Lesson learned about PWM driving a push-pull transistor bridge.

[Triode]

I was going to ask this as a question but I found the answer, so I thought instead I'd share what I learned in case other people didn't realize this.

If you are driving a half bridge, or push-pull output with a PWM signal you need to insert a dead time. I've just been making a brushless motor driver and my first attempt exploded. What I'd built was very basic, based on this type of design.

I used IRF840 mosfets rated for 500V, 8A, 125W. My test unit was a 12V brushless RC airplane motor, and my power supply was an 11.1V battery. I won't get into the code in detail, but I had it running PWM to the three phases with a trapezoidal pattern. I figured a really basic start would be good. I had checked the resistance and the switching time to make sure the frequency wouldn't have it spending too much time non-saturated. I hit run and my board went *pop* *pop* sqeeeeeeeeee *fizz* The picture below shows what I failed to account for. (still not sure what that "squeeee" came from)


I had the PWM running in a complimentary mode, simultaneously switching the bridges for each phase. So when the high side turns off the low side switches on instantly. I had realized that switching time meant that I couldn't use too high of a PWM frequency, but I forgot that it meant if I switched high off and low on and the same instant that while the high side transistor is ramping down, and the low side is ramping up, the two will overlap, connecting your positive supply partially to your ground.

So in my case, after a lot of searching, I happened to find this concern in a tutorial on building a digital audio amplifier. I updated my code to have a gap between the high and low side PWM signal slightly greater than the (rise time)+(fall time) of my transistor, and the motor turned without even heating up the transistors.

*feel free to correct my explanation of this (or my whole solution if I'm wrong), I'm still not an expert in this stuff

 

[ronsimpson]

my board went *pop* *pop* sqeeeeeeeeee *fizz*

I know this sound.
With out a dead time you get a small amount of (top and bottom transistors are on at the same time). This shorts out the battery. Even 100nS of shorting out the battery causes heat in the transistors.
*pop* *pop* sqeeeeeeeeee *fizz* Then the magic smoke escapes the transistors.

 

[large_ghostman]

Some pics have a dead time setting for this, I also learnt the hard way........

Once it happens it becomes obvious why its needed! on H bridge, but I think most people learn this the hard way .

 

[crutschow]

Dead time prevents dead transistors.

 

[KeepItSimpleStupid]

The correct FET driver will help minimize dead time because they actively pull the charge out of the gate, but it still takes a finite time to charge and discharge the gate.

Experience is proportional to the value of the parts cost.

 

[schmitt trigger]

This is called "cross conduction" or "shoot thru".

Joking aside, this is a very very common epidemic that has claimed the lives of countless young and healthy Mosfets the World over.

 

[MikeMl]

I have been beating the drum about this issue for several years. Try this search on these forums:
Keywords: shoot-through h-bridge
Posted by: MikeMl

 

[large_ghostman]

I have been beating the drum about this issue for several years. Try this search on these forums:
Keywords: shoot-through h-bridge
Posted by: MikeMl

But you just cant beat 'THAT' smell!

 

[KeepItSimpleStupid]

Sure you can. The smell of a bad Selenium Rectifiers

EDIT: Completed sentence. Thanks ST.

 

[schmitt trigger]

Sure you can. The smell of a bad Se

Selenium rectifiers!!

Those were really nasty stuff. Highly poisonous too.

 

[Triode]

I'm thinking I should gradually compile a book of reasons boards exploded when I didn't expect them to. Then when I'm retired I can publish it, it will probably be at least 150 pages by then.

 

[MrAl]

Hi,

You have not lived until you have seen a bank of 10000uf 200v caps blow up because someone set the test variac too high in voltage for the bridge rectifier and caps, then hit the power breaker switch. To put it mildly, "The shot heard 'round the block".

There was enough energy happening so fast it bent the 1/4 thick copper buss bars up on a sharp angle. The caps tops blew off like a cannon going off.

Yes dead time is mandatory. Transistors dont like to be the only things trying to short out the large capacitor bank power supply in a big hurry

I always try to tell people that when testing, run the DC power supply up a little at a time while you check for faults like that. The current will go up way too high before you get up in voltage and that will tell that something is wrong already, before anything blows. To perform this test, the control circuit is powered up fully with a separate, smaller power supply while the main power buss voltage is cranked up slowly after the control circuit is up and running normally.