Moog VCF (TomG's VCF2B)

[Martin Czech]

> I use the diodes when I want to attenuate CV signals like envelopes.
> That's because I want to quech the last bit of SNR out of it, given the fact
> that I have input signals of  fixed amplitude. If the envelope goes
> up to 10V, I'll make it linear for 12V, for example. So I'm using the
> diodes especially for *low* dynamic signals !
> 
> >OTOH if you want a filter to deal with any high dynamic input (CD,
> >DAT, whatever) you will surely want something better, in terms of S/N
> >ratio. I'm not sure if the 13700 is good enough in these situations.
> 
> Exactly. For signals with high dynamic range like 80 or 90 dB, the
> 13700 is simply the wrong device. I once tried to build a dynamic noise
> filter for a mixer input with a (linearized) 13700, but mostly it would add
> noise rather than reduce it. There are nice VCA chips for these
> applications around - no need for a 13700 here.
> 
> So far I've covered the 1 out of 10 case, and the cases where I don't use
> the chip at all.
> What is it good for ? Synth sircuits where you want to live with a certain
> amount of distortion, as long as it is the *right*  kind of distortion.
> I happen to like how a 3080 (or 13700 without diodes) goes into overdrive,
> and I do not like the abrupt clipping that occurs when the diodes are used.
> 
> So as a bottom line, my rule is only use the diodes when you're sure that
> the input will not clip, and use a non-gilbert input when there is the
> slightest chance for overdrive.
> And what if you don't like the noise of a 3080 ?  Go for low noise devices,
> that still have the same overdrive courve. I think the SSM2024 is (was?) 
> of this kind, and a matched pair of discrete transistors is fine, too.
> 

If I understand you right your point is that the Gilbert network will
not give much S/N improvement. Theoretically it does but in practical
casese you have to keep more headroom because the network will lead to
a sudden, unpleasant rise in distortion.  So the additional headromm
will eat up all of your advantage and even more.  Using the ota without
Gilbert would have some distortion, but with a slow rise , no knee.

?

Makes sense to me. I listened to both circuits (13700 with and without,
and 3080). I noticed some noise. But I didn't ckeck them side by side which
would be necessary in order to get the "pleasantness of distortion".

It is true that S/N must be always measured together with THD for some
meaningfull result.

I'll have a look at Jorgen Bergfors site, for vca data.

If I can't find what I need, well, I could ask one of my analog friends
here in the company about the distortion/noise analyser...


m.c.