Peltier Controller Schematic

Sloman A.W., Buggs P., Molloy J., and Stewart D. =93A microcontroller- based driver to stabilise the temperature of an optical stage to 1mK in the range 4C to 38C, using a Peltier heat pump and a thermistor sensor=94 Measurement Science and Technology, 7 1653-64 (1996)

It's fairly detailed. E-mail me for a reprint - snipped-for-privacy@ieee.org is a real address.

Analog Devices sell the ADN8831 integrated circuit controller for this job

and the data sheet includes a circuit diagram. It won't do as well as my circuit, but it is a lot simpler.

The Linear Technology LTC1923 is comparable, and the data sheet is more informative

When I last looked, neither of them included my formula for calculating the heat transferred by the Peltier junction (in watts per amp)as a function of the temperature difference across the junction and the thermal resistances in the path.

I've recently sneaked that into the Review of Scientific Instruments

Sloman A.W. =93Comment on =91Implementing of a precision fast thermoelectric cooler controller using a personal computer parallel port connection and ADV8830 controller=92[Rev.Sci. Instrum. 74, 3862 (2003)]=94 Review of Scientific Instruments, 75 788-9 (2004)

so that Americans - who tend not to have access the the British journal "Measurement Science and Technology" - can take advantage of it.

-- Bill Sloman, Nijmegen

AFAIK the important point to bear in mind is that the Peltier device acts like constant-heat pump, driven by junction current, in parallel with a resistor that shows I^2-R losses. So the overall heat pump action will be nonlinear (because of the I^2 factor), and it's really a good idea to not try to PWM the thing (because of the I^2 factor).

Looks like they are finally ditching this chip but if you need a one-off prototype it'll work:

Maybe I am, to a small extent, guilty regarding its demise because I bought the demo kit and then decided I didn't like it that much, and rolled my own.

However, with a sizeable Peltier, one should PWM the driver and then lowpass accordingly so that the Peltier sees DC. Or almost DC.

They are now pushing the ADN8831. I've not been interested enough to work out how it differs from the ADN8830.

-- Bill Sloman, Nijmegen

That exactly what I did. Unfiltered pulse width modulation has been known to melt the solder inside the Peltier cooler, but if you can keep the ripple down to less than - say - 10% of the peak current, the extra resistive losses aren't dramatic. Radio-frequency interference from the leads to the Peltier cooler can be a problem a ripple current levels that won't generate significant extra heating and it can pay to filter this current to reduce the ripple content to much lower levels.

-- Bill Sloman, Nijmegen

My only advice is; If you are going to air cool the TEC then make sure the heat sink is big enough. We use a ~10Watt TEC with perhaps a square inch foot print and the heat sink is 6 inches by 4 inches. My boss was amazed that such a big heat sink was needed... but thermal runaway is not good.

The electronics is pretty simple. Thermistor->bridge->PI(D)->power gain.

You can look at C.C. Bradley etal, Rev. Sci. Inst. 61, 2097 (1990)

George H.

Yup, RF-wise I filtered the dickens out of it. Ferrite bead, cap, another ferrite bead ...

I think that's the datasheet where I found some inconsistency in a formula or two and I let them know. Basically those chips are ok but I found it can also be done for much less money. I mean, over $7 is a hefty price. Even if you need a few dozen jelly-bean parts you can beat that. Or use a low-cost uC. I can't imagine the ADN8831 selling like hotcakes.

I should have been more clear on that -- yes. But it's _not_ like a motor driver where even if you PWM slowly enough that the motor doesn't filter it out you're still (usually) not losing massive amounts to I^2-R losses.

So -- generate the current however you want, but make it reasonably smooth going into the Peltier.

I never understood the need for a specialized Peltier cooler IC, unless it's because they do something fancy to compensate for the nonlinearity.

Unless your problem demands hugely tight regulation you can just design a loop with enough margin to handle the nonlinearity, and if it _does_ need the tight regulation then you either handle it with a few discretes or with a micro.

So why a special chip?

The _only_ advantage those things have is that they are small and have no moving parts. It's interesting to educate someone who doesn't realize how inefficient they are, or how much heat they generate, or how you have to get that heat away.

Oh it's very nice that they can cool as well as heat. Otherwise I'm stuck running something above ambient... And then dealing with the fact that most likely I can get heat into the thing faster than it comes out.... so the control loop has to follow the slow side.

Say with a micro controlling things could you have different time constants for cooling and heating? Do you do that sort of thing?

George H.

Yes and yes.

I'm not sure that is the best way.

If you could measure the temperature at both sides of the Peltier (or if you had a way to estimate the temperature on the heatsink side) you could take a heat flow command and turn it into a current command. This would keep the loop dynamics mostly independent of the temperatures -- except for the fact that you'd always have way more power to heat than to cool.

Saves pick&place costs. But at $7 a pop, nah, that was too much for the privilege.

My case did require hugely tight regulation. Any kind of regulator rumble or other noise on there would have messed up the signal we were after.

Would you shoot me a reprint of each, Bill?

There's also some stuff on how to model TECs based on datasheet parameters in my (free) thermal control chapter,

ftp://ftp.wiley.com/public/sci_tech_med/electro-optical/thermal.pdf.

Cheers

Phil Hobbs

My

That's what my paper describes, amongst other things. E-mail me if you want a reprint.

-- Bill Sloman, Nijmegen

yep, the integrated ones I've seen is basically two buck converters driving each end so it can both heat and cool

-Lasse

re

...

It also saves design and debugging time. Anything that is inside the chip doesn't have to be put onto a schematic or laid out on a printed circuit board, and the draftpersons involved don't get the chance to do it wrong.

The performance offered by the chips isn't impressive. Sticking a pulse-width modulator inside the same chip as a sensitive high gain analog amplifier isn't making life easy for the chip designer (or the amplifier).

-- Bill Sloman, Nijmegen