matched pairs/quads (linear)

[Martin Czech]

:::Hey Martin, "Korg/Haible scheme" is too much honor ! (blush)
:::I just admire what they invented.

You just brought it up to our attention!  I remember a furious debate
in E.N. linear vs expo, it was about a PAIA design.

So Korg/Paia/Haible would hint the reader into the right direction...

:::
:::More seriously: I haven't thought too much about matching in this
:::scheme. I see the temperature coupling as the main achievement
:::of the transistor pair. My guess would be that an offset voltage
:::would not hurt at all, as long as the offset voltage *drift* is low.
:::The drift is probably linked to the initial value somehow, but
:::I don't really know. Can you shed some light here ?
:::


In linear mode (Ube1=Ube2, balanced) it looks much like a differential
amp. You want that the reference current is perfectly mirrored into
the timing cap. Therefore quite high hfe would be nice.  A low offset
voltage will enshure low offset drift, and therefore the base impedances
(resistors) on both sides should also be equal, otherwise base currents
cause offset and drift. If the pair has allready considerable offset
like in usual pairs (not ultramatched) a 10Meg resistor and a trim
pot on one side would help to trim this away and thereby improving
thermal drift also.  Off course, the opamps involved should also have
low bias current, otherwise they steal current which shows up in the low
end. Low bias means low bias current drift, but jfets have considerable
bias current drift vs temperature.  If you want to have it very stable,
well, a bipolar input has bias, but stable bias.

I gues this must be tried, concerning the range of operation you want.

We always look at the differential pair, but for low currents the opamps
become important as well.

m.c.