My electronic load failed. I was trying new hardware and software options. At a certain moment the power MOSFET died. There was no magic smoke, but the source, drain and gate are all shorted (source, drain and gate all measure a few Ohms in both directions). I am not 100% sure of what has happened, but I think that a bad contact at the gate or current sense resistor, while the FET was supplied with 40 V, did it. No sound, no smell. Just an instantaneous death. In this post, I'm checking what's been damaged and - hopefully - how to fix it. Depending on what's defect, this may become an interesting or boring blog. That's how it goes with repairs. (also a shoutout to dougw: this may fit into an e14 repair group)

Known Defects

The power MOSFET is definitely destroyed. The three pins all measure close to short - while they should be very high resistance.

The fact that the drain is almost shortcut to the gate is a worry. High power may have been injected into the control circuit. Let's see...

My process is to first fix the known defect components, then slowly test all other components starting from the source on.

(Sometimes - in particular with low power defects and circuits where most of the functionality is working) I do the opposite. Start from the output and trace back to where the signal is intact.

In this case, I risk blowing the whole thing again when the control circuit is unstable. So I start from the beginning, with the power part disconnected.

Replace the Power Mosfet

The FET is known to be bad, so I have to replace it.

The package is soldered on a big copper fill. With vias, it's connected to an even bigger copper plane on the back side. That plane has the size of the heat sink that's mounted on the backside of the PCB.

To remove the device (the picture above shows the new MOSFET that I placed during the repair exercise) you have to get the whole copper plane area above the melting temperature of (in my case lead-free) solder.

To do that, I first removed the heat sink (and the fan that's mounted on it).

Then I placed the board above a pre-heater for a good 10 minutes. I drowned the FET in flux.

I then removed the solder from the two pads using a 400° C heated solder iron and desoldering braid.

I turned my hot air gun to 500° C (for fast rework without hindsight) and removed the FET.

Then I cleaned up the pads with the iron and the braid. I cleaned the pads afterwards with isopropanol - to remove flux and burn residue.

After that I re-tinned the big copper pour - just enough to have solder plating on it. (I am using non-leaded solder).

I also tinned one of the two FET pins (the right one).

Then, I placed the FET and soldered it in place by heating that right pad.

Next, I properly soldered the left pad and reworked the right one too - to have decent quality joints on both.

I re-applied a royal amount of flux, applied solder with a 400° C iron around the FET and over the edge of the metal back plate, then heated the big pour on the front side with 500° C hot air until the solder became liquid and seeped around and under the FET.

I let the whole area cool down to the pre-heat temperature, and turned the board around.

Then I bombarded the back copper poor (where the heat sink should fit snugly) with 500° until all solder residue had flown through the vias to the other side of the board - ensuring a good thermal bridge from FET to heat sink.

When the back plane was level (no solder bumps left) I let the whole module down.

I re-applied thermal paste and put the heat sink with ventilator back in place.

ADC Board Part 1: Voltage Reference

Because the resistance between gate and the two other FET pins were shot, anything could have happened to the upstream circuit.

I have to check the DACs, ADCs, voltage reference IC and input activation switch (and some other silicon -  I don't believe anything passive has suffered).

The results of monitoring the reference can be seen below.

I have powered the IC from the LaunchPad that controls the load.

In normal circumstances, it's powered from the load's control board, but I haven't tested that yet.

image: capture of the voltage reference. It has been behaving good for two hours (between 2.64230 and 2.64130 V - good enough for Belgium)

In the next post, I'll test the remaining parts of the ADC/DAC board.