[TenTec] Orion 1 & 2 "Talk Power" problem

[Dr. Gerald N. Johnson]

On Thu, 2006-12-21 at 10:19 -0800, Ron Castro wrote:
> I have had a couple of people contact me off-list a problem with the v. 2.x Orions regarding the ability to drive external amplifiers to normal average power levels in SSB.  I have referred to this as the "Talk Power" problem.  A few weeks ago, I looked at the SSB wave form under normal speech on a spectrum analyzer and found that very few voice peaks approached the output level reached when measuring CW output for the same power-level setting.  This appears to be a result of an extremely slow release time in the ALC circuit.  This may be great for preventing CW key-clicks, but looks like very low average power output in SSB.
> 
> I'm wondering if the ALC is controlled by the CPU or if this is a purely hardware-related function.  Looking at the schematic (A5 ALC.pdf in the O2 schematics), it looks to me like R9 and C5 set the timing, and if my math is right, that combination (470 k and 4.7 mf) should result in an 8.7 dB "release" over a period of 2.2 seconds.  If the control is actually digital or combined with a digital override, then the time constants can be completely different.  I sounds to me like very fast attack and very slow release.
> 
> Has anyone else noticed this problem and what thoughts do you have on it?  Also, does anyone know what parameter R-19 in the schematic adjusts?
> 
>       Ron N6IE
>  www.N6IE.com
> (Formerly N6AHA)

Average power may NOT be talk power. The fundamental problems come from
two things: 1. the PA has a peak power limit that if you try to exceed
that you get splatter. ALC works to prevent that over driving condition.
Without being able to control the gains of stages after the D/A it
should be an analog function based on PA output or PA drive. 2. Summing
a gaggle of RF tones of different frequency, every once in a while the
vector peak amplitude to be far taller than the RMS sums. This is true
of voice or multi carrier cell phone base stations. That peak comes when
all the positive or negative peaks of the tones line up. The PA has to
have a peak rating far above the average power. But on CW or RTTY or
PSK-31 there is only one tone or carrier and then the PA can run at full
peak power capability, if it has enough power supply and cooling.

I have thought that the relative phase responses of microphones and
filters can be a contributor to strong or weak talk power, but I think
now that may not be the case. Rather that voice characteristics, and
amplitude response ripples in the microphone and filter serve to either
make one voice frequency component stronger than all the rest so that
the vector sums don't add to the peak so much which gives more talk
power, or the three characteristics make there be two or three
frequencies with about equal amplitude which is the worst case that has
high peak to average ratio. I know that on my of my radios using an EV
664 gets me a higher average power on SSB. Though those in the
neighborhood listening on 2m FM with handhelds having 3/4" diameter
speakers grumble about the bass tones pushing the speaker beyond its
limits. The effects the SSB filter ripple has on voice components is
sensitive to the carrier offset also.

Are there ways to improve the talk power? Maybe. One might shape the
audio response curve to enhance one particular part of the voice
spectrum, trying to turn it into an amplitude modulated tone with just
one carrier frequency and then let the harmonics rattle through at
reduced amplitude to allow better understating. 

One might split the audio/RF spectrum and use a separate PA for each
frequency component so that the peak happened in the antenna, but not in
the individual PA.

One might use a fast acting ALC to limit the peaks but let the average
power come up. Though the RF output may not have a significantly
different characteristic than clipping the RF to limit the peaks.

One might use a peak reading wattmeter and burn the average reading
meter and accept the peak to average ratio governed by the voice,
microphone, filter, PA and PA ALC characteristics and make contacts.
Then it would handy if the receiver S-meter was peak reading, not
average reading on the other end of a path.

One could use a sinewave tone generator in place of the vocal chords and
modulate the amplitude of that single tone and so eliminate most of the
other components of the voice that contribute to the vector sum peaks. I
suppose a DSP could do that too. It doesn't sound natural, but to the
person without vocal chords, any sound is better than silence.

I have thought that SSB might be synthesized by analyzing the audio
waveform each half cycle, and from that half cycle picking an amplitude
and RF frequency so the output of the transmitter had never more than
one tone. I invented this scheme back about 1974 and wrote it down in my
engineering notebook and had an ISU professor understand and witness it,
but in those days it was impossible to accomplish the task with the
available computer and synthesizer techniques. Since I've not yet tried
it, I don't know how natural it might sound or if it would even be
understandable at all. But it might set the peak to average ratio at RF
about the same as at audio, IF the transients from switching synthesizer
frequency and amplitude each half cycle of audio didn't take over.
-- 
73, Jerry, K0CQ,
All content copyright Dr. Gerald N. Johnson, electrical engineer