RTTY filter for K3

[KK7P]

>> Using a excessively narrow filter on RTTY is a balancing act.
>>
>> If the bandwidth is too narrow the tones tend to become "smeared"
>> - the transitions become less easy to detect (in other words,
>> the "eye" begins to close).  However, narrowing the bandwidth
>> also reduces the amount of QRM/QRN making it through the filter
>> and to the decoder/detector.
>
> How about using a Nyquist filter?
>
> A Nyquist filter is one that results in no inter-symbol interference.

One of the difficulties with RTTY is that the tone spacings are so wide.
  For a 45 baud FSK signal, a shift of about 30 Hz would minimize the
occupied bandwidth and have no intersymbol interference.  But standard
practice is to use 170 Hz or 200 Hz shift. [ 200 Hz Shift is about right
for 300 baud FSK, which is what we used (and why we used it) for HF
packet some 25 years ago.  And optimized it for common 400 Hz to 500 Hz
  wide "CW" crystal filters of that era. ]

Thus, the rise of the dual tone filter (K3) and its work-alikes in
various radios and demodulators, with fairly tight filters around each
tone to pass it along with its significant sidebands.  But the overall
IF passband still needs to be wide enough to pass both tones and their
sidebands. I imagine you can observe all of this on your "P1".

Some RTTY decoders take advantage of the tone separation and actually
employ a pair of detectors, one for each tone, and use the outputs
differentially to get some protection from selective fading.  170 Hz
spacing is probably too close for such protection -- 850 Hz spacing
would be better, but that requires an even wider IF passband...

Some soundcard-based RTTY programs support really narrow shifts, but
such signals are not often heard at this QTH.  Maybe it is because
everyone is listening and no one is calling CQ?

Enjoy!

Lyle KK7P

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Alan Bloom]

Re: using Nyquist filters for RTTY.

One complication I didn't mention is the fact that, because angle
modulation (FM, FSK, PM, PSK) is fundamentally non-linear, the
pre-detection and post-detection filters are fundamentally different.

With linear modulation types (AM, on-off keying, QPSK, QAM), the
spectrum of the pre-detection filter will appear the same at baseband.
For example, if the IF filter has a 0.3-alpha Nyquist bandpass response,
the resulting post-detection response at baseband will also be a
0.3-alpha Nyquist lowpass response.

But that's not true with angle modulation.  Using a Nyquist
pre-detection filter does not result in a Nyquist response after
detection.  That is why most FSK and MSK systems use non-Nyquist filters
(typically Gaussian).

To use a Nyquist filter with FSK you would have to use a pre-detection
filter much wider than the baud rate (as is done now for RTTY) and then
add a Nyquist post-detection filter that would further tighten up the
frequency response without causing additional inter-symbol interference.

There may be a good reason why this is not usually done.  When I get
time maybe I'll do some Mathcad simulations to see if it looks like this
could work.

> Some RTTY decoders take advantage of the tone separation and actually
> employ a pair of detectors, one for each tone, and use the outputs
> differentially to get some protection from selective fading.

I think it is not just for selective fading but also for interference
that occurs primarily on one of the two tones.

It might be possible to use separate Nyquist filters and detectors for
the two tones, treating each one as if it were a separate on/off-keyed
signal.  If the transmitted signal is not filtered too heavily (i.e.
jumps quickly from one tone to the other) it seems like that should
work.

Al N1AL


On Mon, 2008-05-19 at 07:00, Lyle Johnson wrote:

> >> Using a excessively narrow filter on RTTY is a balancing act.
> >>
> >> If the bandwidth is too narrow the tones tend to become "smeared"
> >> - the transitions become less easy to detect (in other words,
> >> the "eye" begins to close).  However, narrowing the bandwidth
> >> also reduces the amount of QRM/QRN making it through the filter
> >> and to the decoder/detector.
> >
> > How about using a Nyquist filter?
> >
> > A Nyquist filter is one that results in no inter-symbol interference.
>
> One of the difficulties with RTTY is that the tone spacings are so wide.
>   For a 45 baud FSK signal, a shift of about 30 Hz would minimize the
> occupied bandwidth and have no intersymbol interference.  But standard
> practice is to use 170 Hz or 200 Hz shift. [ 200 Hz Shift is about right
> for 300 baud FSK, which is what we used (and why we used it) for HF
> packet some 25 years ago.  And optimized it for common 400 Hz to 500 Hz
>   wide "CW" crystal filters of that era. ]
>
> Thus, the rise of the dual tone filter (K3) and its work-alikes in
> various radios and demodulators, with fairly tight filters around each
> tone to pass it along with its significant sidebands.  But the overall
> IF passband still needs to be wide enough to pass both tones and their
> sidebands. I imagine you can observe all of this on your "P1".
>
> Some RTTY decoders take advantage of the tone separation and actually
> employ a pair of detectors, one for each tone, and use the outputs
> differentially to get some protection from selective fading.  170 Hz
> spacing is probably too close for such protection -- 850 Hz spacing
> would be better, but that requires an even wider IF passband...
>
> Some soundcard-based RTTY programs support really narrow shifts, but
> such signals are not often heard at this QTH.  Maybe it is because
> everyone is listening and no one is calling CQ?
>
> Enjoy!
>
> Lyle KK7P

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[David Woolley (E.L)]

Alan Bloom wrote:

>
> The problem is that different RTTY transmitters use different amounts of
> filtering.  However I suspect that a "Nyquist" receive filter designed
> to compensate a typical/average transmit filter probably would have
> pretty good performance on most received signals.  Does anyone know if
> this has ever been tried?

With digital filters, one ought to be able to use the sort of adaptive
filters used in modems, where the filter coefficients are adjusted by
comparing the actual signal against the expected signal.  Modems send a
known training signal, to get fast initial convergence, but if the
signal is at least marginally decodable, I would have though it would be
possible get the filter to converge.  One could probably start by making
assumptions about the transmit filter and using known information about
the receiver.

This tactic will work better for synchronous systems, like PSK, but I
would have thought it must be possible to train a filter off an
asynchronous signal as well, although one might have solve for the
signalling unit centre position after receiving the whole character.


--
David Woolley
"The Elecraft list is a forum for the discussion of topics related to
Elecraft products and more general topics related ham radio"
List Guidelines <http://www.elecraft.com/elecraft_list_guidelines.htm>
_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Alan Bloom]

Good point.  It seems like an adaptive equalizer should work so long as
propagation hasn't totally garbaged the signal.  However, it would
greatly increase the complexity of the demodulator.  But maybe not if
the equalizer assumes a simple model of the transmit filter and makes
the a priori assumption that the channel does not change with time (i.e.
ignores propagation).

Al N1AL

On Mon, 2008-05-19 at 14:59, David Woolley (E.L) wrote:

> Alan Bloom wrote:
>
> >
> > The problem is that different RTTY transmitters use different amounts of
> > filtering.  However I suspect that a "Nyquist" receive filter designed
> > to compensate a typical/average transmit filter probably would have
> > pretty good performance on most received signals.  Does anyone know if
> > this has ever been tried?
>
> With digital filters, one ought to be able to use the sort of adaptive
> filters used in modems, where the filter coefficients are adjusted by
> comparing the actual signal against the expected signal.  Modems send a
> known training signal, to get fast initial convergence, but if the
> signal is at least marginally decodable, I would have though it would be
> possible get the filter to converge.  One could probably start by making
> assumptions about the transmit filter and using known information about
> the receiver.
>
> This tactic will work better for synchronous systems, like PSK, but I
> would have thought it must be possible to train a filter off an
> asynchronous signal as well, although one might have solve for the
> signalling unit centre position after receiving the whole character.
>

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[David Woolley (E.L)]

Alan Bloom wrote:

> It might be possible to use separate Nyquist filters and detectors for
> the two tones, treating each one as if it were a separate on/off-keyed
> signal.  If the transmitted signal is not filtered too heavily (i.e.
> jumps quickly from one tone to the other) it seems like that should
> work.

My impression was that that was the expected way of handling RTTY signals.

Transmitter circuits from 1968 simply switched between two oscillators
to key the signal; they didn't try to make a smooth transition in the
frequency domain.


--
David Woolley
Emails are not formal business letters, whatever businesses may want.
RFC1855 says there should be an address here, but, in a world of spam,
that is no longer good advice, as archive address hiding may not work.
_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com