Question about mosfet max Vds for 12V DC motor load switching (no PWM)

[Psi]

I'm trying to figure out the min Vds for a mosfet that will be powering a large 12V DC motor. (winch)
There is no PWM being done, this is a simple forward/reverse H bridge.
The motor can pull up to 150A in extreme cases if the winch stalls (voltage probably like 9V in this situation)
and around 70-100A normally depending on load. Duty time is 30sec max.

There's a cost savings going from a 40Vds mosfet to 30Vds which I can use to get a mosfet with better Rds on, but I'm unsure how close I can ride the limit.

Does anyone have any knowledge in the area of mosfet DC motor drivers?
I imagine the fact that I'm not doing PWM will reduce any voltage spikes a lot and relax the mosfet max Vds requirements?

Any help appreciated.

[T3sl4co1l]

Automotive?

You still get spikes from turn-off.  If you build the bridge tight together and load it up with bypass caps, you can use the body diodes to clamp it for the most part.  Relaxed turn-off time helps, and is the main difference between doing full PWM with it, versus limited switching.

To keep it slow, you may have to add your own Miller capacitors -- transistors these days have so little feedback capacitance, they switch quickly even with very slow gate drive.  An R+C from D to G will do.

I recommend using several transistors in parallel for each leg of the H bridge, to widen the current path.  Keep layout tight, to minimize path length.  Route on top for the most part, reserving the bottom for ground plane, for bypass caps.  The wide paths also allow you to put caps between transistors, keeping loop inductance low.

The alternative to capacitors is TVS diodes, which could be rated for 15 or 18V nominal, which would clamp around 30V (check the datasheet; note and compare test current).  This accounts for not just back EMF, but supply inductance -- whatever the cable length from the battery (or other source) is, it has stray inductance and will store a significant amount of energy at these currents.

And yes, that's multiple TVSs, interspersed just as the capacitors would be.  The switching current can be absorbed by direct clamping, or bypassing.  Bypassing of course being the lower loss option, making PWM an option, even if you don't need it right now.

Exactly how many TVSs and transistors to use, what ratings, and what switching speed, all depends on the layout.  I would do a tentative first layout, estimate its performance, then update parts selection, and so on.  Then make the prototype, and verify those predictions, or correct them if in error.

The transistors can do the clamping directly, but I do not recommend it.  MOSFET avalanche should be a one-time thing, at worst.  Repetitive avalanche ratings, if given at all, are low; and as far as I know, incremental damage is always done, no matter the current or pulse width.  Relying on avalanche, puts a cycle limit on your device.

There were two threads in here recently (a few months ago), on basically exactly this subject, I don't remember the titles unfortunately; see if you can find them?

Tim

[bdunham7]

Is this a PM motor, a series-wound motor or what?  I was presuming PM based on your statement about the H-bridge, but just to be sure...

Presuming a PM motor, do you want the off-state to be open or shorted?  Using another element to short the motor at the instant it is turned off will have the double benefit of braking the motor and dissipating the rather large inductive kick from turn-off.  Just something to ponder.

Is this winch being driven from a vehicle electrical system that is also powered or recharged by an engine driven alternator at the same time that the winch is used?  If so, you are setting it up for the mother of all load dumps.  If anything happens to your battery, you'll get a worst-case sort of surge as you cut the current and it won't be like an inductive turn-off kick that you can snub.  I wouldn't skimp on the V_DS rating at all.

[Psi]

Thanks,

Currently the prototype is using  3x 2.2mR 40V mosfets in parallel for each part of the bridge.
https://assets.nexperia.com/documents/data-sheet/PSMN2R2-40YSD.pdf

They're running around 102C if I push a 100A resistive load through it. So I have some thermal issues to solve.
Ideally would like 0.7mR mosfets to reduce the heat a little, but they are not cheap.

Currently it has no bypass caps, only protection it has is
- one 5kW 24V TVS diode across the main supply rail
- 4 smaller TVS diodes between gate and source to protect all the mosfet gates.

[Psi]

Is this a PM motor, a series-wound motor or what?  I was presuming PM based on your statement about the H-bridge, but just to be sure...

Presuming a PM motor, do you want the off-state to be open or shorted?  Using another element to short the motor at the instant it is turned off will have the double benefit of braking the motor and dissipating the rather large inductive kick from turn-off.  Just something to ponder.

Is this winch being driven from a vehicle electrical system that is also powered or recharged by an engine driven alternator at the same time that the winch is used?  If so, you are setting it up for the mother of all load dumps.  If anything happens to your battery, you'll get a worst-case sort of surge as you cut the current and it won't be like an inductive turn-off kick that you can snub.  I wouldn't skimp on the V_DS rating at all.

Currently it goes open circuit on motor when in off-state. I'd not considered shorting it.

I've not disassembled the winch to check the motor. I did just now move a metal screwdriver around the outside of the motor case and could not detect any magnets.
There's no model number on the winch so I can't check any info.  But I am aware that the motor runs at different speed in each direction.  One direction is intended for slow speed high torque and the other fast speed low torque. So that sounds like it maybe doing something special inside and may not be at PM motor.

The motor is driven from it's own dedicated SLA battery, this battery is charged via a switch-mode charger. The charger gets it's power from another battery which is charged via alternator.

[T3sl4co1l]

Yeah that TVS ain't doin shit out there, flapping in the breeze...

Wait, have we been here before?  Was one of the threads I was thinking of, an earlier one of yours?  I'm bad with names sometimes...  Then, all I can do is repeat advice I've already given; only thing left is to heed it... or ignore it...

Tim

[David Hess]

The alternative to capacitors is TVS diodes, which could be rated for 15 or 18V nominal, which would clamp around 30V (check the datasheet; note and compare test current).  This accounts for not just back EMF, but supply inductance -- whatever the cable length from the battery (or other source) is, it has stray inductance and will store a significant amount of energy at these currents.

Would avalanche rated power MOSFETs be another alternative?

[T3sl4co1l]

Would avalanche rated power MOSFETs be another alternative?

Covered that --

The transistors can do the clamping directly, but I do not recommend it.  MOSFET avalanche should be a one-time thing, at worst.  Repetitive avalanche ratings, if given at all, are low; and as far as I know, incremental damage is always done, no matter the current or pulse width.  Relying on avalanche, puts a cycle limit on your device.

Tim

[bdunham7]

I've not disassembled the winch to check the motor. I did just now move a metal screwdriver around the outside of the motor case and could not detect any magnets.
There's no model number on the winch so I can't check any info.  But I am aware that the motor runs at different speed in each direction.  One direction is intended for slow speed high torque and the other fast speed low torque. So that sounds like it maybe doing something special inside and may not be at PM motor.

And it does all that and still has only two external leads?

The motor is driven from it's own dedicated SLA battery, this battery is charged via a switch-mode charger. The charger gets it's power from another battery which is charged via alternator.

That will certainly reduce the need to accommodate power surges.  Most of your overvoltage risk will be from shutting off the motor.  Is this a trailer loading winch?  Does it have a mechanical brake?

[Psi]

I've not disassembled the winch to check the motor. I did just now move a metal screwdriver around the outside of the motor case and could not detect any magnets.
There's no model number on the winch so I can't check any info.  But I am aware that the motor runs at different speed in each direction.  One direction is intended for slow speed high torque and the other fast speed low torque. So that sounds like it maybe doing something special inside and may not be at PM motor.

And it does all that and still has only two external leads?

Yes, so I was told.
Not sure how, maybe diodes and two field coil windings?

The motor is driven from it's own dedicated SLA battery, this battery is charged via a switch-mode charger. The charger gets it's power from another battery which is charged via alternator.

That will certainly reduce the need to accommodate power surges.  Most of your overvoltage risk will be from shutting off the motor.  Is this a trailer loading winch?  Does it have a mechanical brake?

yes to both

[Psi]

What's the general consensus about having a MCU perform the dead time and shoot through protection in a high current H-bridge?   I mean driving the high and low fet gate from the MCU using opto-fets directly, rather than having an extremal H-bridge IC in between to do that itself?

Obviously I would use a MCU hardware timer Q and notQ outputs with dead time support in the MCU.
But it still seems risky if code is the only thing preventing a dead short

Any thoughts if this is done often? or if it's considered bad practice?

Thanks

[bdunham7]

But it still seems risky if code is the only thing preventing a dead short

Think about all the other things that depend on code, like critical automotive systems.  I think an MCU is not likely to be the weakest link in such a setup and efforts to 'add safety' to it might actually increase the odds of having an issue.  Your most likely cause of a dead short will still be failure of the MOSFETs themselves, IMO.  The only issue I would worry about at all would be the states of the MCU outputs during startup.

[T3sl4co1l]

Depends on hardware support (a deadtime generator helps, aye) and desired reliability vs. software quality.

As they say, "software is eventually right".  How sure are you that it'll never lock up and short out, or do something equally unkind?

Just generically, a good idea?  Probably not...  Possible, definitely.

It's something I'll eventually get around to doing; I want to make a real time control system with that level of software responsibility, some time.  But knowing how to do all of that, so much better in analog (or discrete logic), means it'll probably be a long time before I do.

Tim