Connect with us

MOSFET cannot switch on/off normally.

Discussion in 'Electronic Design' started by Electronic Swear, Aug 10, 2004.

Scroll to continue with content
  1. I have a PWM circuit to control the motor output.
    The Ope-Amp 741 will send out a PWM signal and control
    a MOSFET to chop the DC to control a universal motor.
    Schematic is as following:

    12V 120DC
    | |
    LM741 | (MOTOR)
    Clock Pulse ----|\ |/ |
    | >--| 2N2222 __|
    Control signal --|/ |\> ||
    |--------||<-| IRF740
    | ||__|
    (R=2.2K) |
    | |
    | |

    I have noticed that the IRF740 cannot function correctly
    to switch on/off to control the chopped DC. I have already
    add a fast diode on motor terminal to avoid the back emf
    feeding back to IRF740.

    Is there any connection that I am not correct?
    Thank you very much.
  2. Ban

    Ban Guest

    Almost all:
    741: very bad choice if using as a comparator. Better check LM339, which is
    a similar chip but optimized for comparator use.
    Your transistor can switch the FET on very fast, but it takes long time for
    removing the charge from the gate through that 2.2k. In fact it never gets
    completely removed, so the fet stays mostly on.

    12V o-----+-----------+---+
    | | |100n
    .-. | ---
    | | | ---
    | |1k | | |
    '-' 2N3906 | === |
    |\ | ___ |< GND |
    -|+\ +--|___|--| |
    | >--+ 1k |\ 22R +-----+
    -|-/ | | ___ | |
    |/ | +-|___|-+ ||-+ |
    1/4 LM339 | ___ | ||<- -
    | +-|___|-+---||-+ ^
    | | 22R | |
    | ___ |/ +-----+
    +--|___|--| 2N3904 |
    1k |> |
    | |
    | |
    === ===
    created by Andy´s ASCII-Circuit v1.24.140803 Beta
    Take a fast diode across the fet S/D or use a power schottky.
    You're welcome.
  3. terry

    terry Guest

    not only that, but when you switch your FET off, the drain voltage begins to
    rise. This dV/dt causes current to flow through the so-called "miller"
    capacitance (drain-gate capacitance) and into the gate circuitry (gate
    capacitance in parallel with driver impedance). This current will basically
    "fight" your gate driver, and attempt to turn the FET back on when you
    switch it off. Likewise when you switch the FET on, miller capacitance tries
    to turn the FET off again. This is what gives rise to the flat spot seen in
    gate voltage curves, the width of which is directly related to the rise- or
    fall-time of the drain voltage.

    In your case driver impedance is (perhaps - a 741 is terrible!) fairly low
    for turn on, but 2.2k for turn off. I would expect the FET to take a month
    of sundays to turn off, if at all - if the PWM period is fast compared to
    this, the FET will NEVER turn off.

    Bans circuit solves this problem by providing a nice low impedance for
    turn-on AND turn-off. Using an LM339 comparator provides nice sharp edges
    (certainly compared to the terrible slew rate of a 741) to the
    impedance lowering transistor circuit.

    In this case, Bans circuit can be "improved" thusly:

    12V o-----+-----------+---+
    | | |100n
    .-. | ---
    | | | ---
    | |1k | | |
    '-' 2N3904 | === |
    |\ | |/ GND |
    -|+\ +---------| |
    | >--+ |> +-----+
    -|-/ | | | |
    |/ | | ||-+ |
    1/4 LM339 | | ___ ||<- -
    | +-|___|-+---||-+ ^
    | | 22R | |
    | |< +-----+
    +---------| 2N3906 |
    |\ |
    | |
    | |
    === ===

    The "improved" is due to:
    1) faster operation - the transistors now dont saturate, they are emitter

    2) less bits

    I have built hundreds of thousands of versions of this circuit, driving FETs
    and IGBTs in power supplies ranging from 100mW to 1MW (thats a LOT of really
    big IGBTs in parallel)

  4. Your driver has to be able to Sink and Source current; the one you have
    shown will swith off very slowly so if the PWM frequency is high enough, it
    will not switch off at all! Some mosfet application notes will explain all!!

    Why the 741? -

    If you want a chip in there for some reason, there are dedicated driver
    chips that can drive the mosfet properly. If you want to use transistors, it
    is simple to build a three-transistor totem-pole driver that will work.
    Chips tend to pull in requirements on power supplies and whatnot, discretes
    increase component count in return for circuits that can be made to run on
    *whatever* power is available. Here you have both!
  5. A floating gate doesn't seem a good idea. Can't you move
    that bottom 10K between gate and GND ? Or remove all
    resistors and diode?

    The speed of the motor is not of interest. It's the speed
    of the clock signal that is. I hope it is slow. Is that
    a sawtooth clock? Is the control signal a DC level signal to
    control the width of the PCM?

  6. Greetings.

    Although this circuit is indeed superior to Ban's implementation (namely bad
    things happen when the LM339 output is not actively trying to drive low),
    this circuit too is not without its flaws.

    The 1k pull up resistance is much to low and results in a significant
    probability of MOSFET destruction. The LM339 datasheet is here:

    Unfortunately, the LM339 blows big chunks. Admittedly the device is cheap
    and is still useful, but it has a rather pitiful output driver. It claims
    to be capable only of sinking around a minimum of 6mA with a one volt
    overdrive on the inputs and with the output saturation voltage being less
    than or equal to 1.5V. Obviously this is not good enough for operation at
    12V since the pull up resistor will provide 12mA should the LM339 be able to
    pull the output voltage to ground. As a consequence the MOSFET gate will
    likely be driven somewhere in the linear region and unfortunately get

    This is of course the most severe problem, but other caveats should be
    mentioned when considering using the LM339 in this way.

    The LM339 sucks hard in its instability during output edge transitions. It
    tends to oscillate on output transitions. The application hints of the
    datasheet addresses this issue to a limited extent. Indeed it has been my
    personal experience that the LM339 is quite prone to edge transition
    oscillations even when some hysteresis is used and otherwise seemingly
    reasonable layout techniques are used. These oscillations will naturally
    lead to significantly increased MOSFET and motor antiparallel diode
    dissipation, as well as increased EMI (which is likely of genuine concern
    with a 120V motor).

    For a more robust and trouble free solution it is probably easier to use
    some dedicated PWM controller such as the venerable TL494 ( ) or a microcontroller with
    built in PWM functionality driving a dedicated MOSFET gate driver IC such as
    the TC4427 ( ).
    Of course these types of solutions also have their own problems (such as
    microcontroller susceptibility to upset by EMI) which may or may not need to
    be addressed.
  7. colin

    colin Guest

    dont forget about the ne555, it has high drive op and wil directly drive a
    mosfet quite quickly. it is very versatile and can be used a s a simple
    comparator with built in hysterisis or of course as a pulse width modulator
    / timer etc ...

    Colin =^.^=
  8. You fellers are NOT driving your MOSFETS fast and clean until you replace
    your 2N390x's with a Zetex complementary transistor pack found here:
    Check out the ZXTDB2M832, drives 10A in 30nS with <5nS delay. Small enough
    to place at the Gate-Source connection, with a 0.47uF cap. No gate current
    loops. Controller IC stays on a nice clean ground. Much cleaner than the
    TC442x series over temperature.
    Now that's smoking!
  9. Fred Bloggs

    Fred Bloggs Guest

    In most cases, this will be sufficient:

    View in a fixed-width font such as Courier.

    LM741 | 120DC
    Clock Pulse ----|\ |/ |
    | >--| 2N2222 1N4148 (MOTOR)
    Control signal --|/ |\> +---|>|----+ |
    | | | __|
    | | 1.0K | ||
    +------+---/\/\---+--------||<-| IRF740
    | | | ||__|
    | | | |
    | | e |
    (R=1.0K) | |/ |
    | +--------| 2N2907A AGND
    | |\
    | c
    | |
  10. terry

    terry Guest

    yep. I wuz gonna do a post with decent transistors (aint Zetex great), but
    then I read the latest installment from the OP, who clearly doesnt
    understand whats been said, so didnt bother.

    The last time I did a big inverter I had (IIRC) FZT851 & FZT951 sitting on
    the IGBT itself, with local decoupling, and a single stage of FMMT491/591
    driving the source-terminated cable feeding my buffer amp. The buffer also
    had desat detection with increased Rgoff (the damn IGBTs current limited at
    6,000A so L*dI/dt became a bit of an issue unless Rgoff_fault = 10xRgoff (or
    so). I also had a suitably monstrous transorb network arranged to blow the
    gate bond wires in the event of a collector-gate short. This ensured that
    only the buffer amp pcb (which screwed directly onto the IGBT) died in the
    event of a failure - dynamic thermal stresses mean IGBT failure is
    inevitable, and careful planning can greatly reduce MTTR (and cost). This
    allowed me to integrate pretty much all my electronics onto a single PCB...

    we did extensive short-circuit testing (directly across each half-bridge
    with very low L) and it was solid as a rock. We routinely destroyed (oh
    yeah, smoke, flames, explosions - the whole sheBANG) the prototype
    machine(s) we were driving, but never broke an inverter. We had small output
    chokes (1-2%), and the inverters happily controlled current into a dead
    short on the output terminals, indefinitely.


    oh yeah, its worth mentioning that if you dont have to drive the mosfet
    fast, then dont. lower EMI etc. But dont just increase Rg, shape the
    waveform applied to the emitter follower, and drive the gate with a nice low
    Rg. and/or use analog's dV/dt limiter.
  11. Tam/WB2TT

    Tam/WB2TT Guest

    Outlook Express won't let me see your schematic like I would want to, but I
    see only one driver transistor. You will want to use both NPN and NPN
    drivers. Bases get tied together and connect to the 741. Emitters get tied
    together and connect to the power FET. Collector of PNP goes to ground and
    collector of NPN goes to VCC. You may also want to square up the signal
    coming off the 741 to speed up the rise and fall times. Some even number of
    CD4XXX inverters in series should help.If you use 1 inverter in series with
    5 others in parallel, you might not need the transistors.

    CHECK THIS: What is the low state output voltage of tghe 741? If you are
    using a single supply, it might not be low enough - you want < +1V. The
    4XXX will help that also.

Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day