Building Your Own Mini Fridge?

Ever play with a Peltier plate? They’re these really cool components that kind of look like a ceramic sandwich, and when you put power into them, one side gets hot, and one side gets freezing cold! [Joseph Rautenbach] decided he wanted to try making his own mini fridge out of one — which is typically how most modern mini fridges work these days.

The peltier plate he’s using draws 12v at about 3.5 amps — so about 50W — and if you don’t heat sink it properly you could burn it out in a matter of seconds. Peltier plates only care about the temperature differential between the two sides — if you don’t take the heat away from the hot side, it will soon overheat and destroy itself.

[Joe’s] using a styrofoam cooler for the fridge with a pair of computer heat sinks and fans for the peltier plate, and a temperature PID controller he bought off eBay. The external heat sink sucks away the excess heat generated by the peltier plate, and the internal one helps spread cooled air around the inside of the styrofoam cooler. The PID controller allows him to set a preferred temperature to maintain in the box, which will then control the outputs to keep it that way.

To power the whole thing he’s using a modified computer ATX power supply — but if you’re on the road, you can power it from your car’s 12V socket too!

We’ve seen lots of cool uses for peltier plates before, like this human-energy-sucking LED ring! Or even fancier than a mini fridge — a sous-vide rig!

53 thoughts on “Building Your Own Mini Fridge?

  1. I hate to be pedantic but…
    “The peltier plate he’s using draws 12v at about 3.5 amps” that’s the other way around.
    It draws 3.5 amps at 12v, and it makes all the difference in the world. It will happily take twice as much voltage, draw twice as much current, and fry pretty fast.

    1. Peltiers are actually most efficient at around half their rated power, because the heat transfer scales directly with I while the heat generated in the device scales at I^2

      The largest difference between I and I^2 between 0…1 is at ½ which means the device will transfer the most heat with the least waste at exactly half its rated output. That’s why it’s a good idea to use two peltiers wired in series.

      The less waste heat there is, the cooler the heatsink remains and the smaller the dT between the inside and outside heatsinks, which also improves the heat transfer efficiency. Overall, you can improve the performance 3-4x over the simpler cheaper design.

        1. Best performance is at 15v, that Peltier he’s using is probably a 90W (15v @ 6A) one.

          No sure why the eBay sellers have all decided they’re 12v devices.

          Anyway, that’s 90W of heat you need to get rid of for not much benefit. I was going to build something similar for water cooling, but wound up just getting one of those little bar fridges instead.

      1. I think his point is a bit more Grammar Nazi, current or power is what you draw at a voltage. You don’t draw voltage. I was actually thinking James Hobson did a good job of the introduction, Peltier coolers are misunderstood by most. But yeah, technically that’s a little mistake.

      2. So if I were to build a similar mini-fridge, but with two peltiers running at 1/2 rated power; would it be more efficient to stack them, or place them side by side?

          1. They can be stacked for large temperature differentials, but the system isn’t symmetric because the top peltier has to deal with a higher heat load. It limits the maximum heat throughput to a fraction.

  2. Their was I time when these Peltier elements weren’t cheap to buy, I then took a mini fridge (was cheap and half priced) apart to get one with its heat-sink. Should have kept the enclosure and donate it to him…

  3. That’s pretty nice. I like the foam cooler as the container idea. I would have placed the peltier on the top though, since the “cold” will fall while the heat energy rises.

    I did a similar thing a few years ago with a giant peltier but I made the internal container only as tall as the cold-side heat sink, and used the fins as a pathway to pull room temperature air through with a weak fan, and lightly blow chilled air out, making a grossly inefficient personal air conditioner that actually heated up the room in the grander scheme of things.

    Luckily it was a rather large cubicle farm of a room with a tall ceiling. The room-heating personal air conditioner was a drop in the bucket compared to the stacks of computers and large-volume printing equipment that was running there anyway.

    With the main air conditioner out in the Texas summer, while everyone else was breathing 86 degree air, huddled in my little corner I was sitting relatively comfortably with my light 72 degree breeze. The adapter I was using as a power supply on the other hand, was ready to melt.

    1. Since the cold side has a fan actively circulating air, I’m not sure placing it in the top would make much difference. However, putting it in the top also means you can load the cooler with ice and let it melt, or have a container leak other liquid, without it getting into the fan. So I’m a “fan” of the idea to move it to the top.

    2. I built it that way out of pure convenience, and not having the condensate drip onto the fan. It’s capable of getting down to –2.1°C so it’s fine for my needs!
      Your DIY Peltier A/C sounds pretty fun! What sort of Peltier module were you actually using, what was it from?

  4. I had a problem keeping some meds around 20C in a very warm room (25 t0 27C typically). Air conditioning was out of the question, so I made one of these. I used an MSP430 with a thermistor to monitor temps and a relay to turn the 12V off and on according to temperature. After fiddling with it, I set it to 18C and it worked for years without a problem until the hot side fan died. It was an odd fan and I can’t seem to get a 12V computer fan to get along with the power supply – it just wimps along without spinning much. Back to the drawing board so it will work this summer.

  5. I have been thinking of making something similar but for a small dehumidifier box to keep 3d filament dry The problem is that they are so hopelessly inefficient and it would be cheaper to go a buy filament to print with then bin it till next time you want to print and buy some more.

  6. I usually don’t post, but I thought I could give my two cents here since I know a little bit about thermodynamics and heat-transfer. I don’t know where the numbers about improving the efficiency 3-4x by reducing heat waste or the optimal efficiency at 5volts or whatnot come from? However, Peltier cooling is notorious for it’s inefficiency. In fact they are 4 times less efficient than the conventional compression cycle in your regular kitchen fridge. On the insulation note, sure if you have your beer already at 2deg.C and you have infinite thickness insulation, you won’t need to cool anymore. However, as long as you open the box and want to cool another beer, then you need refrigeration power and you’re still going to be 4 times less efficient than the regular fridge with the same insulation.
    On the voltage note, they’re sold as 12volts because these things are meant to be used in cars. Portability and maximum power (with reasonable heat sinks, not heat pipes, etc.) is key, not efficiency. I found one of these coolers on the street, and I find them to be actually useless for indoor use; the fan is so darn loud.. and it has to be to achieve maximum power. The one I have, doesn’t have a thermostat, and it really doesn’t need it if you use the thing regularly… but I realized it could use one when the outside air is cool.
    In no way I intend to belittle the work nicely explained in the video though. It’s a good make.

    1. I’ve “hotted up” a few of those coolers.
      Step 1 is to take the heatsinks out and ensure they are FLAT on the side contacting the Peltier. Best way is with a flat metal file. Learning how to use a file and keep flat what you are filing is an essential skill. If you have a milling machine that’s perfectly aligned and has a power feed that can run really slow, the surfaces could be cut smooth with a face mill.
      For the little coolers that use an aluminum tub for the inside sink I make an aluminum stiffener plate, thicker than the tub metal, to force the tub into full contact. I use a proper amount of heat sink compound on both sides of the Peltier and between the stiffener and tub.
      Step 2 is rounding all sharp edges on the heatsinks to reduce edge turbulence that may ‘choke’ airflow between the fins. (That worked on the Los Angeles river, on a much larger scale…)
      Step 3 is glass bead blasting all surfaces of both heat sinks (or stiffener plate) except where they contact the Peltier. Apply a piece of masking tape then blast it. A rougher surface has a higher surface area. More area = increased heat transfer. I’ve also bead blasted CPU and video card heat sinks and it helps, especially in marginal cases. Have had some go from shutting down from overheat to staying cool enough after bead blasting.
      Step 4 is replacing the cheap and noisy brush type motor and squirrel cage fan with a 12V brushless DC computer fan. Much more airflow = better heat transfer. Some of the coolers have a hole in the insulation where the motor is. I mount the brushless fan to the outer housing (along with opening up the intake grille) and plug the insulation hole with some expanded polystyrene or polyurethane foam board. Some of the coolers, even some with the thin metal tub inside, use brushless fans. (Coleman used to have a brushless fan upgrade for several of their coolers with a double shaft motor and dual fans but the part is unobtainable now.)
      Step 5 is replacing whatever it had for a gasket (if any) around the Peltier with closed cell craft foam, more than one layer if needed to fill the gap. The gasket is essential to keep condensation from damaging the Peltier. The craft foam is denser than some of the cheap junk used on these coolers.

      Another essential construction detail is to insulate the mounting screws from one or both of the heat sinks. They generally have plastic bushings on the hot side to prevent heat from conducting through the screws to the cold side. If you have a cheap cooler with screws that have direct metal to metal contact with both heat sinks, there’s a large efficiency loss.

      The results of my modifications? A cooler for 9 12oz cans, barely capable of keeping them down to a drinkable temperature, gains enough cooling capability to keep a 1/2 gallon box of ice cream (or a bunch of Schwan’s ice cream cones, item #007 in their catalog) frozen for quite a while.

      1. Here is the performance data (.pdf) for the 12706 modules – looks like you can get a delta T of ~75 degrees if everything is perfect. Heat flow maxes out at around 57W (3.24 btu / minute) which should cool things okay (if insulated well) if all the other details are worked out as you suggest. Round figures suggest cooling a six pack of age-appropriate beverage from room temperature to very cold in less than an hour.

        I also get things as flat as possible then use a little heat sink compound for luck.

      2. right ideas but wrongphysics, for example. it isn’t the surface area increasing heat transfer, it is the roughness encouraging local separation and turbulence. Im skeptical it makes anydifference. we can be sure of these effects if you can calculate some nu and re numbers similar for other mods

        1. I recall seeing a very old video which showed a red-hot chunk of metal being cooled with amazing rapidity, simply by grounding it and aiming an electrostatic emitter at it. The reason given was that there was no need for any turbulence to penetrate the slow moving surface layer of air near the metal (the Prandtl boundary). Every molecule of air that acquired a charge from the emitter, was attracted to and came in direct contact with the metal to give up its charge, carrying away heat with extremely high efficiency in the process. I have never seen this replicated since, in any form. Recently there have been some electrostatic “fans” used to cool CPUs, but are built so the charge in the air is stripped away before CPU/heatsink contact, to avoid damage to sensitive electronics. That wouldn’t be an issue with a peltier. Every time something like this comes up, I wonder if that old video was correct, and if we’ve forgotten a useful heat transfer technique.

    2. ” In fact they are 4 times less efficient than the conventional compression cycle in your regular kitchen fridge.”
      I’ve read datasheets for stacked 1:2 peltier modules with detailed performance curves and at low voltage, like 2v of a 15v rating and low temperature differential, 5-10C, achieve a COP over 3.

    3. I wasn’t sure if the quotes about improving efficiency were right either, so I tried to verify this, and found some charts that were helpful. Here’s a representative:

      The COP (coefficient of performance) of a peltier is dependent on both the dT (differential temperature) between hot and cold sides, and the percent of maximum rated current you’re running it at. Clearly, running a peltier at 100% power isn’t the way to go if you want efficiency!

      Assuming 72°F room temp, a 20°C air dT would place the inside of the cooler at about 36°F, about right for a refrigerator. For peak efficiency, the chart shows to run the peltier at 20% of max rated current, yielding a COP of 1.9. I found a paper where the COP of a “typical” refrigerator was measured and found to be 2.75. So the peltier, used properly, may not be as bad as you suggest – though if it’s 90°F outside and a 30°C dT is needed, COP drops to 1.1.

      Of course, you lose a little more efficiency by using power to run the fans. And if you skimp on the heatsinks/fans, the actual dT of the peltier’s plates will be much higher than the air dT, causing lower efficiency as well. But fortunately, since you’re running the peltier at lower power with longer cycles, rather than 100% power with short cycles, it will be easier for the heatsinks/fans to transfer heat/cold from the plates and keep plate dT closer to air dT, keeping COP high.

      Now if you wanted to get really fancy, you could design a controller than takes multiple temperature measurements, alters peltier voltage/current, controls fans with PWM, and so on to squeeze optimal efficiency out of the system at all times (though you could program in an override if cooler temp needs to be brought down quickly). But just running two peltiers in series seems like a good rule of thumb for decent efficiency, without too much complexity.

      1. It’s always good to see some real charts. Thanks. I couldn’t come up with a similar chart for a modern fridge, but I know for a fact that you can expect much better than the 2.75 that you mentioned. They are also refrigerant and temperature difference, as well as temperature dependent (evaporator and condenser temperatures). Expecting a COP of 6 or even 7 at similar conditions is not unreasonable.

        You mention that;
        “But fortunately, since you’re running the peltier at lower power with longer cycles, rather than 100% power with short cycles, …”

        Just keep in mind that there are two components of load when it comes to refrigeration;
        1-The load required to compensate for the heat loss, or more correctly, heat gain of the stuff inside the cooler due to imperfect insulation.
        2-The load to actually bring down the temperature of the fridge from the room temperature to the set temperature, when new, room-temperature stuff are put in the fridge.

        Now when your design is so stretched that you actually need to run the peltier for longer cycles to improve efficiency, the problem would be that for cooling down a six-pack or worse, turning water to ice-cubes, you’ll have to wait for a very, very long time. How to overcome this “weakness”? Slap a half-dozen peltier modules to the cooler.
        As usual, the remedy for low efficiency is “large” equipment size and cost.
        Don’t get me wrong though, I’m all for solid-state, hassle-free devices and I wish peltier modules had better efficiencies. The same with the opposite application of the peltier, the Seebeck effect, which similar size modules are only capable of charging your phone when placed on a toasty campfire. It could change though with materials of the future.

        1. I can’t seem to find any good info about refrigerator efficiencies. Assuming a COP of 6-7, that’s equivalent to a SEER of 23-27. Compared against A/C units (much easier to find ratings for), that seems unrealistically high, but I don’t know for sure.

          I understand that both #1 and #2 are an issue. They’re also a bit at odds – the more peltiers you use, the more you’re compromising the insulation by cutting out a larger area of it. Personally, I rarely put something room temperature in my peltier fridge, instead relying on it only to *keep* something cold on the road; so if I were to build one, I’d design primarily for #1.

          But I was curious about #2. A typical 12oz. can six-pack weighs 4.5lbs. As an example, let’s say we want to bring it down from 72°F to 36°F in an hour. That requires (4.5*36)=162 BTU of cooling power, or 48W of cooling power applied for one hour. Assuming you can reach a COP of 1.9 by not exceeding 20% of max current as in my earlier example, that requires only 25W into the peltier(s). If all this is right, it seems attainable, especially if running two peltiers in series. Two 60W peltiers comes close on paper. Two 100W peltiers (not actually run at 100W) will probably do it in the real world, with the peltiers costing $10 total from Ebay. This isn’t too expensive, or large, and doesn’t require a half-dozen peltiers.

          Plus, there’s this. Let’s say you start with a six pack precooled to 36°F. You wouldn’t completely empty the cooler before adding a new room temperature six-pack! Instead, remove one cold can, add another at 72°F, and the temperatures will soon equalize to 42°F – even without any assistance from the peltier, which will remove the extra heat in as little as ten minutes. If you can increase thermal mass further by starting off with more cans if space allows, or adding water to fill gaps (which also helps quickly transfer heat, and reduces losses due to cold air escaping each time you open to swap a can), all the better. I’ve used that same technique to keep food in my fridge/freezer cold up to three days without power or spoilage, despite opening it up regularly to consume stuff. Big hurricane coming, start loading up every possible square inch with bottles and baggies of water (incrementally of course, the compressor won’t be happy otherwise).

          1. Interesting thought: They’re also a bit at odds – the more peltiers you use, the more you’re compromising the insulation by cutting out a larger area of it.
            Just a comment on the calculation you made for the time it takes to cool down your six-pack. This is assuming that the six-pack would give up it’s heat at once to the peltier. What one has to remember is that the bottle-neck here is the convection heat-transfer resistance between the six-pack and the cooling element (Peltier or not). So in reality you would see that this won’t happen even in normal fridges in just one hour. It will, if you put the six-pack in the freezer for one hour, which is around 0F.
            This is not a limitation of the Peltier though, it’s the relatively high heat resistance of natural convection. You can speed up the process of course, by increasing the contribution of conductive heat-transfer (making the six-pack touch the cooling element) or using forced, turbulent, convective heat-transfer by recirculating the air inside the fridge and blowing it at high speeds around objects… Just some food for thought :-)

    4. Hi Fletcher,
      I want to rack your brain. If you were to build your own fridge, Im talking a pretty big one, commercial size, to store beer, what components would you use, to be efficient and keep the beer at a constant temperature. If you can help that would be great!

  7. The water collecting on the top fin will run down onto the element. The fins should be vertical so the water will run straight down. Water has a way of getting into “sealed” stuff.
    I still have an old Pentium cooler around, I have used it as a chill pad for swelling. Dripless ice cube.

  8. Thanks for covering this,
    The temperature controller I used isn’t actually a PID controller, it’s just a cheapo $5 eBay temperature controller. Still does its job well.
    And as someone else stated, yes, this Peltier module is stated to draw 6A, however at 12V it’s more like 3.5A – most likely how they are designed for 15V yet eBay sellers say 12V, and for $2 for one it isn’t going to be excellent quality, and even though it is ridiculously inefficient it works fine here – mechanical refrigeration isn’t really the point here.

  9. Im still searching for small and affordable compression coolers because they are so much more efficient than these peltiers… there are some that have the size of a beer can like “LT18DC24H” but 200$ from china is waay too much…
    Does anyone know where to get suhc small compressors cheap?

    1. Those are pretty cool, but it looks like you still need to add evaporator / condenser, pump it down, and add freon.
      Those mini-fridges of the 2 foot cube size are pretty cheap, you may be able to salvage the compressor from one of those.

  10. Ha! Pretty cool. I did almost exactly the same thing during my grad school project to keep a set of load cells and the associated circuitry at a constantish temperature. Neat to see this!

  11. Dumb question for the smart people that read hackaday-How would aluminum or copper anti-seize work for thermal paste-especially on things you weren’t afraid of shorting? Just curious, cause I never have thermal paste, but always have anti-seize.

    1. It apparently works pretty good, Google and you can find reports from some folks who tried it on CPU heatsinks. Though I saw some concerns about the compounds’ tendency to separate. If the mineral oil were to eventually ooze out leaving a dry joint, it wouldn’t work so well anymore.

      Almost anything works, really. I once reattached a stock Intel heatsink to a CPU using axle grease. It actually kept the CPU a few degrees cooler than the compound that originally came preapplied on the heatsink, and continued working without degradation for a year. Peanut butter? That’s a new one on me, but I’m sure it worked fine, at least short-term.

      Though if you don’t feel like experimenting, as [Brian] said, a tube of cheap heatsink compound is $4. And if you want top efficiency, don’t experiment, get a good tried-and-true compound.

    2. I would be more worried about how conductive anti-seize compound would be to electricity than heat. Most all heatsink compounds are designed to be non-conductive (depending on the potentials involved of course).

      There are a lot of people who have done experiments with different things in the CPU/GPU overclocking world, as getting rid of heat is the biggest problem when it comes to running things at higher clocks.
      Deliding the heatspreader off of certain Intel processors is apparently a thing and intel is known for using sub-par heatsink compound. Most people seem to switch it out for gallium based stuff, which is more or less “liquid metal”. The risk is shorting out exposed devices next to the die itself.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s