AW: SEM riddles (was: AW: Opinions: On op amp replacement

[Haible Juergen]

	>Look at the mic-preamp on my homepage! This circuit was invented in
the
	>late 60's by a japanese. It was originally build with triodes.
Triodes have
	>almost a square law like the FETs (I think Ia ~ Ug^(3/2)). With
well
	>matched FETs it is easy to get gains from 50 to 500. Better matched
FETs
	>will allow higher gains.

Ah, I see ! 
I have seen this in tube books, but never looked closer. Must spend
some time on it ...

	>>I wouldn't, at least not generally. If you look at the SEM
circuit, LP
	>>filter
	>>output, you have a two stage buffer made of a FET and an opamp.
	>>But only the FET (Q40) is inside the filter's feedback loop, not
the
	>>opamp (A16). I think that Oberheim did that deliberately, in order
to
	>>avoid the delay of the opamp, which would make the Q factor of the
	>>filter less independant of cutoff frequency. 
	>
	>Nah! I don't think so, since the slewrate and therefore "delay" of
a 3080 
	>is almost a linear function of Iabc. And here the 3080 drives a
capacitive
	>load! And the unity gain bandwidth of a 741 is some 1 MHz! 

Well, the 3080 driving a capacitive load is the integrator, i.e. the desired
effect. What you want is just a single pole at 0. If it's an ideal
integrator,
any change on the input would cause a change at the output immediately.
With the iintegration function, of course. So a step at the input would
result
in a ramp at the output, but the ramp would start immediately when the
step occurs. Now with a non-ideal integrator, you might also have a delay.
In the frequency domain you'd have something like k/s * exp(jwT), which
would have an effect on the overall frequency response of the filter.
The 3080 and the FET buffer surely have some delay as well, but it should
be much smaller than a 741's delay.

Using the word "delay" might be questionable, though. I don't know
if a 741 unity gain buffer is better described by 1/(1+jw/wo) or
by exp(jwT) / (1+jw/wo). In other words, I don't know it the T
in the second expression is big enough to make a difference. I have
not checked it myself.
The whole idea came from an old EN article, where they tested several
different buffers in the SVF circuit. If memory serves, the used the 
word "delay", too.
As a fact, they calculated different capacitor values across various
resistors in the circuit, compensating for the effect of the OTAs
and / or opamps being slow.

JH.