[sdiy] 4-pole SVF theory

[David G Dixon]

> Another is what is called sensitivity -- how sensitive the 
> design is to component tolerances.  The more stages you have 
> stacked up then generally the more sensitive the circuit is 
> to component tolerances, which will degrade the performance 
> of the filter (e.g., the cutoff may not be as sharp as you 
> intended as the poles and zeroes aren't in exactly the places 
> they should be).

That's a very good point, particularly for SVFs and other filters which rely
on summing of outputs and feedback to create various filter modes.  A bit of
frequency response simulation is sufficient to show that such synthetic
filter modes are very sensitive to resistor tolerance.  I have taken to
using hand-matched 1% 30k resistors in all of my own 2164-based filter
builds, and in our commercial (Intellijel) filter designs we are using 0.1%
30k SMT resistors everywhere.  Given the otherwise very tight matching of
VCAs in the 2164 chip, this should be sufficient to ensure very well defined
synthetic filter modes.  Of course, one should also use fairly tightly
matched capacitors, but this has less of an effect than the tolerance of the
summing resistors.