I would suggest that you buy your TECs from Customthermoelectric.
Granted they are expensive especially if you add on the stuff I say below but they are the real McCoy. Ebay TECs are not so good personally I think some of them are "bin rejects" after falling outside the pemissible resistance that's why they are so cheap.http://www.shop.customthermoelectric.com/Standard-TECs_c2.htm;jsessionid=C2891D85794BE05B076E5803E5572B17.qscstrfrnt05
As you are using multiple TECs in one project I would advise you ask and pay for the TECs to be resistance checked.
When TECs are made the resistance varies considerably and when you are expecting them all to be working the same...they arent. Most manufacturers have a relatively wide band of OK otherwise they would have to discard too many...if you get them resistance checked they will all have the same resistance ( sometimes you can specify what you want.) and should theoretically all operate the same.
If you are using a common cold/hotplate I.e they are all bolted to the same piece of copper I would suggest you also ask and pay for them to be height checked.
when TECs are mounted to the same plates even a minute difference in height can affect them. The bigger TECs will more pressure on them which will increase the resistance and leave you with the same problem as above. Bear in mind manufacturers suggest that plates should be flat down to several thousands of an inch.
The fact that the pressure on TECs increases the resistance is why the manufacturers insist on 150-300 p.s.i rather than just screwing them down as hard as you can ...which is what some people do.
If you pay for height checked customthermoelectric will mill them all to exactly the same height.
As far as your project goes your on the right track and so is xtremetechuk. I would say however that undervolting is best done with the biggest TECs you can afford/obtain. This will give you the best efficiency you should also make sure you use TECs with the highest couple count you can get but you with have to bear in mind the input power requirements and choose carefully how you wire them with relation to series/parallel.
As regards to testing water temps etc few if any people have done it properly it requires more sensors than most people are prepared to use...you cant just measure the water temp...you need the ambient temp, if it's in a case the case temp, the temp of the hot and cold sides of the TECs (the hardest ones to achieve.) i.e. you need to measure absolutely every other variable that will affect the water temp...it's a lot more work than most can be bothered. Any less testing than this while making interested, reading is basically no use.Edited by zipdogso - 2/7/11 at 5:17am