Its the corner in the derivative spectrum but its the flattening of the
pseudo sine wave that adds harmonics. Remember a sine wave has no
harmonics, but a very clipped sine wave is a square wave and it has an
infinite number of odd harmonics.
Langevin used to make a "distributed" clipper where the audio went
through a many element low pass filter and at each shunt element of the
filter there were clipping diodes. This rounded the corners. VOA liked
to use it, probably still does. The rounding did tend to remove the
higher harmonics and the harshness to the ear. Yet the clipper was VERY
effective on an AM transmitter. When they cranked in 20 dB of clipping
on a 50 Hz sine wave and cranked the audio output up to 100% modulation
of the Collins 821A-1 (250KW carrier) the scope picture looked like a
nicely keyed CW rig running 1 MW peak and with the keyer stuck on dots.
I don't know how this rounded corner clipper would translate to SSB.
I've never tried it.
The fundamental problem is that while the harmonics are strictly
harmonically related to the fundamental voice pitches at audio, they are
translated to multiple signals, not harmonically related when mixed to
the filter or output frequency of the radio transmitter. This is
important to audio processing because if you take a signal made up of
fundamental and odd harmonics, and the phase of those harmonics is
correct, their sum makes a square wave, or at least a sine wave with the
tops clipped off. But if you invert the phase of each of the odd
harmonics with respect to the phase that made the square wave you get a
very peaked wave, not at all a square wave. Yet to our ears that phase
change is not detectable.
Then at RF, you have these essentially independent carriers each with
peak values dependent on relative amplitudes of the components be of the
original audio. With the small spacing between the carrier components of
the clipped speech, its certain that much of the time the peak output
envelope will average but every once in a while all the carrier peaks
will align and give the PA a sock in the teeth well beyond its linear
range. And then the PA clips and makes splatter limited only by the
antenna bandwidth. This is a current problem in cell phone fixed sites
where they tend to share one PA between a gang of signals. The peak
value the amplifier has to handle is several times the average power and
gets worse the more independent carriers are multiplexed to the single
PA.
In the SSB transmitter the relative phases and amplitudes of the
carriers that represent the audio harmonics are also modified by the
amplitude and phase ripple of the transmit filter.
I have not tried the Heil elements, I found the Shure element a long
time ago. If one was to add processing at audio to compensate for the
transmit filter, and then changed bands where the BFO moved to the other
side of the filter, especially in TenTec ladder filters, the work would
have to done all over because of filter asymmetry.
The reports I see for Heil elements are either loving or hating. Likely
those that love them (and aren't castigated by those listening) happen
to have a filter with responses quite similar to Heil's radio and those
that dislike them have some other filter response and voice
characteristics that takes the Heil's modified audio and makes it even
less intelligible.
Probably some audio processing to emphasize the fundamental with minimal
harmonics, only as much harmonic of the voice fundamental as needed for
understanding (sibilants aren't harmonics but to the spectral display
they get analyzed as if they were, but they are needed for
understanding) plus some of the sibilants, might prove to give more talk
power than any mere amplitude audio processing. Creating that audio wave
form will probably take more than simple filter based spectral
modifications but take wave form analysis with understanding of what
makes speech work. Probably doesn't work out to a simply stated
algorithm.
I proposed (privately) a theory once, probably about 1972, that one
could approximate the voice with a single sine wave at the transmitter
output that shifted frequency and amplitude for each half wave of the
voice wave form. DSP and agile frequency synthesizers weren't even
gleams when I made that proposal and the theoretical minds that I passed
the theory by were not able to shoot it down or to agree it would work.
I should reattack that idea now that DSP and DDS are available. Its now
practical to try. A question could remain about the broad spectrum of
rapidly shifted frequency and amplitude between half cycles of audio.
There have been audio bandwidth compression schemes proposed (one in an
ARRL handbook) based on picking out the fundamental and then a portion
of the harmonics (leaving out the middle) and sibilants. With the
hardware being patented and proprietary, use on the air was very little
and the proposal died. I think the latest ARRL handbook may have another
such scheme.
The ALC in the radio does some compression based on peak drive to the
PA. Some would make the ALC work beyond its capability, then they
produce broad signals.
73, Jerry, K0CQ
--
Entire content copyright Dr. Gerald N. Johnson. Reproduction by
permission only.
|