PEL-8150 Electronic Load Burning up MOSFETs Under Test. Why?

[byte]

I received a MATRIX PEL-8150 Programmable DC Electronic Load for xmas this year. It's great for testing DC power supplies, but for some reason it seems to kill switching MOSFETs and I have no idea why.

I typically just use CC mode on the PEL-8150 because for some reason CR mode seems a bit wacky on the PEL-8150: If I set it for a specific resistance and test with my meter the tested resistance value is not the same as what is configured on the PEL-8150. Probably a separate issue, and I am assuming it's because the PEL-8150 only approximates a resistive load in this mode - but this could be related to burning up MOSFETs so I am mentioning it.

What works is applying a DC voltage to the MOSFET gate first, then enabling the PEL-8150 load input via the menu. Afterwards I can increase CC current value up or down no problem. But if the PEL-8150 load is on before applying voltage to the gate, and then I apply DC voltage to the MOSFET a bunch of clicking may happen in the PEL-8150 - and before I know it the MOSFET has had way too much current pass through it and lets out the magic smoke.

This is of course dependent on the amount of current I have supplying from the power supply, but I wouldn't expect the PEL-8150 to over current the load like this just because the MOSFET switches on. Don't even get me started on PWM output from MOSFETS, that's how I first discovered this interesting feature when testing a PWM motor drive.

Furthermore, in the working case, if remove DC voltage to the MOSFET gate then the PEL-8150 will do a bunch of clicking and disable the load input. No error or anything, it just disables the input for some reason.

I realize programmable electronic loads have active circuitry inside to do their magic - but I just wouldn't have expected this to happen. And because I don't understand why it's happening, I really have no clue as how to safely test MOSFETS (or who knows what else may be susceptible to this) safely using the PEL-8150. I was really excited to have at xmas because I thought it would make testing all sorts of loads much easier.

So any help, pointers, or advice on this topic are highly appreciated.

Thanks in advance!

[Kim Christensen]

Because of the active circuitry, the PEL-8150 isn't going to instantly respond when the MOSFET turns on. This means that there will be a higher current spike than programmed when the MOSFET first turns on. Does the manual say anything about the PEL-8150's transient response time?

[byte]

Does the manual say anything about the PEL-8150's transient response time?

Nothing I can find in the manual:
https://www.manualslib.com/manual/2430121/Matrix-Pel-8000-Series.html

Interestingly, I found on Youtube someone doing a nearly identical test as I was with a Rigol D3021a electronic load and had no issue when they switched their mosfet under test on and off:
Edit: I take this back, on a second view I noticed they had the gate tied to +12V and their switch was just breaking the circuit to the MOSFET's source. In my case, I'm applying voltage to the gate, not mechanically breaking the circuit to the source. Not what I want to do as I wish to test MOSFETS switching with PWM signals.





I also found a video from a different manufacturer, who said their electronic load will go to infinity impedance if enabled and disconnected from the test source, and this can cause a negative voltage into the test source:


Any other insights are welcome. I'm still unsure if I can just setup my load test of switching mosfets differently so I don't burn them out, or if I need a completely different electronic load like the Rigol which vaguely seems not to cause this problem.

[Kim Christensen]

A quick peruse of the manual shows that it has several different Constant Current modes:

4.1.1.1 Standard constant current mode
4.1.1.2 Load unload constant current mode
4.1.1.3 Soft-start constant current mode
4.1.1.4 constant current transfer voltage mode

Have you read through these? Not sure if the soft start mode will be fast enough for your tests, but it should protect the FETs better than normal mode.

[byte]

Have you read through these?

I sure have:

constant current transfer voltage mode

This mode is for testing batteries to ensure current draw doesn't continue below a minimum voltage. Not useful in my situation.

Standard constant current mode
Soft-start constant current mode

Both of these modes tend to burn up a MOSFET under test. However, I did find something interesting, if I use a relatively high PWM frequency of 100KHz, the MOSFET doesn't overheat and the electronic load draw current as if the MOSFET gate had DC applied. If I use something like 1KHz though, a bunch of extra energy is dumped into the MOSFET for some reason causing it to heat up quickly.

Load unload constant current mode

This mode does prevent the MOSFET from burning up, but doesn't draw the full load as programmed. Furthermore, the lower I set the cut on/off voltages, the more energy is dumped into the MOSFET. I can see this happening with my thermal camera. No matter what though, this mode is not drawing the programmed current from the DUT, it's way under by as much as an order of magnitude.

The soft-start and load/unload modes do seem to protect the DUT from excess current draw when just applying voltage to the gate, but they are not working as expected in PWM switching.

Anyone have any idea why a switching MOSFET would be dissipating more energy when switching compared to DC at it's gate? Is this some kind of reactive vs apparent power situation interacting with the electronic load?

[Kim Christensen]

Anyone have any idea why a switching MOSFET would be dissipating more energy when switching compared to DC at it's gate?

Usually that's mostly due to slow rise/fall fall times of the gate drive signal. So if the gate driver circuit cannot charge/discharge the gate capacitance fast enough, the MOSFET stays in it's linear region longer. This power dissipation will become worse as the switching frequency is increased. There are other causes, but that's the main one.

If you want to see what's going on, put a low resistance shunt in series with the source lead and scope the waveform across it to figure out what the current flow through the MOSFET is really doing. I would guess that your E-load  cannot stably operate at these switching speeds and is causing high current spikes at certain frequencies.

Personally, I'd be using a resistor (With snubber if wirewound) instead of an E-load, since the E-load will add it's own flavor to the rise/fall waveform on the drain.
If I had a curve tracer, that's what I would be using for testing MOSFETs and other semiconductors.

[byte]

If you want to see what's going on, put a low resistance shunt in series with the source lead and scope the waveform across it to figure out what the current flow through the MOSFET is really doing. I would guess that your E-load  cannot stably operate at these switching speeds and is causing high current spikes at certain frequencies.

Thanks for the idea, I'll try that.

Personally, I'd be using a resistor (With snubber if wirewound) instead of an E-load, since the E-load will add it's own flavor to the rise/fall waveform on the drain.
If I had a curve tracer, that's what I would be using for testing MOSFETs and other semiconductors.

Unfortunately, a simple resistive load is not programmable. I'm looking for the programmability so that I can vary the load through a range of tests as my power supply is also programmable and will be varying the voltage.

[Kim Christensen]

Unfortunately, a simple resistive load is not programmable. I'm looking for the programmability so that I can vary the load through a range of tests as my power supply is also programmable and will be varying the voltage.

Hmmm... I wonder if your E-load would operate properly in constant resistance mode? I remember having problems with an Agilent E-load in CC mode but it worked better in CR mode for the test I was doing. Might be worth a shot.
(I know you said your ohmmeter disagreed with the E-load's CR mode setpoint, but probably at higher currents it'll work just fine)

[byte]

I gave this a bit more thought, and I considered maybe all I need is a capacitor on the output in order to prevent the e-load from seeing large transients. I started drawing the circuit below, adding a diode to prevent reverse voltage to the MOSFET so I could pull the drain low when the gate is off, and a bleed resistor to discharge the cap over time.

But then I just realized now that the cap is going to pull as much current as possible from the MOSFET unless a resistor is in series, at which point the E-LOAD serves no point.



This was after a long day of work, so my brain is not cooperating at the moment. Anyone have any other ideas on how to protect a switching MOSFET with an E-LOAD, while still preserving the point of using an E-LOAD to control and adjust the load?

Thanks.

[Manul]

Is it N channel mosfet connected as high side switch? How you drive it? You said you just apply DC voltage, so it implies that you are not bootstrapping or doing some level shifting. Then this is wrong. I suggest you post a schematic of your original setup which is burning mosfets.

Regarding the capacitor idea itself, it is bad because capacitor will represent a low impedance at mosfet turn-on, causing huge current pulse, which might by itself overload the mosfet.

[byte]

I don't think a schematic is required as it's pretty straightforward. A MOSFET setup in a high-side configuration with the source tied to 5-24V, and the gate tied to 3.3-5V. It's illustrated in the above schematic without all the other stuff.

Regarding the capacitor idea itself, it is bad because capacitor will represent a low impedance at mosfet turn-on, causing huge current pulse, which might by itself overload the mosfet.

Yep, that's what I said and the same realization I had.

I could put a resistor in series with the drain, but then the e-load is pointless. I suppose I could put a resistor in series with the capacitor - but then it will limit the amount of current transient smoothing. But if the e-load is reacting this way due to the voltage (and not the current) transients, I suppose it could still help.

I wouldn't think I would be the first person to try and test switching MOSFETs with an e-load though. Unfortunately, Google searching is of no help because nearly all e-load use MOSFETs, so all you get are hits on how to build a e-load with MOSFETs, not how to test MOSFETs.

[byte]

So I had another idea:

To prevent the e-load from seeing the current/voltage transients and over-reacting, what about feeding an inverted PWM signal to another MOSFET? My current/power readings on the e-load will be incorrect, but I can just divide the measurement by the duty cycle of the PWM signal. This way the e-load gets a constant source, but the MOSFET under test can still be loaded under switching conditions.



It would be great to avoid this, so I can get accurate power/current readings. But I'm not really seeing another way at the moment.

[Manul]

I still don't get your schematics. You draw N-Ch mosfet with drain to +24, load on the source and drive it with 5V PWM? How that's supposed to work?

[Kim Christensen]

So I had another idea:
<SNIP!>

Your test circuit should look something like this (Figure 11) with the E-Load (RL) between the drain and the +24V (Vdd), the source grounded (Negative supply rail) and the signal fed between the gate and the ground.
It's no wonder you are having problems if you put the E-load between the MOSFET source and ground!

[byte]

I did a bit more research based on the conversation, and found some interesting information (hope this helps others):

• First, it seems people on the internet become enraged when someone suggest trying to use a n-chan MOSFET as a high-side switch. I'm not kidding, 99% of the posts I found on Google are mostly people saying it can't be done, and to instead us a p-chan, without really giving much explanation as to why. Asking more questions usually turns into snarky responses.
• Technically, yes, a n-chan can be used as a high-side switch, as long as the gate voltage is higher than the drain. But due to the high voltage drop across the MOSFET, a lot of energy will be dissipated as heat.
• Not reaching the VGS voltage for RDS(on) (say +10V) means the MOSFET will have higher impedance and produce more heat.
• This makes using an n-chan MOSFET as a high-side switch difficult, as the gate voltage needs to be several volts higher than the drain. But it can be done with a driver on the gate stepping up the voltage. It's much easier to just use an n-chan as a low-side so the source is referenced to ground.
• The transition between on/off for VGS is not instantaneous, which means during constant switching, there are periods of higher impedance - this explains why switching produces more heat than always-on

So, you see n-chan MOSFETs used mostly as low-side switches because of the above, but also because the voltage drop across the load first in series will mean less voltage across the MOSFET. This in turn means less power dissipation given the MOSFET impedance, and therefore less heat.

More importantly, my schematic was wrong. I'm not using the MOSFET as a high-side switch. I'm using some of those little PCB modules with a MOSFET and screw terminals. I thought this was wired up as a high-side switch, but in fact the PCB wires it up as a low-side switch.

This means, that I'm not burning up MOSFETs because I'm using an n-channel high-side without proper configuration. More likely, they are burning up because the e-load can not adjust it's impedance fast enough to match the current transients of the MOSFET transitioning through VGS(th) to fullly on, even using a square-wave PWM.

[byte]

Given what I found above, I have updated the circuit to be low-side switching. I don't think I need a flyback diode across the e-load, but maybe adding one is a good measure.



Again this is not ideal, because I still have to take the PWM rate and divide my readings for current & power on the e-load. This is almost impossible when varying duty cycle within the same test. But I'd think at least by switching between MOSFETS, the e-load will be stable and not over-react to the transients. I suppose I could add a shunt resistor and measure the current/voltage on my meter instead of the e-load.

Any better suggestions are appreciated, especially if it meant that I don't have to do extra arithmetic on the e-load readings.

[Jwillis]

It's virtually impossible to diagnose what the issue may be without knowing what Mosfet you are putting under test. Mosfet characteristics vary significantly from one to the next.
The Vgs(th) is meaningless when sourcing current unless we know the Maximum Vgs (the maximum potential difference ,plus or minus, between the gate voltage and the source voltage).

Also using Mosfets that are designed to be pulsed don't respond well in linear mode. Although the data sheet may have a DC FSOA does not mean that it's an accurate  value on switching Mosfets. Mosfets designed for linear mode have far more accurate values for the DC linear mode. Case in point, I have genuine swithcing  Mosfets that can fail far below the presented DC value in the data sheet when used in linear mode. I also have Linear Mosfet that actually exceed the depicted values in the data sheet. These datasheet values are base on best case average tests since not all components are equal between one production lot to the next nor within the same production lot.

Need more information.Post the datasheet.

[byte]

The first time I encountered this interesting "feature" with the e-load was testing an LED PWM dimmer. Worked fine with LED strip as the load. But even a smallest e-load CC value (500ma @ 24V) ended up frying the MOSFETs on the dimmer if it was anything less than 100% duty cycle. I de-soldered and replaced the mosfets, and it happened again. This was one of those little magic home dimmers rated for 96W with a PWM frequency of 1KHz: https://www.amazon.com/dp/B07JC3L4Z6

I then bought a pack of these HiLetgo MOSFET modules to try and narrow down the issue, as the meltdown was obviously due to an interaction with the e-load:
https://www.amazon.com/dp/B01I1J14MO

The are based on the IRF520N:
https://www.infineon.com/dgdl/Infineon-IRF520N-DataSheet-v01_01-EN.pdf?fileId=5546d462533600a4015355e340711985

Someone was kind enough to trace out the circuit and post it in the reviews:

[Jwillis]

Theirs a difference between sinking current and sourcing current . Very important  when it comes to the Vgs. In the first schematic you posted you are sourcing current through your electronic load which means that the potential difference between gate and source is 24 volts. When you apply a voltage at the gate of 1 volt turning on the mosfet the potential difference is 23 volts which exceeds the maximum Vgs of the Mosfet. Failure. In the the last schematic you posted the load is on the high side and the current is sunk to ground or 0 Volts . So the gate to source potential would only be 19 Volts if  1 volt is applied to gate . This would not exceed the Vgs.

Also the Vgs(th) of the mosfet is 2V min . at 1V you operate the mosfet in the ohmic region which can also be catastrophic to the mosfet. With large currents the mosfet heats up inside really fast . You may not even notice any heat being dissipated before failure. Mosfets need to be turned on hard to avoid operating in the ohmic region. 

[Kim Christensen]

Unfortunately, a simple resistive load is not programmable. I'm looking for the programmability so that I can vary the load through a range of tests as my power supply is also programmable and will be varying the voltage.

What MOSFET parameters are you trying to measure? RDS(on)? Switching time? etc...

[byte]

Theirs a difference between sinking current and sourcing current . Very important  when it comes to the Vgs. In the first schematic you posted you are sourcing current through your electronic load which means that the potential difference between gate and source is 24 volts.

Yes, I said above that I just drew the schematic wrong. I’m testing in a low-side switch configuration.

[byte]

What MOSFET parameters are you trying to measure? RDS(on)? Switching time? etc...

Right now, I’m just trying to test to figure out why an e-load is pulling more current that it’s set point and how to avoid it while still using an e-load. So a bit of everything, thermal, current handling, switching frequency, RDS(on), etc.

Once I can figure out how to test them with a e-load reliably, I want to test a variety of led pwm dimmers and motor speed controllers through a range of voltages, duty cycles, and loads to measure efficiency and thermals.

[byte]

So, at this point I am reconsidering my life choices. I've been trying to design a circuit so that I can just plugin any test source to my e-load, regardless of switching characteristics, and smooth out the input so the e-load does not overreact as it has been. However, all this extra circuitry to make this general purpose, including measurement which can't be done on the e-load in such a case, feels like it is over-complicating the solution.

Does any manufacturer have a e-load that will work for my use case? I'd be surprised if no one has solved such a problem. Also surprised that the switching characteristics of a device under test is expected to be known, before connecting to the e-load.

I'm hoping maybe a higher-end e-load has the features/capability to test these kinds of sources. But before making a new personal investment, I want to be sure.

Thanks!

[Manul]

I will note that sharp unloading (decrease of current) could potentially damage the mosfet too. The reason is inductance present at source terminal, which momentarily brings source potential down and effectively increases gate-source voltage. Gate might breakdown and boom it goes. You might try to eliminate this possibility by adding an adequate zenner very close to mosfet between gate and source.

Another note is that current smoothing is generally achieved with inductors. Think buck converter. You pulse a current through inductor, it freewheels through the diode, you pulse again. If inductor is large enough and pulses fast enough, you can achieve continuous current (with some ripple). So an inductor with added fast diode migh allow you to smooth current changes, so that the e-load does not need to respond super quickly.

So in case that is really a problem with e-load responding poorly, an inductive filter could be a promising option. Problem is, it does not scale very good at high currents. You need big core to prevent saturation, big diode, it will have power loss, also big diodes become slow and so on.

It was mentioned that constant resistance mode could be better.