2040/3320 Differences? was Re: also... Re: [sdiy] SSM2044 Filter schematic??

[René Schmitz]

Hi Jürgen, Scott and all,

JH. wrote:
> It's the nonlinear behaviour that makes a difference, where the "filter"
> isn't
> just attenuating and amplifying harmonics of the input signal, but is adding
> harmonics of its own, and consequently filters theses harmonics with the
> following filter stages. 

Indeed.

 > Add to that a special dynamic behaviour that slowly
> moves the bias point (and thus the type and amount of distortion) as
> a reaction to large input transients, not unlike a tube stage with "soft"
> power
> supply.

I'd say the resemblance is even closer. In AC coupled tube amps the 
overdriven grid can get into conduction, and dynamically shifts the bias 
of that stage. (With the time constant determined by the grid resistor 
and the size of the coupling cap. "Audion rectification".)
Here its not the input of the following stage that generates this effect 
but the driving stage which can only source current, but not sink to a 
voltage less than determined by the Vce voltage of the diff pair 
transistor. But the effect is pretty much the same.

(An interesting aside here, since its the same effect: I think this is 
also responsible for the differences between AC coupled and DC coupled 
linear FM inputs. If you look at the summing node of the expo servo, you 
notice that it can sink current, but not source it (since the 
transistors would be back biassed.) So there would be more flow into the 
node than out of, with consequentially a build up of DC potential across 
the cap.)

> This all comes from an interaction of gm cells with limited output
> voltage swing, and the darlington buffers. I'm pretty sure that part of this
> was a "happy accident", as Dave Rossum certainly was aiming for a
> "high fidelity" 4-pole filter (just as he did with the 2044, and as Curtis
> did with the 3320). It was a certain economy on transistor count that
> led to the 2040 structure (two current mirrors less per gm cell,
> compared to a 3080-based filter), and which created a true classic
> in filter design.

Then again, if he had used integrators instead of buffered caps this 
compliance "problem" would have been cured. (But I guess the pin count 
would be the limiting factor for that approach.)

Cheers,
  René

-- 
uzs159 at uni-bonn.de
http://www.uni-bonn.de/~uzs159