1500W electric trike pwm controller broke for the third time!

Discussion in 'The Projects Forum' started by bob123, Sep 20, 2008.

  1. bob123

    Thread Starter New Member

    Sep 20, 2008
    5
    0
    Hi all,
    I have been building an electric trike and the controller on it is giving me a load of crap. The pwm signal from opamps gets fed into transistors (push pull) and then directly into the gates of 5 IRFZ48N mosfets (64A@55v). The mosfets are on an forced air cooled heatsink and i dont belive that they are over heating.

    I am using diodes both reverse biased parallel with the motor and with the mosfets. There is also 2000uf capacitors on both the supplies. Oh and the supply voltage is +-12v (2 car batteries).

    When i turn it on it works until i turn the power up to high and then mosfet breakdown occurs and the motor spins at full speed and cannot be slowed down until switched off.

    Does anyone know what is going on here?
     
  2. mik3

    Senior Member

    Feb 4, 2008
    4,846
    63
    Can you post a schematic? it will be helpful
     
  3. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Intriguing!
    (doo-doo).
    This doesn't sound good already. Opamps were not designed to be run in open-loop mode, so they really aren't good for square wave generators, which is what a PWM circuit needs. But back to your monologue...
    that really doesn't sound so good either; they may not be getting the gates out of the linear region quickly enough. Transistors will have a good bit of voltage from collector to emitter when they're drawing lots of current. But back to you...
    Now this needs a good bit more detailing. What's the length of the leads to the gates? What gauge wire? Really, that's just the beginning; many more details are needed. The more you supply, the better the result will be.
    They might not be. Then again, the TO-220 package can only take so much. They may be getting "killed" with reverse EMF spikes from the motor.

    Perhaps overkill, but I've always been a firm believer in using a 6x12 when a 2x4 will do. ;) However, reverse-biased diodes won't protect the MOSFETs from overvoltage.
    That's rather like playfully splashing on the Rock of Gibraltar in an effort to dissolve it, but a nice touch nonetheless.
    You do know that car batteries are not suited to such duty, right? Auto batteries are good for perhaps up to 10%-20% discharge on a regular basis; more than that and they'll be dead very quickly. The plates are very thin, and are made from a lead paste. They are designed to start a vehicle's engine, and then to be immediately recharged; not for a deep discharge like what would be seen as a prime mover power source.

    It sounds to me like your MOSFETs are being subjected to more voltage than they can handle, resulting in holes being punched through their layers, shorting them out. This could easily happen due to the back-EMF from a brushed DC motor.

    However, this could also be the result of the gate drive voltage falling off with a large load on the batteries; with a reduced gate drive voltage, the MOSFETs would be operating in a linear mode, generate a great deal of heat (think:instant lightbulb) and fuse into a conductive volcanic blob.

    Higher-voltage MOSFETs might help. Zener diodes in the 34v range might help. Better PWM circuitry would help.

    But really, we don't have enough info yet.
     
  4. hgmjr

    Moderator

    Jan 28, 2005
    9,030
    214
    It sure would help us help you if you can arrange to post a schematic.

    hgmjr
     
  5. bob123

    Thread Starter New Member

    Sep 20, 2008
    5
    0
    Here is the schematic, most of the component part numbers are not really the same.
    I can assure you that the pwm is a very clear signal. The opamps have a high slew rate (20v/us) and the frequency is quite low (5000 - 20000hz). The motor has a stall current of 117A.
     
    Last edited: Sep 20, 2008
  6. hgmjr

    Moderator

    Jan 28, 2005
    9,030
    214
    I have merged the schematic thread with your original thread since they appear to be associated. It is preferred for continuity reasons that related threads by the original poster be kept together.

    Thanks,
    hgmjr
     
  7. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
    That slew rate is going to have a hard time pushing/pulling current onto those 2200uF capacitors. That may be why your FET's get hot - the poor cascode amp just can't swing the gate signal fast enough.

    The only thing necessary to control FET conduction is to put the gate at 0 volts (off) or at 10 volts or above (on), relative to the voltage on the FET source terminal. As an improvement, I would get rid of the op amp section used as a comparator and replace it with a "real" one, like an LM311. The 311 output can handle enough current that you can get rid of the two transistors (and those big caps) and drive the gates off the 311. Put about 27 ohms in series with each gate, and use a 470 ohm pullup on the 311 output.

    Woah! I just looked up the rating on 1N750 zener diodes. They are only 4.7 volt. There is no current limiting in series with them - I don't know how they are supposed to not turn into puffs of smoke. You need to pull the gate down to the FET source voltage to turn it off, but limit the gate to between 10 and 20 volts above the source for on. As it is, when the gate voltage gets 4.7 volts above the source, the zener clamps it, leaving the poor FET's in only partial conduction. You want to clamp the gate voltage to something like 12 volts above the FET source.

    That top zener, D4 is going to clamp the gate voltage when it tries to go more than 4.7 volts below the +12 volt rail. That will leave the FET's conducting when they are supposed to be off.

    You can run the control circuitry with circuit common as the + supply, and -12 as the negative. The 311 will pull down to -12 that way, and, since it has an open collector output, you just tie it to the common rail and it will pull up to 12 volts above the FET source. That will still work for the LM324 circuit, minus the C4 & 5 caps and the D4 & 5 zeners. Do put about 27 ohms in series with each gate, though.
     
    Last edited: Sep 20, 2008
    ajimmym likes this.
  8. bob123

    Thread Starter New Member

    Sep 20, 2008
    5
    0
    i posted a second schematic (sorry there were mistakes in the first one) the capacitors and zeners are between the power supply rails not to the mosfet gates.
    the real zeners are 15v 5w and have not blown. There is no problem with the opamps i promise, I looked at the trace on my scope.
     
  9. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
    Be nice to have an accurate schematic. Your basic circuit is still unnecessarily elaborate.

    The zeners and big caps serve no purpose (the new schematic is even worse in that respect). The gates have no current limiting and can be driven too positive.
     
  10. bob123

    Thread Starter New Member

    Sep 20, 2008
    5
    0
    in fairness the big caps serve a big purpose. When the mosfets turn off the current through the wires cannot stop instantly due to the inductance. The capacitors are there to store this energy and prevent the huge voltage surge that would occur. The zeners are there to clamp the power supplies to prevent damage to the electronics.

    What would the purpose of the 27ohm resistor in series with the gate?
    What is the speed of a diode could it be that the flyback diode is not turning on fast enough? in this case using would using a rc snubber solve the problem?
     
  11. hgmjr

    Moderator

    Jan 28, 2005
    9,030
    214
    While searching around for information on parallel mosfets, I ran into this article. It looks like it might be worth reading so I have supplied a link for you.

    hgmjr
     
  12. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Your Vgs is a good bit too high; absolute maximum is +/-20v.

    I suggest that you use a regulator to step your 24v down to around 12v to supply your electronics/gate drivers. That also means you will not be able to use the connection between the batteries as your "virtual ground"; you'll need to use resistive dividers or Zeners to get a reference level (U3A pin 3, U1A pin 2).
    If the lead from your emitter follower to your gate inputs is of any length, you may be getting an inductive kick that's exacerbated when your PWM is close to 100% ON. It wouldn't hurt to have some diodes clamping the gate voltage to the source and the output of the abovementioned regulator.

    Diodes don't turn on and off instantly. There are fast diodes and Schottky diodes that are mighty quick, but they still take a few nanoseconds to turn on; during which time reverse-EMF spikes can rise pretty high. To slow the rise time of these reverse-EMF spikes, you can use small caps (ceramic is fine) somewhere between 330pF and 1nF across the diodes.

    2N3904/2N3905 transistors are pretty quick, but they're really kind of wimpy for this application; their practical limit is about 100mA, and you're driving the gates of five MOSFETs in parallel. Each gate charge is about 60nC with Vgs=12 and Vdss=27, according to Figure 6 on page 4 in International Rectifier's datasheet for the IRFZ48N.
    http://www.irf.com/product-info/datasheets/data/irfz48n.pdf

    If instead of 2N3904/2N3905 pairs for the emitter followers, you used 2N2222/2N2907, you would have much greater current source/sink capability for your gates. It would be better yet to have individual emitter follower pairs for each gate; if you had one gate get shorted to the drain or the source, or held at some arbitrary level in between, it wouldn't affect the other MOSFETs.

    Using a small resistor between the gate and the emitter followers will help to reduce ringing/oscillations on the gates.
    [eta]
    While the 2200uF caps might help with low frequency transient suppresion, they won't do much for high frequency transient suppression, because at high frequencies their parasitic inductance and resistance comes strongly into play. Using various sizes of smaller caps (say, 0.1uF to 1uF), ceramic or poly, will help a great deal to suppress the high frequency transients. The small caps should be as close to the source of the noise with leads as short as possible.
     
    Last edited: Sep 20, 2008
  13. bob123

    Thread Starter New Member

    Sep 20, 2008
    5
    0
    cheers everyone, i have made amendments to my circuit and would like to see what everyone thinks before i risk destroying my last 5 mosfets.

    i guess the problem could have been that i am using to large a voltage to turn the mosfets on, this has be countered using a potential divider to give outputs of 0v and -10v which should be bang on (hopefully although i am risking leaving the mosfets in there linear region.

    The flyback diodes that i am using are not fast diodes and that could have perhaps been the problem, maybe putting a capacitor across them could solve this problem i am using a 22nf.

    i will add a 10ohm resistor and a 0.1uf capacitor in parallel feeding the the gate of the fets.

    i will also add a zener diode to clamp the gate at 12v above the -ve rail this will hopefully protect it.

    can anyone see anything that may cause my mosfets to die again?
    what effect will increasing, decreasing the frequency have on this circuit? is it better to run at higher or lower?
     
  14. hgmjr

    Moderator

    Jan 28, 2005
    9,030
    214
    I would suggest you use large gauge wire (14 AWG or greater) to bus all the drains together and also to bus all the sources together. Keep the drain to bus wire connections and the source to bus wire connections as short as possible to minimize IR drops device to device. The closer you can simulate the devices being in one package the more equally they will share the load between all 5 devices.

    hgmjr
     
  15. jpanhalt

    AAC Fanatic!

    Jan 18, 2008
    5,696
    904
    I used copper sheet (0.025) for my bus on a couple of high-current controllers. I also like to use a gate resistor for each mosfet, as shown here in an early version:

    View attachment 4501

    John
     
  16. hgmjr

    Moderator

    Jan 28, 2005
    9,030
    214
    Nice job, jpanhalt. That is better than I have seen it done in the past short of a pcb layout.

    hgmjr
     
  17. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Peer review is a good thing, although it may be somewhat agonizing. Bear with us. ;)

    I see that you have added a 10 Ohm resistor and a 10v Zener after the voltage follower. Unfortunately, this will very likely result in R10, D14 or Q3 to be fried to a crisp in a very short period of time, as Q3 will attempt to assert about a +25v potential on one side of R10, and D14 will attempt to clamp the gates to 10V, resulting in 1.5A of current through Q3, R10 and D14. My bet is that D14 pops first, and right after that you'll wind up with another five dead MOSFETs due to overvoltage on the gates.

    22nF is about 21nF too much. You will waste a lot of power charging and discharging the cap each PWM cycle. The caps need to be small, between 330pF and 1nF.
    Good idea, but try cutting that cap in half; around 47nF.

    My crystal ball says that you have smoke in your future. :eek:

    You need to regulate the voltage for your electronics. Right now, you're running everything with +/-12.6v from the batteries. One thing that hasn't been covered so far is what happens if your motor is running wide open, and is drawing so much current that the batteries can't keep up. If your gate voltage drops into the linear range, you will kill your MOSFETs very quickly with the kind of load your motor represents. You might even need a separate gel-cell battery for your electronics and gate drive.

    Lower frequency = less time in transition (linear region) = less heat, but with the current levels you're dealing with, you'll get a good bit of noise. If the noise isn't objectionable and performance is acceptable, keep the frequency low. If you want to go with higher frequencies, you will need much higher current drive capabilities for your MOSFET gates.
     
  18. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    Mosfets need a low value resistor in series with the gate and mounted close to the gate to prevent VHF oscillation which kills Mosfets.

    You are using the lousy old LM324 quad opamp which is about the slowest opamp ever designed (about 39 years ago) because it is "low power". Its output is a triangle-wave (lots of heat in the Mosfets) instead of a fast switching square-wave.
     
  19. mik3

    Senior Member

    Feb 4, 2008
    4,846
    63
    Bob123,

    I dont see the push-pull pair of the transistors Q3 and Q4 useful. It would be better to put a single transistor and a resistor connected as an emitter follower to drive the mos. Also, consider that the parallel diodes across the motor wont conduct all together, only one diode will conduct (the one with the lowest knee voltage) unless they are absolutely matched which is almost impossible in real world. So, if you put them in parallel to divide the whole current in each of them wont work, they will be destroyed. You have to choose a diode which can handle the maximum current the motor draws for some microseconds until it decays exponentially. You can put two of them in parallel in case the one fails.
     
  20. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    No.
    Mosfets have a high gate to source capacitance. Paralleled Mosfets have even more. The push-pull emitter-followers provide a high current to quickly charge and discharge the capacitance quickly (if the opamp was faster) so they do not spend much time as a very hot linear amplifier.
     
Loading...