High frequency MOSFET Driver

[Saitama]

Hello! I have 2xG4PC50FD in half bridge configuration and a low power signal that needs to drive the IGBT. I looked everywhere and couldn't find any MOSFET driver ICs, so I want to make drivers from transistors (I have BD136 and BD139 in mind), but the problem is that the drivers need to operate at high frequency, from 100KHz to 400KHz and I don't know if this will work.

Is this good enough?

[Marco]

What's wrong with existing gate drivers? You're going to need dead time and a bootstrap for the high side ... easier to just  use an IC.

[capt bullshot]

For switching the IGBTs fast enough, you'd want transistors that are rated for maybe 4 Amp collector current. BD136 / 135 aren't strong enough. Diodes and others make special NPN / PNP pairs for this purpose.

https://www.diodes.com/products/discrete/bipolar-transistors/special-function-transistors/gate-driver-transistors/

[Saitama]

I couldn't find any Mosfet Driver ICs in my country.

[SiliconWizard]

Yeah, not sure why you would bother designing a discrete driver here...

MC33153, $2
(there are many others)

[Saitama]

I plan to use this circuit to drive only one IGBT, but if it really needs 4 amp rating should I go for BD 680 and BD 679A?

[Saitama]

I couldn't find any Mosfet Driver ICs in my country.

[T3sl4co1l]

Do you not have access to Chinese suppliers in Georgia?

Tim

[capt bullshot]

I plan to use this circuit to drive only one IGBT, but if it really needs 4 amp rating should I go for BD 680 and BD 679A?

If you plan to use the 22R gate resistor at 12V drive voltage as shown in your circuit, BD135/BD136 or BC639/BC640 will be enough. If this value is good or if you need a lower value depends on your total circuit and application. So, for example, if your circuit requires 4R7 gate resistance and you've got the usual +15V / -5V driver supply voltage, you want a 3A driver / transistor.

[David Hess]

For switching the IGBTs fast enough, you'd want transistors that are rated for maybe 4 Amp collector current. BD136 / 135 aren't strong enough.

Also current gain and Ft drop at higher collector currents so typically the maximum collector current needs to be significantly higher than the intended operating current.

Diodes and others make special NPN / PNP pairs for this purpose.

https://www.diodes.com/products/discrete/bipolar-transistors/special-function-transistors/gate-driver-transistors/

Those types of transistors are ideal although you have to watch out for their maximum power dissipation.  They use something like ring emitter like construction which yields less hfe and Ft dropoff at higher frequencies, higher Ft, and higher hfe which makes them particularly useful for high current switching applications.  I might also use the 2SB1143/2SD1683 which come in a power package.

Is this good enough?

I might use something like that shown below.

[diyaudio]

You might want to look at these babies, have them on order for a gate drive circuit I'm designing.

* 625mW POWER DISSIPATION
* IC CONT 3A
* 12A Peak Pulse Current
https://docs-emea.rs-online.com/webdocs/09c3/0900766b809c30e1.pdf

Also look at ZTX from Diodes Inc they make BJT's for this sort of applications.

BD136 and BD139 are ancient. still cost effective for medium power zener + transistor regulators, better options exists.   

[Marco]

I plan to use this circuit to drive only one IGBT

So what is driving the high side one you have in a "half-bridge"?

[magic]

Pains of BJTs can be remedied by building the driver out of FETs. Here's something I cobbled together from a few jellybean parts. Note that it requires increased supply voltage: -5V/20V.

M1/M2 can be just about any complementary FETs. Reasonable RDS(on), preferably small gate charge, preferably IPAK. Nice if it's rated for 30Vgs, or try with supply reduced to 15V, or add protection zeners.

M3 is used because a BJT was taking a long time to turn off and it looked just lame on simulation Could also give trouble in practice due to mismatched propagation delay of rising and falling transitions.

R1 must be 1W. C1/C2 recommended Make the physical layout compact and ensure that connections are short.

[David Hess]

Pains of BJTs can be remedied by building the driver out of FETs. Here's something I cobbled together from a few jellybean parts. Note that it requires increased supply voltage: -5V/20V.

The higher MOSFET threshold voltage compared to the Vbe of a bipolar transistor requires higher supply voltages.

Common source/emitter output stages swing rail-to-rail but require extra complexity to avoid turning both transistors on during transitions.

M3 is used because a BJT was taking a long time to turn off and it looked just lame on simulation Could also give trouble in practice due to mismatched propagation delay of rising and falling transitions.

Usually the storage time is irrelevant because the driver will be inside of a feedback control loop.  But if it is a problem, then the input transistor can be baker clamped with a schottky diode.  Or a fast saturated switching transistor can be used but their availability is poor.

[magic]

the input transistor can be baker clamped with a schottky diode

I actually know of this trick and tried it. According to SPICE, collector current starts to fall immediately after removal of base current (good) but it takes a linear ramp towards zero which is almost as long as the combined delay and fall time of unclamped and oversaturated transistor. Which is only worse, because now the driven FET is dragged through linear region for half microsecond.

Maybe it was good for speeding up digital circuits, but not so great for analog it seems. A more effective solution was reducing collector resistor to 100Ω, but not practical.

[T3sl4co1l]

I'm fond of this (when I'm not using an IC):



"In" obviously needs whatever level shifting is required to drive a BJT base.

Regarding Baker clamping, it reduces storage time at the expense of more Miller effect.  Since the diode isn't for free, it has some capacitance.  A big ol' BAT54H for example (>10pF) isn't going to do you any favors.  BAS71 would be more appropriate, or preferably anything of similar current rating (and capacitance), but lower voltage rating (and lower Rs, Vf).

If the base drive is from a resistor divider, a "speed-up" cap can also be used there, or the same trick (Baker clamp) done on a tap on the divider, allowing PN diodes to be used instead (e.g. 1N4148 is good enough at ~4pF):



(The BJT variant makes use of the fast recovery and potentially matched Vf of a inverted diode-strapped BJT.)

This driver,



uses the above topology, with a differential input stage to give an adjustable input threshold, and with the output driving a "complementary" pair of small MOS which drive a bigger emitter follower driving the final output (gate drive).

Complementary MOS doesn't really exist, in the same sense that complementary BJTs do; in this case 2N7002 and BSS84 were used, which are similar enough in performance.  Because this MOS inverter stage has massive shoot-thru, there's series supply impedance (R || L), which affects the rising slope (a bit slower) and the settling after the rise (it's clamped to VCC by the NPN, then gently rings down).

The final follower is PBSS4540/PBSS5540.

Performance is pretty good, 8/12ns into 10nF+2.2R.

You wouldn't need to go to such lengths for an IGBT of course; doing the above circuit with PBSS303NX/PX, or ZXGDxxxx something or other, would be more than adequate, and with the current source set for probably <= 10mA and the input transistor's drive set appropriately.

Tim

[magic]

Regarding Baker clamping, it reduces storage time at the expense of more Miller effect.  Since the diode isn't for free, it has some capacitance.  A big ol' BAT54H for example (>10pF) isn't going to do you any favors.  BAS71 would be more appropriate, or preferably anything of similar current rating (and capacitance), but lower voltage rating (and lower Rs, Vf).

Right, I have built an integrator and hence the ramp is so nicely linear.
But even with a tiny diode like BAT81 it is still quite slow if base resistance is a few kΩ. It improves if a diode is added to quickly discharge the base.
Conclusion: 2N7000 is the winner

[David Hess]

Regarding Baker clamping, it reduces storage time at the expense of more Miller effect.  Since the diode isn't for free, it has some capacitance.  A big ol' BAT54H for example (>10pF) isn't going to do you any favors.  BAS71 would be more appropriate, or preferably anything of similar current rating (and capacitance), but lower voltage rating (and lower Rs, Vf).

Right, I have built an integrator and hence the ramp is so nicely linear.
But even with a tiny diode like BAT81 it is still quite slow if base resistance is a few kΩ. It improves if a diode is added to quickly discharge the base.
Conclusion: 2N7000 is the winner

Bypass the base series resistor with a small value of capacitance to remove all of the charge at once.

Or if low voltage compliance is not a problem, drive the input transistor at its emitter instead of base.

Or if low voltage compliance is a problem, drive the base and emitter but this requires a differential signal; bipolar transistors turn off really fast when their base-emitter junction is reverse biased.  This configuration is commonly used to add a fast high voltage level shifter to a fast low voltage comparator which has differential outputs but I have also done it with CMOS logic that provides Q and -Q outputs.  I have never seen it used with a MOSFET driver but I suspect this is because nobody needs a discrete MOSFET driver which is that fast.

The 2N7000 is the winner for simplicity but I have always gotten better performance out of bipolar parts at the expense of complexity in the same application.  Even in single ended circuits, I have gotten better performance because of the high output capacitance of the 2N7000.

[T3sl4co1l]

I've used both before, see the phase inverter here, after IC1B:
https://www.seventransistorlabs.com/Class_D_Amp.pdf
The top right quadrant together is a modest capability gate driver.  One nice feature of this circuit: the current sources only draw current during commutation, allowing them to be set fairly high without overheating.  (Although not in this case due to the limited current draw of the comparator.)

Tim

[David Hess]

I've used both before, see the phase inverter here, after IC1B:
https://www.seventransistorlabs.com/Class_D_Amp.pdf
The top right quadrant together is a modest capability gate driver.  One nice feature of this circuit: the current sources only draw current during commutation, allowing them to be set fairly high without overheating.  (Although not in this case due to the limited current draw of the comparator.)

I have seen similar configurations where the drive to the common base level shifter (emitter of Q4) was also capacitively coupled to the current source (emitter of Q6) so the current source was modulated to follow the output state during transitions.

[T3sl4co1l]

Yup, that's another way to help reduce quiescent current while improving switching performance.  The downside is, supply ripple is coupled in by the same capacitor, so you need good bypassing.

The extreme case of this is a bootstrap gate driver where the level shifter must be inverting, or balanced -- if it's noninverting, it gets stuck as the rising or falling bootstrap supply, plus node capacitance, dominates over the intended signal current.  This actually helps if it's assisting the intended switching edge, but it doesn't take much ripple or bounce on the supply to cause unintended triggering of the high side gate, and then you get mysterious shoot-through events and hair-pulling.



Tim