Envelope follower for my vocoder...

[Martin Czech]

The question is what amount of ripple will be allowed
when extracting the envelope.
Response time and ripple can not be optimised at the same time.

Now, every channel is a bnadpass, this already imposes
some limit for the envelope characteristic.

Eg.: 100 Hz center frequency, 50Hz bandwidth

If I have a sine wave signal with 100 Hz, it can be amplitude modulated
with another sine wave of 25 Hz, this will give side bands of 75 and 125
Hz, these are just in the bandwidth.  If the  modulation is much faster,
the sidebands will be out of band.  At first I thought this argument
with four quadrant multiplication of carrier and modulator. In this
case no energy will be left in the channel. But I assume that natural
sound sources really do amplitude modulation, ie. the carrier always
remains. In this case the argument is not so striking, but still:

The channel energy will reduce, if the sidebands get out of pass band.

Therefore we can assume as a rule of thumb that for the example channel
a modulation faster then 25 Hz will not be expected. This means
about 40ms Period, or 20ms rise time.

Another related argument, or rather the same thing from a different
angle can be obtained via Fourier Transform: A brickwall channel with
bandwidth B will give a time function like si(B*t), ie. the period of
the wiggle will be 1/B. (This is from my head, I don't have the Transorm
table with me now, it must be the same as the low pass case and shift
property applied, hell, I forgot that!). Anyway, one can play arround
with a DFT as well.  

The fastest reponse of a brickwall channel is a
sine like funktion with period about 1/B. And this determines the envelope 
speed.

Now, keeping all this in mind we could come to the conclusion, the a 
peek detector may have a better performance than a FWR with filter.
I'm not sure if the positive or negative wave can be neglected,
for impulse like inputs they will not be the same, I guess
a bipolar peek detector with bleeder according to the foregoing
ideas will do a nice job.

So why do people use FWR?
Why do some designs have the same envelope filter time
for each channel?

m.c.




:::From: "Kevin Peek" <kpeek1 at prodigy.net>
:::To: "Synth DIY" <synth-diy at node12b53.a2000.nl>
:::Subject: Envelope follower for my vocoder...
:::Date: Tue, 7 Nov 2000 05:44:48 -0500
:::Content-Transfer-Encoding: 7bit
:::Importance: Normal
:::
:::Thanks to everyone for helping me out so far with my vocoder project.
:::I greatly appreciate it.
:::
:::I am going to change the VCA to one of Jorgen's LM13600 designs.
:::
:::However, many people suggested a simple diode rectifier for the env.
:::follower.
:::
:::Should this be a half-wave or full-wave?
:::I thought that all full-wave rectifiers need a center-tapped transformer?
:::
:::I have looked at env follower from Craig Anderton's book, DR. Q, the
:::polyfusion, etc.
:::I looked at some half wave rectifiers.
:::
:::It seems they are just :   signal----diode-----cap to
:::ground------resistor-----cap to ground-----output
:::
:::Any further explanation or suggestions for this?
:::
:::Once again thank you all!!!
:::