What the hell is going on with my IRL640A MOSFETS

[Julien3146]

Hello There!

I'm reading this forum for quite a while now, and this time, i decided to register, as I come across something I cannot understand, at all!

So, basically, i designed a Nixie Tube clock, and one of the main switch for the 170VDC was an IRF730, it did a great job except it isn't really logic level gate operation.
The MOSFET never fully turned on and the Rds was quite a problem, as the MOSFET keep heating.

Pretty sure of what I was doing, i decided to replace the IRF730 by an IRL640A, 200V, 18A, 0.18ohm@5V Rds.
As the pinout is the same, it was supposed to be plug and play...

Except it wasn't, at all... Nothing worked, the tubes were off, flickering at best. Probing the gate, it was stuck somewhere at its threshold voltage (±2V), the poor ATMega328p trying to supply 5V in vain.
Desoldering and testing it with an RCL meter, it was fried... appearing as 2 resistances (short between G-S and D-S).

Then i went full retard but I still don't understand how I managed to fried my 10 MOSFETS :

I hooked them up, one by one, to my DP832, 30VDC, 0.5A max on channel 1, between D and S, channel 3 at 5V, negative terminal on the source, and positive on the gate... I FRIED THEM ALL, WTF ?
(they didn't fried at 10VDC, it seems to fry as soon as channel 1 exceeded Vgs max, even if it wasn't hooked onto the gate).
I don't understand what i did wrong, and i HATE that! 

both channels were hooked with the negative terminal to the source, there is now way the gate voltage exceeded 20V, or am I an idiot ?

Thanks a lot and see ya!

Cheers.

[janoc]

Post a schematic if you want someone to help you. From your description that could be pretty much anything, including dodgy FETs (you didn't buy them from eBay/China, did you?) and ESD.

[Julien3146]

Will do tomorrow,

They are from Farnell btw.

Cheers

[floobydust]

The Rigol DP832 overshoots and might have been overvoltaging the mosfet G-S?
https://www.eevblog.com/forum/testgear/rigol-dp832-ovp-overshooting-with-huge-delay/

I'd add a 10V zener and 10k resistor to the gate-drive when testing.
I prefer to not use fancy these chinese power supplies, they are pretty sloppy and I was horrified to scope their output.

[T3sl4co1l]

That's a big fuggin' gate to be driving from a microcontroller pin.

In the PS, they could've just sweated off due to that alone, no idea.

The standalone test, could be ESD, could be stray inductance of wiring/cables, could be thermal, who knows.

Learn to draw the equivalent circuit of the environment the transistor is in, and you will probably scare yourself with what kind of a mess you were doing.   Example, loose cables are about 1 uH per meter of length.  Easy to make a power oscillator or flyback generator that way.  Or the equivalent circuit of the power supply, which I believe has an electrolytic capacitor at its outputs, regardless of how fast the internal current limiting circuitry operates?  It doesn't take much energy to kill a transistor.  Remember, transistors die in 10-100us, fuses open in 10ms, human perception sees ~100ms and below as "instantaneous"!

Tim

[Julien3146]

Wow guys, thanks a lot for the answers!

About the overshooting PSU, I scoped the output and was quite surprised by how much it doesn't.
Anyway, I think I understand now how I killed them with my standalone testing. I basically had a 1µH inductor in series with the 1.5nF MOSFET cap, and a dead short for a non negligible period of time between D and S, before the CC kicked in.

The 10V zener and 10k resistor to test the gate seems perfect, i keep that in mind.

What I don't understand though, is, why they didn't work inside the circuit :

The IRF730 works perfectly fine, and has a similar (while lower) gate capacitance. Even if it's clearly not ideal to drive them without a driver, it's not high frequency at all, it's just an on/off switch.
Maybe it has something to do with the breakdown voltage ? 200V seems quite close to the 170VDC it's driving? maybe with some ringing it exceeded that for a short but deadly period of time ?
Anyway, this clock is a 1 year old design that I decided to improve, and I feel like I shouldn't have : I was even more a beginner than today, and the design is flawed in numerous ways.

I can't find a low Rds logic level 300 or 400V MOSFET, so I think i'll just stick with the IRF730, or maybe build a small breakout board as an addon, with a proper driver and the IRF730.

Thanks!

[Julien3146]

OK guys,
I've imagined some addon board that takes the place of the IRF730.
Bear with me about the crudity of the design, i've drawn it in like 2 minutes, using only parts i had laying around :
The idea is to stick to the IRF730, but using it at 10Vgs instead of 5.



the drain and source pins are still connected to the IRF730, but the gate pin now drives the non-inverting input of a MCP602.
What I inteded to do :

When gate pin is low (it's pulled low onto the original PCB), the non-inverting input is lower than 2V (voltage divider at inverting input), so the MCP602 output 0V
When gate pin is high, the MCP602 output 12V or so, and the divider feeds 10V to the IRF730 gate.
MCP602 are capable of 30mA output, here there's something like 3mA

Well, except the MCP602 won't behave well with 12V... but let's imagine i figured this out earlier and plug another rail to rail opamp that i have laying around and has a wider input range...

Is it an acceptable design ? for me it seems like it.

Thanks and see ya!

Julien

[T3sl4co1l]

There are purpose made drivers for this, logic input, gate output, 12V supply.  FAN3100 comes to mind.

What bothers me about the diagram is the implied length of the source connection.  Is that directly mounted?  On a cable?  No idea...

Better to cut the trace from MCU to gate, and bodge in the driver that way.  Easy enough with some thin wire and glue.

Tim

[Zero999]

That circuit will be very slow, so it's not surprising it's not working properly. The MCP602 has a maximum voltage rating of 6V and won't like 12V.

Use a proper MOSFET driver IC, such as the TC4427 and don't bother with a potential divider on the gate, except maybe for a low value series resistor, to damp any oscillation due to the inductance and high value pull-down resistor, if you're concerned about the gate floating when the power to the driver is disconnected but the MOSFET is still powered. The maximum gate voltage of both of your MOSFETs is much higher than 12V, so there's no need to reduce the gate voltage.
http://ww1.microchip.com/downloads/en/DeviceDoc/21422c.pdf

[Julien3146]

You're awesome guys,

I don't know why i didn't register earlier ^^.

My idea with my bodged breakout board was to be mounted in place of the TO220 of the current mosfet, the D and S connections wouldn't be any longer than now.
The idea was also to bodge something this sunday without ordering anything.
I know the MCP602 won't allow me to do anything with the 12V input, so my plan is going down  .

I've got some enameled wire and a hot glue gun, so I'll have a look at the TC4427 and FAN3100 datasheets, thanks a lot guys.

Anyway, any idea why the IRL640A doesn't work and the IRF730 does ? that i don't understand...

Cheers,
Julien

[Julien3146]

Ok, so I've ordered some TC4427.
I'll bodge that to the circuit and see how it goes!

I'll keep you guys in touch.

Julien

[james_s]

That 1k gate resistor is huge too, 4.7-10 Ohms is more typical. Depending on the frequency you're using, layout can be critical. I built several nixie tube supplies around the MAX771 and while I was able to make them work on perfboard, I had to be very careful with the layout, especially grounding and the feedback loop.

[Julien3146]

Hey!
A little feedback about the design and repair!

Basically the bodge TC4427 did the trick just fine.... and unveiled new design flaws once the FET fully turned on.. the tube were ultra bright, as if the current limiting resistors weren't there... and fair enough it was the case : even disconnecting the anode completely from the 170v rail didn't switch off the tubes... wtf...

I spent hours scratching my hair, before beginning to grasp what was going on... I'm not even sure I have understood that correctly :

My clock is based on HV513 push pull drivers, when a digit is off, the cathode of the digit is pulled high to 170vdc, when it's on, the K is grounded, thus allowing the current to flow... that's a digital era misconception, as tubes aren't diodes:

All the cathodes were supplying 170vdc to the tube, behaving as anodes, and the current limiting resistors were useless... i'm dumb... the only reason i didn't notice earlier was because my mosfet wasn't fully on and acted as a limiting resistor of its own...

Fix was easy once figured out : I lifted all the HV513 high voltage supply pins, to use the ICs as open drain instead of push pull.

This post has nothing to do with the original question, but I think it was a cool issue/trap to share with you guys ^^

Cheers