Cooling (and warming) a container with a TEC (Peltier) - controller suggestion?

[petreza]

Hi,

I am going to use a TEC to keep a container (an ice cooler) at approximately 5C on my terrace. This means that during the summer (30C to 35C) the TEC has to cool the container and during the winter (-5C to -15C) it has to warm it.

1. Could you recommend a controller that has a temperature probe and:
- can lower the voltage as the temperature drops to near the target
- cuts off the voltage at the target
- reverses the voltage if the temperature drops below the target

2. The outside dimensions of the ice cooler are 34"x18"x15" (86cm X 45cm X 38cm).
Link: http://www.amazon.com/Coleman-Quart-Xtreme%C2%AE-Marine-Cooler/dp/B0019QD8HQ/ref=sr_1_1?s=sporting-goods&ie=UTF8&qid=1440543574&sr=1-1&keywords=Coleman+100-Quart+Marine+Cooler
At what rate would you expect the heat to leak into the cooler if the outside temp is 35C and the internal is 5C and the cooler is in the shade? What power rating TEC would you recommend to keep up with the load (I plan to use 4 lower power rating TECs instead of one high power rating TEC for redundancy (if that matters))?

3. What brand of TEC is considered good, reliable one while not overly expensive?

Thank you in advance for your help!!!

[TheElectricChicken]

I am going to use a TEC to keep a container (an ice cooler) at approximately 5C on my terrace. This means that during the summer (30C to 35C) the TEC has to cool the container and during the winter (-5C to -15C) it has to warm it.

Seriously for this application you'll spend a fortune on electricity. TEC's are the worst technology for cooling available. Use a barfridge or even better, a bar freezer. Use an arduino to to switch the freezer on when the temperature reaches 6 degrees and back off at 4, then get it to switch a bank of resistors on when it dips below 4 for a few minutes. The arduino and it's power supply could go inside the freezer in a heated OR moisture proof enclosure if you like. It would help heat it in winter. 

The reason this works better is that a freezer has the best insulation and will cost the least to run. There are many ready made tutorials, and ready-built products and kits to turn freezers into super efficient fridges that keep drinks at 4degC or any other temperature you can think of. As for heating, nothing is more efficient than the cheapest resistors or scrap heating element you can find (short of a heat pump)

[Chris C]

I'm afraid I can't answer most of this, as I don't work with TECs.  But I'll mention one thing I learned about them, which came as a bit of a surprise to me, and so I always remembered it:

In general, the lower the current you run a TEC at, the more efficient it becomes.  Here's a fairly generic chart, typical of most TECs:



Outside temperature of 35°C vs inside temperature of 5°C means dT=30°C, so look at the orange line.  1 on the horizontal axis is 100% of the TEC's maximum rated current.  Notice that if you run it at 50% of the rated current, efficiency doubles!

But that doesn't mean it produces as much cooling as a TEC run at 100% current.  To run it at 50% rated current, you need to reduce the voltage by half.  Power = voltage times current, so at 50% current, you'll actually be running the TEC at 25% power, producing 50% as much cooling due to the efficiency increase, compared to the TEC run at 100%.

A common way to do this is to take two TECs rated for maximum power at 12V, put them in serial, and drive them with 12V.  Each gets half as much voltage, and due to the serial resistances, each gets half as much current.  In the end, you get as much cooling as you would from a single TEC running at 100%, but using half as much power.  Or put three in series and drive from 12V, you get the same cooling, but using a little more than a third of the power.  That might even make [TheElectricChicken] happy.

So there is a good reason to use multiple TECs, but not for redundancy's sake.

And don't downsize the TECs while doing this.  Smaller TECs = less rated current, putting you back on the ugly parts of the efficiency curve.  Use TECs sized so that any one could bear the normal cooling load alone, at 100% current.

[TheElectricChicken]

put three in series and drive from 12V, you get the same cooling, but using a little more than a third of the power. 

Now how many in a row will I need so that it powers itself ? 

[petreza]

I am going to use a TEC to keep a container (an ice cooler) at approximately 5C on my terrace. This means that during the summer (30C to 35C) the TEC has to cool the container and during the winter (-5C to -15C) it has to warm it.

Seriously for this application you'll spend a fortune on electricity. TEC's are the worst technology for cooling available. Use a barfridge or even better, a bar freezer. Use an arduino to to switch the freezer on when the temperature reaches 6 degrees and back off at 4, then get it to switch a bank of resistors on when it dips below 4 for a few minutes. The arduino and it's power supply could go inside the freezer in a heated OR moisture proof enclosure if you like. It would help heat it in winter. 

The reason this works better is that a freezer has the best insulation and will cost the least to run. There are many ready made tutorials, and ready-built products and kits to turn freezers into super efficient fridges that keep drinks at 4degC or any other temperature you can think of. As for heating, nothing is more efficient than the cheapest resistors or scrap heating element you can find (short of a heat pump)

Thanks for the suggestion TheElectricChicken but the landlord in my building does not allow us to have a second fridge/freezer because they pay the electricity. On the other hand I can blast the AC as much as I want. So I am stuck with a cooler with a hidden TEC in the back which I guess will cost them at least 4X as much. Go figure.

Thanks!

[Chris C]

Now how many in a row will I need so that it powers itself ? 

LOL!  You jest, but something like that has actually been documented.  Of course, it only works at VERY low power.  Like in the pico-amps range.  Not exactly useful!

[TheElectricChicken]

sigh. Landlords, say no more 

Other options are evaporative coolers. Placing a cloth or towel over a milkcrate or any kind of box and then keeping it damp will cool the contents on your veranda. I've had friends who do that, and it works extremely well. You won't believe it till you try it. I tried it myself with a bottle of drink in the car, by placing a cloth over it and splashing a little water on it from time to time on a long journey, it was WOW !

and a heater is the easiest thing in the world to hide as it is silent

[petreza]

I'm afraid I can't answer most of this, as I don't work with TECs.  But I'll mention one thing I learned about them, which came as a bit of a surprise to me, and so I always remembered it:

In general, the lower the current you run a TEC at, the more efficient it becomes.  Here's a fairly generic chart, typical of most TECs:



Outside temperature of 35°C vs inside temperature of 5°C means dT=30°C, so look at the orange line.  1 on the horizontal axis is 100% of the TEC's maximum rated current.  Notice that if you run it at 50% of the rated current, efficiency doubles!

But that doesn't mean it produces as much cooling as a TEC run at 100% current.  To run it at 50% rated current, you need to reduce the voltage by half.  Power = voltage times current, so at 50% current, you'll actually be running the TEC at 25% power, producing 50% as much cooling due to the efficiency increase, compared to the TEC run at 100%.

A common way to do this is to take two TECs rated for maximum power at 12V, put them in serial, and drive them with 12V.  Each gets half as much voltage, and due to the serial resistances, each gets half as much current.  In the end, you get as much cooling as you would from a single TEC running at 100%, but using half as much power.  Or put three in series and drive from 12V, you get the same cooling, but using a little more than a third of the power.  That might even make [TheElectricChicken] happy.

So there is a good reason to use multiple TECs, but not for redundancy's sake.

And don't downsize the TECs while doing this.  Smaller TECs = less rated current, putting you back on the ugly parts of the efficiency curve.  Use TECs sized so that any one could bear the normal cooling load alone, at 100% current.

Thanks for the info Chris C. I didn't know that.
I was thinking of running 4 TECs in parallel but looking at the graph I might hook them 2 series-pairs in parallel. Better efficiency with some redundancy - I just need to get TECs with slightly higher max power rating.

I can probably figure most of this stuff out but my main problem is how to control them - there I am a complete newbie.

Thanks again!

[TheElectricChicken]

Like in the pico-amps range.  Not exactly useful!

Oh my calculations show otherwise, an supervillan or superhero may want this setup, many applications I can see.

[TheElectricChicken]

wait, I'm reading the graph wrong !!! argh !!!! plans for world domination foiled again !! 

back to the drawing board.

[Chris C]

wait, I'm reading the graph wrong !!! argh !!!! plans for world domination foiled again !! 

back to the drawing board.

Dude.  I just spit my coffee. 

[TheElectricChicken]

Dude.  I just spit my coffee. 

[LukeW]

"A common way to do this is to take two TECs rated for maximum power at 12V, put them in serial, and drive them with 12V.  Each gets half as much voltage, and due to the serial resistances, each gets half as much current."

Is it true to assume that the V-I curve is linear, like a resistor? At what point on the V-I curve is it most efficient?

These are active devices (although the semiconductor crystals aren't arranged in PN junctions).
Might be best to obtain data (including the V-I curve) for the particular thermoelectric array you're using.

[rs20]

The other thing to keep in mind is that TECs should not be (directly) driven by PWM, for the same reasons outlined above. If you compare using 5V at 50% PWM (i.e. 2.5V average) vs smoothed 2.5V DC; they'll both have the same amount of heat transfer from side A to side B, but the I^2 R heat dissipation in the PWM case will be twice as high. Since I^2R losses are what define the upper limit of TEC performance, this is very significant. Fortunately, sticking an inductor and a capacitor after your power PWM stage will do this smoothing for you (effectively creating an open-loop buck converter).

As for the controller, why not just use an Arduino or equivalent, with a DS18B20+ on some wires for temperature measurement (or whatever standard Arduino temperature bit is standard), and a standard motor "shield"/controller (with the aformentioned L/C stage after it) to drive the TEC?

Remember to put a very large, good heatsink on the hot side of the TEC... the performance of the TEC goes massively downhill if it can't dump the thermal energy on the hot side.

[Chris C]

Is it true to assume that the V-I curve is linear, like a resistor?

I've always heard they were.  But you piqued my curiosity, so I just checked a couple of datasheets, and found they lacked V-I curves.  Although some other graphs not really organized for that purpose seemed to indicate linearity.

In fact I was only able to find one bona fide V-I curve, measured by a hobbyist:



From this page:

http://www.procooling.com/index.php?func=articles&disp=43

Which is close enough to linear, that I'd still feel comfortable informally calling it linear.  And I wonder if this test might have been slightly flawed.  As voltage/current goes up, so will the average temperature of the peltier; unless he has infinite heat sinks attached.  Since resistivity of most metals goes up with temperature, that would account for some, or possibly all, of the non-linearity seen.

Though if you know of any other graphs which show different results, I'm all ears.

[Kleinstein]

Having 2 or even more TEC thermally in series only makes sense if you need a rather large temperature difference. The second lower temperature TEC would need to be larger (e.g. 4 times) to transport the waste heat from the lower temperature one.

Having 2 in series and 2 in parallel is very much as good as using just one. Just the stress to the joints might be little lower.

The curve is correct - at lower temperature difference the best performace shifts to lower currents. However in such an application one needs to choose the TEC according to the highest outside temperature. If you choose it to large, maybe for redundancy, eficiency will drop. Due to the lower cooling requirement at lower outside temperature, the current will be lower in these cases. So its not that bad but sill will not hit the optimum curve - but there is little one can do about it.

Likely the best bet would be going for a larger 1.st freezer inside. If it's a newer model current consumption may not be higher than the old one. 2 nd best option would be offering to pay the typical electricity costs for a second freezer (maybe $50 a year) - usually its noted somewhere in the documentation. If this one is just 5 C, so not makig ice cubes in large quanteties, the numbers are also usually quite reliable.

[Chris C]

Having 2 or even more TEC thermally in series only makes sense if you need a rather large temperature difference. The second lower temperature TEC would need to be larger (e.g. 4 times) to transport the waste heat from the lower temperature one.

Ah, I should have clarified that.  By in series, I meant electrically, NOT thermally.  So the TECs would be side by side on the heatsinks, not stacked on top of each other.  As you say, stacking them is useful for large dT; but only for that, as the efficiency plummets.

[rs20]

some form of H-bridge. like DC motor drives? flips one way, then another ?

As I've already mentioned, you can use a motor driver, but you absolutely must to use an inductor to smooth the PWM into a DC voltage to reduce the otherwise massive I^2R dissipation in the TEC.

[rs20]

i was thinking making 1 from scratch. a simple thermistor-opamp combo, feed some inverting logic, which then flip flops a simple H bridge (in DC, no PWM) like this simulation (sorry its a mess  ).

Again, and again, this is a bad solution. It'll flip-flop between peltier off and peltier 100% ON, which is terrible because the peltier becomes dramatically less efficient as current increases. The control system should be aiming to find a steady-state, DC voltage that maintains the temperature just right. And this is coming from me, I'm normally the one most in favour of rail-to-rail digital solutions, but TECs are absolutely not the place to use this approach.

By contrast, your solution is perfect for a traditional refrigerator.

[Kleinstein]

The size / number of TEC modules should fit the heat load. A TEC that is to large will need more power, as the current will be below the optimum point.  So the TEC(s) schould be matched to ther thermal insulation / size of the box. Neither a small TEC allways running on full power, nor a very large TEC, running on very low current is good. The heat resistance of the TEC is a large part of the heat leakage from inside to out. So one bad thing with TEC coolers is, that they don't hold the temperature very well in case of a power outage.

Also thermal contact / cooling is important. This often means a fan at the hot side at least. If you don't get rid of the heat the TEC has to work much harder.

I would not do the temperature control in analog mode any more. Typical the time constants are rather long , so this requires large caps / high impedance circuit. The TEC is a nonlinear beast - the cooling power is not a simple linear funktion of the voltage / current. Especially at higher temperature differences it helps to let the regulator knows that. At the upper llimit, or with poor heat sink (e.g. blocked air flow) the system could even get unstable and start heating instead of cooling.  If high temperatre stability is not the ultimate goal, there may also be an extended range where the TEC is not needed. For a digital regulator (µC) it is realtively easy to take care of this.

Energy efficiency is allready poor with a TEC, using hard ON/Off or linear regulation something like about doubles the power need. Even if you don't pay for the power, this means more cooling for the unit. A µC can realtively easy provide a PWM signal to drive something like a switching mode output stage.  For switching from heating to cooling a bistable relais might be a good option. Usually there is not need to change that very often or fast.

[petreza]

Thank you all for the replies!!!
As I was expecting it quickly got out of my depth. Let me simplify the requirements so that hopefully the design becomes simple enough so I can implement it.

1. the 5C internal temperature is not a very strict requirement. It can go to, lets say, 2C to 8C. I would like to avoid the lower range though since there would be some humidity in the container and it would freeze and accumulate on the cold heatsink.

2. Forget about the variable voltage. I will use a straight 12V (from a computer power supply) with a variable resistor in series to calibrate the power.

3. Forget about the reversal of voltage. I will do it manually as the seasons change.

QUESTION: Is there some kind of 12V or 5V thermometer device (hopefully with approx. 2ft-3ft thermal probe) that cuts of the voltage (with a relay or something) when the temperature reaches the target 5C.

So my new design is like this:

_____________                                      ________________
|                      |------------------------------|                           |-----------------------------------------|
|   Computer     |     12V                          |                           |          _________________        |
| Power Supply  |                                    |       RELAY           |          |                              |       | <-- To TECs
|____________|++++++++++++++++|                           |+++++|  Variable Resistor     |+++|
    +           -                                          |_______________|         |_________________|
    +           -                                                         |
    +           -                                                         |
    +           -        _______________                      |
    +           -------|                          |                     |     <--- Relay Control
    +                   |                          |                      |
    +  12V or 5V   |   Temperature     |____________|
    +                   |       Control         |                                                                                     |
    ++++++++++|       Device         |_______________________________________________|  <-- Temp. Probe
                         |______________|                                                                                      |

[Chris C]

Go on Ebay and search for "temperature controller".

They are, in general, programmable digital devices with a nice readout, that control a relay.  I know little about them beyond that.  You may (or may not) need to buy a separate thermocouple probe, of specified compatible type.

There are some with separate high and low limits.  This could be used to drive four external relays in a H-bridge configuration, allowing automatic heating or cooling.

I wouldn't bother with the variable resistor, personally.  Using TECs in serial, or serial-parallel, as we earlier discussed will improve efficiency quite a bit.  Not to the degree that something with full analog control of the voltage would.  But a variable resistor itself is inefficient, and you'd have a hard time finding one that can handle the required power too.  I think we should simply consider analog beyond your scope for now, more important to make some progress.

[rs20]

2. Forget about the variable voltage. I will use a straight 12V (from a computer power supply) with a variable resistor in series to calibrate the power.
3. Forget about the reversal of voltage. I will do it manually as the seasons change.

You don't need to make these compromises at all. A dead standard motor driver + a single inductor = a variable voltage, reversible power supply ideally suited to driving a TEC.

[petreza]

Go on Ebay and search for "temperature controller".

They are, in general, programmable digital devices with a nice readout, that control a relay.  I know little about them beyond that.  You may (or may not) need to buy a separate thermocouple probe, of specified compatible type.

There are some with separate high and low limits.  This could be used to drive four external relays in a H-bridge configuration, allowing automatic heating or cooling.

I wouldn't bother with the variable resistor, personally.  Using TECs in serial, or serial-parallel, as we earlier discussed will improve efficiency quite a bit.  Not to the degree that something with full analog control of the voltage would.  But a variable resistor itself is inefficient, and you'd have a hard time finding one that can handle the required power too.  I think we should simply consider analog beyond your scope for now, more important to make some progress.

Thank you Chris C,

This is what I found:

http://www.ebay.com/itm/12V-Double-control-Digital-Temperature-Controller-50-110-Sensor-Thermostat-/321603689558?hash=item4ae112dc56

It even has the the relays built-in. I can short out the positive and the negative of both relays, BUT, of course, make sure in advance that the temperature settings are set properly so that only one relay is closed at a time. That way I can have heating and cooling with only one neat unit between the power supply and the TECs.

What do you think?
Thanks!

[Chris C]

This is what I found:

http://www.ebay.com/itm/12V-Double-control-Digital-Temperature-Controller-50-110-Sensor-Thermostat-/321603689558?hash=item4ae112dc56

It even has the the relays built-in. I can short out the positive and the negative of both relays, BUT, of course, make sure in advance that the temperature settings are set properly so that only one relay is closed at a time. That way I can have heating and cooling with only one neat unit between the power supply and the TECs.

What do you think?

Looks fine, except one problem, which is easily fixed.  To route current forward or backwards through the TECs, you'll need four switches in what is called an H-Bridge configuration.  Wikipedia can describe that better than I can, so:

https://en.wikipedia.org/wiki/H_bridge

Fortunately there are only two combinations of those four switches than need to be produced.  Which means you can use the controller's two built-in relays to drive two external 12V DPST relays:



Giving you the four switches you need.  As per Wikipedia's example diagram, one DPST relay would provide S1 & S3, the other would provide S2 & S4.  You'd still need to set the high/low temperatures appropriately to avoid everything being turned on at once, causing a short.

And this solves one other potential problem as well.  That controller you linked is a no-name Chinese product.  As such, the stated internal relay rating of 10A is probably optimistic, and they wouldn't last very long if actually used at 10A.  If the TECs only draw 5A, that's better, but maybe not enough for years of reliable service.  But since we're going with external relays, you can get quality relays, that will last a long time.  The ones in the controller will then only be switching enough current to drive the external relays' coils, which will be a few tens or hundreds of mA, and should basically last forever.

You don't need to make these compromises at all. A dead standard motor driver + a single inductor = a variable voltage, reversible power supply ideally suited to driving a TEC.

I like that.  But that only solves the driver portion.  Is there a similarly easy controller?  I get the impression [petreza] is trying to avoid dealing with op-amps or MCUs.