As an 'SSB Audiophile', for the past 6 years, I've done much experimenting
with audio compressors, limiters, splitband processing, EQ's and the like.
The goal is to produce the most articulate, pleasant sounding and cleanest
SSB signal as possible. Here are a few rules of thumb that seem to work in
practice:
1. Always strive for maximum linearity in the exciter and linear amplifier
and allow yourself 3dB of headroom. That is what produces a clean and
narrow signal i.e., a signal that takes up no more RF bandwidth then it has
to, for a given baseband bandwidth.
2. ALC should be used to throttle back the drive level when it approaches
say 90% of maximum power. It should warn the operator that he needs to back
off the drive level! ALC should not be used to increase average power!
3. IMHO talkpower is best achieved with a Split-band compressor or Peak
limiter, like the broadcasters use. Dont be deceived! You can have LOUD
without distortion.
4. Audio equalization can greatly increase articulation by emphasizing the
higher frequencies that carry speech silibants, again without that crunchy
odd-harmonic distortion we commonly hear on the air.
Maintain linearity in the RF chain i.e., the exciter and the final
amplifier, by driving it to 50% of maximum power. Allow yourself 3dB of
headroom. The 3dB loss buys you significantly lower 3rd and 5th order IMD
distortion products, thereby narrowing the signal. Ive yet to see a 100
watt RF amplifier that does not benefit from this procedure!
ALC should only be used as a Peak Limiter with the Threshold set about 1dB
below maximum output. ALC should only come into play when the exciter is in
danger of being overdriven, no other time. IMHO. it should not be used to
increase talkpower because theres no way to filter out distortion products
produced with ALC. The Attack time should be AS FAST AS POSSIBLE, otherwise
loud peaks will briefly overload and produce buckshot. The Release time
should be about 200 to 600mS or the syllabic rate of the speech signal.
RF clipping was devised to provide instantaneous attack time by clipping or
hard limiting the low level SSB signal and then filtering it to remove most
of the odd harmonic distortion. This was SOTA 40 years ago, but there are
better methods today that produce talkpower without significant distortion.
However, while IMD products outside the speech passband are greatly reduced
by RF Clipping, inband distortion is not. Although more than 6dB of RF
Clipping is IMHO the point of diminishing returns, most operators use much
more then that! In fact, many of the Dx'ers and contester's believe that
one cannot develop talkpower without distortion and that the distortion
actually improves the readability! This is not true of all methods! It is
possible to produce loud, in your face audio with minimal distortion.
Haven't you ever turned down the TV because the advertisements always seem
much louder then the show? The broadcast boys do this by using
Split-Band-Processing (and other methods) and they produce plenty of
'loudness' AND low distortion. Lets get past RF clipping and be innovative
with the SDR!
In split-band processing for example, we divide up the audio passband
into SIX 1-octave wide bands e.g., (100-200, 200-400, 400-800, 800-1.6k,
1.6k-3.2k, 3.2K-6.4K ) so that no 3rd order or higher distortion can pass
thru any particular band, then we can Clip each filtered signal 6 dB e.g.,
and post filter each band. Each filter output should have no harmonic
distortion and if we sum the filter outputs of all bands to produce the full
bandwidth signal the result will be loud and clean.. This is nothing new
in broadcast audio, but until the interest in 'HiFi SSB' very few phone
operators used these techniques!
Another very important method for increasing articulation i.e., the
ability to be understood in the presence of noise - is using equalization
to shape the audio passband. Although frequencies under 300Hz are not
essential for articulation, frequencies above 2600Hz will greatly improve
articulation! Narrowing speech bandwidth does not concentrate the signal
and cut thru the QRM that is false. What does cut thru the QRM are the
sibilant frequencies from approx. 2000 to 3600Hz. If we pre-emphasize our
speech about 3-6 dB/octave starting at about 800Hz, the resulting speech
spectrum will cut thru the din because we have placed more energy in
frequencies carrying sibilants. The receiving station can narrow his
receiver bw somewhat (de-emphasis) decreasing the hiss and noise and still
understand this processed signal. In 2.8 3.3kc bw.s of course, this
signal would sound tinny, strident and so forth, but when you wish to sound
natural and maybe HiFi, you can switch to a different EQ profile say from
90Hz to 3600Hz and decrease by 6dB the frequency band ranging from 400 to
1800Hz - as an example.
a couple of suggestions -
1. Please keep the SDR as linear as possible.
2. Provide a method for switching to different transmit audio profiles
ie., save the EQ settings, compressor settings and so forth so the operator
can switch from DX-Contest audio to HiFi SSB audio to something else
audio.
3. Provide an ALC THRESHOLD control that is relative to the RF Power
Setting. E.g., if RF Power=50watts and ALC Threshold is set for 90%, then
ALC commences at 45watts.
73, Alan Davis K2WS