[sdiy] why 10V

[Don Tillman]

   > Date: Fri, 19 Sep 2003 18:10:09 +0200
   > From: =?ISO-8859-1?Q?Ren=E9_Schmitz?= <uzs159 at uni-bonn.de>
   > 
   > Why not use a more linear network as the "Weaver" type for a
   > frequency shifter. At least there is no problem with clipping
   > from stage to stage with these. Only at the output you have to
   > watch out for the amplitude.

Well, my answer about the Hilbert transform was interpretting JH's
question as exploring what a worst case situation might be.  I wasn't
assuming any restrictions or application for the all pass filter.

(But of course we know that JH builds frequency shifters!)

This points out exactly why dome filters are used for frequency
shifters.  A Hilbert transform, besides being difficult to
approximate, would create very large spikes from a square wave input.
The dome filter approach is easier to build and well behaved.

By "dome filter", I mean a pair of basic all pass filters tuned so
that their outputs are offset by 90 degrees.  There are several
implementations of dome filters; I can think of three right off:

  a chain of RC phase shifter stages
  a chain of 2-pole active all pass stages
  the web of RC's (in "The Art of Electronics")

  -- Don

-- 
Don Tillman
Palo Alto, California
don at till.com
http://www.till.com