Interesting article. I did experiment further with the processor settings and compared the result with an analogue speech processor that was part of a ssb rig I designed about 20 years ago.  This processor uses variable rf log clipping of an ssb signal (at 10.7MHz) followed by another ssb filter. The audio is pre-processed by a syllabic compressor (a vogad). The clipping is done with two stages of log amplifiers. In the original design the resulting ssb was converted to the required frequency but I had already added a demodulator back down to audio for something else so here the output is audio. As the article points outs, non dc baseband clipping avoids the serious harmonic distortion that baseband clipping induces at all but very light clipping because the harmonics multiply outside the passband. There are still intermodulation products that eventually limit performance.

 

Of course this processor is very component hungry by today’s standards and I happened to have it on the shelf. As the article also points out, the the same result can be achieved in dsp by other means that could not be realised in practice by analogue circuits.

 

I feed the audio output at line level into the pc and have the SDRC processor turned off. The audio processor bandwidth is from about 200Hz to 2.4kHz, determined almost entirely by the analogue ssb filters.

 

I think that overshoot is a slightly different, but related problem. Interestingly though I had a look at this processor’s performance by putting test sine and square-waves frequencies into it and looking at the audio output on a scope. Probably by more luck than judgement, the overshoot is a few percent unless the processing is set ridiculously high. I suspect that a fortuitous combination of filters and time constants, plus the relatively narrow bandwidth requirement helps.

 

I have not checked for overshoot on the rf output from the Limesdr that eventually generates the signal for transmission.

 

My metering indicates that the average to peak power ratio has increased significantly and, although this is pretty unscientific, the “hf pileup breaking index” is markedly improved.  

 

Nothing can be all things to all men (or women). Everything is a compromise of requirements, performance, complexity, time and cost.

 

Mike  

 

Sent from Mail for Windows

 

From: Simon Brown
Sent: 31 August 2021 06:41
To: main@SDR-Radio.groups.io
Subject: Re: [SDR-Radio] Speech Processor

 

Probably.

 

I use a single band RMS compressor followed by a very-well designed limiter which avoids overshoot. I’ve tried multi-band compressors but not achieved much benefit. When I get into my main winter project I may look at this again, having a compressor instance for each octave or half-octave.

 

For winter I want to finally play with my Hermes-Lite 2, with my main amplifier I’ll have ~350 watts so will want to boost the audio.

 

Simon Brown, G4ELI

https://www.sdr-radio.com

 

From: main@SDR-Radio.groups.io <main@SDR-Radio.groups.io> On Behalf Of kb3cs
Sent: 30 August 2021 21:37
To: main@SDR-Radio.groups.io
Subject: Re: [SDR-Radio] Speech Processor

 

Controlled-Envelope good?
http://www.arrl.org/files/file/QEX_Next_Issue/2014/Nov-Dec_2014/Hershberger_QEX_11_14.pdf

 - 73 -

--

- + - + -

Please use https://forum.sdr-radio.com:4499/ when posting questions or problems.