Peltier Cooler Power Supply Design Help

[Dave Wave]

Hi All,

I am building a small temperature-controller box. In this design reliability and long life are important (should run for years)

I estimated the heat loss at about 35-40W worst case.

I plan to use two Peltier coolers to handle the load. One should do fine (rated at around 60W max). Dual units adds reliability.

So far I have the cooler/fan units:

http://www.ebay.com/itm/261792872834

A digital controller (controls a relay that turns the cooler on/off):

http://www.ebay.com/itm/251626927103

And a power supply (12v/15A):

http://www.ebay.com/itm/301424192425?var=600380501799


Now the issue:

The coolers are rated at 5.5 amps max (each). That is below the max of the power supply and the relay in the controller (15A). As reliability is important, the idea of the power supply running flat out when the unit is on does not sound good, nor is it good for the coolers. How can I limit the current to a level so the power supply and the coolers can loaf along at lower power level? Perhaps a constant current supply?  AFAIK voltage is not so important (13V max).

Thanks!

-Dave

[Siwastaja]

You can adjust Peltier power by adjusting voltage. Many of those chinese cheap power supplies actually have a trimpot for voltage adjustment. If that one doesn't, get one that does. It's a very easy way out.

But really, you should control the peltier power by continuously adjusting voltage (or current, doesn't matter, it's the same thing) instead of on/off relay control. While it can still work just fine the way you are doing it, but it's far from optimum. Try it out first.

[ajb]

What are your hot side and cold side temperatures?  TEC performance depends a lot on those factors.  Note that these refer to the hot and cold sides of the TEC itself, so you have to take into account the thermal gradient across your heatsink(s).  Depending on your Tdelta, 2x60W TECs may be just right or nowhere near enough.

I don't see any real specs in the eBay listing, but searching for "TEC1-12706" brings up this datasheet.  Based on your Tdelta you can find a point on one of the lower graphs that will give you the ballpark current you should be operating at, and with that you can determine the voltage that should give you that current.    Note that the power you apply to the TEC appears at the hot side as extra heat that must be removed by the heat sink, so if you want to calculate the actual operating point of your TEC at a given load you'll need to do some back and forth with the datasheet graphs to get the numbers to converge.  The trick is to drive the TEC just hard enough to move the amount of heat you need to move, as the harder you drive it the more heat you're generating in the TEC itself.  The more heat you generate, the hotter your hot side gets.  The hotter your hot side gets, the higher your Tdelta goes.  The higher your Tdelta goes, the lower you coefficient of performance gets. 

[MK]

As said above, you need a clean power supply to provide just enogh cooling to meet temperature requirements, it took a 12V 4A peltier to keep a 4 watt heat leak, and 7 degree module in a 40 centigrade box. a lot gets wasted.

[JohnnyBerg]

I am building a small temperature controlled box also.

Perhap we can share some thoughs and other stuff to get it from the ground?

I have the PSU, and use my own industrial Arduino contoller with PT100 input and in a DIN rail box.

[Dave Wave]

My Delta T is about 13 degrees or so.

To make it a bit more complex... In my application, I am trying to maintain a temperature of about 15c. The environment swings between 8 and 21...so there is not a huge difference. I want to be able to reverse the polarity and heat when it is cold...this controller supports that.


Looking at the chart, that shows 3A at about 7V for 20 watts or so. So at least to my newbie brain, it looks like I should be in the ballpark.

So back to the problem, how do I setup a power supply that will power this without the power supply or TEC's running flat out (running into reliability issues).

The idea here is have something like a 15A power supply only putting out 5-10A max

Thanks for all the help!

-Dave

[nfmax]

There are commercially available TEC controller instruments which are widely used in the optics field, for example https://www.newport.com/Laser-Diode-Temperature-Controller,-350B/358814/1033/info.aspx#tab_Specifications is a low-power model which we use here. Compatible with most types of temperature sensor & TEC. Or you can roll your own using a custom control IC such as the Linear Technology LTC1923, driving a class D H-bridge output stage (which we also use)

[Dave Wave]

There are commercially available TEC controller instruments which are widely used in the optics field, for example https://www.newport.com/Laser-Diode-Temperature-Controller,-350B/358814/1033/info.aspx#tab_Specifications is a low-power model which we use here. Compatible with most types of temperature sensor & TEC. Or you can roll your own using a custom control IC such as the Linear Technology LTC1923, driving a class D H-bridge output stage (which we also use)

Not within my price range....

[nfmax]

I did wonder about your budget...
I would recommend taking a look at the LTC1923 chip, and hunting for application notes from LT and the Peltier manufacturers. The most important things are to drive them with a DC voltage, not switched or pulsed, for two reasons:

• The Peltier cooling is proportional to the average current, but (unwanted) resistive heating is proportional to the mean square current, so to get the best ratio you want a steady current
• Peltier coolers don't like thermal cycling, as it stresses them mechanically and promotes early failure

You must avoid instability in the temperature control loop, as it can damage the Peltiers by excess cycling or simple overload.
You will also want an adjustable current and voltage limiter in your circuit to prevent overloads.
A tip on designing thermal control loops: to get the best dynamic response and tightest temperature control, the temperature sensor needs to be right next to the cooler so it sees temperature changes with the minimum delay. This may not be where you want to control the average temperature. In which case a nested system of a slow, outer control loop sensing near the load and adjusting the set point of a fast, inner control loop driving the Peltier and sensing close to it is the way to go.
The thermal design is also very important. You want a great big heatsink with a fan on the 'ambient' side of your Peltier to keep it close to room temperature, and good thermal conductivity and minimum thermal 'mass' (total heat capacity) on the controlled side. Judicious use of thermal insulation is also beneficial.
In general, it's harder than it looks.

[ajb]

My Delta T is about 13 degrees or so.

To make it a bit more complex... In my application, I am trying to maintain a temperature of about 15c. The environment swings between 8 and 21...so there is not a huge difference. I want to be able to reverse the polarity and heat when it is cold...this controller supports that.


Looking at the chart, that shows 3A at about 7V for 20 watts or so. So at least to my newbie brain, it looks like I should be in the ballpark.

That seems optimistic.  Remember to account for the thermal drop across the heatsinks.  Assuming you use both TECs, if your worst case is T_inside=15C, T_outside=21C, and Q=20W/TEC, and your heatsinks have a thermal resistance of, say, 0.4C/W, then for each TEC you wind up with:

T_cold: 0.4C/W * -20W +15C = 7C
T_hot: 0.4C/W * 20W + 21C = 29C
T_delta: 29C - 7C = 22C

and that's before accounting for heat dissipated in the TEC.  So take those numbers into the datasheet graphs and from the bottom graph, at T_delta=22C and Q=20W, you're at. . .we'll call it 3.2A.  From the top graph, at 3.2A and T_delta=22C, you're at about 9V, so that's 9V*3.2A=28.8W that gets added to the hot side of the TEC.  So if we add that in to the equation from above:

T_cold (same as above): 7C
T_hot: 0.4C/W * (20W + 28.8W) + 21C = 40.52C
T_delta: 40.52C-7C = 33.52C

Now take the new T_delta number and go back to the datasheet, and you'll find that you have actually have to put ~3.8A into the TEC at ~10.8V, so now you have 41W dissipated at the hot side by the TEC.  Keep revising the T_hot equation and adjusting from the datasheet, and eventually you'll find that you either exceed the TEC's ratings, or the numbers will converge and you'll have found the steady-state operating point for your TEC at that heat load and those temperatures.

On the bright side, when you're heating the box, all of that extra heat dissipated by the TEC is moving in the right direction, so you'll need less power given the same heat load and temp differential to maintain your target temperature.

[Dave Wave]

Wow you guys are great!

Thanks for all the information....I really appreciate the help from people that know some of the theory and math behind making this stuff work.

I am going to take a couple of days to chew on what I have and I will check back in.

-Dave

[MK]

Dont forget the heat leak down the claming screws, they need quite a bit of force as they have a large area that you need to take the heat from.

heatsink, peltier, cold-plug, then internal enlosure as you want the peltier right at the heatsink and you need more than 3-4mm of insulation under the heatsink.

[splin]

Dont forget the heat leak down the claming screws, they need quite a bit of force as they have a large area that you need to take the heat from.

heatsink, peltier, cold-plug, then internal enlosure as you want the peltier right at the heatsink and you need more than 3-4mm of insulation under the heatsink.

MK beat me to it - those two screws that hold the heatsinks together and clamp the peltier element will leak a surprising amount of heat. From pictures of other similar/identical units on Ebay, the screws appear to be 4mm steel. Steel has a thermal conductivity of 43Wm/K. The peltier is 3.5mm thick. So assuming that the 4 (in total) screws make good thermal contact with the hot heatsink through the sides of the holes and the total delta - T is 33.52C (from ajb's post).

Thus heat lost through the 4 screws = 43 * .004^2*pi/4 / .0035 * 33.52 * 4 =20.7W!

In reality the screws will probably transfer most of the heat to the hot heatsink through their heads (as the hole will be a clearance hole with only the outer points of some of the threads being in contact with the hole sides) and thus have a rather longer thermal path than 3.5mm. However it is going to be significant. Using stainless steel screws, with a thermal conductivity of 16Wm/K will help a lot. Increasing the hole diameter in the hot heatsink and inserting a plastic bush and using an insulating washer under the screw heads will help even more.

And then there's the leakage through the insulation from the flanges on the small heatsink which I estimate at approx 2400mm^2 for the two units. Assuming the insulation has a conductivity of .035 W/m/K that equates to:

 .035 x 2400/10^6 /.0035 * 33.52 = .8W.

That is only the flow that is directly perpendicular - more will flow through progressively longer, curved paths to the edges of the small heatsink (like magnetic field lines from a magnet pole) so perhaps 1W in total. (Similarly, some additional heat will flow through progressively longer paths through the insulation from the screws to the area of the small heatsink surrounding the screw hole). Putting a 40mm square piece of aluminium between the peltier and a heatsink will allow thicker insulation and longer screws at the cost of slightly lower thermal conductivity between the peltier and heatsink due to the extra interfaces - so best placed on the cold side where the flow is rather lower.

As for ajb's analysis, the numbers seem reasonable except that the inner heatsink being rather smaller will have a larger thermal resistance than the much larger outer one. So while .4 C/W seems reasonable for the larger heatsink (though I'd guess .6 C/W or more), I'd be surprised if the smaller is much less than 2 C/W. This one for example:

www.qats.com/ShowPDF/HSDS_ATS-53400R-C2-R0.ashx is a similar size and specc'd at 1.7 C/W with a ducted airflow of 200 linear feet per minute, but it has 16 fins compared to 9 on the Ebay unit.

So I think you are going to have trouble meeting your requirements with these coolers. It would be instructive if you could let us know how you get on.

Splin

[nfmax]

We always use nylon screws to minimise thermal leakage - but all our TEC's are quite small. Be carefull not to overtighten them, though!

[Dave Wave]

Thanks for all the tips, I will look into the nylon screws.

I am interested in trying to make this work with a PID type controller outputting an analog signal to run through H bridge/amp setup up. I would like to do something ardunio based. However my coding skills are not great, so I have been poking around a bit to find an open source project that could be adapted.

Does anybody have any leads for an aurduino based PID controller with analog output (found quite a few using relays).

Thanks again,


-Dave

[Dave Wave]

So I found a PID type controller that seems to output 0-10V:

http://www.ebay.com/itm/MYPIN-Dual-Digital-F-C-PID-Temperature-Controller-TA4-VSR-analogue-output-0-10V-/291241559520?pt=LH_DefaultDomain_0&hash=item43cf5985e0

Anybody know if this means it can drive a peltier both heating and cooling. It looks like it might...not sure what the double arrow symbol means on the attached diagram (right side...outputs note G)?

I was thinking about using an H bridge motor controller to drive the Peltier.

Thoughts?

Thanks,

Dave

[ajb]

The double arrow thing is the symbol for a triac.  They're trying to show the different types of outputs, but not doing a very good job of it.  What you're looking for is something that can output an analog control voltage to control the power to the TEC, and a relay to control the polarity so you can heat and cool as needed. . . not clear if that eBay PID controller will do that, probably the only surefire way to find out is to buy it and hope it comes with a decent set of instructions.  An H-bridge on its own isn't going to do a very good job of controlling a TEC.  What you ideally want is essentially a class D amp, like nfmax mentioned.  This is basically an H-bridge with some filtering to convert a PWM output into a relatively smooth analog voltage.  The sort of filters you'd need for the size of TEC you're looking at would be pretty big.  You might be able to find an off-the-shelf audio amp that could be easily modified for DC operation. 

This arduino library supports an analog output (PWM in the example provided--you could either filter this to produce a DC control voltage, or adapt it to use a DAC):
http://playground.arduino.cc/Code/PIDLibrary