Discrete driver for N-FET only H bridge - Sanity check

[Psi]

As part of a cost-down project I'm working on an Hbridge redesign to remove the 4x isolated DCDC bricks and 6x opto-gate drivers. I think they're overkill for this application so they are out and i'm trying a discrete non-isolated bridge driver.
The H Bridge controls a 100A DC motor but it's just fwd/rev on/off, no pwm.

Does this discrete H bridge driver circuit make sense?  Or does anyone have tips for improvements in design or protection?
(Schematic only shows half of it, but the other half is just a mirror.)

The "24V_Boosted" rail is the output of a single 5V to 24V DCDC brick.
Max DCDC output current is only 45mA @ 24V, but it has 100uF of bulk capacitance.
The "12V_Boosted" is just a 12V linear Vreg from the 24V above.

I'm a little nervous about the high-side driver and pulling the driver to GND via Q3. Which might try to pull the base below emitter at turn off. So I've added that diode to try address the issue.

The P and N BJT are nothing special BC807-40 and  BC817-40
The main switching N-Fets are TPH1R204PL1 which are serious little things (150A / RdsOn ~1mR )

R6 and R72 are just there to give me an easy way to measure drive current peaks on PCB.

Any thoughts?

Thanks.

[Zero999]

It looks sensible enough to me. The only concern I'd have is the huge starting current damaging the MOSFETs. It's quite normal to monitor and limit the current to large motors. You also need to be sure your controller doesn't reverse the motor, whilst it's already turning.

[Psi]

It looks sensible enough to me.

Thanks

The only concern I'd have is the huge starting current damaging the MOSFETs.

Unlikely, each fet can handle 500A peak and there's 4 of them.  The dead short circuit current of a 24Ah 12V SLA will be well under 2000A.  Also the system does have a Isense system that feeds the MCU comparator which slams the high side fets off if it trips.

You also need to be sure your controller doesn't reverse the motor, whilst it's already turning.

Yeah, that is less of an issue since the motor is a geared winch that stops almost instantly when you disconnect power.
But I have a short delay in the code just to be safe.

[KT88]

Eyeballing the values, I would be concerned about switching losses: To charge up the gates it would reqiure more than 1A peak to drive the MOS-FETs fast enough...

[Psi]

I was trying to decide if I should use a 5V to 24V DCDC brick for the 24V rail.
Or use a 5V to 12V DCDC brick that has its negative connected in series with Vbat to generate 12V above Vbat

Getting 12V above Vbat would be ideal to make sure the high side fets always get 12V to turn on their gates, instead of 9V if Vbatt is 15V

But on the other hand, 9V on the gate should still be ok.
And I really dont like the idea of making a series link between the 24V gate control voltage system and the 12V vbat rail, which is probably going to have at least some ringing on it from the motor kickback. 
Yeah, i don't like that idea at all.  A totally separate 5V to 24V DCDC brick is much safer so i will go with that.

[Psi]

Eyeballing the values, I would be concerned about switching losses: To charge up the gates it would reqiure more than 1A peak to drive the MOS-FETs fast enough...

The DCDC wont be able to supply 1A, it's around 45mA at 24V, but I can have lots of bulk capacitance for a one shot turn on.
And the turn on/off speed is on the 'human operator' timescale, push button on, let go button off.  No fast switching is needed.

I made a couple of mods to the circuit since my first post.
Now the 15R have moved so you only have 15R to the gate resistors from 24V instead of 30.
And each driver has its own 15uF 35V ceramic cap nearby.

I can probably lower those resistors a bit more,  just need to protect from too much shoot through current.

[magic]

These common emitter BJTs will have high cross-conduction peaks when Q3 drain voltage is in the middle.
Wouldn't emitter followers work instead?

You also need to consider what happens when the driver fights against the gate protection TVS.

[Psi]

When dealing with a single turn on, how does one calculate if the switching loss is acceptable for a mosfet to handle.
Obviously it would have to be very very slow for the mosfet to show any heating from a single turn on.

But. Can a turn-on be slow enough that the mosfet is causing itself permanent damage without any obvious signs?
And if so, what datasheet parameters do I need to take into account to confirm the switching speed is fast enough to not cause permanent damage. 

Thanks

[Psi]

These common emitter BJTs will have high cross-conduction peaks when Q3 drain voltage is in the middle.
Wouldn't emitter followers work instead?

You also need to consider what happens when the driver fights against the gate protection TVS.

Could do, the gate threshold is 1.4V So pulling down to 0.7V would probably work.


Yeah, the 24V driver fighting the 12V TVS is something that did occur to me.
But as far as i could see,
1) It's only an issue on the high side driver/fets. (Low side get 12V not 24V).
2) It's only an issue in the moment the high side is commanded to turn on, when the big high side bridge fets are off but the driver transistors change to high on.
For a short time there's a path through from the high side fets through the load to GND. So you get 24V across the 12V TVS through 15 ohms. or 1.6A, just until the high side bridge fets turns on.  Is not ideal, as it robs energy that could otherwise go into turning the fets on faster.  I'm open to suggestions.

[KT88]

The DCDC wont be able to supply 1A, it's around 45mA at 24V

That is not a problem, fortunately. The average current mostly depends on the gate charge and the switching frequency. The peak current doens't affect this.
The switching losses hoever depend mostly on the time it takes to switch the MOS-FET on or off. Switching faster could on the other side result in higher EMI...

[T3sl4co1l]

Transistors are backwards... got a nice short circuit there.

Tim

[KT88]

Transistors are backwards... got a nice short circuit there.

Yes, there should be emitter followers

[strawberry]

The only concern I'd have is the huge starting current damaging the MOSFETs.

Unlikely, each fet can handle 500A peak and there's 4 of them.  The dead short circuit current of a 24Ah 12V SLA will be well under 2000A.  Also the system does have a Isense system that feeds the MCU comparator which slams the high side fets off if it trips.

was that 500A for tmax = 100 μs

[Psi]

The only concern I'd have is the huge starting current damaging the MOSFETs.

Unlikely, each fet can handle 500A peak and there's 4 of them.  The dead short circuit current of a 24Ah 12V SLA will be well under 2000A.  Also the system does have a Isense system that feeds the MCU comparator which slams the high side fets off if it trips.

was that 500A for tmax = 100 μs

Yeah, fair point.

I have datalogged the current (100hz) and got 240A peak. I also had a look on the scope and confirmed it was around that point in case the datalog was missing the peak.

The fets are 150A continuous, 270A Silicon limit,  and 500A 100us.  Plus i have 4 in parallel.
But it will be important to confirm they are all switching on together, but should be ok with 240A peak.

[Psi]

When using emitter follower for the gate driver can i guarantee a FET with a 1.4V minimum Gate Threshold will fully turn off from the 0.7V emitter follower output low?

[Circlotron]

I'd also sus out the cost of a single set of ground referenced MOSFETs connected to a pair of changeover relay contacts. The relay contacts become the H-bridge. Also, the relay would be switching at no load so the contacts will last practically forever, and you only need one quarter of the MOSFETs and the attendant drive circuitry.

[Psi]

I'd also sus out the cost of a single set of ground referenced MOSFETs connected to a pair of changeover relay contacts. The relay contacts become the H-bridge. Also, the relay would be switching at no load so the contacts will last practically forever, and you only need one quarter of the MOSFETs and the attendant drive circuitry.

I agree that would be better, but there's not enough room for that many relays. And the main project requirements are to get rid of all the relays
I managed to secure one relay in the design just to act as a master power enable and prevent reverse polarity to the H-brdige.

[Psi]

Transistors are backwards... got a nice short circuit there.

Tim

Ok,   Have updated to emitter follower.
There are no base/emitter protection diodes, are they needed for this sort of configuration, or are the 1k resistors enough for protection?

[T3sl4co1l]

When using emitter follower for the gate driver can i guarantee a FET with a 1.4V minimum Gate Threshold will fully turn off from the 0.7V emitter follower output low?

Yes, it's fine.  It'll be a little slow getting there, perhaps, but that's fine (as I recall, preferable even, for your application?).

Ok,   Have updated to emitter follower.
There are no base/emitter protection diodes, are they needed for this sort of configuration, or are the 1k resistors enough for protection?

Correct, no need.  Also don't need the base resistors.  The gate resistors handle that, and more accurately (and, individually besides!).  Which since you're not doing much current here anyway, those should probably be more like 100R or something?

Q28 would be better as an NPN with emitter resistor, so that it drops the voltage instead of R212, and less voltage swing --> slightly faster switching.  R525 can then be much smaller, to the same end.

Note that Q142 etc. can turn on pretty hard, making Q21 not very useful: if that's the case, then replace it with a diode.  Then you just have boosted pull-down (R354), which makes sense, a resistor is gonna suck.

R58 and D13 aren't needed.  Make 24V_Boosted relative to source, or Batt_12V, to avoid overvolting the gates if that's the concern.

Also, check R202 vs. R202.

Tim

[Psi]

R58 and D13 aren't needed.  Make 24V_Boosted relative to source, or Batt_12V, to avoid overvolting the gates if that's the concern.

I'm trying to power multiple high-side drivers from a single 24V DCDC module (though not all at once), So I can't get the 24V relative to source.
I have the choice of connecting the DCDC isolated module to GND for a +24V rail completely separate from Vbat
Or, using a 12V DCDC module and grounding it to +12 Vbatt for ~24V.
I can see advantages both ways.

Note that Q142 etc. can turn on pretty hard, making Q21 not very useful: if that's the case, then replace it with a diode.  Then you just have boosted pull-down (R354), which makes sense, a resistor is gonna suck.

R58 and D13 aren't needed.  Make 24V_Boosted relative to source, or Batt_12V, to avoid overvolting the gates if that's the concern.

Is this the sort of thing you mean. (ignore the 24V, it would end up ~12V when used as high side driver, I just GND'ed it in LTspice to have a play)

[T3sl4co1l]

I'm trying to power multiple high-side drivers from a single 24V DCDC module (though not all at once), So I can't get the 24V relative to source.
I have the choice of connecting the DCDC isolated module to GND for a +24V rail completely separate from Vbat
Or, using a 12V DCDC module and grounding it to +12 Vbatt for ~24V.
I can see advantages both ways.

The latter is best: Vgs(on) is always 12V, independent of Vbatt.  No need for a bootstrap supply, or rather, it's a common bootstrap rail for all drivers, and, the driver's output swing is the same, it's just a little slower perhaps due to not being bootstrapped at AC as well.

Whereas the "free standing" +24V rail, if the converter is powered by Vbatt, would lag behind, so you get overvoltage momentarily at shutdown I guess.  Or under high-batt conditions (charging? load dump?), it might not have enough voltage at all.  (If applicable, of course.  And you can have logic to simply not operate into such conditions.)

Is this the sort of thing you mean. (ignore the 24V, it would end up ~12V when used as high side driver, I just GND'ed it in LTspice to have a play)

Yup!

Don't need R6 (it acts in series with R5), and R2 may be optional.

Which, let me see... R2 would be most relevant, I think, when the 24V rail is first starting up, and Vbatt is low/zero, so that the logic level could try and backfeed through Q1 B-C, then... Q2 B-E still blocks it, up to Veb(Q2) + Vbe(Q1) at the input, which, 5V < 7V so it'll never be a problem; yeah, Q2 is optional then.

Tim

[magic]

Your last attempt gave me an idea.

The point here is that Q2 pushes enough current into R4 to pull the drivers +7V above the source, whatever it is, and avoids triggering gate protection. When the GPIO turns off, R4 pulls down the drivers.

And BTW, there are jellybean transistors with more peak current capability than BC547/557

[T3sl4co1l]

Yup, also a good way, double level shifter, doubly free of voltage variations. (A fixed 24V supply would be fine for that.)

Tim