envelope follower

[Don Tillman]

   Date: Fri, 03 Oct 1997 16:11:49 +0200
   From: Rene Schmitz <uzs159 at ibm.rhrz.uni-bonn.de>

   > it seems like it could be used in the construction of a very
   > accurate envelope follower

   Unfortunately not! I had thought about this my self, but recent
   investigation shows, that when you square a signal like
      a1*sin(w1*t)+ a2*sin(w2+t)
   you get

      a1^2*sin^2(w1*t)+2*a1*sin(w1*t)*a2*sin(w2*t)+a2^2 sin^2(w2*t)

   since (a+b)^2 =3D a^2+2ab+b^2  :-< (the old binomial pitfall)
   even if you have the cos by a phase shifting network, this doesn't
   get you where you want.

True, but the extraneous AB term will average out to zero over a few
cycles.  

I *think* (I haven't done the math, but we'll leave that as an
exercise for the reader, okay?) that you can rewrite that extraneous
middle term as a sum and difference, do the same thing for the cosine
part of the circuit, and when you add them together the difference
frequencies will cancel out and you'll be left with only the high
frequency sum frequencies, and they'll be extra-easy to filter out.

  -- Don