The cooler box
An example of a portable cooler box, a popular DIY project
In a previous article, a basic circuit for a cooler box was introduced, which gave only lukewarm results at 11K below ambient, or 14°C in a 25°C room. That said, it’s worth to note that it did so at a relatively modest input of just 13.5W (4.5V / 3A).
This article looks various ways to reduce the temperature inside the box, and with the main objective of learning more about the effects of thermal design. Five different options are given, the best being saved to last.
The full circuit of the thermal cooler box is shown below, and as in the previous article, it starts out with a single TEC 12708 (8A) run at 3A, large heatsinks and fans on both the hot and cold side, and medium sized cooler box. The values in the circuit are typical for this kind of construction. The only difference from the previous article is that the outside points are connected to 25Vdc supplies, which simulates a 25°C ambient. If the use of electrical components to simulate thermal circuits is new, check out this article first.
Circuit based on the previous article, with voltage for temperature, heat flow as current and thermal resistances as resistors. At 3A, the expected temperature inside the cooler box is about 14°C, with 10.8W of cooling and 13.5W input.
Option 1: increase the current
The instinctive response here is to increase the current and work the Peltier harder. At the moment there’s only 13.5W of waste heat, and the heat sink is fairly large and should be able to get rid of lots more heat. But the numbers tell a different story. On the left side is the circuit showing the values from 2, 3, 4, 5 and 6A, and the right side shows a graph with the chamber temperature (inside the cooler box) and “waste” heat:
Despite heaps more power being applied, we still have lukewarm beer!
The problem here is that the waste heat has to travel through the hot side thermal elements, in this case the 0.9K/W for the Peltier itself, and 0.5K/W for the heatsink and fan, giving a total of 1.4K/W. This then reduces the temperature difference that is available for pumping. For example, at 3A, we have a core temperature difference of 60K, but we loose 18.9K from the 13.5W of waste heat that flows through 1.4 K/W on the hot side. The remaining temperature difference available for pumping is then 41K of the original 60K. In contrast at 5A there core temperature difference of 100K, but now we have 37.5W flowing through 1.4K/W reducing the available temperature by a whopping 52.5K, and leaving only 47.5K for pumping. It’s better, but it’s marginal improvement for nearly 3 times the power. At 6A we have a huge core temperature difference of 120K, but we are losing 75.6K from the waste heat, leaving only 44.4K for pumping.
Option 2: An extra Peltier in parallel
One of the common solutions is to add a second Peltier in parallel, since it seems obvious that two is better than one. In the thermal circuit, it’s possible to add a second Peltier in parallel, but to keep the circuit simple an easier approach is to modify the elements to give equivalent results. In this case, Xc and Xh will halve, the waste heat will double, while the core temperature remains the same. We can then run the same simulation from 2A to 6A:
This result seems counterintuitive. We do get a small improvement in the 2-3A region, but it’s hardly worth the effort for adding the extra Peltier. And a 4A and above things not only get worse but useless, despite using significant power to drive the system.
While we do benefit from the lower effective resistance of the Peltier element (since there are two in parallel), we now have double the waste heat which is being funnelled through the same heatsink, which ends up reducing the available temperature. This is very critical point: while paralleling Peltiers can be a good strategy it only works if the hot side heat exchange has very low thermal resistance. This is not just the size but actual constructions that have a low “K/W” figure.
Another downside of paralleling Peltiers on the same heatsink is the risk of a bad thermal joint since there are now four interfaces between Peltiers and metal, and keeping that flat without getting thermal paste everywhere or trapping air bubbles is difficult. Any additional thermal resistance in the circuit will further degrade performance, which wasn’t great to start with.
And if it’s not already obvious, adding more Peltiers in parallel is just going to make things even worse. Unless the bottleneck of the hot side heatsink is dealt with, the extra waste heat from each additional Peltier keeps adding to the pain of lost efficiency.
Option 3: Extra Peltier and hot side heatsink and fan
To avoid the problem of funnelling the waste heat of two Peltiers through one heatsink, and obvious solution is to double the heatsink/fan set up as well. Since it’s difficult to do this on one side only, it makes sense to duplicate the whole set up, essentially two Peltier systems in the same cooler box. Again, we we can simulate in the thermal circuit by just halving Zc and Zh, in addition to the previous changes. In this case we finally see some beer drinking temperatures, and around 4A seems to be ideal Peltier current to drive this system:
While the temperature is looking good, we need close to 48W to get this, which is not so great given the small size of the cooler box. The actual cooling power is 19.3W, so the COP for cooling is fairly weak 0.4. Plus we have the cost of two complete Peltier set ups, and four fans.
Option 4: water based heat exchanges (single fan type)
Water based heat exchanges have much lower thermal resistance, with single fan types around 0.1K/W. Once a water based system is used, it’s very easy to add a second heat pump “cell” in the circuit (second Peltier with two aluminium water blocks on each side). The actual water based system will be introduced in another article, for this article we will just see how effective this is. The following simulation is based on two Peltiers with 0.1K/W heat exchanges on both the hot and cold sides. By playing with the current, it was found that 2.7A was enough for 5°C in the cooler box with a 25°C ambient, with just 22W input.
In the above circuit, the big change is Zh and Zc going from 0.5 K/W (500mΩ) to 0.1 K/W (100mΩ). Even though the Peltier’s internal thermal resistance is significant, reducing the thermal resistance of the outside components allows us to get closer to the raw capacity of the Peltier itself.
Option 5: Better insulation
If the main Peltier system is running efficiently, the main limiting factor is the cooler box insulation. At the start we assumed 25mm thick cheap foam to arrive at the 1K/W figure. If we increased the thickness to say 40mm and used a better material, we could double this to 2K/W. Although this would seem to reduce the cooling power (due to the increased overall circuit resistance), we only need half the power for the same temperature, meaning we can start to see some really cool results:
By playing with the current, the it was found that a current of just 1.82A was enough, with the total input of just 9.94W, running at a COP of 1.0.
Although a COP of 1.0 doesn’t sound great, the actual power of just 10W is amazing. The circuit highlights how in a cooler box application, the main issue is the insulation. With good insulation only a small amount of power is required to overcome the heat losses, which then makes a Peltier perfect for the application. But to get the best out of a Peltier, the heat exchanges must have low thermal resistance, in the order of 0.1 K/W, especially on the hot side.
The caveats
The above models are accurate except for one point: the thermocouple effect which is (or will be) discussed in a separate article. Briefly, the thermocouple effect makes the Peltier appear to have a higher electrical resistance, which not only affects the input power, but also reduces the available temperature difference for pumping. In the final solution (Option 5) this effect will be minor, since the power isn’t large to begin with and the thermal resistance on the hot side is lower. But for other options, the thermocouple effect can result in significantly lower performance than is shown.
Finally, the simulation assume that good thermal contact is made between the Peltier and the heatsink or aluminium water block. Experience indicates that this can be hit an miss, sometimes 0.1K/W, sometimes 0.3K/W. Although these numbers seem small, they can have a big impact in the performance. One of the benefits of a water based system is that it’s (relatively) easy to measure the “success” of the thermal joint between the Peltier and the aluminium. The high degree of variation from this testing is what lead to the “milling solution”.