[sdiy] CEM VCF transconductance stage difference!

[Tim Stinchcombe]

> Did anyone figured out why Doug Curtis used such a (gm) 
> difference between the last transductor and the preceding 3 
> in his filters? Factors of 50, 75 and 100 times can be found 
> in his various designs.

OK, here's my input, though I have to stress it is more in the 'wild ass
guess' category than a highly considered and thought out response.

So, I reckon it is to do with output impedance of the last stage: for the
3320 (all stages the same), we are *explictly* shown an internal buffer
following each gm stage (Fig 1 on datasheet); for the rest (which I make to
be: 3328, 3372 x100; 3379, 3378, 3389, 3391 x75; 3394 x60; 3396 x70) it
looks to me like there is *no* buffer, but in general the output is feeding
an essentially current-input device, a VCA. Now, from the CA3080 datsheet
Figure 29, and App Note AN6668.1 Figure 1, I'm inferring that the output
resistance decreases with increased gm, and that since transconductors are
current output devices, high output impedance is preferred. Hence by
decreasing the gm of the last stage of the filters, the output impedance is
increased which in general I take it as being 'a good thing', and in
particular makes it work well with the in-built VCA which it feeds.

The datasheet for the 3378/3379, page 5, clearly states that since the gm is
decreased, then so should the capacitor value be for that stage, i.e. that
this is nothing to do with pole placement (in that the design aim *is* to
still have symmetry across all *four* poles).

As for the SSM chips that got mentioned, it is not clear to me why the 2044
might have the last cap lower (perhaps just similar reasons of impedance
internally within the chip before the output signal is picked off the last
mirror? - maybe the patent helps?). The 2045 (=2047) is a completely
different topology altogether - 2 cascaded second-order sections, and the
gain of each stage within each section is different too, giving the
flexibility (within reason) to design your own response (and likely
requiring all 4 caps different). But at least they give the design equations
so that you can do as you please, and their example ends up with 3 different
cap values to help make self-oscillation easily achievable.

Absolutely no idea how close to the mark much of this is!?

Tim
__________________________________________________________
Tim Stinchcombe 

Cheltenham, Glos, UK
email: tim102 at tstinchcombe.freeserve.co.uk
www.timstinchcombe.co.uk