quad 1/2 bridge

Half-bridge audio amps can be tricky. At that price, may as well go full-bridge. A half-bridge switcher can pump power uphill, back into the power supply.

I wonder if we can get our little RP2040 cpu chip to do a multichannel closed-loop supply, and maybe generate spread-spectrum PWM on 8 channels.

Phil, the 8962 data sheet discusses driving TECs.

In that package, they’re probably intended for driving LED strings in displays, where board space is not a problem.

Cheers

Phil Hobbs

It's a lot smaller than the four 1/2 bridges I was considering. Each had an SO8 driver chip, two dpak fets, and a bunch of discretes.

I was about to Dremel this:

Thanks.

I rarely have the luxury of using a barefoot switcher for TECs. The capacitance from the supply to the cold plate is tens of nanofarads, and anyway a small TEC runs way under 5V in normal use, so it usually needs a buck.

We have a couple of nice canned designs that use variations on Class H, i.e. a fast buck switcher with a very low-Vsat linear stage that gets rid of the ripple.

The switcher control is stupid simple. The MCU DAC is summed with the switcher feedback network. The MCU very slowly drops the output voltage until the analog loop rails, bumps it up by 200mV or so, then starts dropping it again. You wouldn’t do that with an audio amp, but it works great for thermal control.

For cooling-only, we use a two-pole BJT current source (like a BJT gyrator but with two poles).

For use near ambient, as with diode lasers, we use a slightly more complicated thing, an asymmetrical BJT bridge. One side is a complementary current conveyor (current in, current out) and the other is a regular voltage-output complementary class-B that keeps the load terminals moving symmetrically. (All that is to say that one side comes out the collectors and the other comes out the emitters.)

This approach allows current sensing in all four legs, which turns out to be very handy—besides short circuit protection, it also lets you have different current limits in the heating and cooling directions.

The other trick is choosing the polarity so the voltage-output side is sourcing current when cooling. The control amp runs off the higher rail, so the NPN’s V_CE can go way into saturation. This isn’t an issue on the current conveyor side.

Both fit into under a square inch, including pours, and can drive the usual

Cheers

Phil Hobbs