TEC Optimization

I've been working on a projector that uses a deeply cooled array of red diodes (Mitsu P73's for now maybe Oclaro's soon). Anyway, the three stage TEC stack was designed by me (maybe a bit of a mistake) and the cooling performance was a lot lower than I hoped. With a base plate temperature of 28C I had expected to get between -25 and -35C and saw only -10C. The problem with TEC stacks is a little like multiple element laser beam manipulation. Little errors early on such as miss-estimates of the cooling load from the lasers and the environment are magnified with each subsequent stage. Also if you plan to drive them in series like I did then the specific devices will not necessarily optimize at exactly the same current(must be identical if series driven). Finally, the interface between the three coolers (in my case), the load and the base can't be perfect. That's 5 interfaces that each add an unknown thermal resistance.

So, after a couple of hours of struggle, I managed to unsolder each cooler from it's neighbors and solder on individual leads and feed them all out through airtight holes within the base of the hermetic chamber.

Now as I write this, with the set up otherwise unchanged, in other words this is not a mock up or experiment, but the actual working model with all the variables set, I am slowly adjusting drive voltage (current) to each cooler independently and WOW! I am now at -25C and with continued optimization -30 seems likely. Once all the voltages and currents are known then series vs parallel, add a resistor here and there and finally the necessary power supply.

This is how I will do it from now on.












i

I'm confused - you're saying that just by re-wiring the 3 TECs in parallel (with individual current control on each one) you were able to get better performance than you originally had with all three operating in series (and presumably running all three wide open)?

That's definitely counter-intuitive. Can you post a picture of the rig?

Adam

I'll post a video of the projector that the module is mounted in. I'll do this when I compare the two projectors that will be used to do a side by side of the EMS 8000 scanner and the Saturn scanner from Pangolin.

The reason this is so difficult is that there is no real "wide open" setting for the TEC's unless someone wanted to say it is when it simply melts. The performance curves of these devices in general show a range of cooling that slowly becomes less efficient with increased current. With a single TEC the cooling improves albeit evermore slowly as the current increases until the base plate temperature begins to rise more quickly than the delta T increases due to increasing total thermal load. Better base plate cooling will alter this point. Lower thermal load from the source(diode in this case) will also alter this point.

Once you stack another cooler, these same rules apply, but there may be a counter intuitive interaction. Increasing the current to the top TEC to it's maximum delta may so severely impact the cooling of the bottom by moving it to a less efficient region that the net delta is less than if each were operated at less aggressive currents. Then add in the uncertainty of the thermal interface(s) and additional stages and the optimal current for each stage is difficult to predict.

With direct series operation the different devices are forced to operate at a common current and this may be acceptable, but is unlikely to be optimal. There are only a fixed number of discrete devices available.

The conclusion is that I was able to hit-31C and it actually allowed me to run the top and bottom TECs in series and the middle TEC at about 60% of the current shared by the other two coolers.

The reason I went this route is that the commercial multiple stage stacks that I could find (optimized by the pros) had a very low cooling capacity and mounting 4 of these in parallel would be pretty expensive. Now, that I have a good method, high power multiple stage stacks are available and pretty inexpensive.

Thanks for the extra information. I never would have thought that running the middle TEC at such a lower current would yield superior performance.

How large are the TECs you're using?

Adam

Me either! That's why I designed it the way I did. And, I didn't just screw up the sizes. I rechecked this several times and after isolating each chip I was able to confirm that each chip performed very close to design specs. Also,these weren't just obtained from some sketchy E Bay source. Their specs were accurate. In addition, I could gain a few more degrees if I chose to run the top and bottom independently as well. These optimized at slightly different currents. This property is to be expected as the likelihood of two discrete devices optimizing @ EXACTLY the same current was low. But, the gain was not worth the complexity of providing a third current and so I compromised with series operation here

The top chip was 30mm x 30mm. The middle chip was 40mm x 40mm. And, the bottom chip was 62mm x 62mm. Best cooling that occurred with three operating in series was with nearly 120W of input power while better operation occurred when the middle was driven separately and a total of 90 W was used for the whole stack.

How hard are you comfortable driving your reds at these temps?

Last year,when I tested the P73 at -78C (dry ice temp) and posted the video, I ran it to 2A and made 2W. The music from the boys upstairs is a little annoying, but the test was impressive.

http://www.youtube.com/watch?v=z8EuZwgdqQ4

I ran them at 1.5A down to -190C and a total of about 5 hours of operation without any issue.

http://www.youtube.com/watch?v=0w4czym1FFg

I now run these routinely in the projector @ 1.5A each.

I've learned to run them from a constant current source, not a voltage source. This helps to avoid getting into the region of the device where the Delta T gets inefficient. Its not how the book explanation says to drive them. The manufacturers all seem to quote the same old tired explanations of how to use the response curve data. I find if I starve them for voltage, rather then overdrive them, I can find the operating point easier. I just had to figure out where to drive a unknown large TEC in a DPSS head, cooling a pump diode, and I switched the bench supply to constant current and more easily found the "sweet" spot. The Delta T and stress across the TEC is actually less then it would be if I ran it at some arbitrary constant voltage, for a given diode temperature.

But its easier to tell a non-engineer just to buy a different sized part then it is to get them to design a proper controller. This is why TEC has such a bad reputation.

Your mileage may very, but I've found it to be easier. Granted most people buy OEM power supplies of constant voltage.

Steve

That's a good idea. When I am running a single device, the fact that these act as a pure resistance load means that I can get away with voltage control even if it is not as direct as current control. But, when running a cascade stack the chance that the total series resistance is equal to the optimal current of the combination is very low.