Odp: [sdiy] what is the amplitude envelope of a signal

[Roman Sowa]

how about this:
rectify signal, compare each sample with 2 registers: temp and output.
If it's bigger, replace appropriate register. Every 10ms or so write
temp value to output, and reset temp to 0. During next 10ms if signal
won't reach previous 'output' value, it won't be changed, but after
that 10ms will drop to current 'temp' value which holds the peak from
last 10ms.
This way you have immediate (1 sample) attack response, and decay
delayed by 10ms. No problems with other peaks and multiple zero
crossing in each cycle. Changing this 10m time together with appropriate
corner frequency of smoothing filter will give different responses.
That's just an idea, it's late night. Based on Harry's follower recently
published in EDN.

Roman

----- Original Message -----
From: Ian Fritz <ijfritz at earthlink.net>
To: Scott Bernardi <sbernardi at comcast.net>; Synth-DIY list
<synth-diy at dropmix.xs4all.nl>
Sent: Monday, September 29, 2003 4:46 PM
Subject: Re: [sdiy] what is the amplitude envelope of a signal


> At 07:02 AM 9/29/2003, Scott Bernardi wrote:
> >How about an interpolation between the peaks of a rectified signal?
>
> That's OK if you have just one peak  per cycle.  But if you have more,
> there could be big problems.  Think of the waveform you get if you excite a
> hi-Q filter with a narrow pulse.  This is not too extreme an example -- a
> bassoon give similar effects.
>
> In general the envelope extraction problem is pretty difficult --
> expecially if you want some kind of general implementation.
>
>    Ian
>
>
>