MOSFET intrinsic diode vs. Schottky (PWM motor circuit)

[rhodges]

I am testing a PWM motor control circuit. The motor is DC 12V and about 260mA.

My first thought was maybe a 2N4401 could handle that. I fed the 5V PWM signal to a 510 ohm resistor to base, common emitter. I put a reversed Schottky on the motor and a 0.56 uF as a snubber. I believe I read on a thread that the motor acts like an inductor, so it should have the diode to protect the transistor (just like on a relay.)

The 2N4401 ran too hot, probably because it spent some time in linear mode due to the 0.56 uF snubber. I could see it on my scope quite clearly.

I got rid of the 2N4401 and replaced it with a logic level MOSFET (IRLZ24). Much better, but I did not like the time in linear mode. After removing the capacitor, the traces cleaned up nicely. There is some ringing, but I don't see it needs a capacitor.

First question: Since the IRLZ24 has an intrinsic diode, should I remove the Schottky? The PWM is 10khz, by the way. The datasheet states diode current of 18A at 1.5 volts, so it should be able to take everything the motor can push, and it should be fast enough.

Also, my "gut feeling" is that I should have a capacitor as a snubber, even though my scope suggests it is just fine. Maybe a very small one? Or no?

Sadly, it looks like I lost my Tek 2465 while testing this circuit. I smelled "magic smoke" and thought it was my circuit under test. By the time I decided the circuit was fine, smoke was coming out of the 2465 case

Thanks!

[Benta]

You'll have some stray inductance, so a reverse diode across the motor is a good idea. The rest will be taken care of by the intrinsic diode.
What you normally need are noise caps on the motor due to brush fire/arcing. For a small motor like yours, I'd suggest 10...22 nF across the motor terminals, and 2.2...4.7 nF from each terminal to motor case.

I strongly suggest that you reduce the PWM frequency. At 10 kHz you'll have quite a lot of magnetic losses in the motor, limit is usually around 3 kHz.

[langwadt]

0.56uF at 10kHz is insane.

Your snubber is WAY too large. You are forcibly charging a cap every time you turn on your FET.

Also, keep the schottky, some random 1N5819 or PMEGxxxx should work.

Normally should shouldn't see reverse voltage on your motor due tp the inertia of the rotor keeping voltage potential on the terminals, but if your motor hits a bump and had a sudden stop, the reverse EMF can and need a path to freewheel.

also need a path to freewheel in the off part of the pwm cycle, https://www.4qd.co.uk/4qd-diags/tec/chop.gif

[rhodges]

Great, thanks for the advice! The Schottky is MBR745 (7.5 volt amps). It was what I had. The IRLZ24 is what I had.

Adding 0.01 uF capacitor actually increases the ringing undershoot (switching on) from about -0.7 volts to -1.8 volts. And the ringing time increases from about 0.2 uSec to about 1 uSec.

I will leave the Schottky in and I will think about the 0.01 uF cap.

This is a personal project, by the way. Price is not an issue.

Thanks!

[strawberry]

use ~10R loss resistor in series with capacitor to surpress ringing

[Zero999]

I am testing a PWM motor control circuit. The motor is DC 12V and about 260mA.

My first thought was maybe a 2N4401 could handle that. I fed the 5V PWM signal to a 510 ohm resistor to base, common emitter. I put a reversed Schottky on the motor and a 0.56 uF as a snubber. I believe I read on a thread that the motor acts like an inductor, so it should have the diode to protect the transistor (just like on a relay.)

The 2N4401 ran too hot, probably because it spent some time in linear mode due to the 0.56 uF snubber. I could see it on my scope quite clearly.

I got rid of the 2N4401 and replaced it with a logic level MOSFET (IRLZ24). Much better, but I did not like the time in linear mode. After removing the capacitor, the traces cleaned up nicely. There is some ringing, but I don't see it needs a capacitor.

First question: Since the IRLZ24 has an intrinsic diode, should I remove the Schottky? The PWM is 10khz, by the way. The datasheet states diode current of 18A at 1.5 volts, so it should be able to take everything the motor can push, and it should be fast enough.

Also, my "gut feeling" is that I should have a capacitor as a snubber, even though my scope suggests it is just fine. Maybe a very small one? Or no?

Sadly, it looks like I lost my Tek 2465 while testing this circuit. I smelled "magic smoke" and thought it was my circuit under test. By the time I decided the circuit was fine, smoke was coming out of the 2465 case

Thanks!

Are you talking about an h-bridge or single ended drive: i.e. motor with one end to +V -> transistor -> 0V?

If it's the latter then the MOSFET's internal diode will provide absolutely no protection against the inductive kick-back from the motor. The diode needs to be connected in reverse parallel with the motor. You don't need a snubber capacitor.

[rhodges]

Are you talking about an h-bridge or single ended drive: i.e. motor with one end to +V -> transistor -> 0V?

It is single ended, just a simple low side switch.

If it's the latter then the MOSFET's internal diode will provide absolutely no protection against the inductive kick-back from the motor. The diode needs to be connected in reverse parallel with the motor. You don't need a snubber capacitor.

Okay, that makes three strong votes to keep the Schottky.

But I am curious. The IRLZ24 datasheet says the "Diode forward voltage" is 1.5 volts max at 17 amps. And the IRLZ44 is 1.3 volts max at 25 amps. To me, this suggests that the reverse diode is very strong. And the "recovery time" is 260 nS and 120 nS for those. I would think that is fast enough for a 10khz PWM.

Thanks!

[langwadt]

Are you talking about an h-bridge or single ended drive: i.e. motor with one end to +V -> transistor -> 0V?

It is single ended, just a simple low side switch.

If it's the latter then the MOSFET's internal diode will provide absolutely no protection against the inductive kick-back from the motor. The diode needs to be connected in reverse parallel with the motor. You don't need a snubber capacitor.

Okay, that makes three strong votes to keep the Schottky.

But I am curious. The IRLZ24 datasheet says the "Diode forward voltage" is 1.5 volts max at 17 amps. And the IRLZ44 is 1.3 volts max at 25 amps. To me, this suggests that the reverse diode is very strong. And the "recovery time" is 260 nS and 120 nS for those. I would think that is fast enough for a 10khz PWM.

Thanks!

doesn't help when the diode is in the wrong place

[Siwastaja]

People tend to put the correct components in the wrong places. Be careful. A diode across the MOSFET is not the freewheeling diode. The body diode cannot do this job, it's in the wrong place.

The correct circuit (just a random Google image search) is:
https://cdn.sparkfun.com/assets/9/8/6/0/f/51770c5bce395f3149000000.png
(Add DC link capacitors. And of course, MOSFET is used instead of BJT here.)

I prefer this schematic drawing style which shows that the mosfet-diode-bridge is the shortest path, and the motor is "behind the wires". This gives a better hint for correct layout.

The freewheeling/flyback diode needs to be of ultrafast recovery type (tens of nanoseconds, preferably; not thousands). Schottky works for low voltage bridges.

1) Minimize the physical loop size from DC link cap, through the freewheeling diode, and MOSFET. The diode is not placed "at the motor". The diode, MOSFET and the capacitors should be close to each other.
2) Use a combination of electrolytic and small ceramic caps on your DC link
3) Remove any snubbers. Measure the switch node (connection between the diode and MOSFET) vs. GND with an oscilloscope (preferably > 100MHz). If you have excessive voltage overshoot (say, over 30-40%) or ringing, start designing in an RC snubber. Your snubber capacitor is many orders of magnitude too large, producing too much power loss, and I'm assuming you don't have R as you are only talking about C. The R is essential. The snubber goes right across the MOSFET, another snubber can be placed right across the diode, if needed.
4) You won't likely need any schottky diode in parallel to the FET.


Nitpick: nS is nanosiemens, unit for conductivity.

[rhodges]

People tend to put the correct components in the wrong places. Be careful. A diode across the MOSFET is not the freewheeling diode. The body diode cannot do this job, it's in the wrong place.

Yes, I see it now. Right idea, wrong picture!

My schematic does have the diode across the motor. But my mental picture was different from the paper!

My scope shows very minor ringing (almost none), so I will leave out the snubber for now.

Thanks!

[T3sl4co1l]

Note that you need the diode beside the transistor, and what the diode returns to (+V), needs a bypass to FET GND in the same location.  The diode shouldn't go directly across the motor.  The FET is causing the flyback, the diode needs to be at the FET.

Tim

[MagicSmoker]

Note that you need the diode beside the transistor, and what the diode returns to (+V), needs a bypass to FET GND in the same location.  The diode shouldn't go directly across the motor.  The FET is causing the flyback, the diode needs to be at the FET.

Tim

More specifically, any stray inductance between the freewheeling diode anode and the MOSFET drain causes a delay in the transfer of current from one to the other and also induces nasty little semiconductor-destroying spikes. Similarly, stray inductance between the input capacitor and the switch/FWD will also induce nasty little spikes and ringing every switch turn-off.


EDIT - this reply is intended for the OP's benefit; Tim likely doesn't need it 

[jmelson]

0.56uF at 10kHz is insane.

Your snubber is WAY too large. You are forcibly charging a cap every time you turn on your FET.

Yes, but the whole idea of a snubber is to DISSIPATE energy, so you need a resistor in series with it.
A very low resistor of maybe 10 - 25 Ohms might be around right.

Jon

[Zero999]

0.56uF at 10kHz is insane.

Your snubber is WAY too large. You are forcibly charging a cap every time you turn on your FET.

Yes, but the whole idea of a snubber is to DISSIPATE energy, so you need a resistor in series with it.
A very low resistor of maybe 10 - 25 Ohms might be around right.

Jon

But a snubber is probably not needed here, certainly not such a large one anyway.