Thermo-Electric Generator - TEC1-12706

[rickman]

I am looking at powering a very low power device from a TEG and can't
find hard data on the I-V characteristics. I found one graph someone
posted of his own measurements where it looks like the internal
resistance is around 10 ohms. I tried measuring it myself and it
appears to be closer to 1 ohm. Does that sound right? The temperature
difference was in the ballpark of 15 °F.

I expect to use this at a lower differential, likely just 2 or 3
degrees. I will also characterize it at those temperatures.

--

Rick

[Bill Sloman]

The manufacturers sell Peltier junctions as heat transfer devices, and characterise them for that application. Exploiting them as Seebeck generators is perfectly possible, but not popular.

Internal resistance depends heavily on the size of the device and the way the junctions are assembled. Lots of junctions in series have a higher series resistance than the same junction is parallel.

http://www.marlow.com/thermoelectric-coolers.htm

The tendency is to characterise them at the highest temperature difference they can sustain (while transferring no heat) and at the higher heat flow they can offer (while sustaining no temperature difference). Turning that data into Seebeck parameters isn't easy, but if you think about the thermodynamics you can get some of the way.

Price typically depends on the size of device you buy. My guess would be that you will have to buy one and measure it before you can get anywhere.

Farnell/Newark stocks a fair number of Marlow parts, at intimidating prices -
$50 each and up.

--
Bill Sloman, Sydney

[rickman]

Yes, I have two and have done some characterization as I said in my OP.
My initial measurement was a bit on the crude side. I used clip
leads, but when dealing with mV levels and 1 ohm or less loads that
isn't the best way to go. I improved my connections and used some
resistor values closer to the expected peak and found an internal
resistance in the 5 to 6 ohm range which is not far off from the page I
found that showed 10 ohms. Even though we were both testing the same
capacity module, it may not be from the same maker and so may vary a bit
in the specs.

I'm just surprised the internal resistance is so low, but I guess I
shouldn't be, it is rated for 6 amps. At that current it would be
dissipating over 200 Watts just in the internal resistance. I should be
surprised the internal resistance isn't lower!

I'd like to connect it to one of the LT chips designed to turn this
output into useful power. But their eval module is $150. I may make my
own board and do some testing. My initial measurements indicate I will
need more than 2 of these TEG units to work with the LT devices at just
a couple of degrees delta.

--

Rick

[Bill Sloman]

It might make more sense to buy a single bigger unit. That usually means more junctions in series, thus more volts for a given temperature differential.

The bigger units are more expensive, but you get more junctions per dollar that way. I doubt if you will be drawing enough current for the series resistance to be of anything but academic interest.

For heat transfer, Peltier effect is measured in watts per amp, while the waste heat you generate in the junction is proportional to amps squared, so the series resistance defines your operating range.

--
Bill Sloman, Sydney

[Martin Brown]

On 04/11/2014 05:17, rickman wrote:

> I'd like to connect it to one of the LT chips designed to turn this
> output into useful power. But their eval module is $150. I may make my
> own board and do some testing. My initial measurements indicate I will
> need more than 2 of these TEG units to work with the LT devices at just
> a couple of degrees delta.

Before you go too far down that route check that you cannot more
sensibly power the device for years with a button cell.

I have done candle powers a white LED once using a high temperature TEC
in series with a low temperature TEC and then a DC to DC converter from
that combined output. It was a lot harder to make it work than I
expected and I had a huge temperature difference to play with. To make
it work really well the lower plate had to be at the melting point of
solder and the upper cold heatsink fresh out of the deep freeze.

You can get rather more junctions on a larger plate but check first that
the same volume of batteries won't be a *LOT* more cost effective!

--
Regards,
Martin Brown

[George Herold]

We did this a year or so ago... someone had a nice link to old Russian TEG's that had some efficiency numbers. As I recall they had a few percent (5%?) of the theoretical max. efficiency. (The max is Thot-Tcold/Tcold) where all T's are in degree's K. You can see right away that a few degrees of temperature difference is paltry. I think you will also struggle with a low voltage. Putting my hand across a ~128 element TEC I get ~100 mV of open circuit voltage. (My hands are always cold though.) Do you have a way to gain up such a low voltage?

George H.

[rickman]

I've already looked at devices to capture the energy from a TEG. Energy
harvesting is not a brand new area and there are a number of chips
available. LT has a couple that seem to work well and one even has a
pair of inputs which allow it to capture TEG power independent of
polarity. So you can get energy on both the ebb and the flood tide.

I think I am going to take the LT eval board design and turn it into a
unit that I can make up cheaply using oshpark.com. I'll need to ramp up
on KiCAD. I've been using FreePCB which works just fine, but I have
wanted to move toward running Linux for some time and working with dual
platform tools is the first step. So this can be my test/learning project.

The LT parts use a 100:1 transformer to step up the voltage and claim to
work with voltages as low as 30 mV. That may mean the loaded voltage
which is a lot lower than the open circuit voltage of course. Even so,
my measurements show there is "significant" power. Significant is
relative of course. I'm looking for a few hundred uW.

Along with that will be my first rPi effort. I need to measure the
relative temperatures accurately as my meter with 1 °F resolution just
doesn't cut it. So I'll be adding electronics to interface some
thermistors and display the temperatures.

--

Rick