[CQ-Contest] SSB Waveforms

[Jim Brown]

On 2/27/2023 7:32 AM, Joe wrote:
> I'm looking for a nice set of what the audio of a SSB signal looks like 
> processed vs non-processed.

There are two forms of audio processing that matter. The most important 
is equalization -- that is, elimination of low frequency audio 
contributes nothing to speech intelligibility but wastes half of 
transmitter power. Optimum settings kill everything below 400 Hz, and 
provide a small peak (3-6dB) around 3 kHz. With an octave-band 
equalizer, this would be max cut of 50, 100, and 200 Hz bands and 6 dB 
cut of the 400 Hz band. This EQ increases talk power by about 3 dB, 
making a 100W signal the equivalent of a 200W un-processed signal.

On a spectrum scope, un-processed audio (that is, WITH the audio lows 
present) shows a broad peak near the suppressed carrier frequency. With 
equalization applied, the peak is far less pronounced, and may not be 
visible at all.

The other form of processing is loudness, and usually takes the form of 
compression and/or peak limiting. Human speech is dynamic -- that is, it 
varies in loudness between syllables, words, and phrases. Dynamic 
processing in ham rigs reduces the loudness of the strongest peaks so 
that when adjusted below the point of clipping in RF stages, the weaker 
parts of speech are louder. How much compression can be applied before 
it starts making things worse depends on how good circuitry or firmware 
is at doing its job, but a good rule of thumb is to stop somewhere in 
the range of 10-13 dB. This results in that much increase in talk power, 
in addition to the 3 dB provided by equalization. 10 dB of compression 
added to 3 dB from EQ makes a 100W rig the equivalent of 2kW, or a 5W 
rig the equivalent of 100W.

Dynamics (loudness) processing is nearly always done to the audio 
signal, so any distortion products are confined to the bandwidth of the 
sideband filters (that is, no splatter), but distortion will reduce 
copyability and will sound bad.

Splatter occurs in the RF chain, including circuitry and firmware used 
to generate the SSB signal. Relatively recent Yaesu rigs produced very 
strong splatter on both sides of the signal that had nothing to do with 
overdrive or other operator adjustments. These rigs are so dirty that 
adjacent channel sidebands are only 20 dB down from voice peaks of the 
legal signal. I first ran into this when I was trying to work a station 
with a pretty decent signal 3 kHz from one of these dirty rigs. (I use 
K3s with 8-pole 2.8 and 2.1 kHz roofing filters).

On a waterfall, a clean signal, even with a lot of dynamic processing, 
will look like a vertical bar. On a slow waterfall, splatter will show 
as horizontal off-shoots on both sides of that bar corresponding to 
voice peaks. A good place to see this is on any phone band that's not 
too busy, especially with multiple stations in a rag chew. The splatter 
from these rigs looks just like that produced by overdriving an amp or 
running the rig at full power and using ALC between rig and amp to set 
power.

73, Jim K9YC