Source for tec calculations/datasheets.

[zipdogso]

Good morning. Despite my low post count on here I have some idea about computers, and have been looking into thermoelectrics for a while. Finally reached the point of running some numbers for a potential project, and discovered data sheets to be difficult to source, and the tecs themselves only seem to be available in the UK through ebay.

Yes charts can be a pain to get the best ones in my view are ferrotec - unfortunately they recently crippled their TEC Calc site - you can no longer set the hotside temp you want to work with it is set to 50ºC now and you can't include heatsinks anymore bit of a shame.
You don't actually need the exact chart for a TEC.
What you need to observe is the number of thermocouples and the current.
Find a chart with the correct values for those and you can use that and it will be as near as.
So long as those values are close the chart wont be much different.
If you cant quite get the right number of thermocouples so long as the difference is not much divide the number of thermocouples you have by the number of thermocouples on the chart and on the volt/amp chart pick your operating point (current) read off the voltage then multiply it by the figure you originally got.

TECs aren't made in the UK unfortunately.... customthermoelectric, melcor, marlow, ferrotec, and TE Technology are quite good USA suppliers. Kryotherm are russian but have a USA based internet sales page. The only Far East manufacturer willing to deal with small order is Hebei IT, I have ordered from them and they are quite good and surprising quick from China too. I did order some TEC's from some Hong Kong supplier on Ebay but one did not work on reciept and one only lasted an hour (I DID NOT ill-treat it !) one more I used for a while but it and the other two I have to admit I have side-lined.

The other problem is the TEC weren't originally intended for cooling PC CPU's virtually 90% or more of every TEC manufacturers standard catalogue is NOT suitable for computer use...just not big enough. It tends to be the Far East manufacturers that have watched the market grow and tried to fill it.

TEC's draw a certain current, which I believe to be temperature dependent. The power consumed is related to the heat transferred, it seems at low voltages the heat wattage transferred can be higher than the electrical power consumed, whereas at high voltage the opposite is true. So multiple tecs undervolted are the way to go.

No the max current rating is determined by the size of the thermocouples. Toperating point current is determined by the voltage you apply.
The power consumed is a simple I x V calc based on the input.
The wattage transferred is determined by a number of factors but mainly the input power at the operating point.
The coefficient of performance (COP) rises as the input power goes down till it reaches a point roughly 30% Imax when it comes down sharply. between 100% power and 60% power the COP varies between - 0.5 and 1. 1 being the point where the input power = the cooling power. It is not uncommon for the COP to rise to 3 or so at 30% power. so the cooling power is 3 times the input.

Your right about the undervolting, multiple large undervolted is the way to go but there are a few things to remember 1) Most people buy a TEC rated at the heat they want to cool and are forced to run it a high power and this is the generally accepted way of doing things. 2) While undervolting is obviously advantageous particularly with regard to efficiency you must remember the Dt problem. A heavily undervolted TEC will have a small Dt (temp difference between hot & cold sides.) will not cool very well if you have a high hotside temp.

Geometrical constraints suggest at most half a dozen 50x50x3.6mm tecs, ebay offers the 12726. Does anyone happen to have a link kicking around for the datasheet on this? I have found the undervolting chart in the sticky, but would like to know more.

Take the undervolting chart with a bit of salt it was calculated using Ohm's Law.
For basic info on TEC's I use this site from Teilung, a Far East manufacturer (but in my experience don't bother trying to buy any.) for charts I just hunt around the USA sites starting with Ferrotec looking for a chart similar to the thermocouple + current values I have.
I couldn't actually find a 12726 chart but customthermoelectric have a 12724 - http://www.customthermoelectric.com/...W_spec_sht.pdf the current is quite high so 2 amps won't make a massive difference. The Qmax of a 12726 is around 230-240w and the Qmax of 12722 is around 190w so using 12724 chart for a 12726 will get a good ballpark figure. just strip the the figures of the volt/current graph and replace with the 12726 figures, strip the current axis off the Qc/current graph and replace with 12726 current the read the Q for your operating point. Heavily undervolted will only be a few watts different and max will get a max difference of about 20watts....not really enough to argue about !

The intention is to run them at 6V, where the table suggests a current draw of 9.5A and a power of 35W. 9.5A drawn at 6V is 57W by my reasoning, so what does the 35W refer to? Hence the request for data sheets and a push in the direction for calculations. I'm aware that using an atx psu is frowned upon, however were I to use a reliable one with 57A spare on the 12V rail I fail to see the issue.

The reason for the discrepancy is probably becuse your using Ohm's Law, aren't you ? TEC's are semiconductors and as such do not follow Ohm's Law strictly....basically if the data sheet says 35w then that's it.
Another point - there is one thing you cannot calculate - heat transfer loses - not correctly tensioning the TEC, incorrect TIM, wonky mounting, copper plates not flat enough etc etc all these things can and will affect the amount of cooling you get some by a considerable amount.

The reason ATX are frowned on is because 1) they need modding slightly to work without a mobo, 2) the wires and molex sockets are only rated for 8 amps max - 7 amps for safety. to use with TEC you need to cut off the molex and solder wires together. If you need 12 amps you solder two 12v wires together and 2 black wires - if you need 24 amps you solder 4 12v wires and 4 black etc. 3) they are intended for internal mounting and look naff outside a case and 4) they are not a variable supply - it's 12v/7v/5v/3v or nothing.
That's not very useful if you are trying to tune the efficiency of a TEC and certainly is no use if you are serious about undervolting. My favourite supply goes 3.5v - 16v at 22 amps (25a peak but I think it shuts down before that.) and I find it perfect for undervolting and it even has a voltmeter and an ammeter - all for the same price as a meanwell which only has a +/- 10% variation and no meters.

To clarify. I am interested in calculating heat transfer across the peltier as a function of voltage, current and temperature either side. I am also interested in calculating what additional heat must be disposed of if it is not equal to IV. I'm failing to find this in the literature, I suspect because I'm looking in the wrong place.

The formulas for this are a headache they can be found along with a lot of very good solid info in the Melcor Thermoelectric handbook - http://www.melcor.com/handbook.html it is downloadable free. If you haven't read you should.

Charts are quick and easy once you understand how to read them they certainly give you a good ballpark figure.
Calaculating it is OK if you understand the formulas and can be bothered with them but you have to remember one thing...Nothing about a TEC is set in concrete...with the exception of physical size of course. You cannot set a TEC to run at a particular temp or Dt or anything except voltage/current everything else varies whilst the TEC works it can just be simple thing like an increase in ambient temp and it will change everything particularly on a variable heat source like a CPU. That is why stats generally just state maximums. So really calculating everything exactly is a bit pointless...just use a chart and get a ballpark then run the TEC and see...

As a rough figure just double the Q, the cooling wattage, at the operating point your using to get a good estimate of the heat coming off the hotside.

Intended use is as a heat exchanger. Cold loop with various components in, hot loop with radiator. The current objective is above dew point cold loop temperatures (I believe this can be achieved by varying the voltage across the tecs downwards from 6V and/or turning off pairs when idle) using a smaller radiator than is required to cool the components normally. To determine whether this is viable for six, four or even two tecs I need some means of calculating the maximum heat they can move. The hope is by running the hot loop at 60 centigrade or so, radiator efficiency will be improved and counteract the additional heat dump.

TEC's generally get quite cold even a dinky 40w will go sub zero never mind sub ambient with a very small load and the correct operating point. Achieving a small temperature difference to stay above Dew point, while possible is certainly more involved than just a simple TEC setup. Like I have already said calculate the heat moved if you want but in practice it will constantly vary so even an exact calculation is just a ballpark figure...might as well use a chart.

The bottom line is TEC's are not an exact science..you could spend a long time calculating things exactly for it to all unravel in practice. My advice is put your calculator and pencil to one side, buy some cheap TEC's - under 100w, the 80w ones are quite jolly pick up a few heatsinks (vapochill micro's can be had quite cheap now.) get a couple of cheap digital temp sensors and some copper plates and just run some TEC's and see. You will have a far better idea and with real obtainable results. Not a fast fix I admit but you will be in a far better position to do what you want.