MOSFET Power Switch occasionally fails - why?

Tip / Sign in to post questions, reply, level up, and achieve exciting badges. Know more

cross mob
mandrews
Level 2
Level 2
5 replies posted 5 sign-ins First solution authored

Hi,

I'm using back-to-back 100V / 40A IRF5210PBF power P-channel MOSFETs in a 48V power switch with a short time duration soft-start. Occasionally a MOSFET dies. Would you please tell me why?

The input and output MOSFETs of the switch are protected by

  • 48V TVS's (Infineon, P/N 824520481, 600W, Vclamp = 77.4Vpk @ 10/1000usec 7.8A) to handle voltage pulses on turn-On, turn-Off, and with motors turning On and Off. To prevent avalanche faults. 
    • The load is a bank of motor driver boards with a heavy capacitive load of 10,500uF total max, all MLCCs.
    • Average motor current is less than 7A with random load spikes lasting microseconds up to 16A peak.
  • 400V/10A power diodes on the power output connectors to the motor boards, cathodes to the +48V supply input
  • The soft-start is a very short duration right now since I didn't want to risk over heating the power MOSFETs and didn't know everything I'd eventually be powering with their own power reset periods. In other words, a normal design. 🙂 

Any suggestions for improvements? The back-to-back approach was used so I didn't back-power something else on the same +48V supply when shutting things down.

Thanks.

 

mandrews_0-1644458474372.png

 

0 Likes
1 Solution
mandrews
Level 2
Level 2
5 replies posted 5 sign-ins First solution authored

Hi AndiM,

Thank you for your response, too. I really appreciate you taking your time to look at this.

You and Meghana mention inrush as the potential problem. Very possible. I can slow the soft-start by increasing C18 to reduce inrush. Right now it's well under 1ms turn-On which is likely too fast, effectively no inrush reduction. Could bump it to 5ms or so, analyzing what that does to the MOSFET's Pd and the reduction in inrush. Definitely worth looking at. Thanks.

You're right that D18 and D23 bypass Q12 when they're conducting. However they only conduct when +48V is over +48V from an inductive kick or whatever. Once the signal over +48V is gone, Q12 will still block the remaining "switched +48V" as the output capacitors discharge. Wait a second - you have an excellent point. Because as the master input +48V supply drops, the protection diodes in parallel with the power MOSFETs as they turn Off and increase their RDS(on), will pull the output filter caps with it within reason. This leads to the question can the OTS power supply absorb the caps' discharge currents or will it just take a while?

This is why I enjoy seeing things thru other peoples' eyes - easy to see something so much that you 'stop seeing it' after a while. 🙂 

View solution in original post

0 Likes
11 Replies
Meghana
Moderator
Moderator
Moderator
50 likes received 100 solutions authored 10 likes given

Hello Mandrews,

Thank you for posting on Infineon Community. 

From the schematic shared and the application details, I see that there is very high capacitive load of 10,500uF. This can cause very high inrush current which can be a reason for the MOSFETs failure. 

Can you please tell us which of the two MOSFETs fail here. Is it Q12 or Q7 ? 

Can you also please share the waveforms of Id, Vds & Vgs of both the MOSFET's. To suggest any improvement, it is important to know the failure cause. As such, the schematic seems ok. We will be able advise any change after analysis these waveforms. 

Regards

Meghana

Hi Meghana,

Thank you for your response.

Q7 appears to be damaged, along with one or more of D21, C18 or R41 which are shorted. I'm pulling them out to determine which one or more died an ugly death.

Once I get the latest hardware & software configurations together I'll provide the latest waveforms. 

This is part of an Emergency Stop circuit. I felt sorry for the Technician who tested it 150 times in a row. No failures but could be accumulated damage from inrush as you suggest.

The entire system was powered down for a time. After they turned the system back On "the smell" occurred. No Emergency Stop button press. The Emergency Stop buttons continued to work for the other 2 units.

Additional notes: 

  1. There is about 20' of cable connecting 3 to 4 modules each with this circuit in it. One large loop, aka antenna.
  2. Each module has its own separate 48V power supply to be switched locally for its loads. Powered off the same VAC input, also chained between the units.
  3. The first module provides the +12V to the loop
    mandrews_0-1644523612258.png

     

  4. I used an ADG1402 as a digital relay. A simple BJT or MOSFET would have done the same thing possibly taking up more board space.
    1. The 30 Ohms on the input and output of U1 provide 'sorta' ESD protection - better than nothing
  5. The TVS's clamp the voltage of the Drains of the 100Vds(max) power-MOSFETs to 77.4Vpeak to prevent avalanche failure of the MOSFETs.
  6. The TVS's can handle 7.8A peak for 10/1000us, repetitive, 100A non-repetitive

The Emergency Stop buttons control the "Return" signal: 7.5VDC when open (loop closed) to 0.00VDC when latched closed (loop open). The load is the series of 30 Ohm resistors with the 2K parallel resistors to ground. This is used to turn On/Off MOSFET Q4 which in turn hits the Vgs of Q7 and Q12 with 12V from zener D21 which also protects the power-MOSFETs from exceeding their +/-20V limit.

0 Likes
AndiM
Employee
Employee
First like received First reply posted Welcome!

Hello Mandrews,

I have a short off topic question concerning your schematic. You mentioned to use back to back configuration to make sure not to back-power. Isn’t D18 and D23 bypassing this function?  So, I think Q12 can be removed?

As Meghana question is important I have a second question. How is the input P47 buffered? With a load capacitor of 10mF it is possible to create a very high inrush current during turn-on the Input Switch: Worst case… (48V / 2 times Ron > around 1600A

 

Regards Andi

mandrews
Level 2
Level 2
5 replies posted 5 sign-ins First solution authored

Hi AndiM,

Thank you for your response, too. I really appreciate you taking your time to look at this.

You and Meghana mention inrush as the potential problem. Very possible. I can slow the soft-start by increasing C18 to reduce inrush. Right now it's well under 1ms turn-On which is likely too fast, effectively no inrush reduction. Could bump it to 5ms or so, analyzing what that does to the MOSFET's Pd and the reduction in inrush. Definitely worth looking at. Thanks.

You're right that D18 and D23 bypass Q12 when they're conducting. However they only conduct when +48V is over +48V from an inductive kick or whatever. Once the signal over +48V is gone, Q12 will still block the remaining "switched +48V" as the output capacitors discharge. Wait a second - you have an excellent point. Because as the master input +48V supply drops, the protection diodes in parallel with the power MOSFETs as they turn Off and increase their RDS(on), will pull the output filter caps with it within reason. This leads to the question can the OTS power supply absorb the caps' discharge currents or will it just take a while?

This is why I enjoy seeing things thru other peoples' eyes - easy to see something so much that you 'stop seeing it' after a while. 🙂 

0 Likes
lock attach
Attachments are accessible only for community members.
mandrews
Level 2
Level 2
5 replies posted 5 sign-ins First solution authored

Meghana and AndiM,

Thank you both so much for your suggestions. I think you hit on it. 

I had left the soft-start turned-Off thinking the power dissipation would be fairly low with everything turning On very quickly, very solid, and very low impedance. I know things can get hot really fast but this surprised me.

Here’s the simulation model. The TVS model may or may not be exactly right. I have an email into Littelfuse’s Tech Support. I’m using their SMDJ48A vs. the 8.0SMDJ48A I plan to use. Much heftier TVS: 8kW vs 600W. Not much physically bigger, though, and still SMD. The simulation doesn’t show it’s needed but real-world loads will likely make a big difference.

Here are the simulations for various soft-start cap values (C1). If you see any issues or have other suggestions I'm happy to hear them. Simulations as you know are only as good as their inputs.   🙂 

mandrews_0-1644565805272.png

Snag_4817f78.png

Snag_4817fa7.png

Snag_4817fc6.png

   

Snag_481bdd9.png

Snag_481be08.png

Snag_481be27.png

   

Snag_481dbd1.png

Snag_481dbf0.png

  

0 Likes
mandrews
Level 2
Level 2
5 replies posted 5 sign-ins First solution authored

And following up with your idea, AndiM, with the kick-back diodes at the +48V outputs, I can remove Q12 (M4 in the simulation) since I have those diodes going back to the power supply output in parallel with the MOSFETs during discharge.

The fun never ends ... 🙂 

0 Likes

The simulations show removing the input MOSFET Q12 further increases the power dissipation on Q7.

I found some power-MOSFETs with 10 milliohms of RDS(on). Life is wonderful.   🙂 

0 Likes
AndiM
Employee
Employee
First like received First reply posted Welcome!

Hello mandrews,

Some more things worth to have a look:

Maximum ratings of the IRF5210PBF, as the maximum current shall not exceed -140A

If turn on time is increased please have a look to the “Safe Operating Area” described in Fig 8 of the datasheet. The switch may be outside this SOA. Pre-charge is an interesting topic for semiconductor switches.

I can’t find R2 on your simulations inside the initial schematic. Is this the connection cable? If so I would propose to add the inductance of the cabling as well. (Roughly 1uH/meter) It is important to have correct L/R values.

If Q7 thermally died, it’s likely all the pins are shorted. D21 should be ok, please check.

 

Sorry I did not have a look to everything in detail.

 

Regards

0 Likes
mandrews
Level 2
Level 2
5 replies posted 5 sign-ins First solution authored

Hi AndiM,

Yes, D27 was fine. I appreciate your help.

Mike

0 Likes
mandrews
Level 2
Level 2
5 replies posted 5 sign-ins First solution authored

All great feedback, AndiM. Thank you. I will look at SOA. R2 represents cables and multiple connectors.

0 Likes
mandrews
Level 2
Level 2
5 replies posted 5 sign-ins First solution authored

I also increased the soft-start timing capacitor C1 to 2200uF. Reducing the Pd in the power-MOSFETs and the inrush current to the heavier load. And made the +48V power-switch single fault-tolerant.

mandrews_0-1645088003553.png

 

0 Likes