P Channel MOSFET Gate Drive Question

[petersanch]

Ive read to control P channel MOSFETs with a microcontroller to use a N channel MOSFET and pull up resistor on the P channel's gate like in the picture.


But lots of P channel MOSFETs like this DMG2305UX have VGS thats much smaller in amplitude than VDS.


If Source connects to 20V and the N channel MOSFET grounds the gate, will this exceed the VGS limits? Then how can you control this circuit using this circuit?

Cheers

[Whales]

Add a resistor in series with the drain of the lower mosfet.  Combined with the other resistor this makes a voltage divider.

(There are other ways too.  This style of mosfet driving isn't any good for very quickly turning the P-mosfet on and off as you need for PWM, but it's fine for simple ON-OFF control)

EDIT: I'm uncomfortable seeing a 20V Vds part being used to switch 20V on and off.  Too close for comfort!  Any inductiveness in your load and that 20V will turn into a lot more during turn-off.

[sandalcandal]

Another way if switched voltage is variable would be to use a Zener diode parallel to R1 (cathode to P-MOSFET source). Will still need to add a resistor on the drain of the control N-MOSFET to limit current.

Also seconding the opinion you should typically derate your MOSFET V_DSS so pick something a good bit higher than the expected nominal 20V. 30V is probably good as long as you don't have a situation where you'll see excess transients e.g. long wires.

[petersanch]

Another way if switched voltage is variable would be to use a Zener diode parallel to R1 (cathode to P-MOSFET source). Will still need to add a resistor on the drain of the control N-MOSFET to limit current.

Cheers. This sounds helpful for battery applications!

Add a resistor in series with the drain of the lower mosfet.  Combined with the other resistor this makes a voltage divider.

(There are other ways too.  This style of mosfet driving isn't any good for very quickly turning the P-mosfet on and off as you need for PWM, but it's fine for simple ON-OFF control)

EDIT: I'm uncomfortable seeing a 20V Vds part being used to switch 20V on and off.  Too close for comfort!  Any inductiveness in your load and that 20V will turn into a lot more during turn-off.

Thanks.  I wanted to give an example to show extreme cases. I wont want to switch 20V with this part either.

What circuit do you recommend for switching the P channel MOSFET for PWM? I read about MOSFET drivers but I would like to learn other ways too.

[T3sl4co1l]

You can take the resistor divider and/or zener thing, and simply add a complementary emitter follower to it.  Gate voltage will swing in about t = hFE * Rthev * Qgs(tot) / Vgs(on).  A series gate resistor may be desirable, typically being sized a bit below the peak applied voltage and current.  That is, if you're delivering 1A peak for a swing of 10V, less than 10 ohms would be fine, say 4.7.  This suppresses oscillation, and shares current between gates if you're using multiple in parallel.

There aren't really P-ch drivers, but you can hang a regular driver below the supply rail and use it the same way.  Mind the input voltage threshold and range; you can easily pull its input below VSS in this configuration.

Better is to skip the P-ch entirely, and use N-ch which performs about 2.5x better (lower Qg*Rds(on) figure of merit for same voltage rating).  Use a bootstrap driver, and either a charge pump supply to keep it fed, or switch often enough to keep the bootstrap supply refreshed, or use an isolated supply.  If supply is isolated, it's a small step further to isolate signal as well, and then you have a fully isolated gate driver that can be connected any which way.

Tim

[petersanch]

You can take the resistor divider and/or zener thing, and simply add a complementary emitter follower to it.  Gate voltage will swing in about t = hFE * Rthev * Qgs(tot) / Vgs(on).  A series gate resistor may be desirable, typically being sized a bit below the peak applied voltage and current.  That is, if you're delivering 1A peak for a swing of 10V, less than 10 ohms would be fine, say 4.7.  This suppresses oscillation, and shares current between gates if you're using multiple in parallel.

There aren't really P-ch drivers, but you can hang a regular driver below the supply rail and use it the same way.  Mind the input voltage threshold and range; you can easily pull its input below VSS in this configuration.

Better is to skip the P-ch entirely, and use N-ch which performs about 2.5x better (lower Qg*Rds(on) figure of merit for same voltage rating).  Use a bootstrap driver, and either a charge pump supply to keep it fed, or switch often enough to keep the bootstrap supply refreshed, or use an isolated supply.  If supply is isolated, it's a small step further to isolate signal as well, and then you have a fully isolated gate driver that can be connected any which way.

Tim

Thanks Tim.
Excuse my novice questions. If using a N channel for high side switching, if the supply is 5V for example then is the voltage on the N channel gate 5V + threshold voltage? What happens if the supply voltage is 20V but the maximum rated VGS for the MOSFET is 8V like in my originally P channel example but with a N channel instead?

[T3sl4co1l]

You use a bootstrap driver, which drives the gate relative to the source.  Supply voltage doesn't factor into it.

With respect to ground, what you'll see is the gate being driven to Vsupply + Vgs(on), for a total swing of say 30V for a 20V supply and 10V drive.

Remember that voltage is always and only a difference.  We measure voltages with respect to ground, for a variety of important reasons, but devices only see whatever voltage is between their terminals.  To the transistor, "ground" is the source terminal.  (Or gate or drain; source just happens to be more common.)  We can build circuits around whatever ground reference we like.  To the bootstrap driver, its ground is also the source.  We can't measure this so easily (with our ground-referenced instruments), but it most assuredly works.

Tim

[petersanch]

Thats very helpful, thanks!

Makes me wonder how how level translate chips work internally while being able to switch high frequencies in the MHz? There are SOT23 components translating from 1.8V to 5V in the MHz. Wonder if they have charge pumps internally and do something similar with the N channel MOSFETs as you explained

[Peabody]

There are power switches, also known as "magic mosfets", which are N-channel with charge pumps built in for use as high-side switches.  They use a little current to do that, and can be expensive.  See the BTS50025-1TEA for example.

[T3sl4co1l]

Nothing fancy is needed when the supply voltages are low, and common ground.

Level shifters typically use a pair of NMOS driven complementary, pulling on a cross-coupled pair of pull-ups (PMOS).  This gets full CMOS voltage levels, with consistent timing and low supply consumption.

That is, there are two transistors pulling up, one gate connected to the other drain and vice versa, so one is pulling up strong while the other is off; yank down on the pulling-up one, and the other turns on and pulls up, then the first turns off.  It's bistable, and only draws current when changing state.

Or they may simply burn quiescent current on a more analog (CCS into floating resistor?) approach.  Some level shifters draw relatively high currents (fractional mA) -- the transistors in the middle of the circuit are quite small (100s nm) so it doesn't seem like much on the board level, but it's decidedly higher than proper CMOS.

Schmitt trigger circuits can do a similar thing, they might not draw much current most of the time but for intermediate input voltages, they can draw quite high currents (low mA), something to watch out for on low-power circuits.

Gate drive ICs almost uniformly use an input circuit with an NMOS, into resistor or CCS pullup, into a Schmitt trigger.  The NMOS gives a TTL compatible input threshold.  The pullup accounts for much or all the quiscent current drawn by the things, annoying if you want a low power gate driver.

Bootstrap gate drivers do it up to 600V with pulsed CCS level shifting.  The high side stays latched in one state or the other, and receives pulses through a high voltage transistor straddling the barrier.  The downside to this method is, it ignores inputs if the high side dips below ground (the HV transistor can't "transist" in that direction, indeed it acts as a diode instead), and if the latch's state gets corrupted, the whole thing probably destroys itself.  Fortunately, being on-chip, it has pretty good immunity, and this isn't a problem in practice.  There's also additional conditioning circuitry to ensure consistent operation during dips or rapid swings.

Still others do it by differential capacitance, through small on-chip capacitors; these can even offer full (galvanic) isolation, like TI's ISO family.  Others use onboard transformers(!), like Analog's ADUM series (was the first I think, and everyone else has them now as well).  Transformers aren't easy to make this small; they typically run at a few 100 MHz, and have poor efficiency (~50%).  Still, it's enough for a watt or so, and that's enough for a lot of common applications, like isolated serial ports, or small gate drivers.

Tim

[petersanch]

Nothing fancy is needed when the supply voltages are low, and common ground.

Level shifters typically use a pair of NMOS driven complementary, pulling on a cross-coupled pair of pull-ups (PMOS).  This gets full CMOS voltage levels, with consistent timing and low supply consumption.

That is, there are two transistors pulling up, one gate connected to the other drain and vice versa, so one is pulling up strong while the other is off; yank down on the pulling-up one, and the other turns on and pulls up, then the first turns off.  It's bistable, and only draws current when changing state.

Or they may simply burn quiescent current on a more analog (CCS into floating resistor?) approach.  Some level shifters draw relatively high currents (fractional mA) -- the transistors in the middle of the circuit are quite small (100s nm) so it doesn't seem like much on the board level, but it's decidedly higher than proper CMOS.

Schmitt trigger circuits can do a similar thing, they might not draw much current most of the time but for intermediate input voltages, they can draw quite high currents (low mA), something to watch out for on low-power circuits.

Gate drive ICs almost uniformly use an input circuit with an NMOS, into resistor or CCS pullup, into a Schmitt trigger.  The NMOS gives a TTL compatible input threshold.  The pullup accounts for much or all the quiscent current drawn by the things, annoying if you want a low power gate driver.

Bootstrap gate drivers do it up to 600V with pulsed CCS level shifting.  The high side stays latched in one state or the other, and receives pulses through a high voltage transistor straddling the barrier.  The downside to this method is, it ignores inputs if the high side dips below ground (the HV transistor can't "transist" in that direction, indeed it acts as a diode instead), and if the latch's state gets corrupted, the whole thing probably destroys itself.  Fortunately, being on-chip, it has pretty good immunity, and this isn't a problem in practice.  There's also additional conditioning circuitry to ensure consistent operation during dips or rapid swings.

Still others do it by differential capacitance, through small on-chip capacitors; these can even offer full (galvanic) isolation, like TI's ISO family.  Others use onboard transformers(!), like Analog's ADUM series (was the first I think, and everyone else has them now as well).  Transformers aren't easy to make this small; they typically run at a few 100 MHz, and have poor efficiency (~50%).  Still, it's enough for a watt or so, and that's enough for a lot of common applications, like isolated serial ports, or small gate drivers.

Tim

Thats very helpful. Thanks for your explanation Tim! Theres a ton to learn about this topic.