MOSFETs heating up (Gate driver? Or maybe a snubber)

Thread Starter

KaloyanP

Joined May 27, 2012
5
Hello,

I have been working on a line follower robot as a part of my university project for quite some time now. We recently upgraded the motor to a much more powerful one, which made it horribly noisy. This was due to the low driving frequency, around 2KHz, which wasn't a problem with the previous setup. I tested the circuit at 20KHz and the MOSFETS started heating up really badly really quickly and I can't quite figure out why.

I am only switching the bottom transistors of the H-Bridge since the top ones have a very slow driving circuit and heat up immensely if switched faster than a few hundred times per second. The bottom (N channel) transistors were driven from a signal generator with an output impedance of 50ohm, 12V pp voltage.

Now, I can only see two reasons for the transistors to heat up so much:

a.) Avalanche breakdown occurs on each switch on/off cycle. This is very unlikely, as there are schottky diodes in place to prevent it. The circuit even worked with regular diodes in place, so this can pretty much be scratched out

b.) Large switching losses due to poor gate driver. Maybe the gate isn't being driven hard enough by the signal generator. I was thinking of augmenting the gate driver in some way.

For the gate driver, I was considering a totem pole with pnp/npn transistors, or something like a NOT gate made of logic level MOSFETs. However, I don't know whether shoot through wouldn't be a problem with them.

What should I check next time I am in the lab?
 

Attachments

#12

Joined Nov 30, 2010
18,224
Gate driver. MOSFETs can use incredible amps for nanoseconds at Megahertz frequencies. I'm glad you only need to get out of audible range. Unfortunately, I tried to advise one of these a few days ago and fell on my face. Somebody else is going to have to help with a driver chip.
 

antonv

Joined Nov 27, 2012
148
You should definitely use a gate driver. You can make your own as you suggest or use something like the L6384 in the attached data sheet. Logic gates alone will not drive enough current.

Change the top two transistors to N-channel ones like the lower ones since that is what this IC is designed to drive.

You will need two of the L6384s, one for each half of the H-bridge.
 

Attachments

Jaguarjoe

Joined Apr 7, 2010
767
Like #12 sez, gates can consume lots of current for a short time. The L6384 part flows 650ma at the most. I don't think that will solve your problem.

The MOSFET data sheet should specify the value of gate charge (Q). This charge must be overcome for success. Using Q = IxT, you can plug in Q along with your desired switching time and solve for the amount of current needed to get there.
 

takao21203

Joined Apr 28, 2012
3,702
I use a resistor divider + a discrete BJT to control MOSFET gate. Works well into the 100KHz range, and as much as 100V. So the divider defines -12V from the input voltage, and the BJT is switching the bottom resistor. The gate will either float to Vcc, or reach -12V defined by the bottom resistor. The smallest possible time is a function of the gate capacity and the bottom resistor.

I use 200 Ohms + 800 Ohms which is good enough for a power MOSFET.

I do not know the wattage of your motor, but upto 50W, and upto 100V, resistor and BJT can be used.

Also depends on the frequency of cours, but for 20 KHz should not be a problem. Or do you use very large MOSFETs with big gate charge?
 

Thread Starter

KaloyanP

Joined May 27, 2012
5
Hm... I will try a NOT gate made up of 2N7000 and BS250, fed from a 5V supply. If this is still too little, I will try the Microchip TC4426 for the low side and see how that goes. The high side is fine as it is f since we are not switching direction often.often. Really, there is no need to modulate the top tansistors, so I would prefer to go without the additional costs. The budget for the project is very limited on purpose as a part of the exercise, so we are trying to maximize every penny.

Takao21203, that's exactly what I am using for the top side driver and I have found the MOSFET heating up too much. The P channel is IRF 9024N (despite the N, its a P channel). Even though it has a very small gate capacitance for a power transistor (19nC), with a 1000 ohm resistor it used to heat up at 2KHz, I havent even bothered to test it at 20. I will try it on the bottom side, though, because this is where I need small switching losses.
 
Last edited:

takao21203

Joined Apr 28, 2012
3,702
I don't know the spec. for that IRF MOSFET, I used them on a prototype, driving 50W halogen bulbs which of course needs cooling. Later I changed the design to use a large power PNP, they are more rugged, while the IRF exploded very easily (I used more than 100V sometimes during testing). They also have a much larger case than these IRFs.

What are the motor specs?

I did not run into problems even well beyond 100 Khz.

What is important is a good saturation of the gate with full voltage.

Of course even smaller loads need a cooling grid. The max. current you can run through a IRF without cooling is about 2 Amps., which will make them turn pretty hot.

When the MOSFET is not fully on, this will increase loss.

I have never used motors, they are inductive load of course.

If the heating is not caused by improper driving a driver IC will improve absolutely nothing.

Instead of a cooling grid you can use a small VGA card cooler, with the IRFs mounted upside down like DPAKs, and the VGA cooler back pressing downwards on the metal tab, with added thermal grease.

It is a good solution, and very workable. So make sure it is not just heat developement from the current draw.
 
Top