Message written at 26 Sep 1999 15:54:06 +0100
--- Copy of mail to f00f@freemail.hu ---

regarding tone, if the half-power frequency is set to 1, then the
algorithm needs to calculate cos(1/(2pi8sr)).  For your sample rate
this is cos(0.000142) which is 1- 3.8*10^(-12), which to single
precision is 1.  Then the filter coefficient c1 is 0,(and c2 is 1) and
so there is no sound.  
  This looks like a numerical problem rather than a bug as such.

However working in double precision improves this, but I assume there
is still a value for which teh numerics will break down.