Ideas for driving high side NMOS

Discussion in 'General Electronics Chat' started by Robin66, Jul 19, 2016.

  1. Robin66

    Thread Starter Member

    Jan 5, 2016
    102
    3
    I'm developing a 3-phase brushless motor controller and have been pondering ways to drive high-side NMOS. I stumbled across an ingenious neat trick (left pic). However since my power-supply can be up to 60V I run into trouble with the max Vgs of Q1 (typically 15-30V) and the V tol. of C1. Therefore I have made a modified version (right pic), which now includes 2 zener diodes (~20V). This necessitates the inclusion of R5 and R6. Now my circuit is a lot less neat and R5 will slow down the transitions. Does anyone have a better suggestion to protect the gate of Q1? I could use 2 caps in series to double the V tol. of C1 so I'm not so worried about that problem.

    PS. I should have mentioned that this circuit is the top-part of half an H-bridge. I've replaced the lower-part with R1. Therefore R1 can be thought to vary between Inf and ~5ohm. It's important that the top of R1 is allowed to float when Q1 is off since I'll be sensing back-EMF from this point

    Capture.PNG
     
    Last edited: Jul 19, 2016
  2. GopherT

    AAC Fanatic!

    Nov 23, 2012
    5,986
    3,730
    Why not use a p-channel mosfet? Great p-channel mosfets are available with good switching speed and low R(on).

    Why not use a mosfet driver designed to do what you want (if you want to stick to an N-channel).
     
    shortbus likes this.
  3. AnalogKid

    Distinguished Member

    Aug 1, 2013
    4,516
    1,246
    If your goal is to "saturate" Q1 or Q3, neither circuit will work. Yes, you do need to clamp the gate voltage so it never is more than 20 V above the source. But the gate must be greater than the drain for saturation, and since both are driven from the same 60 V source, this can't happen. This is a common problem when using an N-channel MOSFET as a high side switch. There are many gate driver circuits on the web, and isolated high voltage gate driver chips specifically for this. Or, invert the drive logic and use a P-channel FET. You get slightly less FET performance-per-dollar, but need a much less complex driver circuit.

    ak
     
  4. Robin66

    Thread Starter Member

    Jan 5, 2016
    102
    3
    That's not correct. When Q1 is off, C1 is charged to 60V (thru R1). When Q1 is on C1 reverse biases D1 and takes the gate voltage to 60V above the source.
     
  5. Robin66

    Thread Starter Member

    Jan 5, 2016
    102
    3
    I'm using n-channel for the usual reasons: far superior devices, lower cost, lower unique component type count...
    Yes I might end up resorting to a dedicated chip, but where's the fun in that?
     
  6. benta

    Member

    Dec 7, 2015
    101
    24
    Look at IR2111 or other devices from the same series (IR21xx).
    They'll do exactly what you want.
     
  7. Robin66

    Thread Starter Member

    Jan 5, 2016
    102
    3
    IR2111 isn't actually suitable because it doesn't allow the output to float ie. both high and low side off.
     
  8. benta

    Member

    Dec 7, 2015
    101
    24
    Then a IR2101 or IR2102. I said "from the same series".
     
    Robin66 likes this.
  9. Robin66

    Thread Starter Member

    Jan 5, 2016
    102
    3
    Yes that looks suitable. They're fairly cheap too: £1.50 for 5 from China. I feel like a cheat but I might take this route.

    PS. I've found spice models for these. I'll simulate their performance, but if they live up to the datasheet it's gonna be a no-brainer
     
    Last edited: Jul 19, 2016
  10. crutschow

    Expert

    Mar 14, 2008
    12,986
    3,226
    If you still want to go discrete, here's an LTspice simulation on my take of your circuit.
    I added a transistor to provide a faster gate risetime, and diode to provide a faster fall time.
    You can, of course, use a MOSFET for Q1 if you like.

    upload_2016-7-19_8-57-4.png
     
    Robin66 and cmartinez like this.
  11. cmartinez

    AAC Fanatic!

    Jan 17, 2007
    3,555
    2,513
    Question, what pwm range are you planning to use? for instance, 0-50%, or do you need the full 100% range?
     
  12. Robin66

    Thread Starter Member

    Jan 5, 2016
    102
    3
    Nice. That's the next step up from my mods. Thx for the ideas guys. I think the take home is going to be go down the IC driver route.
     
  13. Robin66

    Thread Starter Member

    Jan 5, 2016
    102
    3
    This is for the H bridge of a 3 phase motor so the duty cycle is 33%. I realise my original solution wouldn't work for a 100% duty cycle
     
  14. cmartinez

    AAC Fanatic!

    Jan 17, 2007
    3,555
    2,513
    If that is the case, then I suggest you take a look at this thread. I'm currently building the thing, and plan to post my results when I'm done. It's not a discrete component circuit, since it uses a driver IC, but it does have its advantages.
    I understand you want to measure back EMF (possibly for speed control purposes), I believe that is possible using my design.

    BTW, if you wanted to go full 100% duty cycle, then I suggest you take a look at this. It's an opto isolated floating supply driver, it's drawbacks are that it's a bit more expensive, and the PWM frequency would not be as high as with other methods.
     
    Robin66 likes this.
  15. benta

    Member

    Dec 7, 2015
    101
    24
    The IR21xx devices have been on the market for 30+ years and generally work every time with no issues. Lots of application notes are available.
    I'm not certain, but seem to remember there was a device with 3 + 3 drivers available. Or it could be one part with three high-sides, and one part with three low-sides.
    Check the Infineon web-page, you'll see what they have (no affiliation).
     
    Robin66 likes this.
  16. Techno Tronix

    Member

    Jan 10, 2015
    140
    10
    I agree. IR2101 or IR2102 are high speed power MOSFET and IGBT drivers with independent high and low side referenced output channels.

    Are you using separate supply for the high-side driver? Or depending on the output switching to provide a bootstrap?
     
  17. Robin66

    Thread Starter Member

    Jan 5, 2016
    102
    3
    The supply is effectively common, so I'll be using output switching to bootstrap. The battery powering the whole rig will be 24V, but I'll incorporate a boost converter along the motor path which when active can step-up the H-bridge supply to a max of ~60V. The current plan is to control/monitor everything with a pic16f882
     
  18. Robin66

    Thread Starter Member

    Jan 5, 2016
    102
    3
    Thanks a lot Benta. I'll have a dig around. I'm sold on the FET drivers. I was hoping that I was missing an easy win with the discrete method but it seems not. From the datasheet it's not clear to me how the IR2101 will respond to changes in the H-bridge supply (as mine will vary 24-60V) or how it will allow the mid-point to float when both FETs are off. Will the SPICE model be accurate in those regards? I'm not sure.
     
  19. Robin66

    Thread Starter Member

    Jan 5, 2016
    102
    3
    Thx for that cmartinez. That thread goes thru a lot of issues that I've encountered.
    Here's a little observation: I always start off trying the simplest discrete implementation and then modifying as I come across its limitations. The disadvantage is that it's a slow process. The big advantage is that by the time I throw in the towel and get an IC I appreciate what the IC is doing internally. eg. in this example I'm aware that the boast cap needs time to charge.
     
    Last edited: Jul 20, 2016
    cmartinez likes this.
  20. Robin66

    Thread Starter Member

    Jan 5, 2016
    102
    3
    This is simulating nicely. I'm quite impressed with the max dVs/dt rating (slew rate of high-side reference). The low-side driver is actually performing worse than my discrete set-up, which consisted of a 2N7002 in class A config ie. inverting, optimised for fast switch-off. I chose this discrete design to allow for as fast switch off as possible, at the cost of slower switch on. Comparing the max pulsed sink of IR2101 to the expected Id of 2N7002 (Vgate = 5V), the 2N7002 should outperform IR2101 on the low-side, so my sim is consistent.
     
Loading...