High power MOSFET switching

[fenclu]

Hi,

I am designing a high power MOSFET switch for a battery disconnect. I have my design requirements set at around 90 V and 150 A. I realize, that the continous operation of the MOSFET would require a proper thermal design, i.e. a good heatsink and low Rds(on).

The trouble I have now is switching the goddamn thing. The switching speed doesn't actually matter all that much, I need to find a good balance between switching losses and inductive spiking.

I've tried the gate driver IC NCP81074, with the typical circuit schematic, driving the gate at 12 V (gate charge and discharge resistors both equal to 2.7R, then i changed them to 10R, didn't seem to make much difference):
https://www.onsemi.com/pub/Collateral/NCP81074-D.PDF

I tried the circuit with several MOSFETs - IAUT300N10S5N015, IRF150P220 and the IXFH220N20X3, all of which should be easily able to switch this sort of load. I don't have a PCB made yet for this project, I am using a combination of a bare copper laminate with a prototype board for the gate driver circuit, the whole thing looks like this:

I know, the grounding for the gate driver looks horrible, but the actual gate drive waveform looks quite nice.

Anyway, I am loading this transistor with a high power PSU switching it at 40 V and 40 A (1/8th of the required load). If I dare to go closer to 50 V Vds, the transistor fails short after turn on.

The turn on waveform looks like this:

I can see that there is some body diode reverse conduction, the Vds dips below 0V for a brief moment, then it spikes up and swithes on very slowly. I am not 100 % sure what is causing this effect, possibly the parasitic inductance of the PSU? Remember, after cranking up the Vds, the transistor fails short on turn on.

The turn off also looks terrible with a huge spike of the Vds (above 100 V):

Thing is, this spiking looks a bit different to what I saw online, usually the spikes settle down on the high Vds voltage (the transistor is off by then). However, here the spikes are damped after a short time and only then the Vds starts to rise (it is not visible on the images).

I managed to cure the spikes by adding a 1 uF capacitor between the drain and source, but also radically increasing the turn off time:

This circuit will be quite safety critical, since the control circuit has to be able to disconnect the battery at full load, the bad thing about MOSFETs is that they usually fail short. Basically, I am trying to build a semiconductor contactor (a semicontactor? ).

I have thought about using multiple MOSFETs in parallel, however, for now, the big problem seems to be the Vds voltage, not the drain current. I am leaving the thermal design in continous conduction for later, first I need to have good confidence in the switching circuit.

TBH I am not sure what is causing these effects, since altering the switching speeds doesn't seem to do much to this effect (could some sort of snubber help?). I never had to worry about damaging MOSFETs, since I was switching a couple of watts at best. Also, I intend to measure the drain current, I should have a current probe available in a few days time.

[PartialDischarge]

Hi,

There is something very wrong with that circuit or with the measurements, ie makes no sense for the Vds not to rise as soon a Vgs is 0.

Anyway, I can tell you one thing for sure, you are missing the most important parameter which is the current Id. You absolutely need to measure it if you want to make anything work, for example the PSU transient response is not current limited, so you may be well switching 200A peak at turn on with 40V, you won't know until you measure it. You cannot develop or troubleshoot a power system without measuring the currents

[fenclu]

Damn, of course, I didn't take into account the fact, that the PSU has a software current limit.

Probably time to get some proper high power load and a big honking battery.

I'll try to get hold of the current probe and come back with some results.

[PartialDischarge]

I'll try to get hold of the current probe and come back with some results.

My guess is that those mosfet you've used are all fried. Measure the current and start little by little with the voltage, you'll be surprised at how fast current rises.

[hoebagg]

Have you had any more progress with this?
I'm thinking about making a similar device to switch on batteries for an e-bike project
The batteries provide 50V and draw as much as 180A. (a few minutes of ride time, stunt bike - not your traditional e-bike.)

[sandalcandal]

Wasn't here for when this thread was first posted but agree with MasterTech's assessment that the issues with failure are likely due what to seems to be a test PSU being used without a dummy load and relying on the PSU's transient current limiting.

For this sort of battery safety disconnection it is much better to use a contactor. Lower loss and much safer failure modes particularly due to MOSFETs failing short circuit. The benefits of MOSFET based "disconnection" (won't qualify as disconnection in safety critical applications) is cost (at lower voltages), packaging and maybe resistance to shock. If you can fit contactors into your application then use those.
Example https://www.digikey.com/en/products/detail/te-connectivity-aerospace-defense-and-marine/LEV200A5NAA/2362828

[jonpaul]

  design

Assume the load has inductance and stores energy.

Every motor and solenoid in a car has substantial energy storage. Try to break the connection with a relay or knife switch: You will see an arc.

any disconnect must redirect the stored inductive energy.


Impossible to comment further on the waveform  until you can get a well specified current probe.

Check your scope probe's ground connection, seems some ringing due to ground lead L.

Jon

[hoebagg]

I agree with the safety implications of MOSFETs failing short.
I was thinking more along the lines that contactors being mechanical will also eventually fail and degrade over time due to arcing.

I'm currently trawling through this thread https://endless-sphere.com/forums/viewtopic.php?f=3&t=40142 where someone has made something similar using parallel mosfets but i'm yet to discover any math behind their mosfet and component choices.
Following the thread so far it seems there is a lot of trial and error which I'd like to avoid.

[sandalcandal]

I agree with the safety implications of MOSFETs failing short.
I was thinking more along the lines that contactors being mechanical will also eventually fail and degrade over time due to arcing.

Failure and degradation of the contactor is quite unlikely to be an issue, particularly since I'm guessing this is just a safety cut-off so shouldn't be switching under load regularly nor will be be switching on and off constantly. The previously linked contactor has a rating of 10,000 cycles at 120VDC 500A so the cycle life should be much higher for your application even with a "dumb"/barebones implementation though improvements can still be made by making sure any inductive kick or capacitive inrush is managed properly. Even then I think a contactor is going to be far more robust, reliable and longer lived than a solid-state solution for this application.

I'm currently trawling through this thread https://endless-sphere.com/forums/viewtopic.php?f=3&t=40142 where someone has made something similar using parallel mosfets but i'm yet to discover any math behind their mosfet and component choices.
Following the thread so far it seems there is a lot of trial and error which I'd like to avoid.

That's quite a large thread so unless you can provide some specific key points, then might not be able to give much feedback. Not to be disparaging but solutions from hobbyist forums are a bit likely to be less "professional" so you'll probably see some disconnect in the reliability of solutions proposed here compared to on Endless Sphere.

It looks like they were trying to use MOSFETs to do precharging by operating them in a linear mode. I think that's a poor idea since you'll need to have MOSFETs that can handle a large transient of power for a short time during the precharge so an overall reduction in cost and complexity is unlikely compared to the "normal" solution. What's normally done is a low value resistor is used to limit current that way the resistor is being loaded rather than your switching device so you can use a small MOSFET or relay without issue to do precharge. You can then switch on your main disconnector and switch off the precharge circuit after the precharge is complete. The "Solid State Precharge CoolMOS solution for automotive HV battery" posted on page 30 of that thread you've linked is a MOSFET doing PWM on a precharge resistor which is similar to what I've described but the PWM seems unnecessary and will require a beefier MOSFET as well as producing lots of electrical noise.

As for design methods, the main thing is calculating the power each MOSFET is being made to dissipate and making sure that's within spec of the MOSFET. You'll just need to read datasheets and do some basic calculations, if unsure how then some basic MOSFET tutorials should sort you out (ask if you want some suggestions). I think I saw some "design by sim" in that thread which can be ok but you need to ensure your simulations and input values/models are setup realistically.

Edit: Perhaps some reduction in cost would be possible if the precharge MOSFETs were also used for "disconnection". Would need to check if the power dissipated during precharge necessitates an upgrade in MOSFETs used.