Hi, all,
I'm designing a diode laser controller board for a customer in Scandinavia. It has a sub-Poissonian adjustable current supply, a resistor-linearized thermistor temperature sensor, and of course a thermoelectric cooler driver.
The TEC requires different current limits in different quadrants. (A TEC is actually a four-quadrant device since it generates electrical power from the temperature difference as well as the other way round.)
With a bridged current driver, that can run into quite a few parts--you need one current limiter per polarity, or if you want any short circuit protection, one per leg (4 in all). So simplicity is at a bit of a premium.
The usual two-terminal, two-BJT current limiter suffers from a fairly gross temperature coefficient, like 3000 ppm/K. Protecting a TEC isn't a super high precision application, but something a bit better than that would be good.
I've sometimes got round that by putting a Schottky diode in series with the sense resistor, like this. (The base current for Q1 actually comes from a RRO op amp via R1--it's a class B complementary bridge.)
Q1
0-*--------- -----*---->|---R2R2---*------0 | \ A | | | ------ ------- | | | / V Q2 | *--R1R1-------*---- -------------*Trouble is, Schottkies don't have the same dV/dT as BJTs--it's more like
1 mV/K at high current, vs. 2 mV/K at low current, so you don't gain that much--about 1500 ppm/K.A PN diode connected as above plus something like a TLV431 in series with Q2's emitter will do a good job, probably 300-500 ppm/V, which would be fine. However, that costs four more ICs and another 2.4V of headroom (1.2V per side in each polarity).
Seems like an arbitrage opportunity. ;)
My current thinking is to put the diode in series with the base of Q2 and use Q1's V_BE to provide a reasonably stable bias current, like this:
Q1 1ohm
0-*--------- --------*-------R2R2-----*------0 | \ A | | | ------ A Schottky | | | | | *---R1R1------*--R3R3----* | | | | | ------ | | / V Q2 | *------- --------------*The diode runs at much lower current, and so has a higher dV/dT, and by choosing R2 correctly, I can adjust it to match Q2's a bit better.
One wouldn't want to bet one's firstborn child on the unit-to-unit consistency being too wonderful, but ISTM that I ought to be able to get a factor of maybe 6 this way.
Anybody else done this sort of thing? Better suggestions welcome!
Cheers
Phil Hobbs