DIY Temperature Chamber

[jmaja]

12/24 V Danfoss BD35F has a built in RPM control, but only from 2000 to 3500 rpm. The cooling power drops only about 30-40%, thus it doesn't help much for controlling the temperature. They are also rather expensive compared to single phase AC freezers you can by very cheaply. http://www.novakool.com/support/bd35f_compressor_data.pdf

[3141592]

I'm building one now too, managed to get -8C at zero load (no heatsource), with a pretty cheap heatsink, although it took hours to get there. The limiting factor, as noted before, seems to be the cooling of the hot side. An interesting thing that I've found is that the temperature-power curve seems to have a peak. Depending on the performance of your hot side cooling solution, there'll be a sweet spot of supplied power. Run it with a bit less or more and you'll see a much worse performance from the unit.

I'm gonna try some bigger fans, I want a repeatable -10C performance for some passive components.

[mzzj]

Peltier is very inefficient at low temperatures. Thus if you need to get below zero C compressor is the only option.

Single-stage peltier cooling is viable alternative if you need best stability and wont need high cooling rate or big volumes.

IE.  1 liter box to -20C  is doable with peltier(s) 
even -40C is doable but then you need multi-stage peltiers and your efficiency REALLY sucks.
For small volumes this might be still viable alternative to compressor cooling, ie: cheaper to build and much easier to control.

[TerraHertz]

Rather than building the 'cold generating' part yourself, why not just obtain a cheap 2nd hand freezer chest? Possibly even free.
Then fill it with thermal mass (bricks? sand? glycol?) and run some heat exchanger pipe through that mass.
Have your environment chamber adjacent, with a small easily speed-controlled pump circulating the transfer fluid or gas between the freezer chest thermal mass, and piping on the walls of your test chamber. Add electrical heating for fine control, and positive temperature range.

Typically domestic freezer chests go down to -40 C, and lab ones to -86 C. Or you could substitute a liquid nitrogen dewar if you needed to go lower. Though then you'd have to be careful what you used for a heat transfer medium.

Anyway, this way you can design the test chamber with far fewer design restrictions. Since you have a huge heat sink in the freezer chest, with potentially very high heat transfer rates. And you can just set the freezer chest thermostat to the bottom of its range and forget about it. Let the freezer chest refrigeration pump run at the rate it was designed to.

You'd only need a temperature sensor in the test chamber, and one on the incoming coolant pipe, for your PID controller to achieve good regulation.

[mikerj]

although it has been years since I dealt with any Peltier's, everything I remember reading suggest PWM power to supply the current. although I may be wrong about the full rated voltage. in fact I briefly looked up a company that makes Peltier's and they indeed recommended PWM power. And although I did not read the entire article. Way too long for this late in the day. It did discuss PWM power. I think I'd rather go by what the experts say that makes the Peltier's.

Unfiltered PWM reduces the efficiency of a Peltier heat pump by a significant margin unless you are running at either 0% or 100% duty cycle..  This should be obvious if you think about it; the heat pumping ability will be proportional to the average of the PWM waveform, but the heat generated within the TEC itself is proportional to the square of the current (or voltage), and with PWM you are applying maximum current/voltage for part of the waveform.

The net result is that using PWM increases the amount of heat generated within the TEC for a given average pump current/voltage, so the TEC needs to be driven harder for a given heat load.

[kony]

Carefull with that heatpipes fitted heatsink! As the heat transfer in those things operates based on phase change heat, hotside temperature will be set based upon material used to fill the heatpipes, usually being too high for any reasonably low teperature to be achieved on cold side of the TEC. Nothing beats water, when maintaing low temperatures at moderate to high powers is required.

[3141592]

I'm building one now too, managed to get -8C at zero load (no heatsource), with a pretty cheap heatsink, although it took hours to get there. The limiting factor, as noted before, seems to be the cooling of the hot side. An interesting thing that I've found is that the temperature-power curve seems to have a peak. Depending on the performance of your hot side cooling solution, there'll be a sweet spot of supplied power. Run it with a bit less or more and you'll see a much worse performance from the unit.

You're absolutely right, the heatsink characteristics have an effect on the current/cooling power optimum. Achieving low temperatures on the cold side requiere to take a lot of care on the insulation on the cold side and on the heatsink performance (Rth to ambient) on the hot side.

Have a look at these papers for more matter :
http://electronicdesign.com/components/simple-design-equations-thermoelectric-coolers
https://www.tetech.com/wp-content/uploads/2013/10/ICT94MJN.pdf

The best price/performance ratio I found for low Rth heatsinks is CPU coolers. You can find nice used assemblies on eBay, with a copper heatpipe, lots of thin fins plus a fan for <30 Euros. On my personal temperature chamber I use an 'Arctic Cooling Freezer Pro'. Rth < 0.5°C/W is possible.

Example : http://www.ebay.co.uk/itm/Freezer-Pro7-AC-Arctic-Cooling-/361446480482

Yep, CPU coolers are awesome price/performance compared to heatsink prices from part vendors. I'm actually using an Arctic Alpine 11 plus. I should've opted for something bigger, but still, I was quite amazed at the effectiveness of this sub 50$ (total) setup I'm using now. I'm going to replace the fan on the top with two on the sides in a push-pull configuration and see how that goes. Could you post some photos or data of yours? How low can you get with it, assuming a 23°C room temperature? Thanks for the links, I'll read them through!

[nctnico]

Rather than building the 'cold generating' part yourself, why not just obtain a cheap 2nd hand freezer chest? Possibly even free.
Then fill it with thermal mass (bricks? sand? glycol?) and run some heat exchanger pipe through that mass.

That is not a good idea. You will always have a temperature difference between the two (2nd freezer and the actual chamber) so you will not be able to reach very low temperatures. IMHO a freezer is still de best way to go because peltiers are just to inefficient. If temperature stability is an isue you can always add some thermal mass in the form of water. Bricks or glycol are not a good choice because of their poor heat capacity (compared to water).

[TerraHertz]

That is not a good idea. You will always have a temperature difference between the two (2nd freezer and the actual chamber) so you will not be able to reach very low temperatures.

Of course, there has to be a difference, to be able to maintain temp regulation in the chamber by varying the flow rate of the transfer fluid. But with fluid circulating at full flow in insulated pipes, there's not going to be much difference.
That required difference is why I mentioned lab freezers, supposed to go down to -86 deg. C. This method is still going to beat the pants off any peltier system, in efficiency, speed of temperature slew, thermal capacity, and lowest reachable temperature. Plus, it's simple to construct, and the whole freezing system is off the shelf and easily replaceable with any roughly equivalent scrounged freezer.

IMHO a freezer is still de best way to go because peltiers are just to inefficient.

That's why I suggested it. Except with a method to avoid having to mess with the freezer control loop and motor, and just let it run at it's own pace maintaining the lowest temp it can reach.

If temperature stability is an isue you can always add some thermal mass in the form of water.

Thermal mass in the actual chamber will slow down achievable slew rate.

 

Bricks or glycol are not a good choice because of their poor heat capacity (compared to water).

For the thermal mass in the freezer, used as a cold sink for heat transfer fluid circulated to the chamber, it's also important to maintain high thermal conductivity. Water will freeze, limiting heat conduction from the piping. Depending on lowest temp needed, different things could be tried.
And heat capacity probably wouldn't be so important, due to chest freezer high volume relative to the mass of the test chamber, and that there is the freezer running to keep it cool. It's just acting as a low pass filter and transfer medium between the chamber coolant piping, and the freezer coolant piping.

[splin]

By the way, I measured the heatsinks' Rth from this cheap thermoelectric cooling kit : http://www.ebay.com/itm/261792872834. Rth = 0.41°C/W on the hot side, and Rth = 1.8°C/W on the cold side. Both with no dependance in temperature.

The listing is no longer available unfortunately, but assuming you mean a peltier sandwitched between two heatsinks, how did you use it to measure yours? The usual way to test a heatsink is by applying power and measuring the temperature rise.

How can you be sure of those heatsink Rth values? Surely they depend on orientation, airflow and the vendor not making up the numbers?

[CatalinaWOW]

Peltier coolers have no moving parts and are electronic which appeals to people on this forum.  Other than that they have little going for them.  There is a reason that several people have recommended not using them in this application, and why I am not aware of any commercial temp chambers based on this technology.

Refrigerators and freezers are good approaches.  One other thing which may appeal to you is occasionally used in the commercial world.  Liquid nitrogen is cheap and readily available in many locations.  Venting the boil off from a dewar through a solenoid can be used to achieve a controlled temperature.  It also has the advantage that the boiled off nitrogen is quite dry and helps reduce the problems of condensation and frosting that occur when you go below the dew point.  Be sure you know what you are doing if you go this way.  There are many hazards in using cryogenic fluids, including the fact that the boil off is not life supporting.  People have died from lack of ventilation.

[splin]

The listing is no longer available unfortunately, but assuming you mean a peltier sandwitched between two heatsinks, how did you use it to measure yours? The usual way to test a heatsink is by applying power and measuring the temperature rise.

How can you be sure of those heatsink Rth values? Surely they depend on orientation, airflow and the vendor not making up the numbers?

Sorry for the link, it's corrected now.

Source of the Rth values = My measurements, the same method as you described. And as both heatsink works in forced convection mode/with a fan, the Rth value should not vary much with orientation.

Ah apologies I thought you meant you used the peltier module to measure your heatsink - doh! Interesting numbers so thanks.

I used a free program 'TEC Calculator': selecting a 40mm, Qmax = 65W, 16.7V max peltier running at 50% Vmax (8.5V) with a hot side of 32.3C gives 30W total heat load, 5.5W heat moved from cold side, delta T = 50.3C and a COP of .22.

With Rths of .41 and 1.8C/W that means 20C ambient, hot side at 32.3C (20C + 30 * .41), cold side at -8.1C (32.3 - 50.3 + 5.5 * 1.. 5.5W isn't much and in practice the overall performance will be much worse due to heat flowing through the clamping screws and the relatively thin insulation between hot and cold heatsinks. Using 2 x 3mm stainless screws for the clamping screws, with a delta T of 50C I calculate that you'd lose 2.8W of the 5.5Wthrough the screws alone (and three times as much for carbon steel)!

Increasing the power to the peltier doesn't help much as the efficiency drops rapidly and the resulting higher temperature of the hot side heatsink offsets most of the gain in delta T. Eg @ V = 60% Vmax, heat moved kept at 5.5W, delta T increases to 59C but total heat load increases to 42.7W. So hot side = 37.5C, cold side -11.6C. And the parasitic heat flow from hot to cold increases as well so you might not be any better off at all.

5.5W of cooling with -8C internally, 20C ambient equates to a 100 x 100 x 100mm (internal) box with 75mm of polystyrene insulation!

[splin]

I used a free program 'TEC Calculator'

I didn't know that one. Nice find, thanks!

About thermal leaks through the box and through the clamping screws, my results are different than yours :
- If we assume for polystyrene lambda = 0.04W/(mK) and a 100mm side, 75mm thickness cubic box. Rth [W/°C] = 0.04 * 6*0.1² / 0.075 = 0.032 W/°C. For DeltaT = 30°C the heat flow is 1.0W.

Your right - I changed my mind on the dimensions and forgot to recalculate.

- For stainless steel (304/316 variety) lambda = 16W/(mK). Two 3mm diameter, 5mm plain screws. Rth [W/°C] = 2 screws * 16 * pi*0.0015²/0.005 = 0.045°C/W. DeltaT = 30°C cause a 1.4W heat flow.

But delta T is 50C as the screws are connecting the heatsinks which are hotter/colder than ambient. I also assumed the screws were 4mm long to match the peltier thickness of 3.5 to 4mm.

I estimated the heat flow through the insulation around the peltier and directly between the heatsinks in reply #12 here:

https://www.eevblog.com/forum/projects/peltier-cooler-power-supply-design-help/?nowap

.8W was for a 33.52C delta T so  1.2W @ 50C delta T

Interesting how close AJB's estimate of .4C/W was for the large heatsink (unlike my .6C/W guesstimate) and my "I'd be surprised if the smaller is much less than 2 C/W" for the smaller one were to your measurements.

Essentially the actual equilbrium temperature would struggle to reach the -8.1 and probably 0C would be more realistic. Pu foam or even aerogel would help, as would plastic screws.

BTW on the cheap thermoelectric cooling kit from eBay, there's 2 stainless steel clamping screws (d=4mm, l=10mm) with thick plastic washers. Not bad for the price! (I paid mine $12 including S/H).

Amazing price for what you get. It's on my shopping list.

[Zeranin]

my limited experience with Peltier's tells me that they work best using a PWM modulated voltage. at full rated voltage. When you vary the voltage you Reduce their efficiency.

That's exactly the opposite of how TEC efficiency works.  As you increase the current, Ohmic losses increase with the square of the current, so while more current means more heat moved, more and more of the heat you're moving to the hot side is heat generated in the TEC itself!  If you're controlling the cold side temperature, then more heat on the hot side means a higher delta T, which means the TEC's coefficient of performance decreases, meaning more waste heat is generated for every watt of heat removed from the cold side, so you have a negative feedback that means that the amount of heat you move increases sub-linearly with current. 

So for maximum efficiency you would want to run the TEC at the steady-state current that moves the amount of heat you need.  If the TEC is lightly loaded relative to its capacity, you could get away with cycling it between off and its maximum COP point and get reasonable efficiency, but at the cost of inflicting thermal cycling stresses on the TEC.

It's more complicated than that. I can see what you are getting at. Just say you want to charge a battery, in which case only the average current matters as far as getting a certain number of AH into the battery, so in that case it is more efficient and produces let I^2R heating if you use a steady current rather than PWM modulated.

However, Peltiers are not analogous. If you run a Peltier at a steady voltage and current below rated voltage, then the thermal performance is quite different than being PWMed at the same average current. Specifically, at low voltage and current, the efficiency is extremely poor at higher temperature differentials, and compromised as to the magnitude of differential that can be reached at all. However, when PWMed, the Peltier always operates at 'full voltage', and can therefore still shift heat at a large differential, though of course the amount of heat shifted is scaled down by the duty cycle, and this is usually what you would want - the ability to shift less heat, but still at a large differential. Thus, in many instances, the original claim that Peltiers work best using a PWM modulated voltage, at full rated voltage is correct.

[Kleinstein]

For the Petier effect it's very much like the battery charging analog, maybe even worse. Usung PWM is way less efficient than a DC current of the right size. The petier effect is very linear and the loss is mainly I²*R type, with much of the cooling is needed to compensate the loss.

To a very good approximation the cooling power follows a parabola, with the highest cooling power at the so called optimum current, which is usually very close to the rated current. At zero current there is obviously no cooling - just the background heat conduction. As an example at half the optimum current you get 3/4 of the maximum cooling effect and only need 1/4 of the electrical energy. If you use PWM at 1/4 the time with full power you have 1/4 of the maximum cooling, or 1/3 the efficiency you get with a DC current.

So PWM for the Peltier is a very poor idea.

As a second effect, a low PWM frequency causes termal cycling and can this way cause faster aging. So a simple on/off regulator is also not that good.

[Zeranin]

For the Petier effect it's very much like the battery charging analog, maybe even worse. Usung PWM is way less efficient than a DC current of the right size. The petier effect is very linear and the loss is mainly I²*R type, with much of the cooling is needed to compensate the loss.

To a very good approximation the cooling power follows a parabola, with the highest cooling power at the so called optimum current, which is usually very close to the rated current. At zero current there is obviously no cooling - just the background heat conduction. As an example at half the optimum current you get 3/4 of the maximum cooling effect and only need 1/4 of the electrical energy. If you use PWM at 1/4 the time with full power you have 1/4 of the maximum cooling, or 1/3 the efficiency you get with a DC current.

So PWM for the Peltier is a very poor idea.

As a second effect, a low PWM frequency causes termal cycling and can this way cause faster aging. So a simple on/off regulator is also not that good.

I am not saying that PWM is always the most efficient way to operate, but sometimes it is, while at other times it is more efficient to operate at a lower, smooth DC voltage. To this extent we are 'both right'.

That said, it is not as simple as you make out, and not entirely analogous to battery charging.

The problem with your reasoning, is that you are considering only 'cooling power', without also paying attention to cooling differential. Here is an example to illustrate the point. Just say we have an application that calls for a cooling power of 20W, at a differential of 40K, and it just happens that a particular model of Peltier provides exactly this requirement at full rated voltage and current, say 12V at 5A.

But think of all the I^2 R loss running the peltier at full current! By your reasoning, if cost was no object, it would probably more efficient to buy 4 such peltiers, and operate them in series/parallel, so that each is just idling along at half rated voltage and half rated current. I used to think that too, and by the battery charging analogy, it would be a good solution. In reality, this solution would either not work at all, or would be less efficient that a single peltier run 'hard', because although peltiers run at partial current are very efficient at shifting large quantities of heat at a low differential, they are inefficient or incapable of shifting heat at a high differential. There is literally an extra 'dimension' to the problem that you have not been considering.

When determining the optimum operating current for a peltier to match a given thermal load, we need to consider not just 'cooling power' as you are doing, but the required cooling power AND the differential. In most applications, the required differential is more-or-less fixed, being for example the difference between the desired temperature inside a fridge, and the outside temperature. After the drinks in the fridge have been cooled, the thermal requirement will likely be for a small cooling power, but still at a large differential. What, then, is the most efficient way to operate the peltier to produce a large differential at a small cooling power? Unlike with battery charging, where a low current always produces useful charging, what we find here that at a low current, we don't get any cooling at all at the required differential, in fact even worse, at a low current we are still putting electrical power into the peltier, but this is heating up the fridge, not cooling it! As you increase the peltier voltage in this example, you reach a point where the efficiency is zero, that is, you are pumping in considerable electrical power, but producing zero cooling, though to add insult to injury that completely wasted input power is heating up the hot side, increasing the differential still further. If you then increase the operating current slightly more, then you match the thermal requirement, producing just a small amount of cooling power at a largish differential. However, the efficiency is crap, because you have moved only slightly from the zero-efficiency point just discussed.

Now consider what happens in the same example, if full voltage (and therefore current) is applied to the peltier. It will operate at high efficiency, because when the required differential is high, a high voltage and current produce the best efficiency. That's great, except that at full input voltage, the amount of cooling being produced is excessive to match the requirement. But by operating full voltage with PWM, you get to have your cake and eat it too, operating at high efficiency, while still producing the required small amount of cooling.

The more general conclusion is that the most efficient way to operate a peltier depends on the required combination of cooling power and differential, and in many instances such as the example above it is found that PWM is more efficient.

Agreed that thermal cycling is a serious issue if the PWM frequency is too low.



   

[Zeranin]

Further to my previous posting, here are some real-world examples to illustrate what I was saying about Peltiers, and how PWM tends to be more efficient when the requirement is for a small cooling power at a largish temperature differential.

Attached is the performance curves for a typical peltier device, a TEC-12705 from Everredtronics, designed to operate at about 12V and 4A.

To demonstrate a situation where PWM provides the most efficient operation, assume that the thermal requirement is a cooling power of 2W, at deltaT=40K, for Thot=27 DegC, so therefore Tcold=-13 DegC.  The required plot for Th=27 is on the left of the page, and deltaT=40 can be found on the lower X-axis. From this plot, it is seen that an operating current of 2A exactly provides the desired 2W of cooling, at dT=40, so this is the point at which a non-PWM driver would operate. This operating point is :-

deltaT = 40 DegC
Cooling = 2W
I = 2A
V = 5.5V
Pin = 11W
Efficiency = 2/11 = 0.18

OK, that’s pretty crap efficiency all right, but peltiers are always inefficient when producing large deltaT. Now let’s look at a different operating point, that would be applicable to PWM operation.

deltaT=40 DegC
Cooling = 12W
I = 4.0A
V = 10V
Pin = 10x4 = 40W
Efficiency = 12/40 = 0.30

An efficiency of 0.3 is still not great, but it’s a damn site better than 0.18, so this is a much preferable operating point, except that the 12W of cooling is excessive, as we want only 2W. No problems, use a 10V supply and set the PWM  duty cycle to 1/6 and we get to have our cake and eat it too, with 2W of cooling power produced at 0.3 efficiency rather than the 0.18 efficiency that we would get with a non-pwm driver set to a steady 5.5V as in the first example.

I’ll follow up with a posting illustrating the converse case of large cooling power at low differential, where PWM is less efficient, and efficiencies of 300% or more can be achieved with a deltaT of 10 DegC. 

[Zeranin]

Here is the follow up posting illustrating the case of larger cooling power at low temperature differential of 10 DegC, where PW<, if used  at all, produces pooer thermal efficiency. 

Refer again to the performance curves for a typical peltier device, a TEC-12705 from Everredtronics, designed to operate at about 12V and 4A, attached to the previous posting.

In this case, assume that the thermal requirement is a cooling power of 6.7W, at deltaT=10K, for Thot=27 DegC, so therefore Tcold=17 DegC. 

The required plot for Thot=27 is on the left of the page, and deltaT=10 can be found on the lower X-axis. From this plot, it is seen that an operating current of 1.0A exactly provides the desired 6.7W of cooling, at dT=10, so this is the point at which a non-PWM driver would operate. This operating point is :-

deltaT = 10 DegC
Cooling = 6.7W
I = 1A
V = 2.4V
Pin = 2.4W
Efficiency = 6.7/2.4 = 2.8 (280%)

Whoa, now we're talking seriously good efficiency of 280%, so for every 1W of input power, we create 2.8W of cooling! Part of the reason, of course, is because effciency of any refrigeration system is improves for a small temperature differential, deltaT. Even so, this is better than most people associate with peltier devices, almost as good as some compressor systems, and the reason it's so good is because the Peltier is just idling at only 1A, which is the optimum way to operate a peltier when the required deltaT is small. Operating at only 2.4V and 1A does mean that the cooling power is modest at 6.7W, but that can always be scaled up by using more or bigger peltiers. For really good efficiency at low deltaT, this is clearly the way to operate. If more cooling power was required, a good solution would be to wire 4 peltiers in series, with a 9.6V rail, so that each peltier was operating at 2.4V. Then we would get a respectable 26.8W of cooling for only 9.6W of DC input. I have used up to x16 peltiers in series/parallel to shift significant amount of heat with bugger-all DC power input, in applications where deltaT is modest.


Now let’s look at providing the same deltaT, but pushing the peltier much harder at 4A.

deltaT=10 DegC
Cooling = 29W
I = 4.0A
V = 9.60V
Pin = 9.6x4 = 38.4W
Efficiency = 29/38.4 = 0.755 (76%)

Ouch! That's a horrible efficiency, 3.7 times worse than we got above  by 'idling' the peltier at 1A, though in absolute terms we are getting much more cooling, of course. If the requirement was for only 6.7W, as per the first case, then we could PWM the 9.6V rail to get the 6.7W, but the efficiency would still be crap at only 76%. If the requirement is for 29W, then it's still a crap solution, as we have already shown that we can get 26.8W by using x4 peltiers at 280% efficiency with a total input power of only 9.6W. 


So in the examples of this posting, with small deltaT of only 10 DegC, using a relatively high voltage rail with PWM produces poor efficiency compared to using a lower, continuous voltage. This is the exact opposite of how best to operate your peltiers when a large deltaT is required, as per the example in the previous posting.

[Kleinstein]

In all cases using PWM directly with a Peltier will give lower efficiency. In Zeranins calculations above the loss due to thermal conduction in the off times of the PWM is ignored, so efficiency is calculated to high. So this type of PWM would only work with thermally disconnecting the element - not at all practical. 
The comparison is independent of temperature difference: The ohmic loss is proportional to the current square, in the approximation of constant temperature profile, which is a very good approximation for fast PWM too. The heat flow due to the peltier effect is proprotional to the average current. The heat flow due to thermal conduction only depends on the temperature difference, so it's just a constant background. So the best ration of wanted peltiier heat pumped ( ~ I_average) to ohmic loss (~I²_average) is at a constant current. Especially if the needed power is low (and thus possible only at a low temperature difference), the constant current can be much more efficient.

Even if the current is not high enough to reach the current temperature difference, a low current will still produce some heat flow that counteracts thermal conduction inside. So less heatflow from hot to cold than without the current.

One problem with a peltier cooler compared to a compressor type heat pipe is the rather high thermal conduction between the 2 sides (this is the main reason for the poor efficiency). One consequence is that the size of the Peltier cooler should be carefully matched the power requirements - to large an element will cause more heat conduction and thus lower efficiency. So while it is really easy to reduce the cooling power by changing the current or using PWM, this reduces efficiency. It is somewhat OK if the lower heatflow is needed because of a lower temperature difference, but not if it is in something like a fridge to fast cool a new contend and than keep it just there. Usually the Peltier coolers are also dimensioned that they have just enough power to keep the temperature - adding a warm bottle takes a very long time to cool it down. This also makes regulation easy -there is essentially no active regulation, if at all its just an on/off type to prevent to low temperature is very rare cases in winter. 

The nominal (sometime called optimum) current is the current that gives the highest cooling power at any given temperature difference and thus the highers possible temperature difference. The best efficiency is reached at a usually lower current, that is the needed temperature differential divided by the maximum differential times the nominal current for that element. So if an element is rated for 50 K differential and only 10 K of difference is needed, the best efficiency is reached at 1/5 the rated current.

[splin]

To demonstrate a situation where PWM provides the most efficient operation, assume that the thermal requirement is a cooling power of 2W, at deltaT=40K, for Thot=27 DegC, so therefore Tcold=-13 DegC.  The required plot for Th=27 is on the left of the page, and deltaT=40 can be found on the lower X-axis. From this plot, it is seen that an operating current of 2A exactly provides the desired 2W of cooling, at dT=40, so this is the point at which a non-PWM driver would operate. This operating point is :-

deltaT = 40 DegC
Cooling = 2W
I = 2A
V = 5.5V
Pin = 11W
Efficiency = 2/11 = 0.18

OK, that’s pretty crap efficiency all right, but peltiers are always inefficient when producing large deltaT. Now let’s look at a different operating point, that would be applicable to PWM operation.

deltaT=40 DegC
Cooling = 12W
I = 4.0A
V = 10V
Pin = 10x4 = 40W
Efficiency = 12/40 = 0.30

An efficiency of 0.3 is still not great, but it’s a damn site better than 0.18, so this is a much preferable operating point, except that the 12W of cooling is excessive, as we want only 2W. No problems, use a 10V supply and set the PWM  duty cycle to 1/6 and we get to have our cake and eat it too, with 2W of cooling power produced at 0.3 efficiency rather than the 0.18 efficiency that we would get with a non-pwm driver set to a steady 5.5V as in the first example.

Sorry Zeranin but Kleinstein is right - that 40mm square peltier will have a thermal conductivity of around .3W/K, so for 5/6 of the time when the power is off approx 12W will be flowing back to the cold side which means it couldn't maintain a 40C deltaT for very long under those operating conditions.

The thermal conductivity can be calculated from the information on page 13 of Laird Tecnologies handbook THR-BRO-Thermal Handbook 0110.pdf. Their CP14,127,10 module is similar to the one you quoted so it has a geometry factor of .077 and 127 elements.

[Zeranin]

Sorry Zeranin but Kleinstein is right - that 40mm square peltier will have a thermal conductivity of around .3W/K, so for 5/6 of the time when the power is off approx 12W will be flowing back to the cold side which means it couldn't maintain a 40C deltaT for very long under those operating conditions.

The thermal conductivity can be calculated from the information on page 13 of Laird Tecnologies handbook THR-BRO-Thermal Handbook 0110.pdf. Their CP14,127,10 module is similar to the one you quoted so it has a geometry factor of .077 and 127 elements.

Kleinsten's original posting wan't clear to me, but his later posting, and your posting, are absolutely correct! Thank you. As you say, the correct reason that PWM falls down is because of  ‘thermal leakage’ when the peltier is OFF. Everything I said was impeccably correct, except  that I neglected to consider what happens when the PWM peltier is OFF. Ooops.

The correct conclusion then is that for any given Peltier, non-PWM operation provides the best efficiency for all combinations of required cooling power and differential. That said, it would be instructive to look at just how much worse is PWM operation under typical operating conditions. The good news is that Linear or PWM operation converge to the same thing at full cooling requirement, 100% duty cycle, which is arguably where efficiency matters most. If PWM turns out to be significantly worse only at very low duty cycles, then it’s no big deal. Unfortunately though, the following example, chosen to be a real-world as possible, shows that PWM gives really horrible efficiency under typical partial-load conditions.

Referring to the Peltier plots attached previously, I have chosen an operating condition of Thot=50 DegC, deltaT=40, so Tcold=10 DegC. For this particular Peltier, typical full-power cooling would be at 4A, producing 16W of cooling, and I will analyse the efficiency at 5.33W of cooling power (1/3 of full) using both Linear and PWM voltage drive. The peltier thermal conductivity is taken as 0.3W/K, for a back-leakage of 12W when peltier is OFF. From the peltier data :-

deltaT=40 DegC
I=4A
V=11.2V
Pin = 44.8W
Cooling = 16W
Eff = 36%


Now here is the operating point that gives 5.33W of cooling :-

deltaT = 40 DegC
I = 2A
V = 6V
Pin = 12W
Cooling = 5.33W
Eff = 44%


Now for the crunch, to see how PWM compares at producing 5.33W of cooling :-

Duty cycle = 0.62
Cooling = 0.62 x 16 = 9.92W (while peltier is ON)
Back leakage = (1-0.62) x 12 = 4.56W (while peltier is OFF)
Net Cooling = 9.92-4.56 = 5.36W
Pin = 0.62 x 44.8 = 27.8W (average input power)
Eff = 5.36/27.8 = 19%

Yuk! That is really horrible. To get our 5.3W of cooling, we are pumping in 28W of electrical power, compared to only 12W using a non-PWM input of 6V. To put it another way, the efficiency is worse by a factor of x2.3 I am an efficiency nut, so I do find this very offensive, but we need to keep in mind that the cooling performance of the fridge or whatever won’t be any worse because the 2 drive methods provide the same maximum cooling power, so the only consequence, actually, is that the PWM implementation will consume more electrical power under partial load. Whether this matters depends on the application.   

Another interesting point arises on account of that damned back leakage. Usually, it’s good engineering to ‘oversize’ within limits, be it a resistor, power semiconductor, heatsink or whatever. Peltiers are cheap, and we know they are not very efficient, so there is a great temptation (at least for me) to use more or bigger of them than strictly necessary. At worst, it might be thought, you spend a little more on the peltiers than was needed, but in terms of performance and efficiency, intuition suggests you can’t go wrong with using more or bigger peltiers, so each one doesn’t work as hard. That’s the way I always design things! In this case though, for any given requirement of cooling power and differential, there is an optimum size of peltier that maximizes efficiency. Of course, you need to allow some reserve capacity, but if you greatly oversize the total area of Peltiers, then the efficiency goes to pot, essentially because it takes a significant overhead of electrical power input just to stop the heat flowing backwards through the peltier thermal resistance, before you even get to the stage of pumping the heat in the correct direction. There is no analog of this with compressor cooling systems, and it’s not a nice feature of Peltiers. Edited in later: Note to Kleinstein. I hadn't read your posting #47 when I wrote this, but we are independently saying the same thing.