I'm trying to make a BLDC controler, and I'd like to use an atmega168 as microcontroler and N-FETs (x3) to control the motor. The microcontroler will output a PWM that will control the motor's position.

I've seen several Brushless controlers and the one I have here seems to have several transistors before driving the actual FET. (see image) bldc controler

Why is that ? Thanks for your help

  • \$\begingroup\$ The caption of that image is "When Photoshop Attacks". \$\endgroup\$ – Ignacio Vazquez-Abrams May 22 '15 at 13:11
  • \$\begingroup\$ didn't put any caption ! \$\endgroup\$ – nairyo May 22 '15 at 13:18
  • \$\begingroup\$ I was being facetious. \$\endgroup\$ – Ignacio Vazquez-Abrams May 22 '15 at 13:18

The controller you show is a full H-bridge controller.

Here's the rub: The motors will be running on a voltage higher than the microcontroller's voltage.

The two extra transistors are to transition from the microcontroller voltage to the motor voltage, allowing the MOSFETs to fully open and close through the voltage on their gates.

Imagine this:


simulate this circuit – Schematic created using CircuitLab

Imagine the lamp is a motor, because I didn't find a motor.

The gate of the MOSFET is now switched between 0V and 5V, but it's a P mosfet, so 5V is still: Vgs = 5V - 12V = -7V. -7V will open up the P-MOST very easily, even if it's a very insensitive one.

So what they do is:


simulate this circuit

The MOST is now being pulled closed by the 1kOhm R3, until Q2 turns on, then the MOST turns on as well. (Note it's good practise to add a 1kOhm into the gate of the MOST, to protect from the strong flank created by the transistor).

This is actually all you would need to drive it.

So then why the extra transistor? Why! WHYYY?

Easy! They are saving microprocessor signals by adding an inverter to the P-MOST. If you invert the on-switching of the P-MOST, then the N-MOST and P-MOST in the same arm of the H-bridge can be connected to one signal.

I find it arguably better to control each MOST with the controller if you can, because this allows a very important concept of dead-time to be used. Look it up if you're interested.

Please note that many MOSTs will have a maximum gate-source voltage, usually in the 12V to 20V domain, so doing this hard-switching at 24V may well be bad for a P-MOST. Then you'd have to protect the P-MOST as follows:


simulate this circuit

The two resistors now divide the main voltage, the lower resistor being lower keeps the gate at 12V under the positive rail or above, using a 36V supply, keeping the gate safe.


Power MOSFETs normally have relatively high Vth, which requires voltage higher than MCU will output. Also the gate capacitance will normally require high current (sometimes even 5A!) for like tens of nanoseconds each PWM cycle, which is way beyond what MCU can do. And of course the high side MOSFET's source voltage rises to Vbus, so to drive it's gate you really need some driver that will rise with the MOSFET.

Take a look at Silabs isolated gate drivers. I use them for the last couple of years, they are just perfect.


  • \$\begingroup\$ Power MOSTs with Id > 80A and Vth < 2V are very much affordable these days. Also, they don't require 5A gate current, if you can source it it will flow, due to the gate capacitance. But lower grade power MOSTs may be destroyed by such currents, so blindly dimensioning any circuit with a 2A+ MOST may not work out well. It's safe with the real bulky MOSTs, but the SO8 Si packages that go to 8A Id are much less robust on their gates. \$\endgroup\$ – Asmyldof May 22 '15 at 14:43
  • \$\begingroup\$ I'm just curious, could you point out some of those? I mean, if you surce 5mA like normal mcu would, no chance you ca drive a 10khz chopper to the motor. \$\endgroup\$ – Gregory Kornblum May 22 '15 at 14:52
  • \$\begingroup\$ It's not 5mA from an Atmel, it's 25mA initial as the differential is high and 15mA when the gate gets closer to the rail voltage. The internet is brimming with those Si types (the above shown BLDC controller, for example!): farnell.com/datasheets/1866542.pdf the millions of others I leave as homework. This one has a WORST CASE gate cap of lower than 3nF -> 3V rise with 10mA fixed takes less than a microsecond. But pumping 5A into its gate WILL seriously harm it. And these kinds of types come with much smaller gate capacitances as well. \$\endgroup\$ – Asmyldof May 22 '15 at 15:06
  • \$\begingroup\$ Cool, thanks for the homework. What powe do you expect to be dissipated by the MOSFET with 1 microsecond rise/fall time under, say, 24V bus voltage, 2A load, 10kHz PWM? Or in other words, why the hell would they still sell gate drivers of all kinds? \$\endgroup\$ – Gregory Kornblum May 22 '15 at 15:20
  • \$\begingroup\$ Because sometimes that dissipation and/or the rise time is not satisfactory. My point was merely to show that assuming 5A into a gate is not always needed, nor is it universally smart design. Since the gate rises faster than 1us and the turn on time is different than gate rise time and this being 100% hypothetical with a MOST I just quickly pulled off Farnell, I'm not really in a calculating mood. Hope you forgive me. \$\endgroup\$ – Asmyldof May 22 '15 at 15:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.