I don't understand thermostats

Not yet. A few watts likely.

I can't afford to let the loop go linear here, or I'll toast the amp.

Before I read the other posts, I want to say I was thinking about this on the ride home. The bang-bang control, comes at the oscillation of the P-only control.. but from the other side. (The other side of stability space...) Which is cool. As long as you don't care about the time constant, and some oscillation, then you've got a useful system...

I still say you've got a knob that is the max power of your bang-bang. (and too much power will give oscillations.)

I've been dreaming about building my own thermal test bed. I'll now have to include bang-bang.

George H.

One thing (I think) I've found out about TEC's is that how tightly they are clamped in place, can change the gain. (little pieces of crud between the surfaces can also ruin a day.)

George H.

I use the electrical model for thermal stuff all the time. It works just fine (to the ~20% level).

I didn't find the modelling very hard at all. I wanted to define a new unit, the thermal ohm... Tohm (degree K [voltage]/Watt[current]) I'm not sure if it should have a "th" sound or just "tee".

George H.

We tend to call that "theta."

Would a heat pump count as a thermal gain element?

My 1996 micro-degree paper (of which you've got a copy) has got an equation for TEC performance, where the summed thermal resistances between the TEC and the object being controlled as well as between the TEC and the output h eatsink influence performance - if either is significant they can really l ower performance. We used graphite cloth as a tidier and better-performing) substitute for zinc-oxide loaded silicone grease.

I've got a Honeywell bimetal thermostat in my junkbox with a 1K hysterisis I use it as a room themometer.

Imagine trying to build an engine with 5% tolerance parts.

Yup. For one-offs I sometimes lap the TEC and heat sink together with fine valve grinding compound. Thins down the bond line pretty well.

If you get the hard-solder ones (Ferrotec/Tellurex/Marlow), you can soft-solder them to the load and heat sink, at least for sizes less than

With larger TECs, or arrays of them, the shear from CTE mismatch will crack the bismuth-telluride pillars. (The alumina is in compression, so it doesn't usually crack.)

Cheers

Phil Hobbs

Chip design isn't so bad. The whole engine might be too large or too small by up to 30% but at least the parts usually match each other within 0.1% or so.