Hmm, where did all my words go? I could've sworn they were on the screen a minute ago.

Trying again...

That's better!

The whole 'don't boot until everything is cold' is easy to accomplish. Attach the thermistor to the cold plate, and use the output (high = cold enough) to drive a npn transistor spliced into the power_ok line, or a pnp transistor across the reset switch.

For a turn on temperature of 0C, the pot would be set to around 40k. The 150k resistor will give about 5 degrees of hysteresis; smaller = more hystereisis, larger = less.

A second copy will give overheat protection. Thermistor on the hot side, pot at 2-3k (60-70 C), output drives a pnp transistor across the ps_on line.

Soft shutdown - any ideas on how to get the temps into the computer?

I haven't read the tread, but for the picture you want some termal-protection curcuit.
The peltier is runing from 12V on the psu, by using the power_off, there is only 5VSB, so you would be waiting a long time for that computer to turn on :) Be better off to go for the reset instead.

Why don't use RS232, I heard (haven't testet it) that with most computers you can use +5 and gnd instead off +-12V, take one of the hand-shaking pins, and have it turn high/low.

two pages......?

Yea, you could use the serial ports. I've never fed data into a PC that way, but it shouldn't be hard to figure out. Software is my big impediment - I've zero programming skill. Who will write the code?

I'd be more likely to add an LM83 to the mix. It's a SMBus temperature chip (one local, three remote) supported by MBM5. It just seems kinda silly to have both a thermistors and a thermal diodes attached to the pelt.

Guys, if I may interject a simple comment here:

No offence meant, but building a RS232 interface is not exactaly simple, and I do think that we should be concentrating on the focus of this design, instead of nice extra details.

I think the hardest thing to do is get the temp sensor interfaced, why not work on that?

I've bread-boarded and bench-tested this one. It will do the essentials as long as your pump and pelt are activate by relays.

At a cold side temperature above that set by the 50k pot, the lower comparator has a zero volt output, which keeps the optocoupler off, which keeps the motherboard from getting the power-ok signal. After a little thermistor chilling, the comparator goes high, turns on the opto, and the board boots. With the comparator high, the 150k resistor provides positive feedback, putting the reset temperature about 5 C higher.

On the hot side, a similar thing. If the temperature is too high, the upper comparator goes low, turning off the optocoupler and interrupting the powersupply-on line, which shuts off PSU.

OK. What've ya got?

Hmmm, are there any reasone for using those optical buffers (in lake of a better word)? Need to use mony? :)

The opto's are a result of a failed first try. I had planned to just use transistors to turn things on and off, but when I was playing with test versions the results sucked.

Even a 'rail to rail' comparator or op-amp won't quite get all the way down to zero volts when low. The output was enough that the absurdly sensitive computer was turning on and off randomly.

With the opto-isolaters I got consistant switching, due to their lower gain and higher turn-on voltage.

As for the $$$, Digikey has 'em for under a dollar.

Very nicely done. I second the use of opto-couplers here. There are a ton of reasons for using them. First, they physically separate the two circuits if you power the comparator circuit from an external source. Second, they avoid the "chattering" and other bad properties of relays if you were to go that route. Third, as was noted, transistors often seem to have minds of their own.

Always handy for testing. And, if the circuit is powered by the pelt's supply, it makes it easier to add in pulse-width-modulation like the SubZero4G's.

Mmmm, pulse width...