Announcement

Collapse
No announcement yet.

Reason for shorted FETS etc

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    Reason for shorted FETS etc

    Do we have any evidence that shorting capacitors are responsible for the shorted FETs and resulting smoke leakage? And not something like the switching frequency of the buck converter increasing to compensate for the reduced capacitance until the mosfets spend too much time switching and overheat, or some other similar failure?

    Just pondering.

    --Randy

    #2
    Re: Reason for shorted FETS etc

    i split the thread cos that is such a good question, we need to get into these topics cos it is getting boring.

    i am interested in that also. I am wondering if shitty design of psus are a factor also. i dont have such great knowledge of electronics though. i know some people here do...

    anybody?

    at least we can start with the fact that there is an event that can result in shorting fets and/or vrm chip damage. This event involves the machine shutting down unexpectedly, never to post again before repair. However this event can also lead to no damage to board components. The second is due to a cap failing open because if there was a short, there would be damage.

    we need to know what generally makes fets short and what damages vrm chips.

    can cheap fets be a factor or would a condition make even high quality fets fail.
    capacitor lab yachtmati techmati

    Comment


      #3
      Re: Reason for shorted FETS etc

      duty cycle goes up to keep vcore up as the caps get very leaky.fets run hot which heats the ground plane of the board thus heating the caps further and accelerating their failure.this gets worse till poof!

      Comment


        #4
        Re: Reason for shorted FETS etc

        I'm thinking more of a non-leaky-type failure...

        If the capacitors were not leaky, but just had very high ESR and very low capacitance? How would the VRM respond to this? The second the buck controller switches the FETs on, the voltage would shoot way up, as a combination of the high esr causing jumps and the low capacitance causing the voltage to rise much quicker. Then, would the controller imediately switch the FETs off to prevent overvoltage? If so, the voltage would immediately drop back down faster than it should, again due to the high esr and the low capacitance. Or with a fixed-on-time controller, the duty cycle would probably bounce randomly, as the fixed on pulses would probably cause excessive voltage...

        In any case, it seems the switching frequency of the mosfets is likely to increase as the capacitors fail... and since mosfets usually dissipate the most heat during switching (full off is no heat, full on is very little heat... but switching with current flowing through them is lots of heat), this might cause them to overheat, even if the capacitors are neither leaky nor shorted.

        Anyone have a board with failing (preferably very toasted) caps they could clip their scope onto the gate drive and see what (if anything) changes after they recap it?

        I don't have any boards with bad caps at the moment, and not going to start fucking one up to test it.

        A failure related to leaky caps just seems unlikely... with the cpu drawing easily 20 amps, a capacitor would probably need to leak 10 amps before the VRM might have any problems... and 10 amps, even at 1.5v, is 15 watts... a capacitor this leaky would let the magic smoke out, and we'd be seeing melted wrappers and other problems.

        A sudden short would probably just shut down the VRM, psu, or in some other way would cause a shutdown instead of fried FETs... even if just by the cpu voltage going to 0. not to mention many sudden shorts might clear themselves as the current spike (helped by the remaining capacitors) blows away the shorted area. (think screwdriver across the terminals of a large cap. we've all tried this, right?

        --Randy
        (who's probably thinking more about this than is healthy)

        Comment

        Working...