stupid envelope follower idea

[Martin Czech]

:::Nahhhh, not if you're clever.  
:::
:::It's a simple matter to split the input signal into several bands,
:::perform a separate envelope detection on each, and sum the results.
:::That way each frequency band can be optimized and you can have both
:::low ripple and high speed.

Ok for a sine wave, now, what happens in the case of complicated
mixtures of partials and noise? I think the total energy would
divide into the seperate channels, if you pick now the largest channel output,
you'll loose all other energy. This means that broad band signals would have a
too low envelope output compared with some sine wave of the same magnitude.
So, how do you combine the different channel outputs? The phase information
is lost.

Or do I miss something?


:::I'm not getting this; how exactly would AM modulation help?

Balanced modulation should give the same envelope, but with less
low frequency energy, only 50%. Since the lower side band is mirrored,
even less. But I admit that I did not think about what happens after
peak picking that modulated signal, in time domain with "pencil simulator"
it looked ok. Someone wrote it would be too expensive.
Just one 1496 and a simple sine wave generator, never mind distortion
or intermodulation.

m.c.