Temperature regulation

[BadNeurons]

My aim is not to create a sub-zero monster, but to create a quiet, efficient heat extraction system.

I've worked with peltier cooled scientific lab equipment before (my old company made an instrument that had to regulate a small aluminum block at 15C +-0.1C). At first glance, I thought I'd just slap a peltier on the CPU, apply a voltage and VOILA.

The more I thought about it, the worse an idea that sounded though...what about condensation? So I considered a thermistor and opamp driven relay. Melcor, however, strongly urges against pure "on/off" states in their literature. Therefore I've come up with the idea of driving the peltier through bipolar power transistors. A bridged pair of thermistors (one for the cpu, one for ambient temp) feeds a signal to an opamp, which feeds that voltage to the transistors, regulating the voltage seen by the peltier between 1.5V and 23V. The bigger the delta T between the peltier and the ambient, the bigger the voltage seen by the peltier. The upshot is, of course, that you can aim for some (relatively) constant delta T from the ambient.

Has anyone tried this? Apparently some 2N3055's available through Radio Shack are able to take a constant current of 7A at 24VDC. (Running in parallel would, of course, increase the amperage capacity. I'm anticipating using a Drift 0.8)

I've tried simulating the circuit on MicroCap (demo, of course) software, and everything seems OK; with my expected heat output, the voltage should settle at about 12VDC out of a 24VDC linear supply. I am a bit nervous about blowing one of the transistors and/or starting a fire.

Any advice, comments, etc. would be helpful.

[Butcher]

you'll probably need to heatsink the transistors, they'll have to sink any power the pelt doesn't get, so at 12V from 24V you drop 12V over thr transistors, at 7 amps that's 84W on the transistors...
70C is the approximate target for thast sort of electronics so a HSF with a rating of say 0.7C/W would do on the transistors, that's fairly hefty if passively cooled but something with a fan it's similar to a low spec cpu cooler.

[BadNeurons]

Originally posted by Butcher:
you'll probably need to heatsink the transistors, they'll have to sink any power the pelt doesn't get, so at 12V from 24V you drop 12V over thr transistors, at 7 amps that's 84W on the transistors...
70C is the approximate target for thast sort of electronics so a HSF with a rating of say 0.7C/W would do on the transistors, that's fairly hefty if passively cooled but something with a fan it's similar to a low spec cpu cooler.

OK, let me see if I've got this straight. An electronics genius, I'm not. I'm assuming a relatively linear relationship between the voltage the peltier sees, and the amperage it draws. Probably not the best of all assumptions, but in the absence of curves, it's all I have.

If my controlling voltage to the transistors yields about a 6v output, then I've got the following to deal with:

Peltier (11.3A @ 24.6V) wants about 3amps (more or less).
That means there is a drop of 18volts x 3 amps across the transistors, 54W of heat I need to get rid of.

If, on the other hand, my controlling voltage allows 22V to pass:

Peltier wants @ 10.5 amps, therefore, the drop across the transistors is 2V x 10.5 amps, 21W.

I need keeping in line on this, as it's counterintuative to me that the heat dissipated in the transistors will start low, go high in their "sort of on"/linear response state, and then drop again when they're saturated.

So far, my calculations suggest that the peltier will be running at about 12-15V on a normal basis.

Thanks for your help, Butcher

[Butcher]

yeah that's pretty much how it works

[imagex]

I started to do the same thing and it's not easy, at my day job i build rf amps and can get my hands on just about any parts "2907's, 2222's,etc" and have the help of there main tech, it was very hard to find anything that could handle that kind of load, with out big $$$$. if u find a cheap simple way to adjust 24 volt 10 amp I would love to know.

[Butcher]

Yes it's pretty easy really, I built a 15A psu not so long ago. The trick is to use multiple transistors in parallel with some load balancing resitors (typically 0.1 ohm 5W+ type). Put a resistor in series with each of the transistors and away you go.

[BadNeurons]

Thanks for your help, Butcher. I thought I'd probably have to put my fire extinguisher to the test once or twice during this little adventure. I'll still keep it handy, of course, but I think I have a much better grasp on the potential problems now.

BadNeurons

[imagex]

if you ever get it made and would like to produce them let me know i have access to everything needed can brun boards and have pick and place machines.

[Butcher]

Actually I have another idea I'm throwing about right now. My 15A psu is on the heavy side - the transformer weights 3Kg (6.6 lbs), and after talking to a friend's dad who knows of electronics things he suggested just rewinding the transformer in a comp power supply to give all the power at one voltage... Could be damn handy if it works, I'll let you know