Fail Of The Week: Flaming Brushless Motor Controller

Yep, smoke and flames are usually a sign that your electronics aren’t functioning as expected. This is actually the second failure encountered while learning about brushless motor controllers.

[Michael Kohn] purchase the motor while working on a different project and it went unused for quite some time. When he came across it again he decided he should learn the not-so-dark art of BLDC control.

The first hurdle was to figure out how to drive the three-wire motor when he had been expecting only two. The answer required him to come up with switching mechanism that allows three states for each wire: positive, negative, and not connected. His solution was to use MOSFETs. It’s a good idea, but unfortunately during the first iteration they were under-spec’d and he scared the crap out of himself when one of them blew up during testing (clip #1 below). After sourcing a more robust set of MOSFETs [Micheal] went back to testing which is when this little fire broke out. The 22 gauge wires connecting the Lithium battery to the driver just couldn’t cut it. See for yourself in the second clip.

It’s been awhile since we’ve said it: Please remember the Fail of the Week is not about ridiculing the hacker who was gracious enough to document his or her failure. It’s about learning from the mistake and discussing alternatives that can help others in the future. For instance, in this case some advice in determining MOSFET specs and wire gauge for any type of motor would be quite helpful. Have at it in the comments.


2013-09-05-Hackaday-Fail-tips-tileFail of the Week is a Hackaday column which runs every Wednesday. Help keep the fun rolling by writing about your past failures and sending us a link to the story — or sending in links to fail write ups you find in your Internet travels.

41 thoughts on “Fail Of The Week: Flaming Brushless Motor Controller

  1. What you need is real gate drivers, and much smaller gate resistors. The idea is that the MOSFET needs to quickly switch from ‘off’ to ‘on’ state, and back. The more time it spends in the halfway territory, the more power it starts to dissipate.

    Also, the gate needs some real voltage. A 3V3 output from a controller pin is usually not enough to fully turn the MOSFET on. Most of them need ~10V to work at their best.

    1. Agreed.

      When you turn off the bipolar gate drive transistor, you have three things slowing down the transition: the slow turn-off of the bipolar itself, the large capacitance of the gate, and the pull-up/down resistance.

      An dedicated IC gate driver will fix this, plus also enforces deadtimes, and prevent “FIRE!” states should the MCU glitch.

      But if [Michael] wants to stay in the realm of discretes, a fairly respectable gate driver can be built from two NPNs, a diode, a few resistors, and an optional cap. It acts as an inverter, with voltage level conversion, active drive for both low and high transitions, and switching times below a microsecond. I think I originally saw it in a TI app note some time ago, but I forget which one. I can post my personal schematic later today if anyone requests it.

      1. Page 24 figure 48 in TI’s TPS2811-2815 (Dual MOSFET Driver) datasheet.
        Passives CR2, CR3, R11, R12 and the input capacitance of the drivers
        adds a delay. Unfortunately, it is dependent on the layout / wiring and
        your circuit, so the R values needs to be adjusted. Alternatively, most
        PWM timers allow you to generate non-overlapping gate signals.

        http://www.ti.com/general/docs/lit/getliterature.tsp?genericPartNumber=tps2812&fileType=pdf

        1. http://img853.imageshack.us/img853/5509/cqjm.png

          VCC isn’t critical, it still works good down to 5V, or up to 24V (assuming R1 is 1/4W). Input voltage also isn’t critical, it will work with 3.3V or 5V, so long as the input can source and sink current (any typical MCU pin). I added one optional diode (D2) beyond my original description, which speeds up Q1 turn-off time a bit more.

          I consider the level-shifting inverter a fundamental building block, and it’s useful for other things besides gate drives. I have a few pre-built on perfboard, with a male header so I can use them on breadboards.

  2. Yeah people forget that mosfets act as resistors when between there on and off states and not just get very hot but give strange outputs and waste a lot of power … probably would not have solved this that SOUNDS like he is not triggering the fets right and using way to think of wire but its somthing that he is also doing wrong … a simple totem pole driver with some NPN and PNP resistors can do wonders in a proper brushless and stepper controller! … not to mention on powersupplies or most things that control large fets!

  3. check stats on the motor & power setups (batt, ESC, motor). a lot of the hobby RC motors do draw 30 – 60 amps peak depending on load (prop). check wattflier or any other electric RC forum for help determining batt, ESC & motor combos….

  4. Yeah. I had a heck of a time trying to explain to a professor that a standard brushed dc motor controller is not as a brushless dc motor controller. BLDC is 3 phase ac. As such, the simplest method to do this is how commercial 12v DC to AC inverters work (but 3 phase instead)

    http://en.wikipedia.org/wiki/File:3-phase_inverter_cjc.png

    That method takes care of switching as well as shunting the back EMF that can possibly destroy your FET/BJT. If you are going to build something robust, that is absolutely necessary. Yes, it can be a headache to drive a bridge like that with a uC, but it gives you some amazing control (variable frequency, etc).

  5. A good rule of thumb, look at the leads on your other RC toys. 22awg is for something like a nano heli that charges off the controller. The motor you have there is looking large , substantially larger than the micro heli’s pencil eraser sized motors. You will need to consider C rating with such a large motor and the leads, and connectors, batteries and solder should be chosen accordingly.
    Silver solder for high current applications, it stays solid to higher temps. Be sure your connectors , wires, and lipos are capable of the currents you need.
    I would have expected the second video to end in a lipo fire, it is probably fortunate the leads vaporized first.

    1. C rating is the capacity of a battery – doesn’t really have much to do with wiring, solder or anything else. The max amperage of the battery is quoted as a product of the C rating but the amp draw is dependent on the motor (11 amp max for that particular motor, according to the specs). 11 amps isn’t going the bother most LIPO batteries in the size range of the battery in video. GIven that he was using 22 gauge wire, which according to the awg guidelines is rated for 7 amps a LIPO fire was even more unlikely – he unwittingly included a fuse in the circuit.

  6. The BUK9509 part has input capacitance of 2700pF and total gate charge
    of 32nC. So to fully turn on/off the MOSFET, you have to charge up /
    discharge the gate capacitance very quickly. Typically you want that to
    be <100ns so that you MOSFET isn't wasting power heating itself up in a
    switching circuit.

    When making a H bridge, you also have to make sure that one of the
    MOSFET get a change to be turned off before turning on the other one.
    Slow drivers limits your motor speed/power and heats up your MOSFETs.

    10k resistance on open drain driver means that PMOS will take a few RC
    time constants and probably around to turn off. Probably around 50us for this!

    Use a proper IC gate driver rated for 1 – 2A for that size and treat it as a high speed circuit i.e. low inductance decoupling, short wire and proper return path.

    1. In addition to the gate charge, the driver also needs to overcome the Miller effect. As the FET switches, the output voltage swings rapidly, and this creates a negative feedback on the gate, due to parasitic capacitance inside the FET. If you look at the gate voltage on the scope, you can see a clear plateau right in the area where the FET is dissipating. A strong gate driver helps to minimize the length of this plateau.

  7. Moreover, according to the webpage, Michael needs to use at least 6-step commutation (tables on the page show 3-step commutation) for smooth and efficient motor operation. In addition to gate drivers, of course.

  8. Having spent a lot of time fickin’ about with MOSFET gate drives (for an induction heater, 3 or 4 years ago. Ended up building nearly 7 different topologies), I would avoid any discrete driver like the plague. Not to say it can’t be done, there’s just far better solutions readily available.

    Some drivers I like: UCC37322/1 (Texas Instruments) is a 9A peak device. TC4420/4429 (Microchip) is 6A peak. Both are available in inverting and noninverting configurations, which makes it easy to drive a gate-drive-transformer (push-pull) should you ever need isolated drive. Gettin’ bigger, TC4451/4452 are 12A peak, and come in a TO220 package. Gettin’ real big, IXYS do drives up to crazy currents like 40A peak.

    For things like half-bridge drives, International Rectifier do some neat chips like the IR2110 which is a dual high/low side driver with a bootstrap supply up to 600V.

    1. What’s the point of building anything yourself when you can just buy it already made?

      On this premise, let’s start up a business that sells pre-crashed RC toys. No need to source the parts, build it, drive it or crash it. Just buy it built, driven and crashed, all done for you!

      1. +1 what is the point of designing something building it and then destroying it when somebody has already done it? where is the fun in that? Is there a “buy it now button” for the burned build?

  9. The first design actually has another FAIL which I haven’t seen mentioned here, or in the article; the power supply for the motor is 7.4V, the microcontroller driving the MOSFET runs at 3.3V, so it’s output voltage will be a little less than that, if it’s a bipolar output stage. This leaves the gate-source voltage at -4.1V or more, while the MOSFET he used has a gate-source threshold voltage of -0.4/-0.7/-1 (min/typ/max). The P-channel MOSFETs simply can’t be turned off this way!

    The REAL lesson from this is obviously DO NOT TEST NEW STUFF WITH BATTERIES AS YOUR POWER SUPPLY!

  10. If you feel the need to keep reminding everyone not to make fun of the fail posts, why do you even continue featuring them? Honestly, it comes off as a bit condescending, like “We know this guy is an idiot so everyone else is obviously going to rip him up.”

    And if those kind of comments actually are a real problem, once again I say: Why do you keep featuring fails? I enjoy trying to diagnose the issues with these projects, but begging us not to make fun of the maker seems insulting to them, you, and us.

  11. Nice fail. Hey, we’ve all done it…let the magic smoke out. I’ve burned up some $100 modules by connecting them backwards, toasted a whole motherboard once by forgetting to pull the line cord before trying to swap memory modules (toasted the brand new memory module, to boot).

    The important thing is that we can learn from others’ failures (while still laughing at them, of course). Failing is part of making. It’s even better if it’s someone else’s fail that makes your design better. But don’t get too cocky — there’s always something you haven’t thought of. Just make sure you have a video running when you first put power to it — that’s the *real* lesson here!

  12. nobody mentioned that bldc controllers use current regulation on the coil drivers? thats why they dont blow up….

    Driving mosfets properly is important,
    Flyback diodes are also really important to keeping a mosfet cool, dont use “that internal diode”
    Anytime your doing power switching that needs a mosfet, put a fuse on the power supply.

    The commutation is an independant process, the current set point on the coils determines the speed the motor will be running at.

    1. current regulation is not required, though nice to have if for nothing else short circuit protection and start up.

      The intrinsic diode works just fine it is usually rated the same current and power as the FET, it is potentially slow but the FET should be on during most of the fly back period to avoid the diode loss anyway.

      speed is proportional to voltage, torque is proportional to current

Leave a Reply to not importantCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.