I was here wondering what can we discover about why a MOSFET failed. It could have been because of too high voltage on gate, due to too much current for the bounding wire / die, due to repeated avalanche, due to... I've noticed that sometimes a failed MOSFET seems ok except the ON resistance is too high, at another times a MOSFET is shorted, also from gate to drain/source (a few Ohm), so I wonder, what can we learn about the failure mode of a MOSFET by inspecting its dead corpse?
How to know why a MOSFET failed
- Thread starter Njay
- Start date
SamTexas
1 MW
Water! I got caught in a torrential rain, water got into my controller, probably flooded the whole thing. I heard many loud (relatively speaking) explosions. Got home (pedalled), opened the controller and found 6 of 12 exploded.Njay said:
Sorry, I was actually speaking from a controller development point of view 
Mines always blow silently and suddenly, no smoke, no plasma, no bang-sound, no visible physical damage (I don't have current limiting active yet, but I know it's not always the current).
Mines always blow silently and suddenly, no smoke, no plasma, no bang-sound, no visible physical damage (I don't have current limiting active yet, but I know it's not always the current).
If a item is operated with in its parameters it should last for a very long time! SO if you have a good quality fet and you have it fail then one or more of the parameters must have been exceeded. If you run it at the max voltage hot off the charger then it will likely see voltage spikes above the max voltage rating. If you are using a china controller it will likely be using battery amps to guess at a phase amperage which is what the fet has to deal with. It is also possible the Caps are not located very well or they are china knock offs and are not doing the right job. The only way to find this is to test with the proper equipment and test in different operating scenarios until you find the parameter being exceeded. INjay said:Sorry, I was actually speaking from a controller development point of view
Mines always blow silently and suddenly, no smoke, no plasma, no bang-sound, no visible physical damage (I don't have current limiting active yet, but I know it's not always the current).
But when you say you don't have any exciting explosions like I do and they just "go to sleep" then I would bet they are just barely being pushed past one or more limits and might not be often. Or they are not a quality part or the inductive paths are slightly different. Is it the same fet always blowing???
I didn't do the previous repairs so I really don't know if it's always the same FET, but I do know the same type of FET (I have P's and N's) has been swapped in another occasion. Actually after I removed this last "blown" FET I could see that it suffered physically; the back plate has a small area that seems to have "boiled" and it looks like the silicon has tried to escape through the line joining the plate with the black epoxy . I'll post some photos later.
I don't have harm voltage spikes (unless they only happen sporadically) even when I stall the motor axis and turn the throttle a up (not very up), although I know the layout is bad (~2 inches from MOSFETs to input caps, 5mOhm shunt in the middle next, to the caps, and no further caps). The FETs don't seem hot (only slightly warm) after the incident. What I do see is regular, predictable and huge current oscillation spikes (something like 10x the controller's max current), measured at the shunt, at the MOSFET ON->OFF OFF->ON transitions; I'll put some scope snapshots later. I'm going to increase the switching speed (it's around 250ns now) to see what effect it has. This is a brushed motor H-Bridge running on 1 direction, the measures I mention are just the lower MSOFETs (N) switching while the top one (P) is constantly ON. The blown P is the one on the opposite bridge, the one who's internal diode is used for current re-circulation while the N's are off. This is something I''m currently changing, as this dissipates a lot of energy on that P and that makes it really hot under load - however I had the failure while speeding it without any load.
. I'll post some photos later.I don't have harm voltage spikes (unless they only happen sporadically) even when I stall the motor axis and turn the throttle a up (not very up), although I know the layout is bad (~2 inches from MOSFETs to input caps, 5mOhm shunt in the middle next, to the caps, and no further caps). The FETs don't seem hot (only slightly warm) after the incident. What I do see is regular, predictable and huge current oscillation spikes (something like 10x the controller's max current), measured at the shunt, at the MOSFET ON->OFF OFF->ON transitions; I'll put some scope snapshots later. I'm going to increase the switching speed (it's around 250ns now) to see what effect it has. This is a brushed motor H-Bridge running on 1 direction, the measures I mention are just the lower MSOFETs (N) switching while the top one (P) is constantly ON. The blown P is the one on the opposite bridge, the one who's internal diode is used for current re-circulation while the N's are off. This is something I''m currently changing, as this dissipates a lot of energy on that P and that makes it really hot under load - however I had the failure while speeding it without any load.
dnmun
1 PW
.
I don't think you can tell much from the failure mode. In most cases they short. Sometimes the short circuit causes enough current to make them fuse open. Knowing the conditions right when they failed would tell you more. Over voltage, over current, over temperature or over dissipation can all cause failure.
Jeremy Harris
100 MW
Njay said:
They fail for a host of reasons, usually associated with the FETs being poorly specified for the requirement (because of cost pressures, I suspect), because the controller doesn't have adequate protection for the FETs or because the controller has been modified to run outside its safe limit.
If the FET fuses, because of a short duration over current event, then they will usually blow the internal bonding wires, or maybe a leg, and the failure will be pretty obvious (often the package will explode if the bonding wires fuse). The FET will usually, but not always, end up open circuit after an event like this. This shouldn't happen if the controller is unmodified, as the controller peak current shut down (if it's a common type like a Xiechang) should activate and shut the controller down. The most common reason for this type of failure is probably people lowering the shunt resistance to increase the current limit, as this will also increase the peak current protection circuit threshold, which isn't a good thing to do normally. Increasing the current limit by altering the controller programming is safer, as it keeps the FET peak current protection shut down working normally.
If the FET overheats from prolonged use at high current (the worst case case being when the motor is running slowly, but not stalled, under high load) then the result may well be a shorted FET with little sign of external damage, other than, perhaps, a damaged gate drive resistor. This shouldn't happen if the controller over-current protection (the main current limit) is set to a safe value. The most common reason for this type of failure is probably people re-programming the current limit to a value that's higher than the FETs can safely operate at and then pushing the controller hard by riding up a steep hill, or maybe by doing a lot of hard acceleration.
If the FET goes open circuit, with no sign of external damage, then the cause may be an over-voltage punch-through event, maybe caused by operating at a supply voltage that's close to the maximum rating of the FETs, or by something like inadequate commutation capacitors on the supply, perhaps as a result of increasing the controller current to a level beyond the ability of the capacitors in the controller to adequately control supply ripple close to the FETs.
There are a host of other reasons for FETs failing, but I'm going to hazard a guess that these are probably the most common.
The promised photos of the blown FET:
May have been a combination of using this FET's internal diode as freewheel device (heats alot) and not having the current limiter ON while speeding up abruptly.
Seems like there isn't much info to gather from a post-morten FET analysis. Thanks for your comments guys.
May have been a combination of using this FET's internal diode as freewheel device (heats alot) and not having the current limiter ON while speeding up abruptly.
Seems like there isn't much info to gather from a post-morten FET analysis. Thanks for your comments guys.
There are many reason a MOSFET failed.
The failure mode can be Design related or product Reliability (quality) related.
There are many Failure analysis methodology to decipher the root cause. It is sometime possible to determine the root cause just by seeing the burnt area. Some failure mode produce a consistent failure signature. However, it is also possible for a secondary failure to occurred masking the primary root cause signature. The secondary failure is usually overcurrent where you get a blob of carbonised compound. However, ususally it make sense to look at the cct diagram and to understand the behaviour of the MOSFET operation + the failure signature. (Bit of a CSI work)
The first sign of stress on a MOSFET would usually appear as Idss leakage current or Igss leakage current. If the Rdson has risen, that can related to several cause, such as die attach wear out i.e. the solder has reflow due to massive amount of heat but short of killing the MOSFET. Rise in Rdson can also be cause by the bond wire contact wear out due to thermal cycling.
In principle, you can define MOSFET failure under the following:
Electrical Stress
. Over Voltage (avalanche- single or repetitive)
. Over Current (Kaboom)
. Linear Mode (thermal runaway)
. ESD (zapped!)
. EOS (Electrical overstress)
Environmental Stress:
. Moisture ingress
. Package delamination
. thermal fatigue
. high mechnical pressure - usually associate with retention clip pressure for TO220
Mechanical issue:
. fracture or crack die
. single/cluster cells defect
. contamination e.g. ionic contaminant
. Silicon to package or package to PCB interface
You'll be surprise alot of the problem reside on the poor PCB cct. layout and dodgy gate driver or poor thermal budget.
The MOSFET parasitic diode is extremely robust hence, unlikely the cause of a typical failure. Though they have higher VI power dissipation vs MOSFET I^2R, it is the designer duty to ensure the thermal consideration are taken care off.
This is not exhaustive and serve as a intro. It all depends on a case by case. Hope this is useful
Note:
1. Not all new and latest super dooper MOSFET is best for your application.
2. Beware there are huge amount of knock-off copies floating in the market. some product has been discontinued in early 2000 but product are still available with 2010 datecode.
The failure mode can be Design related or product Reliability (quality) related.
There are many Failure analysis methodology to decipher the root cause. It is sometime possible to determine the root cause just by seeing the burnt area. Some failure mode produce a consistent failure signature. However, it is also possible for a secondary failure to occurred masking the primary root cause signature. The secondary failure is usually overcurrent where you get a blob of carbonised compound. However, ususally it make sense to look at the cct diagram and to understand the behaviour of the MOSFET operation + the failure signature. (Bit of a CSI work)
The first sign of stress on a MOSFET would usually appear as Idss leakage current or Igss leakage current. If the Rdson has risen, that can related to several cause, such as die attach wear out i.e. the solder has reflow due to massive amount of heat but short of killing the MOSFET. Rise in Rdson can also be cause by the bond wire contact wear out due to thermal cycling.
In principle, you can define MOSFET failure under the following:
Electrical Stress
. Over Voltage (avalanche- single or repetitive)
. Over Current (Kaboom)
. Linear Mode (thermal runaway)
. ESD (zapped!)
. EOS (Electrical overstress)
Environmental Stress:
. Moisture ingress
. Package delamination
. thermal fatigue
. high mechnical pressure - usually associate with retention clip pressure for TO220
Mechanical issue:
. fracture or crack die
. single/cluster cells defect
. contamination e.g. ionic contaminant
. Silicon to package or package to PCB interface
You'll be surprise alot of the problem reside on the poor PCB cct. layout and dodgy gate driver or poor thermal budget.
The MOSFET parasitic diode is extremely robust hence, unlikely the cause of a typical failure. Though they have higher VI power dissipation vs MOSFET I^2R, it is the designer duty to ensure the thermal consideration are taken care off.
This is not exhaustive and serve as a intro. It all depends on a case by case. Hope this is useful
Note:
1. Not all new and latest super dooper MOSFET is best for your application.
2. Beware there are huge amount of knock-off copies floating in the market. some product has been discontinued in early 2000 but product are still available with 2010 datecode.
Very interesting topic. Especially thanks to SiC and Jeremy.
I have this situation though... 18fet (IRFB4110) Infineon ("LYEN") controller. No mods. Hooked to Crystalyte X5 hub motor. Battery voltage 66V (20S A123), controller's components rated at 100V.
I have connected the hall sensors a bit wrong and the phase timing was advanced a bit. Well, the motor was spinning ~20% faster than it should (but no weird noises, everything as usual). At full speed, if throttle is released quickly, the motor would go to strong regen because it's BEMF is higher than battery voltage. I did the spin-up&release test few times and the controller died. Silently, no smoke, nothing.
After opening it up, there was no visual damage. After measuring stuff, one high side mosfet on one phase was shorted out (all three pins). Other two parallel mosfets seemed to be OK (no GS/DS/DG leakage, but I have not measured the Rdson below 0.1Ω). No sign of high temperature on any of the three. I have also inspected the driver circuitry (3 BJTs, diode, resistors, caps) - all OK. So I just replaced the mosfet and assembled everything back and reconnected the sensors in correct order.
The controller seemed to be working OK, motor spinning with no weirdness... Did some spinup&stop tests... Assembled everything on the bike and tried to accelerate. First two attempt were successful (track lenght 5 meters).
On third attempt controller died again. Same situation: no bang, no smoke, no nothing. Again, one mosfet died in the same position.
Current limit is not very high, its like 60A or so. I was guessing that this could be caused due to resonance between battery wires (1 meter long in total, so its like 1μH), controller's caps and 16kHz PWM. But since battery voltage (66V) is quite far from 100V rated limit, this seems unlikely.
Any thoughts?
I have this situation though... 18fet (IRFB4110) Infineon ("LYEN") controller. No mods. Hooked to Crystalyte X5 hub motor. Battery voltage 66V (20S A123), controller's components rated at 100V.
I have connected the hall sensors a bit wrong and the phase timing was advanced a bit. Well, the motor was spinning ~20% faster than it should (but no weird noises, everything as usual). At full speed, if throttle is released quickly, the motor would go to strong regen because it's BEMF is higher than battery voltage. I did the spin-up&release test few times and the controller died. Silently, no smoke, nothing.
After opening it up, there was no visual damage. After measuring stuff, one high side mosfet on one phase was shorted out (all three pins). Other two parallel mosfets seemed to be OK (no GS/DS/DG leakage, but I have not measured the Rdson below 0.1Ω). No sign of high temperature on any of the three. I have also inspected the driver circuitry (3 BJTs, diode, resistors, caps) - all OK. So I just replaced the mosfet and assembled everything back and reconnected the sensors in correct order.
The controller seemed to be working OK, motor spinning with no weirdness... Did some spinup&stop tests... Assembled everything on the bike and tried to accelerate. First two attempt were successful (track lenght 5 meters).
On third attempt controller died again. Same situation: no bang, no smoke, no nothing. Again, one mosfet died in the same position.
Current limit is not very high, its like 60A or so. I was guessing that this could be caused due to resonance between battery wires (1 meter long in total, so its like 1μH), controller's caps and 16kHz PWM. But since battery voltage (66V) is quite far from 100V rated limit, this seems unlikely.
Any thoughts?
Looks like something you deal with on a regular basis, SiC (Sillicon Carbide ?). What is "CCT"?
Indeed the body diodes seem quite robust. I've done mistakes where I connected the battery (small lead) with polarity reversed and they held on 10-15s (saving the rest of the circuit).
?). What is "CCT"?Indeed the body diodes seem quite robust. I've done mistakes where I connected the battery (small lead) with polarity reversed and they held on 10-15s (saving the rest of the circuit
).
whatever
100 kW
- Joined
- Jun 3, 2010
- Messages
- 1,297
Many moons ago.... I've come across a few motors that a certain combination of hall/phase wires you would get this odd very fast rpm motor situation,
it was in my case obviously not the correct combination....I've not seen it mentioned much as i dont think all motors show this odd combination effect.
It may well be you've just had the wrong combination and thats somehow led to the fet failure.
I would be very interested if anyone has an understanding of this odd high rpm combination that occasionally is seen in some motors when trying to find correct
hall/phase combinations.
From memory in that high rpm incorrect combination it was accompanied by very high amp draw even at no load
it was in my case obviously not the correct combination....I've not seen it mentioned much as i dont think all motors show this odd combination effect.
It may well be you've just had the wrong combination and thats somehow led to the fet failure.
I would be very interested if anyone has an understanding of this odd high rpm combination that occasionally is seen in some motors when trying to find correct
hall/phase combinations.
From memory in that high rpm incorrect combination it was accompanied by very high amp draw even at no load
