2040 filter clone with unmatched transistors ?

[Magnus Danielson]

>>>>> "HJ" == Haible Juergen <Juergen.Haible at nbgm.siemens.de> writes:

 HJ> Hi,

Hi Juergen!

 HJ> I was wondering lately how precisely the transistors in my SSM2040-style
 HJ> filter must be matched. (http://www.synthfool.com/diy/jh2040.gif, or
 HJ> http://www.it.kth.se/~e93_mda/synths/friends/haible/jh2040.gif)
 HJ> What I have written in the schemos is what I have *done* when I built mine,
 HJ> i.e. matched the transistors in pairs to delta Vbe < 1mV @ Ic=100uA.
 HJ> But I am wondering if this is really necessary.

Ah. One thing that I would match would be delta Rbe, something that is
important in many cases, this is in effect what you are matching since
match the voltage Vbe at a known current. Juergen will recall a
private email chat about the importance in Rbe matching for the use in
exponential converters. In a way is a diff-pair working in a quite
similar way even if it is not obvious. The common emitter current
should be split equal if the Vbe1 and Vbe2 is equal (thus the base
inputs have equal input voltage), however the currents is important
and the difference in bulk resistance will manifest itself as a
different current/voltage relation and the resuling splitting of
common emitter current will not be equal. This will directly effect
the CMRR capability of the diff-pair. I think that this should show up
in Spice simulations (BTW, read in Bob Peace book about Spice simulations).

The active loads that the current mirrors will act as will improve the
raw gain of the cursuit alot.

The current-source for common emitter will greatly improve CMRR as
compared with just a resistor based current source (a resistor to
negative supply).

BTW. Did you clone the SSM2040 transitor setup out to BC550/BC560?

 HJ> Think of this: both transistors of each npn differential pair are working at
 HJ> very different Vbc. So far, I have ignored this effect, but Spice
 HJ> simulations
 HJ> tell a different story. With one collector near 15V, and the other one
 HJ> near 0V, the whole thing must be rather unbalanced. A Vbe difference
 HJ> as high as 10mV would cause a similar effect. One could possibly
 HJ> cure this "problem" with some additional transistors to form a cascode,
 HJ> but I'd rather look at the whole thing from this point of view now:
 HJ> It does work in spite of this effect, so why shouldn't it work with
 HJ> unselected
 HJ> transistors, too ??
 HJ> I have not tried this on the real thing, and probably don't find the time to
 HJ> do it
 HJ> in the near future, so I'd like to hear comments if you can confirm this
 HJ> from
 HJ> theory, or even better, from building such a filter with unmatched
 HJ> transistors.

 HJ> Any ideas ?

I hope so ;)

Cheers,
Magnus