AW: Re: [AH]: Questions regarding S/H

[Haible Juergen]

>What are you expecting to get by running other sounds (rather thatn noise)
>into the s&h?
>[...]
>Sampling an oscillator would not give anything useful, except that
different
>shaped waveforms would give somewhat different distributions.(in the
limiting
>case, a square wave gives only 2 levels, with the distribution varying with
the
>pulse width modulation).

Hi Paul and list,

haven't got the start of this thread, but what you describe here, changing
the
distribution of random signals, can be quite useful indeed.
Sometimes you want a "strong" S&H type modulation, but you don't want
the random values go out of a certain limit (for example to avoid a note
being completly muted by a low filter cutoff). Amplifying the noise input,
and clipping the S&H output may be one way, but it might not lead to the
desired
result, as values that would be normally out of range are now just
accumulated
at the borders of the limited range. 
Creating a limited range of random CV steps with a *chosen* distribution
is a useful thing. Sampling an oscillator waveform, with the oscillator
frequency
modulated by noise is a good idea. Not sure if, and under which boundary
conditions, the oscillator waveform represents the distribution of the
random
steps *directly*, but it might be a good approximation at least.
I think the Steiner-Parker Synthacon (sp?) has its S&H input hard wired
to one VCO. And the famous Buchla "Source Of Uncertainty" (sp?)
has an LFO which is modulated with noise, and the LFO is then sampled.
(It's not exactly FM of LFO with noise, it's more like a 1st order PLL,
which might not make much difference, but simplifies the circuit.)
I *guess* that the aim was to get a uniform distribution (from the linear
slopes of the triangle LFO), but as I said, I can't be sure, because the
slopes
of the modulated triangle aren't precisely linear anymore ...
But let's assume that this is a good starting point. Then there is a
potentiometer
to make a mix of this noise/LFO signal and the output of the S&H. This mix
is used as S&H input signal, so you can control the *correlation* of 
successive output samples with the turn of one knob.
I have not built this yet, but it's high up on my list, because it's one of
the
most interesting S&H circuits I've seen so far.
There is this other method of course, artificially increasing the
on-resistance
of the FET switch with a potentiometer connected in series (will limit the
speed of camacitor charge, and thus increase the correlation, too), but
here the effect is not independent of clock rate.

JH.