Tom Holden wrote:
[snip]
I concluded the following to be good targets for AGC
behaviour after surveying a number of radios -
corroboration or otherwise appreciated.

Attack: 1-13ms
Release:
- fast: 25ms
- medium: ~300ms
- slow: 1.8-3 seconds

I thought the fast release to be too fatiguing for human
listening to SSB speech and ICW code but desirable for
machine decoded data formats to minimise loss of data.
Also, with audio derived AGC, the distortion on heavy
bass modulation of all AM modes would be excessive.

In applying mods to the DX-394 by others and some
designed by myself, stretching the release time towards
the 'slow' target has the side effect of lengthening the
attack time to potentially a few hundred milliseconds. My
version is the fastest so far with an attack of about 100
ms on a release of 2 seconds. I'm wondering if there is
much to be gained by struggling to bring that down to the
target of 1-13 ms.

Comments on my assumptions, logic, conclusions and
questions most welcome!

Still looking for feedback on these targets.

Had one private reply that raised an interesting point about noise blanker
operation and 'delayed' RF AGC:
"You might want to test out a system of this sort in the face of a
significant amount of broadband impulse noise - e.g car-ignition noise
or something like that.

If the attack is made too fast, then impulse noise at a rate of, say,
60-100 Hz would tend to force the RF AGC into its low-gain state, and
this could allow weaker RF signals to be buried in the noise of the
first-stage mixer.

Using a longer attack time constant will render the RF AGC a lot less
vulnerable to the effects of impulse noise. You can then use a noise
blanker, located prior to the first IF AGC, to trim out this noise.
As long as the first-stage mixer and the RF amp stage aren't forced
into excessive intermodulation by the content of the impulse noise,
I'd expect that this approach would give you the best set of
behavioral tradeoffs."

73, Tom