TEC lifetime

[zappedagain]

Does anybody have any experience with the lifetime on a thermo-electric cooler (TEC)? I've designed circuits for small ones in the past (controlling the temperature on photodetectors) and now I'm tasked with designing a large one (30W, 30C delta) that we want to run for 100000 hours (about 10 years).

What are the failure mechanisms?

How much margin do I need? I doubt running at 100% of the rated capabiltity is a good idea.

I've checked the Melcor site (

http://www.melcor.com/mtbf.html)

and it indicates I'm in a workable ballpark (depending upon your definition of MTBF). Are there any other good site recommendations?

Thanks,

 

[IRstuff]

TECs are made from some rather fragile materials that are also somewhat susceptible to moisture effects. My recollection from one of the 2 or 3 manufacturers is that reliable operation in a non-benign environment requires sealing the edges of the TEC structure.

Vibration is supposedly also bad for it.

TTFN

 

[logbook]

That is a seriously large TEC to give 30W of cooling and a 30degC differential. I wouldn't like to bet on 10 years of life in terms of a warranty period or any such thing.

The real killer is that there is a shed load of junctions in series, so when one goes the whole lot has gone. The ones I have used, and use now, are rather brittle, so when you apply any sort of mechanical stress you are asking for failure. This is probably a bigger cause of life expectancy loss than any theoretical MTBF figure. By the time you find out that your mounting method was slightly wrong you may have accumulated an unnecessarily high failure rate.

 

[BobbyNewmark]

I have been told by someone who used them that thermal cycling is a popular way to kill them. I would do accelerated life testing by running them across the full warm-up/cool-down cycle over and over.