Thom Kouwenhoven wrote in a message to Greg Mayman:

 TK> A very good afternoon to you Greg

 TK> On 30 Dec 02 22:04 you wrote something about *I/F*...

 TK>> Another one: Why not taking a 4066 (make it two) and switch witin
 TK>> a resistornetwork....a bit more noise and distortion than the
 TK>> dedicated chips but still usable.

 GM> Or use a 4066 as a switch on the signal and drive it at some fast
 GM> rate, say 50KHz, with a variable duty cycle square wave.

 TK> Hmmm...i can see where this is going....a PWM regulator.
 TK> Don't forget to filter...just to keep artifacts from the poweramp
 TK> and the speakers :-)

So if,  for example,  you had a resistance connected in series with your
audio signal,  and a pair of 4066 or 4016 gates,  one grounding the signal
and one connecting it to the output,  a couple of nontrivial questions jump
right out -- those being the question of clock frequency and output
filtering.

I'd tend to assume that the clock frequency would need to be up there in
the tens of KHz at least,  probably around 50 KHz if you wanted to process
the whole of the audio band,  though it could perhaps be a little lower for
other applications.

And assuming that the output filtering would consist of a capacitor across
the signal,  how would you determine the best value of that part?  Would it
also probably be a good idea to have a resistor across that cap,  as well?

And,  does dc biasing or levels come into the picture with this at all? 
I'm thinking it probably would,  unless you had the CMOS powered with both
positive and negative power supplies...

---