So here's another update - and a request for opinions, because I'm doing this for you chaps, and I want to give you something good and useful.
This evening I was hammering the unit with CW signals of various intensities from 20m and 40m bands, as the evening drew on, and later in the evening I got some horrendous high level bursts of modulated squawking on 40m which I presume to be transmissions from one or more of the digital modes as I was tuning up and down looking for QSOs in progress: eeek! this is one of the main reasons why we need AGC!
So again, the clamping is working really well - it smacked that darned computer mode stuff down with no problems. As I mentioned before, I think, given that I'm working in a digital buffer, I can push the onset of the gain reduction factor (derived from the averaging function - which takes a number of samples) back before the tone started, which means I don't get the chirp or chiff that a loaded fast-acting analog AGC would deliver. But the reason for bringing this issue up this evening is not the attack but the decay. As anyone who has played with AGC circuits and studio audio compressors will know, any background noise will get clamped and replaced by the dominant sound source whenever the latter hits, and if that dominant sound is percussive, staccato-like, the background noise will go up and down in opposition to the dominant sound, giving a sort of "breathing" effect, which can be distracting.
Now there's no way of avoiding the sharp reduction of background noise when a loud tone comes through (the "attack") - that's the whole point of the AGC - but what to do about the release after the tone disappears?
There are at least two things I can do: I can hold the gain reduction on for a programmable and adjustable amount of time that is just longer than a "dit space", and adjustable for WPM, so that the gain stays reduced for the length of a whole character; and I can allow a ramped release, such as is used in normal audio conditioning circuits. I suppose I could do both, but I want to get the distinction between the concepts out there. Remember that any user adjustability brings with it the headache of controls which, for the QCX version, might be challenging. For a freestanding version, of course - which will have a display just like the QCX - all adjustments are possible. And for the minimalist SOTA version, it will all have to be preset, or keyed in using Morse sequences, which sounds rather clumsy. I could expose a serial port, but that seems counter to the SOTA philosophy.
Input is enthusiastically invited on this issue.
Lastly, I'm going to fire up my website so I can share sound files, screenshots, etc., for anybody who is interested. More mews on that to follow presently.