Ideas on how to efficiently power TEC modules

Thread Starter

PaulEE

Joined Dec 23, 2011
423
I am working on a project that involves the use of high-power thermoelectric cooler modules for use in cooling metallic material in a circuit. The TEC modules need to be able to move up to 30 watts of heat energy, and since TECs aren't that efficient, I need a circuit that can drive 60 or so watts into them. (Hand-waving going on there. This isn't the meat-and-potatoes of the post. Please continue on...)

My question:

What's the most efficient circuit that would allow a control signal to throttle the power through the TEC modules? The signal is coming from a 0-+5v DAC but it can be multiplied, divided, or shifted as needed.

Ideas I've tossed around thus far:

1) Linear/power op-amp/variable voltage regulator with series pass transistor
-->Each of these dissipate more power between "full on" and "full off" than the TEC modules themselves. Grossly inefficient.

2) Half-bridge/H-bridge/pulse-width modulation/variable switching supply
-->The PWM H-bridge is currently what I'm using. With low Rds-on MOSFETs, I can get FAR better efficiency than the linear stuff in 1).

I also considered using a national semiconductor (now TI) "simple switcher" variable switching regulator, but could not figure out how to control the output voltage with something besides a resistive divider for voltage feedback. There were other issues with the switching regulator as well
-->lowest output voltage was above zero.

If anyone has any other suggestions regarding a better/improved way to do this, or if you feel this is already the best way (the PWM and H-bridge), I would appreciate any feedback. Along the lines of PWM, I built circuits to discretely generate the PWM signal (ramp, comparator, level-shifting, the works.) If anyone has any suggestions on an already-made pulse-width modulation IC, I'd appreciate feedback on that, also.

Thanks in advance.
 

Thread Starter

PaulEE

Joined Dec 23, 2011
423
P.S. - It was implied but not explicitly stated...the power needs to be variable from full cool to off to full heat...hence the H-bridge topology at the moment.
 

wayneh

Joined Sep 9, 2010
16,114
You'll want a full-on, full-off design as you suggested. Getting proportional control with a linear system is just too inefficient, and HOT! from the current involved.

I built a calorimeter with a simple (ie. not PWM) thermostat control system. You can find it in the completed projects forum. The control I was able to achieve was actually quite good, something like ±0.1°C, so I don't much regret not using PWM.

One problem you'll want to design for is ∆T/∆t. That is, you don't want to damage the TEC with massive thermal changes. You don't want the control to go from full heat to full cool without a dead band or some ramp-up of the duty cycle.
 

Thread Starter

PaulEE

Joined Dec 23, 2011
423
wayneh,

The device in question is cooling a current shunt. It is going to dissipate a lot of power, and the TECs are going to keep the temperature low such that the shunt characteristics stay within predetermined limits.

I figured that it would be a switching design in the end, and I realize that abrupt changes in temperature, as well as large gradients across the module, will destroy the module. I know this because I've already made the mistake. :)

Any thoughts on the intricacies of the switching circuit topology? Thoughts on PWM chips, if that solution is the most practical?

As from the above thread-starting post, I need to have proportional control. It can't be just on or off in this particular case.

Thanks again for the suggestions.
 
Last edited:

n1ist

Joined Mar 8, 2009
178
If you are both heating and cooling, you would need either an h-bridge or bipolar supply. Otherwise, a single switcher would suffice. You can vary the output of the switcher by summing in the dac voltage to the voltage divider at the adjust pin with a resistor.

/mike
 

crutschow

Joined Mar 14, 2008
23,500
If the PWM meets your needs than I would stick with that. I don't see any other approach giving better efficiency or being significantly better in any other respect. You don't even need a bridge, just one transistor to switch the current on and off.
 

Thread Starter

PaulEE

Joined Dec 23, 2011
423
n1ist, I think I'll stick to the H-bridge/PWM design, then. Thanks for the feedback on that and the feedback (pun intended) on the switcher.

crutschow, I can't change direction of current flow through device with one transistor..?

Thanks for feedback on the PWM. I figured that's about as efficient as I could get without getting into really complicated things, but I just wanted to put a feeler out to see what others thought.

Thanks everyone
 

crutschow

Joined Mar 14, 2008
23,500
.......................
crutschow, I can't change direction of current flow through device with one transistor..?

.....................
Why do you need to change it's direction? If it's only going be used for cooling then current only needs to go one direction.
 

steveb

Joined Jul 3, 2008
2,436
Just a quick comment.

I didn't see any mention of PWM frequency, so I thought I'd mention something. I'm not in any way an expert on TECs, but I did work with them in various control applications years ago.

I remember talking to an expert over at Melcor and his recommendation was that PWM frequency should be greater than 2 kHz, and preferably greater than 10 kHz. I don't remember the exact reasons, but I have a vague recollection that it has something to do with thermal/mechanical stress that impacts the long term reliability.

Just thought I'd mention it in passing. I'd recommend calling the manufacturer to verify this. Times change, - technology advances with time - and my memory degrades with time.
 

wayneh

Joined Sep 9, 2010
16,114
I'll reiterate something also: The biggest problem with using a TEC for cooling, is removing heat from the "hot" side. This is made more acute by the fact that the TEC adds ~9 heat units to each 1 unit it moves from the cold to the hot side. So adding the TEC increases your heat dissipation problem ten fold (OK, 9 fold in my example). Of course, the reason you use a TEC is to drop the cold side temperature lower than you could otherwise, even with infinite dissipation of the heat.
 

Thread Starter

PaulEE

Joined Dec 23, 2011
423
Thank you all again. As per the last post, I've looked at a scientific article that confirms this. I believe the article used something like 23khz.
 
Top