High-side N-Channel MOSFET getting hot only when driven with PWM

[Muny]

Hi all,

I'm working on a high-powered N-channel MOSFET-only H-Bridge for driving a 200W servomotor.  I decided to use only N-channels because: 1) they are apparently more efficient, allowing higher current capabilities; 2) I have a bunch of them (http://www.ti.com/lit/ds/symlink/csd18536kcs.pdf)

I knew I would be needing a driver chip for the MOSFETs because two of them will be on the high-side, requiring a gate voltage higher than the source.  I decided on sampling one of these (http://cds.linear.com/docs/en/datasheet/1161fa.pdf) since it seemed to fit all of my requirements.

I wired up the following schematic to test the functionality of the chip:


When I drive the input to the driver HIGH @5v from the Arduino, it runs the motor at full speed completely fine with no heating of the MOSFET whatsoever, even under a mechanical load.

However, when I drive the input with a PWM signal from the Arduino, it gets extremely hot.

Am I doing something wrong?  Could it be the PWM frequency? If so, what would the optimal frequency be?

Also, if I should post my code, let me know.

Thanks for reading!

[danadak]

A MOSFET has very high gate C and that needs to be driven by a fast high current driver.
Eg. the MOSFET stays too much in the linear region and no longer acts as a switch because
the Arduino cannot source much current to drive the Gate C.

MOSFET Ap Notes

http://www.infineon.com/dgdl/an-937.pdf?fileId=5546d462533600a40153559ea1481181

http://www2.st.com/content/ccc/resource/technical/document/application_note/68/cd/c6/ab/ef/17/41/06/CD00003900.pdf/files/CD00003900.pdf/jcr:content/translations/en.CD00003900.pdf

http://www.radio-sensors.se/download/gate-driver2.pdf

http://www.ixys.com/documents/appnotes/ixan0010.pdf

https://www.fairchildsemi.com/application-notes/AN/AN-6069.pdf

http://www.ti.com/lit/an/slua054/slua054.pdf

http://ww1.microchip.com/downloads/en/AppNotes/00898a.pdf

http://www.onsemi.com/PowerSolutions/supportDoc.do?type=appNotes&category=809



Regards, Dana.

[EntropyWizard]

I think you need a sense resistor (0.01 ohm) from drain to the voltage source according to the data sheet. The driver is trying to sense current throught the mosfet and reading zero voltage drop from V+ to the sense pin.

[Muny]

I think you need a sense resistor (0.01 ohm) from drain to the voltage source according to the data sheet. The driver is trying to sense current throught the mosfet and reading zero voltage drop from V+ to the sense pin.

If I understand the datasheet correctly, it says that's only required if you need the current limiting feature.  Otherwise,

If current sense is not required in any channel, the sense pin for that channel is tied to supply.

[Muny]

A MOSFET has very high gate C and that needs to be driven by a fast high current driver.
Eg. the MOSFET stays too much in the linear region and no longer acts as a switch because
the Arduino cannot source much current to drive the Gate C.

MOSFET Ap Notes

http://www.infineon.com/dgdl/an-937.pdf?fileId=5546d462533600a40153559ea1481181

http://www2.st.com/content/ccc/resource/technical/document/application_note/68/cd/c6/ab/ef/17/41/06/CD00003900.pdf/files/CD00003900.pdf/jcr:content/translations/en.CD00003900.pdf

http://www.radio-sensors.se/download/gate-driver2.pdf

http://www.ixys.com/documents/appnotes/ixan0010.pdf

https://www.fairchildsemi.com/application-notes/AN/AN-6069.pdf

http://www.ti.com/lit/an/slua054/slua054.pdf

http://ww1.microchip.com/downloads/en/AppNotes/00898a.pdf

http://www.onsemi.com/PowerSolutions/supportDoc.do?type=appNotes&category=809



Regards, Dana.

The gate in my case is being driven by a separate gate driver chip, but I assume the same issue could exist?

[moffy]

The data sheet shows that the TURN ON time is about 250us at 12v, and TURN OFF is about 45us. Lots of transition time spent in the linear region of the FET when you use PWM. You need a faster high side driver chip.

[Muny]

The data sheet shows that the TURN ON time is about 250us at 12v, and TURN OFF is about 45us. Lots of transition time spent in the linear region of the FET when you use PWM. You need a faster high side driver chip.

Thanks for the insight and recommendation.  How low of a turn on/turn off time should I be looking for? Something in the hundreds of nanoseconds?

[Mechatrommer]

Am I doing something wrong?  Could it be the PWM frequency? If so, what would the optimal frequency be?

i agree with moffy, you pwm freq is high enough making percentage of transition period high enough to make the effiency low enough. optimal freq? zero = 0. ie at PWM 100% as you have verified. but as my general rule of thumb, transition period is less than 1/10 of fully ON/OFF period. from moffy's figure, transition period is 295us, so ON/OFF period should be minimum 2950us, hence total period = 3245us = 308 Hz, sounds too low, no wonder why by looking at the datasheet, the chip only capable of sourcing 50mA pretty low for a quad dip/soic hi-side mosfet driver chip. why dont you put the N-FET at low side and make discreete push pull transistor as driver? this is the closest thing to the ideal efficiency of mosfet gate driving... like so... with 4.7ohm Rgate, it can source/sink a bit more than 1A, several magnitude better than 50mA of your hi-side driver chip.

[rx8pilot]

The driver chip says it can turn on in 220us per 1000pf and the MOSFET is 9000pf. That is a SLOOOOOOW rise time for PWM - nearly 2ms. The driver may work ok with ultra low input capacitance or used as a DC switch. It is a terrible option for even slow PWM.

[Circlotron]

You need a reverse connected schottky catch diode across the motor. As it is now, when the gate goes low the motor inductance drags the source below ground (i.e. below gate voltage) just enough to make the MOSFET conduct, and it is pulling the full motor current through the MOSFET at 12V drop + whatever gate-source voltage needed to turn it on sufficiently to supply this current, a total of say 15.5 volts. That spells lots of heat.

[rx8pilot]

You need a reverse connected schottky catch diode across the motor. As it is now, when the gate goes low the motor inductance drags the source below ground (i.e. below gate voltage) just enough to make the MOSFET conduct, and it is pulling the full motor current through the MOSFET at 12V drop + whatever gate-source voltage needed to turn it on sufficiently to supply this current, a total of say 15.5 volts. That spells lots of heat.

The voltage spikes could easily be in the full destruction level too.

[ZeroStatic]

Circlotron is right, add a FAST diode or Schottky diode ASAP or you will kill your mosfet, I killed about $100 worth of fets before working this out, many moons ago.

[Mechatrommer]

As the figure i attached, thats pretty much the complete protection g-d,g-s,d-s n motors leads, that one can think of...noticed for the motor there are 2 diodes? In reality one small one big,that hell lot of diodes. if one want to minimize the board size that can be a conflicting situation. that however is the protection subject, but to the op problem, slow gate charging is still slow gate charging...

[garyww]

My guess it that it could be the ac loss. Essentially, the nmos spends lot of time to behave like a high resistance switch(since its gate is 5V yet, the switch has not fully turned on yet).  It becomes less efficient and gets hot.