz. PDM  amplifiers
are smaller, more efficient and cheaper than linear
amps, while  maintaining excellent audio quality.
 
I had been wondering about how to modulate a strobe
tube with PSK  somehow. And since you're pretty much
stuck with one pulse width and  relatively slow pulse
rates with strobes, how to convert PSK into  something
that could easily modulate them, and at the same time,
also be  able to use the same standard psk software to
receive. 
A PDM version of  the PSK31 waveform at 'baseband'
rates seems perfect since you can get  maximum power
output from the strobes at low pulse rates (31.25 Hz)
-  that's a problem though without modified PSK
software.
 
 
 

Tim Toast
 
Re: Pulse Density Modulation  -  PPM to  PSK31