How to increase mosfet switching speed, and decrease switching losses?

Question

I hope this is not a too broad question, but what are the best practices to achieve fast switching on a MOSFET driven by a PWM signal?

My current knowledge tells me I can do two things:

1 - To use the lowest possible PWM frequency, because switching losses are higher at higher frequencies.

2 - Drive the gate with the maximum possible current, to overcome gate capacitance as soon as possible. To do this, I avoid adding a resistor between MCU and gate, or add a general purpose transistor between MCU and mosfet, so I can drive the gate with higher current.

Currently, I have a PWM that must run at least at 100kHz using a N-channel IRLZ44 mosfet, so first point is not applicable, and the second point is not enough to give me acceptable switching losses. My mosfets are overheating and I would like to find a better solution than using a bigger heatsink.

Should I look for a better mosfet? Or perhaps, should I try adding a capacitor somehow to kick in when PWM signal rises, boosting current through the gate? Or are there other ways to achieve faster switching?

Update:

I thought the question didn't need an example circuit diagram, but here goes it:

I got to this circuit based on other questions I asked in here. I'm using 5V and the load is about 1A. As you can see, I'm driving a transformer. In this configuration, I have 10 Vpp on transformer primary, and secondary elevates this to 1500 Vpp.

Based on current comments and answer, it's already pretty clear to me that using a driver is the easiest, cheapest and simplest way to achieve lower swirching losses. But if there's a way to improve the circuit without a driver, I would be interested on learning about it.

Answer 1

1. provide a suitable gate drive circuit that can sink/source a high enough current and at a decent slew rate (others have posted about a dedicated gatedrive)

2. Correctly choose your gate resistor w.r.t. gate charge curve (or total gate capacitance). Too high and you will switch slower and more switching losses. Too low and there is a chance of power cct ringing (increases your losses) and worse-case... setting up a pierce osc

3. If you are switching an inductive load KEEP the the stray inductance between the cathode of the freewheel diode and the FET very, very low (not as low as convenient as low as you can - re-layout if needed)

4. Again, if you are switching inductive load, do not overlook the reverse recovery of the diode. choose an appropriate diode

5. Minimise the gate-source lead inductance (twisted pair, short), again not short for convenience, short as possible.

6. if you are power switching, minimise stray inductance to the bulk DClink capacitor. Again not short for convenience, but as short as possible.

7. consider some form of lamina busbar w.r.t. 5

Answer 2

Either choose a better MOSFET or use a push-pull driver like this: -

Notice that this chip uses identical MOSFETs in the output stage. Here's another using the FAN7842 from Fairchild: -

You should also make sure there is enough deadtime between one turning off and the other turning on.

Both devices can be used to drive single MOSFET outputs if needed. Here's one that drives a highside MOSFET: -

Avoiding P channel devices will earn you a couple of percent more efficiency (genralism alert). This is a useful set of images to give other ideas.

Answer 3

As Andy aka advises, there are tons and tons of integrated MOSFET drivers available, and they work really well with a minimum of parts.

But in case you want a one-off design with discrete parts, here's a starting point: (The switch represents your microcontroller, or whatever is driving this arrangement)

^{simulate this circuit – Schematic created using CircuitLab}

Q1 and Q2 are a push-pull pair of emitter followers. Their output (at M1's gate) is held at approximately the same voltage as the input (modulo the base-emitter voltage), but the BJT's current gain multiplies the current available from the input.

Consequently, you'll need something connected to the input which can get up to the gate voltage you'll want to use. If you are using a microcontroller its output voltage will probably be 3.3V or 5V. You can find MOSFETs designed to work at these gate voltages, but most power MOSFETs work best with something more like 12V, so you'll want to add additional circuity to perform the voltage conversion. See driving low side of a mosfet bridge with 3.3V which also includes a more complex discrete MOSFET gate driver.

Answer 4

Good gate drive is a step in the right direction and has been stated in other answers. Now it is time to look at T1 .There will be some leakage inductance between each leg of the CT primary.When you turn off Q5 or Q6 the current is broken .Energy stored in leakage inductance will go into horrible high voltage spikes in your circuit .You must deal with this to stop Mosfet failure .When you plug in ballpark figures for this inductive energy that on your circuit is wasting and multiply by frequency to estimate power loss you will find that these losses are bad .So try to recover the wasted power to limit the voltage spikes and keep the mosfets cool.One straightfoward way to recover this energy is to build your passive snubber that burns power into a resistor so the fets do not blow anymore.Then optimise the waveforms .Now decide if you want to put the energy into the input or the output or some aux device like say what I did was the cooling fans .Now all you need to do is build a small DC/DC convertor to do this .You should be able to get 90% back without too much effort.You could also try an active clamp system .Active clamps are easy to drive .I have not implemented an active clamp.