Liquid Cooling Keeps This Electronic Load’s MOSFETs From Burning

Problem: your electronic load works fine, except for the occasional MOSFET bursting into flames. Solution: do what [tbladykas] did, and build a water-cooled electronic load.

One can quibble that perhaps there are other ways to go about preventing your MOSFETs from burning, including changes to the electrical design. But he decided to take a page from [Kerry Wong]’s design book and go big. [Kerry]’s electronic load was air-cooled and capable of sinking 100 amps; [tbladykas] only needed 60 or 70 amps or so. Since he had an all-in-one liquid CPU cooler on hand, it was only natural to use that for cooling.

The IXYS linear MOSFET dangles off the end of the controller PCB, where the TO-247 device is soldered directly to the copper cold plate of the AiO cooler. This might seem sketchy as the solder could melt if things got out of hand, but then again drilling and tapping the cold plate could lead to leakage of the thermal coupling fluid. It hasn’t had any rigorous testing yet – his guesstimate is 300 Watts dissipation at this point – but as his primary endpoint was to stop the MOSFET fires, the exact details aren’t that important.

We’ve seen a fair number of liquid-cooled Raspberry Pis and Arduinos before, but we can’t find an example of a liquid-cooled electronic load. Perhaps [tbladykas] is onto something with this design.

21 thoughts on “Liquid Cooling Keeps This Electronic Load’s MOSFETs From Burning

  1. A good friend of mine has an even less conventional water cooling setup for his load testing. Its 1800watt static load, using 50watt resistors. Bolted to an old aluminum toyota cylinder head, thats sitting in a kids little red wagon with the garden hose running over it. Its ghetto but it works great for testing power supplies for large sustained loads. I think he built it in 2011 around the same time TiN (of evga and K!NGP1N fame) was building a dynamic tester.

  2. When I need some relatively low voltage load for testing, I string together some resistors form old TV’s until I get the needed resistance, then solder on two long leads and drop the whole contraption in a bucket of water to make up for the totally inadequate power rating of the used resistors.

    1. I have an old tv with a power supply like that. Just resistively drops the mains voltage with an absurdly wasteful string of gigantic ceramic-encased power resistors. Barbaric.

      You could dunk it in oil instead and get more voltage range out of it.

      1. Heh, I had a kluged up motherboard for socket 7s, it almost had a split plane supply, it was supposed to be future upgradable when the MMXes came out, but nobody ever had the module in stock if it ever actually existed. Anyway, big 10W resistor dropped the 3.3V enough for a Cyrix MX-166 not to cook. It bumbled along okay, was really to keep the kids off my mighty K6-333

        1. OT, but: The best CPU I ever had up until that point was a K6-450. Then it was the Phenom II X2 that I unlocked to X4, and now it’s the Ryzen 5 1400. I’ve always been an AMD fan, and the Ryzen’s were worth the wait. Thanks for the reminder about the K6’s. They were a great processor!

          1. That was a great time for personal computers. I had a K6-2 300 and a K6-2 450. One running Windows for gaming and one running Debian for my little forays into home servers. They were cheap to build and ran great for years.

  3. Interesting re-use of a liquid CPU cooler. But if on a budget and a used one is not available then an alternative “liquid” heatsink could be something similar to the HEATHKIT HN-31 Cantenna; That popular kit was built from a gallon paint can filled with mineral oil.
    https://beachpackagingdesign.com/boxvox/the-heathkit-cantenna

    The original design had a 50 ohm RF load rated at 1KW. But feel free to change the load value or paint can size to satisfy your requirements.

    – Thomas

    1. The best water cooler I made was a repurposed tea’s maid.
      The heating plate had a built in water pipe and temperature sensor. Fairly trivial to feed it with cold water. A lot of sanding to get it flat enough to sit well on a processor. But getting it clamped down evenly was the tricky part.

    2. Initially the idea here was to repurpose a stock aluminum heatsink I used for an old load, tap it, and use thermal paste as the interface. However, that one and most I saw online only went to 90W TDP ratings. I looked at copper-based coolers with heatpipes or a vapor chamber, but I really was not comfortable soldering a closed gas pipe with the possibility of damaging things. Hence, this solution!

      A bucket of water or mineral oil would perfectly suffice with a tiny pump and heatsink. I was able to score the AiO (Corsair H55) for about $30 shipped. It’s also possible with AliExpress/eBay watercooling gear, but I really hate to make a bad loop.

  4. > This might seem sketchy as the solder could melt if things got out of hand
    Solder melts above 200C and is beyond boiling point of water at 100C. MOSFET probably rate at 125C
    Melting of solder is the least of your problem as the tubing would burst first spilling hot water and steam if the person operating it is asleep.

    Dan should learn some basic science before writing nonsense comments to sound cool.

      1. There is no way to get >~180C on the thermal interface between the mosfet and the copper heat exchanger while there is still water present on the other side. The thermal resistance will not be high enough to result in a 100K delta T. It would require a thermal resistance greater than ~0.3K/W. For a copper plate like this it would mean a thickness of a few centimeters…

    1. Dan was actually thinking of a failure mode where there was a leak in the plumbing. Pumping all the cooling fluid out of the system would make the boiling point of water irrelevant in that case, leading to a thermal runaway that could easily exceed the softening point of solder.

  5. I needed a 10KW 220V AC load to test my generator. MEP-803 (look it up, you really want one and sometimes you can find them cheap!) load banks are slightly less expensive than unicorn tears. Turns out you can buy a 1200 watt ceramic heater from ama$on for $20. 1200*10 = $200 and I have the most useful load bank in the world, it doubles as a Y3K emergency heater. Love non conventional solutions to everyday engineering problems!!

  6. Some Questions on the soldered connection

    1. How was it made? Propane torch?
    2. What kind of solder?
    3. Why not cpu thermal paste and a metal plate to sandwich it?

    The soldered joint is the coolest part of this hack, insights please :)

    1. Hey! I have an 858D clone hot air station that I used to solder the MOSFET to the plate. The procedure was pretty simple, I took the copper cold plate off of the AIO, removed the gasket, applied some solder paste, and heated it up like any other PCB. It’s just 60/40 solder paste too.

      Thermal paste was debated but that would require me to drill into the AIO (wrecking the coldplate fins) or making a weird mounting contraption that may or may not work. Soldering the MOSFET on was perfectly fine for me because I could just throw it on, and if it broke, it was relatively trivial to fix/replace with my setup.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.