[Elecraft] oh no, another crazy audio idea from Holland

[Leigh L. Klotz, Jr WA5ZNU]

I thought of this recently as well, in another context.  With PSK and 
other digimodes, the wide AGC doesn't do the right thing.
Especially with modes which aren't amplitude-sensitive, an AEC 
(automatic equalization control) would be better than an 
across-the-board AGC.
I think it should be possible to do a crude AEC by monitoring the power 
spectrum and sending RS232 commands to control the RXEQ buttons.
An on-device version would be better, though, as it would be be able to 
use smaller bin sizes.
Of course, this wouldn't help digimode reception when the AGC action is 
strong enough to cause hardware AGC, but when the dynamic range is such 
that it's within the range of the K3 DSP, but the generated signal 
dynamic range exceeds that of the PC sound card, it could be useful.

You can decide yourself if this would be a good idea for your 
computer+rig by performing the following test:
1. go to 14.070 on a busy day
2. turn off the speakers and turn off AGC
3. ride the RF gain control and see if you can find one spot that shows 
all weak signals without driving your sound card into distortion from 
the strong ones.

It'd be nice if Lyle KK7P or Wayne N6KR would comment on whether they 
think this is valuable.

Legih/WA5ZNU
> Hi all,
>
> There have been some reports (f.i by Rob Sherwood) of AGC's in modern DSP
> receivers being a bit "restless" (being over-responsive or "dancing on the
> noisefloor")  The K3 AGC is most certainly on the right track but has had
> some work in the past.
>
> Could the following be the case:
>
> Besides prevent clipping in RF/IF stages, in the end of the day our AGC is
> mainly there to serve our ears. Since the AGC detector stage is triggered by
> "white" bandnoise, is it not relatively(!) more sensitive to sounds near the
> high flank of our passband? In other words being triggered relatively more
> by static and pulse noises, which undermines it's control over the lower
> voice-content (the vocal-cord part of the spectrum).
>
> Up till now I've only seen approaches in the time and gain domain (delaying
> the attack,modifying the attack/decay timeslopes and onset threshhold) but
> not in the frequency domain, adjusting some frequencies to have more grip on
> the agc and some less.
>
> In audioland it's quite common to "side-chain" a dynamics processor
> (compressor/limiter) with a filtered version of the program material you
> wan't to treat. If you offer the AGC detector an input with a HPF inline it
> will become less sensitive to bass content. If you offer it an input with a
> LPF inline it will become less sensitive to high tones. In other words you
> get "frequency-dependent dynamics processing".
>
> If you would use some sort of sloping (pink? ;-) LPF in front of the AGC
> detector in a DSP receiver, could you prevent it from being overly sensitive
> to white noise content? It might do a better job on tracking voice and be
> less responsive to pulse and static type noises near the high slope of the
> passband.
>
> Would this (combined with Rob Sherwoods "delayed attack" transient approach)
> be an option to really tame the AGC in a near perfect DSP receiver?  Excuse
> me if i'm kicking in open doors...
>
> 73'
> Paul
> PD0PSB
>