H-Bridge burns out from one side without load (Randomly)!

Thread Starter

HelpMe2020

Joined Feb 24, 2021
22
Hello All,

I am new to the community and this query would be my first post on the forum. So, apologies in advance if I not strictly adhering any guidelines.

I have designed this H-Bridge and corresponding PCB was also fabricated. I simulated it on LTSpice, too. It has two variations with +12V and +24V supply. I have tested it and it has been working fine driving various DC motors. However, sometimes (randomly and sporadically on both +12V and +24V) with no load i.e no motor connected, the one side of H-bridge (both P-MOSFET and N-MOSFET) would fry out together very bad, and then those two MOSFETs are actually gone bad. I'd replace those two MOSFETs with new ones and it would start working again. Next time, randomly, other side would burn out. However, I never saw it failing or heating out whenever it starts working once and load is connected. It would work seamlessly. It randomly fries out only when no load is not connected. (Always right away when I insert +12V or +24V supply into it --> Again, randomly burns out). It would fry out right away after voltage supply is inserted if it has to otherwise it wouldn't. It never fries once the voltage is provided and circuit is stable. It would fry only randomly and immediately when +12V/+24V supply is inserted. Could this be random inrush current? Could this be mismatch of Rds? Let me re-emphasis that both the MOSFETS from one side fries out together.

Please recommend changes that might fix this solution. Any recommendation will be greatly appreciated!

Note: The above experience happened when M1 and M2 were AOD4189. I have changed the P-MOSFETs to the new ones (IPD068P) with matching Rds that of IRLR8726 to avoid shoot-through. I haven't tested the changes. I want to be double sure that I am not missing something else.

Screenshot 2021-02-24 at 12.58.06 PM.png
 

crutschow

Joined Mar 14, 2008
34,285
You need some small dead-time before the bottom MOSFET turns on after the top MOSFET on the same side turns off, and similarly before the top MOSFET turn on after the bottom MOSFET turns off (non-overlap dead-time).
I do not see that in your circuit.

Matching Rds has little to do with shoot-through, it's the overlap in switching times that causes that.

Also the 1K pull-up resistors likely result in a slow turn-on for the MOSFETs, which increases any shoot-through current duration in the linear region, causing high momentary power dissipation.

If you look at the MOSFET source shoot-through current in LTspice and I think you will find it's quite high.

Are you using the bridge just to reverse the motor (at a low frequency)?
If so, one easy way to avoid shoot-through is to have each control input drive the transistors on opposite sides (i.e. top left with bottom right, or top right with bottom left).
That way as long as the control signals some deadtime, there can't be any shoot-through on the same side MOSFETs.
 

Thread Starter

HelpMe2020

Joined Feb 24, 2021
22
@crutschow

First of all thanks a lot for your valuable inputs. Need little more clarity.

"Are you using the bridge just to reverse the motor (at a low frequency)?"

Yes, that's true. I just have to reverse the motor and it doesn't need to be high speed switching. Just regular change in motor direction.

I am bit confused. First of all, you are talking about switching the MOSFETs need dead time for each MOSFET. However, the MOSFET fries out at the start of supplied voltage when there was no switching taking place and no load connected. Probably, MOSFETs are in one state only. When they work, they work perfectly well. If they fry out, it happens only in the start when Vsupply is just provided. Maybe I missed something!

"You need some small dead-time before the bottom MOSFET turns on after the top MOSFET on the same side turns off, and similarly before the top MOSFET turn on after the bottom MOSFET turns off (non-overlap dead-time)."

How to achieve it? So, shall that be fully achieved if I have each control input drive the transistors on opposite sides? I'd surely implement that way.

I can definitely increase pull up resistors. I have to reconsider some of the parameters though. I am not aware of how to MOSFET source shoot-through current in LTspice!


P.S: I am a software professional with some background in electronics and this is my first circuit. I am kind of learning.
 

crutschow

Joined Mar 14, 2008
34,285
the MOSFET fries out at the start of supplied voltage when there was no switching taking place and no load connected.
When power comes on the top MOSFETs turn on and the bottom MOSFETs turn off.
Any overlap during that sequence could lead to large shoot-through, leading to the observed failures.
How to achieve it? So, shall that be fully achieved if I have each control input drive the transistors on opposite sides? I'd surely implement that way.
Yes.
If each control signal activates MOSFETs on opposite corners of the bridge instead of on the same side then. if you have some dead-time between the two signals, than cannot be any shoot through current.
And all MOSFETs will be off when power is applied.

Below is the LTspice of a simple concept H-Bridge that does that:
The gate-source overvoltage protection Zeners and any needed resistors for operation at 24V are not included.
I took the opto output from the transistor emitter so that the MOSFETs are off with no input.

Note that the Schottky diodes you added for inductive spike suppression are generally not needed for your application since the MOSFET intrinsic substrate source-drain diodes will perform that function.

1614234864306.png
 

AlbertHall

Joined Jun 4, 2014
12,345
At power up and before the controller gets running, what happens on the control pins for the bridge?
The hardware needs to ensure that the MOSFETs are off until the controller is up and running.
 

Thread Starter

HelpMe2020

Joined Feb 24, 2021
22
@AlbertHall

The micro-controller controls the control pins of h-bridge and it also controls the power supply to h-bridge via a 12V/24V DC relay. So, I have option to set the control pins at any state and then switch on the power supply to h-bridge. I believe that it is somehow a shot through problem (I am not 100% sure though) because when it happens my relay also goes bad - suggesting that one side of the bridge was overloaded, damaging the both the MOSFETs. Thus, the relay goes bad that was also overloaded because of shot through! (FYI, my power supplies are regular power supplies, not the batteries). Sometimes, I think there could be some kind of inrush current that might need choke or some power supply isolation, but only H-Bridges are failing sporadically. I have other FETs/BJTs/IGBTs running without any failure on the same board with same power supplies. So, I am heavily inclined towards the issues around the design of H-Bridge. (Just my 2 cents!)

Any other improvements/suggestions in the circuit will be appreciates. I have included my relay/power circuit below that powers up the two H-Bridges.

Screenshot 2021-02-25 at 1.27.11 AM.png
 

Alec_t

Joined Sep 17, 2013
14,280
R66/67 have unnecessarily high values and will slow the MOSFET switching, increasing the likelihood of shoot-through. Are they redundant?
 

ApacheKid

Joined Jan 12, 2015
1,533
I think what others' have said is likely the case, both transistors are dying, not one, but always both together and there's no load.

Very clearly a large current is flowing at that time, through both devices, so they are - by definition - both conducting at that instant.

A very simple initial option is to insert a fuse in that conduction path, this will solve the problem of destroying the devices in this circuit. You can then continue as-is yet be safe and maybe replace the fuses as and when you accidentally connect the device with no load, not a solution to the flaw but a decent workaround.

This same problem is what causes complex CMOS circuits to consume more power as the switching frequency increases, this is why many devices have low power modes, they automatically reduce the clock frequencies when idle and the total time spent in that "limbo" zone (where devices are neither on or off) is reduced and so power is conserved.

I'm certainly interested in what others more expert than I, might suggest as a design improvement, its an interesting problem.
 

ApacheKid

Joined Jan 12, 2015
1,533
I guess another slightly better solution than fuses, is to put a low value, high power resistor across the load connections. Choosing a reasonable value will ensure that there's always a load, yet when the motor is connected the losses in the resistor will be a small % of the total power.

Get some idea of the "resistance" (I say that because it's a reactive, inductive device) when the motor is running and choose a resistor that is say 10 times higher, then see if that is enough to prevent the destruction with no load, eventually you could get a value that doesn't waste much power but still draws enough to avoid the destruction.
 

Janis59

Joined Aug 21, 2017
1,834
When power comes on the top MOSFETs turn on and the bottom MOSFETs turn off.
Any overlap during that sequence could lead to large shoot-through, leading to the observed failures.
Yes.
If each control signal activates MOSFETs on opposite corners of the bridge instead of on the same side then. if you have some dead-time between the two signals, than cannot be any shoot through current.
And all MOSFETs will be off when power is applied.

Below is the LTspice of a simple concept H-Bridge that does that:
The gate-source overvoltage protection Zeners and any needed resistors for operation at 24V are not included.
I took the opto output from the transistor emitter so that the MOSFETs are off with no input.

Note that the Schottky diodes you added for inductive spike suppression are generally not needed for your application since the MOSFET intrinsic substrate source-drain diodes will perform that function.

View attachment 231361
Throwing the first sight I just jumped up from joy how wonderful circuit is this, but then the bold pich amount made all things impractical - upper shoulders have different polarity mosfets what is completely unpardonable, simply it not exist in well power and well frequency sectors. So, circuit may be used only (but with success) for tiny objects with tiny aims.
 

crutschow

Joined Mar 14, 2008
34,285
Throwing the first sight I just jumped up from joy how wonderful circuit is this, but then the bold pich amount made all things impractical - upper shoulders have different polarity mosfets what is completely unpardonable, simply it not exist in well power and well frequency sectors. So, circuit may be used only (but with success) for tiny objects with tiny aims.
I have absolutely no idea what that means. :confused:
 

Thread Starter

HelpMe2020

Joined Feb 24, 2021
22
I have taken everyone's feedback and incorporated the changes overall as follows:

  1. Implemented Cross-Control of MOSFETs to sideline the shoot-through problem.
  2. Increased Pull-Ups values all over for faster transitions.
  3. Increased Rb of a transistor to increase switching speed and current.

I feel there is a disadvantage in this circuit. It is not capable of applying brakes for the motor. We know H-Bridge is capable of four different states of motors. It manages only three I guess. If both control signal is zero, motor will coast. If either one is high motor will rotate in one direction. We can't make both high because it can literally blow out the whole circuit on fire.

Below is the circuit diagram and I have also attached the LTSpice schematic for your reference.

Screenshot 2021-02-25 at 10.32.24 PM.png

Looking for comments and feedback from everyone before I release it.
 

Attachments

crutschow

Joined Mar 14, 2008
34,285
I simulated the circuit and it works, but the P-MOSFET turn-on times are quite slow because of the large value for R4 and R14.
Why are they so high?
Changing those to 10kΩ greatly reduces that time.
 

Thread Starter

HelpMe2020

Joined Feb 24, 2021
22
So, here is once again an updated schematic with all the changes requested as follows:

  1. Cross-Control of high side MOSFETs to sideline the shoot-through problem.
  2. Increased Pull-Ups (R3, R7, R9, R12) to 10K values for faster transitions.
  3. Decreased R5 & R6 to 100ohms for transistors (Q2 and Q3) to increase switching speed and current.
  4. Decreased R4 and R14 to 10K for decent speed and limit gate current into the P-FET.

Screenshot 2021-02-25 at 11.59.30 PM.png

I'll wait for a day or two if someone suggests any other cool idea to improve something or add a feature (coasting the motor by some means). Otherwise, I'll release the design and go for PCB fab. (Rest of the circuit on PCB works fine - As per tested on last PCB board)

 

Attachments

kennybobby

Joined Mar 22, 2019
75
With respect to the original schematic posted:

What are you using to switch on the +24V power--how is that being done?
[edit] i see you are using a relay

Do you have a precharge resistor for filling the big capacitor? or a drain resistor to bleed it when it is OFF?
[edit] i see no precharge resistor.

There is no path to bleed off the charge on the gates of the FETS, once they are turned ON they will stay ON until the gate charge has been removed. i think your problem is due to stored charges holding FETS ON, and then the top P-FETs get turned ON when the +24V is applied.

The relay contacts are bouncing and arcing due to the high inrush to C42, and the contacts are likely frosted over with carbon causing intermittent contact, and this is causing the state of the FETs to be uncontrolled.

i've never seen an H-bridge using p fets and n fets such as this. i suppose it could be made to work, but as it is now there is no way to turn all the fets OFF.
 
Last edited:
Top