Why does this MOSFET keep burning out?

N5KS

Joined Apr 17, 2016
5
2 things come to my mind. Is the 1N4006 correctly wired as shown? And, are the grounds common between the 180V supply and the 5V supply?
 

rphare

Joined Nov 20, 2015
9
Careful here. 180V and 1111 ohms = 29W of power. It might glow in the dark.
What kind of solenoid are we talking about here?
There's something wrong about the numbers.
 

ebp

Joined Feb 8, 2018
2,332
I have absolutely no reason to disbelieve any of the numbers. The 1111 ohms is almost certainly from an actual measurement, since no manufacturer is going to write a resistance spec to 4 significant digits, especially given the large temperature coefficient of resistance of copper. The power level is not unreasonable. 180 V is roughly what you would get from rectified AC at nominally 120 VRMS if were a bit high.

It is unlikely the freewheeling diode is connected in reverse since that would almost certainly have killed the FET in the 12 V test and none would have survived even a single switching at 180 V - and some did. An open freewheeling diode would result in a dead FET, though it is extremely unlikely any would have survived their first turn-off (noting, as previously, that the slow rise and fall of the gate drive would be a mixed benefit; prediction is difficult without an inductance spec for the solenoid, but it would almost certainly be at least hundreds of millihenries).

Inductance in the loop containing the driver and the gate and source of the FET could be the culprit.
 

Thread Starter

Involute

Joined Mar 23, 2008
97
It may sound a silly question, but are you QUITE SURE that the freewheeling diode is connected the right way round?!

If it isn't, that would be a very good way to fry multiple MOSFETs!
Not sure of anything at this point, so I'm carefully starting over and testing each part of the test circuit (at low voltage) as I build it back up. I even have two massive 30W resistors that, in series, will give me 1220 ohms (the closest I could get to 1111 at that wattage). This will let me test current and power without the flyback issues. I also just received 10 of the original MOSFETs plus four of a couple of other similar ones, so I have a lot of parts to play with (including some 200V zeners).
 

Thread Starter

Involute

Joined Mar 23, 2008
97
And another silly question: Are you sure about the solenoid resistance? Can you measure it? Where did the suspicious value of "1111" come from?
Yes, that's the measured value in the off state. I don't think the magnetic field affects it much in the on state, but correct me if I'm wrong. I guess I could infer the on resistance by measuring the current draw in the on state, if necessary, but the datasheets I've seen for solenoids and relays have always specified the off-state resistance measured through the coil, so I figured that was the figure of merit..
 

Thread Starter

Involute

Joined Mar 23, 2008
97
Try double vds rating or 400v irf740
I'd really like to do this with an SMD part, if possible. Thanks for the pointer, though. Ultimately, it just has to work, so I'll go with a through-hole part if that's the only way to get it to. Of course, I'd still like to figure out why the current part isn't working.
 

Thread Starter

Involute

Joined Mar 23, 2008
97
I have absolutely no reason to disbelieve any of the numbers. The 1111 ohms is almost certainly from an actual measurement, since no manufacturer is going to write a resistance spec to 4 significant digits, especially given the large temperature coefficient of resistance of copper. The power level is not unreasonable. 180 V is roughly what you would get from rectified AC at nominally 120 VRMS if were a bit high.

It is unlikely the freewheeling diode is connected in reverse since that would almost certainly have killed the FET in the 12 V test and none would have survived even a single switching at 180 V - and some did. An open freewheeling diode would result in a dead FET, though it is extremely unlikely any would have survived their first turn-off (noting, as previously, that the slow rise and fall of the gate drive would be a mixed benefit; prediction is difficult without an inductance spec for the solenoid, but it would almost certainly be at least hundreds of millihenries).

Inductance in the loop containing the driver and the gate and source of the FET could be the culprit.
My power supply is exactly as you suggest: 120 VAC into a full-wave bridge with the output filtered through a 470 uf cap. The DC voltage actually drops to 161 VDC under the load of the solenoid (or, under my latest test, 1220 ohms of 30W resistance). It actually starts at 180 VAC, no-load, because I think my isolation transformer is a little off and outputting a little higher voltage than it should. Ideally, I think the no-load output should be closer to 168 VDC, but I don't think the difference is material here.

I'm not using an actual gate driver. Just the output of the 555 connected directly to the gate.

I've acquired multiple parts to experiment with and hope to report back with more granular data in the next two or three days.
 

Thread Starter

Involute

Joined Mar 23, 2008
97
2 things come to my mind. Is the 1N4006 correctly wired as shown? And, are the grounds common between the 180V supply and the 5V supply?
I believe the diode was wired correctly. Things would have gone south faster and more dramatically otherwise. My test setup is disassembled, however, and I'll be slowly rebuilding it as I test each step of the way. The grounds are common.
 

Thread Starter

Involute

Joined Mar 23, 2008
97
Careful here. 180V and 1111 ohms = 29W of power. It might glow in the dark.
What kind of solenoid are we talking about here?
There's something wrong about the numbers.
This kind (the "37" row under "AWG Number"). In operation it will only be on for 400 msec. In my testing, it's on for longer, but less than 10 sec.
 

Jack Yu

Joined Dec 9, 2017
4
I'd really like to do this with an SMD part, if possible. Thanks for the pointer, though. Ultimately, it just has to work, so I'll go with a through-hole part if that's the only way to get it to. Of course, I'd still like to figure out why the current part isn't working.
I just went through similar mosfet failures like you did. I was switching 120vdc using 200 vds rated mosfets. I burned about 8 of them before I finally settled on much higher vds version. Now my circuit is working without any issue. mosfets are very sensitive to vds limit. Exceeding this limit just few ns can burn the mosfet. Transient voltages can arise from this kind of circuit which can lead to higher vds than expected. Maybe you can try putting a capacitor as snubber across the solenoid if that helps.
 

Thread Starter

Involute

Joined Mar 23, 2008
97
I just went through similar mosfet failures like you did. I was switching 120vdc using 200 vds rated mosfets. I burned about 8 of them before I finally settled on much higher vds version. Now my circuit is working without any issue. mosfets are very sensitive to vds limit. Exceeding this limit just few ns can burn the mosfet. Transient voltages can arise from this kind of circuit which can lead to higher vds than expected. Maybe you can try putting a capacitor as snubber across the solenoid if that helps.
I’ll keep higher voltage parts in mind if all else fails. How would you size the cap? Why would that be better than a plain diode or zener?
 

Jack Yu

Joined Dec 9, 2017
4
I’ll keep higher voltage parts in mind if all else fails. How would you size the cap? Why would that be better than a plain diode or zener?
I used 10nF to 47nF capacitors. It helps to absorb and dissipate transient voltages. You still need to keep the diode in the circuit.
 
Last edited:

MaxHeadRoom

Joined Jul 18, 2013
21,633
My power supply is exactly as you suggest: 120 VAC into a full-wave bridge with the output filtered through a 470 uf cap. The DC voltage actually drops to 161 VDC under the load of the solenoid (or, under my latest test, 1220 ohms of 30W resistance). It actually starts at 180 VAC, no-load, because I think my isolation transformer is a little off and outputting a little higher voltage than it should. Ideally, I think the no-load output should be closer to 168 VDC, but I don't think the difference is material here.
It is not really necessary to run a DC inductive device such as a solenoid on filtered DC, a bridge is usually sufficient and the voltage is closer to correct.
Brakes, solenoids and clutches all over the world run this way.
Max.
 

ebp

Joined Feb 8, 2018
2,332
Transformers usually have the turns ratio adjusted so that the output voltage is at nominal at the full rated current of the transformer. This compensates for voltage drop due to the winding resistance. An isolation transformer is likely to put out a voltage a bit on the high side when lightly loaded.
 

ebp

Joined Feb 8, 2018
2,332
If you try omitting the filter capacitor, I would definitely recommend a TranZorb type suppressor to protect the FET. Although the load current is small, it is still sufficient to produce a voltage spike in the secondary winding "leakage inductance" when the load is turned off. The distributed capacitance of the winding may be enough to "eat" much of the energy in the leakage inductance, but it may not.

In a transformer, leakage inductance can be modeled as a small inductance in series with a winding. It is due to magnetic flux that is not "coupled" to what might be regarded as the "main" flux. Like any inductance, it will "try" to keep the current through it constant - flowing in the same direction at the same magnitude. If 150 mA current had been flowing into a 1100 ohm load and suddenly the load is removed, the voltage will rise as high as necessary to keep the 150 mA flowing - just as with the solenoid coil itself.

A capacitor directly across an inductance will produce resonant ringing that can make problems worse. With the high resistance of the solenoid, there will be substantial "damping" so the ringing might not be too horrible. You must also consider the fact that the capacitor will allow a large and ill-defined current spike at switch turn-on - which can play against you in inductances in wiring. A resistor in series with capacitor will limit the current spike and can also dampen ringing. RC snubbers can work well, but they do have to be designed properly. Zener type suppressors are fast and effective without the ringing problem.
 

Thread Starter

Involute

Joined Mar 23, 2008
97
It is not really necessary to run a DC inductive device such as a solenoid on filtered DC, a bridge is usually sufficient and the voltage is closer to correct.
Brakes, solenoids and clutches all over the world run this way.
Max.
I'm energizing 40 solenoids in rapid succession, 40 msec. apart, each one for 400 msec. The timing is important. The unfiltered DC has 160V of ripple around 110V at 60 Hz, meaning it's below 110V for 8 msec. of every 16 msec. cycle. I'm concerned the solenoid wouldn't have enough force to do its job with this much ripple (or, at least, to do it as quickly). It works unloaded with unfiltered DC, but it's hard for me test it with its operating load. I wouldn't mind eliminating the cap, but its cost (both for the cap and its associated PCB space) isn't consequential in the overall BOM.
 

Thread Starter

Involute

Joined Mar 23, 2008
97
If you try omitting the filter capacitor, I would definitely recommend a TranZorb type suppressor to protect the FET.
I don't recognize some of the specs for those things in the Digikey part selector, Can you point me to any discussions on how to select one for a particular application?
 

Thread Starter

Involute

Joined Mar 23, 2008
97
Well, things are behaving more predictably now, due, in part, to a cleaner test setup:
Test setup 1.jpg
I replaced the 555 with an Adafruit Trinket to reduce clutter. Pressing the button to its right triggers a 400 msec. pulse (.5 usec. rise and fall time) to the MOSFET's gate (orange wire). The 1N485B is a plain, 200V diode, connected cathode to drain and anode to source. Unfortunately, it's not terribly effective, as you can see in the trace of Vds, below:
plain_diode.png
That spike is a lot uglier without the diode, but since the rebound exceeds the MOSFET's Vds max by 72V, I suspect that's why my previous tests failed, and did so erratically (some devices could withstand the abuse longer than others). Still, I thought suppressing this sort of thing was the whole point of using a diode with an inductor. As soon as the flyback voltage starts going negative at shutoff, the diode should start conducting to ground. What am I not understanding? Is a plain diode too slow?

Anyway, since I now had some zener diodes on hand (based on ebp's suggestion; thanks), I swapped out the plain diode for a zener, producing this Vds trace:
zener_diode1.png
Much better! Here's a closer look at that rising edge, where the messiness should be:
zener_diode3.png
No problem. This looks like the way to go, though now it has me wondering how to calculate the wattage necessary for the zener. The one in this test is rated at 4W and is a large, through-hole part. I'd like to switch to an SMD, but the least expensive ones are only rated for 800 mW. Any tips or further comments appreciated. Thanks for everyone pitching on this, btw.
 
Last edited:

Jack Yu

Joined Dec 9, 2017
4
Well, things are behaving more predictably now, due, in part, to a cleaner test setup:
View attachment 163083
I replaced the 555 with an Adafruit Trinket to reduce clutter. Pressing the button to its right triggers a 400 msec. pulse (.5 usec. rise and fall time) to the MOSFET's gate (orange wire). The 1N485B is a plain, 200V diode, connected cathode to drain and anode to source. Unfortunately, it's not terribly effective, as you can see in the trace of Vds, below:
View attachment 163084
That spike is a lot uglier without the diode, but since the rebound exceeds the MOSFET's Vds max by 72V, I suspect that's why my previous tests failed, and did so erratically (some devices could withstand the abuse longer than others). Still, I thought suppressing this sort of thing was the whole point of using a diode with an inductor. As soon as the flyback voltage starts going negative at shutoff, the diode should start conducting to ground. What am I not understanding? Is a plain diode too slow?

Anyway, since I now had some zener diodes on hand (based on ebp's suggestion; thanks), I swapped out the plain diode for a zener, producing this Vds trace:
View attachment 163087
Much better! Here's a closer look at that rising edge, where the messiness should be:
View attachment 163088
No problem. This looks like the way to go, though now it has me wondering how to calculate the wattage necessary for the zener. The one in this test is rated at 4W and is a large, through-hole part. I'd like to switch to an SMD, but the least expensive ones are only rated for 800 mW. Any tips or further comments appreciated. Thanks for everyone pitching on this, btw.
Here're some SMD TVS.

https://www.digikey.com/products/en?mpart=SMCJ188CA-TR&v=497
 
Top