ESSB user:

I'm not sure what transmitter sweep you are referring to,
but all my sweep tests are generated via Cool Edit 2000
using my sound card. The sweep test is a single tone that
modulates from 20 Hz ~ 6500 Hz in a 30 second linear sweep.
The sweep tone was sampled at 44.1 kHz, 16 bit mono.>>

W8JI response:
That's what I was afraid of when I looked at your site.

You aren't going to like to hear this, but the results of
your sweep tests are NOT the amplitude of your off-frequency
products. They might not even be indicative of in-passband
amplitude response. What you claim is your measured occupied
BW is certainly not even close to what it is, because that
method does not test IM distortion at all, and even excludes
in-channel harmonic distortion data. As a matter of fact a
rancid sounding wide band splattering class C amplifier
would actually look significantly narrower using your test
method than a linear amplifier with minimal distortion
products!

Sorry to bring such bad news. I don't know who was giving
you technical support, but they sure missed something very
obvious.

Next exchange:

ESSB User:
      Your question did not indicate that you were
interested in how I measure any distortion products or about
anything regarding my test methods. I assumed that you
wanted to know how I measured the audio bandwidth of the
transmitter.>>

W8JI:
When I read your web pages, I formed the strong conclusion
you were claiming your signal was transmitted with the bandwidth
shown in the spectrum graphs with the slope you plotted.
That isn't true, and I'll explain why.

ESSB user:
      If you were referring to the 850/DSP-100 transmitter
bandwidth as published on my Apologetics page, I used a
commercial Ten-Tec RX-340 receiver opened up to 10k and 15k
of receiver bandwidth, feeding the soundcard and using the
excellent Spectra Plus audio analysis program. As you know
Tom, if a transmitter is properly adjusted, clean and
linear, the AF and RF bandwidths will be virtually the
same!>>

W8JI:

No, that is not correct. RF amplifiers are considerably
worse than class A audio amplifiers, especially those audio
amps designed with attention to distortion. The typical SSB
transmitter in a two-tone test only has IM3 down about
30-35dB below PEP, and that test does not fully reflect the
distortion since dynamic variations in bias and power
sources are not tested by the steady high frequency load.

Say we inject two tones spaced 2kHz in a SSB transmitter,
one at 3900kHz and one at 3902 kHz. The load variation on
the power supply and bias is at a rate of  2000Hz, so bypass
and filter capacitors tend to stabilize the voltage. When a
voice is added, the lower frequency variations cause supply
and bias voltages to vary and this adds distortion that is
not indicated in a conventional two-tone IM test. SSB
transmitters are typically worse for actual voice
performance than a two-tone test indicates. Even a notched
noise test won't be accurate, because it lacks syllabic
variations in power and bias loading.

Since you use a single tone sweep test, the effect is much
worse. That test not only fails to include any distortion
contribution to BW, it worse yet makes things look *better*
when the transmitter is non-linear! As a matter of fact the
more non-linear near zero crossing, the better your
bandwidth slope will appear. That means if you ran a class C
PA it would look narrower than a class A PA!

All you really plotted was (assuming you measured fully out
of ALC and without gain compression or expansion in any
stages) the passband of the selectivity, NOT the transmitter
BW with a complex signal like voice.  Any claim that the
signal is only that wide and has the dB per Hz slope shown
is absolutely not correct.>>

ESSB user:
      For occupied RF bandwidth, which can be much more
difficult to determine, 850 linearity and amplifier
linearity, I use two channels of my scope looking at both
pre and post transmitter signals simultaneously, observing
the scope trapezoidal images. While it is certainly true
that an audio analysis program (like Spectra Plus) will not
indicate distortion products when performing a real-time
audio analysis, a scope will certainly tell me when I have
exceeded the desired modulation or pushed the RF signal into
non-linearity... and I observe mine constantly, on both SSB
and AM. By the way, Cool 2000 and Spectra Plus both have IMD
measurement tools built in, although not a s useful as a
true RF spectrum analyzer.>>>

W8JI:
Doesn't matter. Subtle differences in gain slope we can't
detect by eye on a CRT amplitude display cause the IM that
is worrisome in SSB rigs. Remember we can bother people with
"junk" -30 to -40dB. We know from a two-tone test most rigs
are only -30 to -35 dB PEP for low order IM products. That
means when we process a 6 kHz audio signal and a 3kHz signal
the IM3 of that combo is out 15 kHz at the widest IM3 combo
(two times 6 plus 3) of the two. We know that product is
only down 30 to 40 dB below PEP, because that's how the rigs
measure in easy two-tone tests.

By the way Eimac rates distortion dB below one tone instead
of the ARRL's silly "dB below PEP". This means an Eimac tube
at -35dB IM3 is actually -41 dB IM3 PEP, or the way modern
rigs are measured by the ARRL. My old Collins KWM-2 measures
20dB better than my new 756ProII for IM3 and IM5 in a
heads-up test!

When we add bass the dynamic regulation of bias and power
sources is much worse, so distortion levels can be and
almost always are higher.

ESSB user:
      IMD, THD or any kind of distortion for that matter,
sounds bad... That's why those of us who are playing with
higher quality audio absolutely do NOT want these types of
distortion!!! This is TABOO with the true audio guys!>>

If we did a blind A-B test and added IM3 down 30dB to a SSB
signal, the guy listening would never know. That's true no
matter how golden his ears are. I've performed tests like
that before, and confirmed IM3 and IM5 at -30dB is
impossible to detect. Heck, most rigs are that bad or worse
already even when operated "linear". If it made a
difference, all the hifi users would be running FT1000MK V's
in class A instead of ratty IM3 rigs like 756PRO's.

ESSB user:
            Now if I could just get these pesky guys off my
back who are so quick to criticize due to misunderstandings,
instead of offering support, I could really enjoy myself...
HI.>>>

W8JI:
Well, they may have a good point you are missing. I'm sort
of in the middle on this. I have no objections at all to
people running wide-fi on bands that are not busy, but using
two or three communication channel widths on a crowded band
or near weak signals is (in my opinion) very selfish. This
is especially true when it is done without consideration of
where in the band it is placed. We probably need to
establish a wide-fi window like an AM window. Almost
everything in the world has its place, and I certainly would
not want to kill the fun of what you are doing if you enjoy
it. On the other hand I also know from first hand experience
that enhanced SSB even without wider filters increases
adjacent channel QRM. That isn't an issue on an uncrowded
band.  It certainly is on a crowded band.

73 Tom