stupid envelope follower idea

[Ian Fritz]

----- Original Message -----
From: "Harry Bissell" <harrybissell at prodigy.net>
To: "Ian Fritz" <ijfritz at earthlink.net>
Cc: "Terry Michaels" <104065.2340 at compuserve.com>; "synth-diy"
<synth-diy at node12b53.a2000.nl>
Sent: Wednesday, November 22, 2000 10:54 PM
Subject: Re: stupid envelope follower idea


>
>
> Ian Fritz wrote:
>
> > Hi Terry --
> >
> > > RMS detection is a means of calculating the "area under the curve" of
the
> > > waveform.
> >
> > Huh?  :-)
> >
> > The area under the curve of a waveform is zero. (Assuming ac coupling,
or,
> > equivalently, ignoring the zero frequency Fourier component.)
>
> No fair... if we are talking envelope detection, we assume absolute value.
Area
> under
> the curve is ... er...area under the curve.... ;^)

Well, yes indeed it is! But I still don't see what it has to do with RMS.
It's the square of the function that you integrate for RMS, not the
function.

>
> >
> >
> > > It is commonly done by squaring the instantaneous points of the
> > > waveform with an analog multiplier.  The peak voltage squared yields
the
> > > RMS power.
> >
> > Well -- it's (1) square the voltage, (2) average over a cycle, then (3)
take
> > the square root.
>
> RMS detection is usually used in things like compressors to the the human
> perception
> of equal amplitude... IMHO it does not work well and I never personally
use that
> detection method.  It is a nice way to have an "average" level detection,
but in
> real life
> it is slow. If I wanted to ride gain on a complete audio signal (like a
CD...)
> it would be good. But on individual waveforms I think its too slow. I
can't wait
> full cycles. (or even half cycles...
>

I've never tried it for envelope detection, and I agree that it sounds like
it might be slow. But again, if the signal is being split into bands maybe
it could be sped up.

> >
> >
> > But since the instantaneous power is independent of the phases of the
> > various Fourier components, the phase shifts you are concerned about
don't
> > matter. They change the instantaneous waveform amplitude, but not the
RMS
> > value.
> >
> > > I hope that's right, I'm going by memory here.   The problem
> > > with that approach might be the time altered waveform might not have
the
> > > same "area under the curve" as the original waveform.  It should be
> > > possible to do a mathmatical analysis or simulation to answer your
> > > question, maybe someone on the list has the means to do so,
unfortunately
> > I
> > > don't.
> >
> > I think the math is simple. The RMS voltage is the square root of the
sum of
> > the squares of the Fourier amplitudes. Phases don't matter.
>
> OTOH: I bet after you break the signal into bands, RMS each, and sum them
> together...
> that the envelope follower looks a lot like the original signal...
unfollowed
> (so to speak)

Well, RMS implies an averaging, so the output of each band would be
smoothed.

> But its worth a try. Now I'll have to try and do a simulation on this...
that
> should keep
> me busy for a while...

We'd all be interested in the result of such a study, I'm sure. Let us know
what you find.

>
> H^) harry
>
> >
> >
> >   Ian
>
>

  Ian