How did I fry these MOSFETs ?

[MathWizard]

I have some plate steel that I thought might be ok as a heatsink in a big load tester, each 1 is 12"*12"*3/8", and it was easy to drill through, as some on here said.

So I was just trying to see how much heat, 1 plate could handle with 1 mosfet. So right in the centre, I screwed on an IRZ44N, with some thermal paste under it. The plate was sitting up on some wood blocks, the fet and resistors were on a PCB, 90deg to the plate, nothing shorted.

They are rated 55VDSmax, 16VGSmax, 47A, JtoC=1.4C/W. I got them from Tayda-electronics., not their ebay site anyways.

I've tested similar mosfets on a much smaller HS before, with 30-40W, so I figured I'd try ~50-60W, by applying 30V@2A, from my Siglent 3303X-E PSU, hacked to the 1mV-1mA version. Test circuit is below.

So with me doing the gate regulating, starting at 2.5V, and getting it running up to 2A, as it heated up it needed less VGS, and the PSU max current set to 2A, the PSU was voltage limiting to ~25-28V @2A.

Anyways it was doing ok, ran for about 5-10min, I kept turning the gate voltage down a few mV at a time, under 4V, just trying to get the DS voltage to stop limiting lower and lower. If I dropped the VGS way down, it's the current that kept creeping up, so I just let it sit at 2A and Vlim.

The edges of the plate were still cool, but right near the chip was getting too hot to touch, the average wattage from the PSU was ~55W. So I think it's not a very good heatsink that way, and I might have just been cooking it already. But I put a fan over it 1/2 way through.

I turned on/off the gate voltage 1-2 times. Then when it came back up, DS around 26V, the PSU changed trans. taps for the DS rail, switching a relay. I think it went to the tap for under 22V or 24V as it voltage limited, then immediately, it limits to 1V @2A....because the mosfet fried. R-DS is now 0.5ohm, GD and GS are now 66ohm.



So what did I do wrong? Maybe overheat it, the steel is not a great thermal conductor, 55-60W is getting up there. The center of the plate was burning my finger, but a few inches away was just fine, and almost roomtemp on the outer edges.

Or are these counterfeit fets? Or was it the relay, and back voltage, but surely the PSU has snubber dioides, and the fet has a body diode.

Maybe turning on/off the gate, with only 100k to gnd, was bad ? The design I'm going to copy, uses that ~1k, I added the 100k, to allow the gate to discharge better. It's the Array 3700 on the bottom of the site below. Earlier I fried another 1, without having that 100k there, and the gate was not discharging either, when turned off. But with a 1k between the gate and PSU, that should handle anything with the PSU at ~4V.



http://www.kerrywong.com/2018/11/05/teardown-of-an-array-3711a-300w-dc-electronic-load/

[magic]

At the very least, you should have calculated what maximum case temperature the FET can take with 55W of dissipation (thermal resistance specs) and then monitor case temperature during the experiment.

Another option is to monitor junction temperature by connecting reverse voltage through a current limiting resistor and, during the test, briefly disconnecting the PSU and measuring forward voltage of the body diode. This measurement would need to be calibrated beforehand, for example with boiling water. And I'm not sure how briefly is briefly enough, maybe it would require some electronic switch and a fast sample-and-hold.

[MathWizard]

Ok I never heard of that before, I do have thermocouples, and actually a cheap IR sensor, I never thought I'd need them.

I want to try again, but with a lot less power, I'm still hoping I can get 4 fet's on 1 plate. But yeah I probably just cooked them.

I should breadboard an op-amp section too, for some control, with a shutoff. I made a simple NTC shutoff before.

[thm_w]

Relay in the PSU is not an issue yeah.
You could measure the Rds on when you are doing future tests to verify its near the rated 18mR, but that would not be entirely conclusive. Tayda should be reasonable reputable.

One issue is the IRFZ44 has no rated DC SOA, so there is no real safe spec for the way you are using it.
Of course, the heat probably didn't help, so getting an idea of the FET case temp as mentioned would be a good idea.

https://www.infineon.com/dgdl/Infineon-IRFZ44N-DataSheet-v01_01-EN.pdf?fileId=5546d462533600a40153563b3a9f220d
https://electronics.stackexchange.com/questions/361407/dc-operation-for-irfz44-mosfet-soa-curve-max-current-at-40v

[MathWizard]

Ok, yeah and I got these mosfets, just because they were cheap. The main PSU's I want to test are some computer PSU's, so LV anyways, but super high currents from the 12V.

I better get a thermocouple or 2 on there today when I try again......and start low...I don't want to waste anymore of these mosfets.

Or I need more expensive ones, more suited to the role. In those threads, it doesn't sound like they can handle much of anything I wanted to use them for. I bought 20 of them. Oh well.

[MarkF]

IMHO your overall design (in your PDF file) is a mess.
Also, I would use a MOSFET in the large TO247 package (for example a IRFP250) for larger heat dissipation.
You might want to take a queue from the HP6060 Electronic Load:

[mindcrime]

Steel has very low thermal conductivity, especially compared to materials like aluminum or copper that are more commonly used for heat-sinks. See:

https://thermtest.com/thermal-conductivity-of-steel

Steel has one of the lowest thermal conductivity values of all metals, making it an ideal material to use in high-temperature environments such as vehicle or airplane engines. Thermal conductivity describes the rate at which thermal energy is transported through a material.
...
Poor thermal conductors, such as steel, carry heat very slowly and are ideal materials to use as insulators.

Emphasis added. Note that an insulator is the exact opposite of what you want here.


I suggest re-running your trial with some thermocouples taped to your MOSFETs and see what actual temperature values you're getting on the IC itself. Even if it means risking burning up a couple more in the process... at least then you would know if temperature is what's killing you or not.

[WatchfulEye]

Heatsinking is critical when operating at high powers. TO220 are not ideal for high powers, because their small contact patch means that good thermal contact and excellent thermal conductivity of the heatsink material are critical and dissipation capability of the heatsink is very important.

If you want to dissipate 60W from an IRLZ44N and keep junction temperature safe, the heatsink temperature must be kept below 60C at the point of attachment. Even a copper block the size of your steel plate won't achieve that, it just doesn't have the surface area, unless you have a lot of forced airflow. Additionally, because steel is such a poor conductor, even a 10 mm thick plate has minimal spreading capability.

I played around a bit with a heatsink calculator https://www.heatsinkcalculator.com/free-resources/flat-plate-heat-sink-calculator.html and came up with a junction temperature of approx 240 C for your setup. Although the IRLZ44N as old 5th gen HEXFET planar FETs are pretty tough and even though not DC rated, are fairly tolerant to DC loading - but not once you go outside the temperature limits.

If you want to deal with the power levels needed for a PC PSU tester - then you need to use large high surface area heatsinks (preferably copper base) with plenty of fan assisted air flow.  You should also use MOSFETs with very high continuous power dissipation rating in the largest practical package (i.e. TO247). The IRFP260N and IRFP260M (they are basically the same part from different factories) are good choices and reasonably available, and they're also IR 5th Gen planar FETs so pretty tough, and the high voltage rating means that at 12 V you'll be miles away from any thermal runaway zone. For best reliability, aim for junction temperature somewhere below the max - maybe 135 C. So, if you're aiming for 100W loading for each MOSFET, then aim to keep heatsink mounting point temperature below about 60 C at 100 W - this should be achievable at 100 W with many off-the-shelf PC CPU heatsinks intended for modern AMD or Intel CPUs. To aid mounting, look for ones without heat pipes integrated in the base, so that you can safely drill and tap a mounting hole (Something like an "AMD Wraith Spire" looks to be readily available used on ebay, and looks generally suitable).

For higher powers, you'll need to parallel several MOSFETs. When doing this, make sure that each MOSFET has individual current control. Use a current sense resistor on each MOSFET source, and an op amp to drive the gate. The Schematic posted by MarkF is a very nice example of how to do this.