[sdiy] tanh distortion in a filter

[David G. Dixon]

> The 'sine shaper' only works at one particular input 
> amplitude. Too low, it does not distort, too high and you get 
> a square wave anyway. I was thinking it could be 'good' for a 
> filter but then changing input amplitude changed my mind.

Right, but isn't this still more or less what you want a distorting preamp
to do: distort high-level inputs and leave low-level ones alone?  This
sounds like an asset to me, rather than a drawback.

Simulations suggest that this would actually work pretty well, because the
sine shaper boosts the signal at low levels and limits the signal at high
levels.  If, for example, a 10Vpp input signal is amplified at a gain of 1.5
to 2, then attenuated with a pot, the signal would come out of the sine
shaper largely undistorted at about 6Vpp at a pot setting of about 20%, at
50% it would be "sine-shaped" at 10Vpp, and at 100% it would be well and
truly distorted at about 11Vpp.  Since perceived volume is exponential with
amplitude, there is not that much audible difference between 6Vpp and 10Vpp.
Also, constant-level signals are typically patched to filters and AM VCAs
are typically patched downstream, this would be OK.  You would set the
distortion level as desired, adjust any input attenuators on the filter to
get the right level or mix, and Bob's your uncle.

> In fact, I think I have seen (am seeing) an effect where an 
> LPF works 'better' with a higher input, and as the input 
> decays, the overall cutoff seems to go up (higher center 
> frequency). I theorize its that 'sine shaping' that is 
> somewhat inherent in the OTA filter makes the LPF better at 
> higher amplitudes, worse at lower amplitudes.

Again, wouldn't the sine shaper in front of, say, a 2164-based filter (which
seems not to distort at any amplitude unless it clips) give it that OTA
flavour (i.e., distorting at high amplitudes)?

> anyone adept at math-fu please propose any alternate 
> explanation for my observation... 
> 
> The Moog ladder is of course famous for its sine-shaping 
> overdrive characteristic...

This could be a simple way to make any filter sound like anything from
super-clean to OTA to Moogy.  I'm gonna try it this weekend, if I have the
time.