"Telamon" wrote in message
Basically the radio has to adapt to the signal level you are tuned to by
changing its sensitivity.

.... in order to avoid overload distortion at any stage in the receiver by
excessive signal and to even out the loudness between weak and strong
signals so that the listener does not have to adjust the volume control.

For any kind of digital mode you want fast AGC ....

Telamon, I've read that elsewhere and it seems to be embedded in specs but I
don't understand why after having experimented with modifying a DX-394A. My
impression from the literature I have read is that the 'ideal' receiver is
supposed to have fast attack (under 10 milliseconds) and variable release
(slow 1 second, medium ~ 100's of milliseconds, fast ~ 10's of
milliseconds).

1. A problem is if the attack is under 10 ms and the release is slow, then a
very short noise burst reduces the receiver gain for a long time. Seems to
me that the attack should be proportional to the release to avoid that
problem but that might not be ideal for SSB or Morse. I can see that, for
most data, a fast attack/release combo would likely minimise dropouts. Maybe
the ideal has independently variable attack and release to accommodate all
conditions.

2. Many DRM experimenters have found that the maximum DRM SNR is achieved by
turning AGC off but that may result in a higher long term error rate due to
signal fades and surges after the gain has been set. The reasons why the Max
SNR is reduced by AGC are not fully understood - one, specific to the
DX-394A and original DX-394 but possibly a factor in other radios, is that a
local oscillator was pulled as the mixer gain varied, whether by manual RF
Gain control or by AGC. This results in a phase noise and, in extreme swings
in gain, in the loss of lock. There are other more subtle factors degrading
the Max DRM SNR that remain that I don't know about. (Note that the DRM SNR
has nothing to do with the perceived audio SNR but is a simple
representation of the quality of the transmission channel calculated from
the scattering of the transmission coding constellation.)

Tom