Gain tweak--needs to be medium fast and very very quiet

So, a matched pair of JFETs, one sets gain and the other for feedback. This trick only works on low potential swings, but a few percent of a 250 mV swing qualifies nicely. Alas, the occasions to do this kind of thing are few, and matched-pair JFETs are not a common item. Digikey gives you the choice of PMBFJ620 or nothing.

That's the whole point of the circuit. It's an integrating temperature servo that forces all three arrays to have the same temperature.

If one of the NPNs saturates, its dissipation goes to near zero, and the others warm up to match. In steady-state, the V_CEs adjust themselves so that the dissipation in each array is just what's required to bring them to the same temperature. That's why it's so cute.

Early voltage limits the low-frequency gain--a 2-mV V_BE error produces about 9% change in collector current. With a V_A of 50V, that corresponds to about a 4.5 V change in V_CE, which at 5 mA I_C corresponds to 22.5 mW change in dissipation at DC. So it isn't quite an integrating servo--it needs a bit of help. A bit of positive temperature feedback would work, but it's probably easier and more stable to us an auxiliary temperature control loop to supply the missing low-frequency gain.

The servo actually drives the array temperatures to the point where the Delta V_BEs are exactly the same. Due to slightly different emitter areas and R_EE' values, that won't quite correspond to Delta T = 0. Elsewhere in the circuit I'm measuring Delta V_BE as a function of collector current on the other array transistors, so it isn't too hard to put a millivolt or two on each base to account for that.

What I like about the circuit is that it can force as many arrays as needed to have the same temperature, with almost no additional circuitry, and that it responds fast enough to compensate for the dynamic dissipation in each of the arrays.

Cheers

Phil Hobbs

ance

ext -

OK. there's always at least one wiper in the low impedance path.

George H.

One other thing--there is one possible start-up problem. If one NPN starts out significantly colder than the others, its I_C will be too low, and its corresponding PNP current source will saturate, so that its I_C will drop, so its dissipation may not be enough to bring it back into range. The auxilary slow loop will take care of that too.

Cheers

Phil Hobbs

Even so, as long as the dpot in the low impedance path can drag the tap point at least 8% low, I can get the fine adjustment from the other one, which is why it's clever. (I missed it the first time round, as I said.)

The only problem is whether the variation of the wiper resistance is something manageable.

Cheers

Phil Hobbs

Well, it works, with a thermal control bandwidth faster than 1 kHz. See "Gain Tweak from SED" on alt.binaries.schematics.electronic.

Cheers

Phil Hobbs

output.

net

in

Its in the fridge waiting for people to show up!