2.5A full FET bridge driver - who needs so much current?

[nuno]

Nice drivers.  I use something almost the same.  I haven't yet figured out how to make the schematic large enough to read the part labels,

Thank you . It's the "Raw" button on the right on top of the image, but here's a direct link:
https://github.com/vnevoa/DiffTrike/raw/MarkIV_RasPi_NJAY/Electronics/PowerBridge/hw/bridge-fet-drivers-schematics.pdf
I just noticed the schottky diodes's symbols somehow ended up the same as PN ones, I'll have to change it when time permits. The github project under ".../PowerBridge/..." has a 25V 25A DC motor controller project with all design files for schematics (KiCAD but there are PDFs of all schematics), pcbs and fw for whoever wants to look.

There are a few "problems" with these drivers, but I also wanted to keep component count down, as much as desperately possible. For example, the low driver doesn't keep the FET off when the input is floating (in my application this isn't a problem because the high side ones are guaranteed off) and the top output transistor is always dropping 12V (and it needs a constant, although tiny, load, guaranteed by the FET's gate-source resistors).
Thanks for the suggestions below.

but see that first resistor, through which the input signal passes to the pull-down transistor's base?  I add a Schottky diode in parallel with that.  This accomplishes two things:

1) When connected to an MCU or other device which can sink current, and it switches to logic low, the Schottky conducts and bypasses the resistor.  Allowing the full available sink current to drain the base charge, and improving transistor turn-off time.  It does so even though there's an anti-saturation diode between that transistor's base and collector, they work great together.

I currently don't have any of those drivers with me, but I was pretty happy with their switching latency, basically "zero" (switching time in my app is decent too). I have some scope shots, including comparison with no anti-sat diode at all and with 1N4148 as anti-sat, but not here with me. When I get my hands on the drivers again I'll try that 2nd schottky to see what's the difference - as it is, the driver's output (loaded with the FET) starts reacting "instantly", but from your description I guess it goes up/down faster (it is actually irrelevant in my application, because the phase's di/dt is being limited by parasitic inductance at the FET's source (which is bad and good) - hey, I'm still learning!).

2) When it switches to logic high, the Schottky is reverse biased and does not conduct.  Instead, it acts as a capacitor, in the neighborhood of 10pF.  This is a significant portion of the the transistor's base capacitance.  And so, by again bypassing the resistor, it improves turn-on time some too.

I see. The input's 0->1 transition, being fast, goes through the "schottky's capacitance" bypassing the resistor.

[nuno]

Unfortunatelly, I can't see any of them capable to provide negative Vgs, which should help overcome many problems eg. in half-bridge in synchronous buck converter 
Isn't negative Vgs a must in more demanding applications?

Nope, they do not provide negative turn off voltage. It wasn't necessary in my application, and as far as I've seen, that is used, typically, only with IGBTs (because they are more stubborn to turn off I guess, with that "current tail" or "trail" or whatever).

For those who might now know: the key term is 'Miller capacitance' or 'Miller effect'. (...)

Just to add another Miller term: "Miller plateau"

(...) I have once designed a driver which had to provide over 40A of peak drive current (and yes,  the EMI from this was extreme). (...)

Now that seems like a lot of fun!

[Chris C]

I have some scope shots, including comparison with no anti-sat diode at all and with 1N4148 as anti-sat, but not here with me.

I only recently found out a Schottky was an option for the anti-sat diode.  I'd always assumed it would limit the base voltage too much, and therefore limit current through the transistor.  But I have not yet had a chance to actually try it and see.  It sounds like you've performed this experiment.  Don't go through too much trouble finding scope traces, but since you went with the BAT54, I assume it did in fact work out better than the 1N4148?

Also, now that I can see the schematic (thanks for the help on that), and realize the anti-sat is actually a Schottky, then the Schottky I suggested adding is not the optimal choice.  The base voltage will never be high enough that my Schottky will significantly assist with turn-off, other than acting as a bypass capacitor.  Better to just use a capacitor instead!  If you get a chance to try it, I'd recommend starting with a 22pF.

[nuno]

I have some scope shots, including comparison with no anti-sat diode at all and with 1N4148 as anti-sat, but not here with me.

I only recently found out a Schottky was an option for the anti-sat diode.  I'd always assumed it would limit the base voltage too much, and therefore limit current through the transistor.  But I have not yet had a chance to actually try it and see.  It sounds like you've performed this experiment.  Don't go through too much trouble finding scope traces, but since you went with the BAT54, I assume it did in fact work out better than the 1N4148?

Much better, no comparison possible as far as I remember. No visible latency at all with the schottky (no time spent removing junction charge). I have the shots easy to reach, it's just that I'm not at home, I'll post them later.

Also, now that I can see the schematic (thanks for the help on that), and realize the anti-sat is actually a Schottky, then the Schottky I suggested adding is not the optimal choice.  The base voltage will never be high enough that my Schottky will significantly assist with turn-off, other than acting as a bypass capacitor.  Better to just use a capacitor instead!  If you get a chance to try it, I'd recommend starting with a 22pF.

Ok, thanks. I'll sure try it. I have seen the cap usage to assist removing the bjt from saturation but I think when I saw it it was something like "you can use the cap or you can use the diode", never gave it a thought about using both methods.

[T3sl4co1l]

For those who might now know: the key term is 'Miller capacitance' or 'Miller effect'. In datasheets it's usually referred to as Cgd. In great simplification: as your fet starts to turn on or off,  the voltage at the drain changes rapidly and that high frequency signal propagates through Cgd and counteracts the change of gate potential.

This is absolutely true, and a large part of the operation of MOSFETs!

But be careful with the quantities.  Measuring MOSFETs by capacitance is highly ineffective...

The charge equivalent capacitance is typically more than quadruple the value of Cgss!

The reason for this is dual: not only do you have to charge Cgd (Miller effect), but all terminal capacitances (Cgs, Cds and Cgd) vary considerably with drain and gate voltages.  Take careful note of what voltages Cgss is measured at, then refer to the plot of capacitance versus drain voltage.  Ask yourself... is that *really* useful?

So, while you can use capacitance, do make sure it's the charge equivalent: Ceq = Qg(tot) / Vgs(on), as in my example.

Tim

[nuno]

3 shots, without any anti-saturation diodes nor other prevention techniques (note the time scale, different from the next 2 shots), with a 1N4148 and with a BAT54, on the PMOS driver (this one is non-inverting). I can't remember exactly but some of the interference is probably measurement artifacts, is not physically possible to measure these 2 signals with short GND clips. Yellow is driver input, can't remember anymore what the blue trace is, the hump at the end looks like the Vcc + Vfw(intrinsic-diode) which should be visible at Vd, but dropping only 9V suggests it's Vg. Anyway, for the comparison at hand I think Vg or Vd is irrelevant, it represents the driver's output.

[Chris C]

Oh yeah, that BAT54 is performing very nicely.  Really about as good as I've seen with a standard switching diode for anti-sat, and either the Schottky or capacitor bypassing the input resistor.  Thanks!