Generally the divide by 10 circuit looses amplitude information. And the
spectral compression depends on considering that the SSB signal can be
approximated by sampling the instantaneous frequency every ten cycles.
In one of my witnessed entries to engineering notebook I have made such
an allegation. Then even if we put back the amplitude information we
tend to loose it in the multiplier, most effective multipliers being
very nonlinear for amplitude (e.g. drive it hard and get harmonics,
drive it gently and get nothing). Though I suppose the receiver
processing could recover the amplitude before expanding the frequency.
Both processes might be accomplished by DSP with adequate programming to
get around the limitations of analog hardware. I suspect quality would
suffer. I suspect we'd need to clean up the modulation waveform with a
narrow bandpass filter to keep down the splatter created by the
frequency division process.
I know that the inverse has been proposed, maybe used in Europe, for
microwave SSB. Where multipliers, mixers, and linear amplifiers are a
great bother and multipliers and class C amplifiers are easiest, it has
been proposed and maybe used to take SSB at some low frequency, divide
it by the multiplication factor to follow, modulate the multiplier's
driver then accept the constant amplitude results. I've not tried it,
but I suspect the bandwidth after the multiplication is wider due to the
unfiltered effective RF clipping.
Again it depends on the premise that SSB can be adequately emulated by a
sequence of selected signal frequency signals. I think it can and have
proposed in my notebook that one determines the amplitude and frequency
for each of those sequential frequency components by measuring the
amplitude and period of each half cycle of audio. When I wrote out that
idea in the early 70s, DSP wasn't even on the horizon and the analog
processing to try the idea was overwhelmingly difficult. PLL
synthesizers didn't have the resolution though DDS now do.
I'm fairly sure that the simple divide and modulate then multiply scheme
leads to generally understandable SSB. Whether its natural and has
decent quality will depend on how particular the listener is. I suspect
some care will be required to clean out artifacts from the division and
multiplication processing.
Then to simplify my scheme one might analyze the audio into a sequence
of numbers for each half cycle of audio (low pass filtered probably),
one number for amplitude, one number for period or frequency, then send
just those numbers, and let the receiving computer use those to create
audio again. Something that needs trying. Though it can't be as
spectrally efficient as sending phoneme numbers and amplitudes. It might
sound better than the results of the phoneme numbers scheme.
As one of the public TV science shows used to say, "TRY IT!"
73, Jerry, K0CQ
--
Entire content copyright Dr. Gerald N. Johnson, electrical engineer.
Reproduction by permission only.
|