[Elecraft] Re: Displaying the keying waveshape on Spectrogram.

Mon Sep 15 17:43:00 2003

Michael, with all respect, I think you need to review the mechanics of FFT
processing. What you propose as a mechanism for Spectrogram operation would
result in a single data point for the energy integrated over the entire FFT
bin width which you would set equal to the receiver bandwidth..

I don't know your background in this area, but evidently either there is
some considerable confusion or I am not understanding what it is that you
are trying to convey.

As I posted before, there is no meaningful relationship between FFT bin
width or resolution  and the bandwidth of the receiver in examining the
spectrum of a keyed waveform.  The resolution or bin width used in
Spectrogram is not " a narrow filter" as you state. I should not be compared
with the receiver bandwidth in determining the validity and accuracy of the
spectral estimate.

You are correct in that what the spectral estimate looks like depends
heavily on the receiver bandwidth involved in the conversion to audio for
the spectral estimate to be made. Guy is correct in that the bandwidth needs
to be large in order to capture the high-frequency components that
contribute to keying artifacts.

In most cases, the conclusions derived from spectra of keyed waveforms are
as much a matter of receiver bandwidth as it they are of the nature of the
waveform itself.

Please accept these remarks in the spirit in which they are intended and
offered.

73/72, George
Amateur Radio W5YR -  the Yellow Rose of Texas
Fairview, TX 30 mi NE of Dallas in Collin county EM13QE
"Starting the 58th year and it just keeps getting better!"
[email protected]

----- Original Message ----- 
From: "Masleid, Michael A." <[email protected]>
To: <[email protected]>
Sent: Monday, September 15, 2003 1:14 PM
Subject: [Elecraft] Re: Displaying the keying waveshape on Spectrogram.

-----Original Message-----
From: Masleid, Michael A.
Sent: Monday, September 15, 2003 12:12 PM
To: Guy Olinger, K2AV
Cc: [email protected]
Subject: RE: [Elecraft] Re: Displaying the keying waveshape on
Spectrogram.

Hello Guy,

I seem to have caused some confusion about what I
am worried about.

>What we did with the wide SSB RX filter was to try and take the
>receiver OUT OF PLAY. For moderate cases we succeeded quite accurately
>at handing the problem uncluttered to Spectrogram. From there on it's
>just math in a CPU.

>Your question, or argument, at this point deals with whether
>Spectrogram can do what it advertises.

The fast Fourier transform used in Spectrogram returns two sets of
numbers.  These are related to the sin() and cos() functions.  Engineers
tend to think of these as the real and imaginary parts of the waveform.
In practice, a quadrature detector can manipulate both parts.  Spectrogram
presents the power spectrum only.  That is the sum of the squares of each
pair of bin values.  Some information is lost at that point in the process.

The quadrature information is useful in predicting how the click will sound
when it goes through a wide filter.

>By listening through a receiver you have converted the signal to be
>examined into a fairly faithful audio parallel to the RF. If the
>received audio sounds clean, full range and undistorted to your ear,
>it's likely close enough. (Note here that limiters, signal processing,
>etc can add distortion/change not present at the RX input, and all
>such should be off.)

Yes.

>It's a frequency shift to a range for which you have a convenient
>analog to digital conversion device:

Yes.  I used to process the quadrature data streams off of a Le'Croy
'scope to figure out what multipath would be when we were developing
spread spectrum technology.  It took a lot longer to process the
data back then.

>...your sound card, which converts that analog to digital, which
>Spectrogram processes. The program's doing it's own thing. The narrow
>resolution possible simply means that any odd artifacts to the
>received signal will likely be presented quite clearly and faithfully.
>The MP has some interesting artifacts.

What is "The MP"?

>I would have killed to have a setup like this back in the 60's when I
>was working AT&T long distance plant.

>Just about everything I had back then (including some really pricey
>multi-multi-K$ analog stuff) is rendered obsolete and fairly crude by
>a modern SSB RX, PC sound card and Spectrogram.

Back in the '80's I was figuring out what the noise floor should be
for baseband signaling over coaxial cable.  I thought the committee
had set rather optimistic values.  So we rented some really expensive
analog/digital stuff, strung coax all over the place, and got down to
business measuring field strength and ingress.

The important thing that come out of this is that basic shape of the
power spectrum depends on the bandwidth of the filter used to measure
the spectrum.  A filter responds differently depending
on the bandwidth of the signal being measured.

>We are living in a wonderful time for techno-tinkerers.

Agreed.

What I'm trying to point out, an not doing a good job at, is this:

By using a narrow filter in spectrogram, you are under reporting clicks.

The clicks will look worse if you increase the bin width.

Ok so far -- Now, the next statement might seem a bit strange.

The clicks should look much worse if you increase the bin width.

To correctly report the impact of clicks on real receivers, you must use
the same bandwidth that the receiver uses.

After that, I think I lost everyone.  To measure clicks, you must
set the bin width in spectrogram to the same bandwidth that the
hams will be using in real life.  I don't know if spectrogram can
do that.  I'm pretty sure you can't integrate the power per bin
over the desired bandwidth to get the peak power at a wider
bandwidth.

And finally, where I start to have trouble:  Is it possible that
two filters with the same bandwidth will not respond to clicks the
same way?  When I designed filters to correct for signal dispersion
I used a time reversed conjugate match to the signal dispersion.
I think there was something called a R.A.K.E. filter that was
similar.  The implication is that a filter can be built to
accent or attenuate the click, and still have the same apparent
bandwidth.  A good job for a DSP no doubt.

73, Michael, AB9GV
_______________________________________________
Elecraft mailing list: [email protected]
You must be subscribed to post to the list.
To subscribe or unsubscribe see:
http://mailman.qth.net/mailman/listinfo/elecraft
Elecraft Web Page: http://www.elecraft.com
Also see: http://www.elecraft.com/elecraft_list_guidelines.htm