So I built this h-bridge driver for a solenoid valve, and it worked beautifully for a while.

After my third circuit, suddenly things went boom and magic smoke came out of the p-fets (and sometimes of the n-fets too)... I don't know what changed. Maybe I've been lucky and the circuit's been working on the transistors' tolerance thresholds until now.

My take is that for a brief moment M1 and M4 (or M2 and M3) are short-circuited and that's why things are burning up ... I can find no other reason.

​

CW and CCW are 0 to 5V signals, btw. Zeners are rated at 18V.

What am I doing wrong? Is there a way to add simple parts to it so as to somehow de-phase M4 and M1 (and M3 and M2) when they're triggered on/off so that they'll never short circuit?

I was thinking of placing a diode and a cap and resistor in series in inverse parallel at each transistor's gate to delay their turn-on time whilst keeping their turn off time almost unaffected.

Something like this:

​

Is this a good idea? Anyone have a better one, without requiring that each transistor be switched individually?

 

My take is that for a brief moment M1 and M4 (or M2 and M3) are short-circuited

That's correct.
And with a 10kΩ pull up resistor, they spend a fair about of time with both being on as it transitions from one state to the other..
If you make R3 and R4 1kΩ or less it will greatly speed up that transition and reduce the length of the shoot-through current.

But if you want to totally eliminate this shoot-through, you can use a clock-nonoverlap circuit, example below:
The cross-coupled latch insures that the N and P channel devices on the same side (P1-N1 and P2-N2) are never on at the same time.0
The RC time-constant determines the amount of non-overlap.

(Incidentally, you likely don't need the Zeners across the solenoid, since the MOSFET body diodes will conduct the transient to ground and V+.)

 

Thanks for the advice, Crutschow. It's always thoroughly appreciated.

The clock non-overlap circuit is quite interesting, but it's a bit too much for what I want to implement at the moment. But I'll keep in mind for my next version of the circuit. Thanks.

I'm going to get rid of those zeners, and see how things turn out.

And here I thought that increasing the value of the pull up resistor maybe would improve things ... you've just saved me a headache.

 

You need to separate the gate control signals and introduce turn on delays.
Here is my idea, not tested.

U1 circuit delays the rising edge but not the falling edge.
U2 circuit delays the falling edge but not the rising edge.

Modify R and C values to obtain the desired delays. R values of 1kΩ were used for 7400 TTL devices. These can be changed to 10kΩ while lowering C values for CMOS gates.

 

he clock-nonoverlap circuit is quite interesting, but it's a bit too much for what I want to implement at the moment.

Just two resistors, two capacitors, and two IC's.
Far fewer than the 14 discrete parts plus the two driver MOSFETs, you have in your second circuit.

 

Just two resistors, two capacitors, and two IC's.
Far fewer than the 14 discrete parts plus the two driver MOSFETs, you have in your second circuit.

​

 

Unless Your Solenoid actually contains a Magnet,
You don't need a Full-Bridge to drive it.
One FET should do the job.
.
.
.

 

I can’t provide too much help for your circuit, @cmartinez, but I do have a hard earned tip from long experience:

keep your face out of circuits when you energize them

I first learned this when, as a 12-year old I got a flying capacitor can to the eye (thankfully no real injury) when the poor little thing got mains voltage applied to it for a while.

 

Unless Your Solenoid actually contains a Magnet,
You don't need a Full-Bridge to drive it.
One FET should do the job.
.
.
.

The solenoid is part of a 3 port latch valve, and thus it's bidirectional. So the use of an h-bridge is a must.

 

There are many threads on this site about H bridges and all of the have the same issue. I don't know where people get these ideas to connect Gates together for P- and N- MOSFETS.. Internet is full of example circuits to do things correctly and they warn shoot-through problems.
Here are few.....

 

I can’t provide too much help for your circuit, @cmartinez, but I do have a hard earned tip from long experience:

keep your face out of circuits when you energize them

I first learned this when, as a 12-year old I got a flying capacitor can to the eye (thankfully no real injury) when the poor little thing got mains voltage applied to it for a while.

No, of course it didn't literally blow up in my face... ... I used an eye-catching title to see if I could draw more attention to my thread... And it worked

But you're right ... I guess that'll teach me not to rely too much on hyperbole ...

OTH ... it did literally burn my fingers. Twice!
That happened when I accidentally touched one of the bridge's pFets and they were somehow triggered and short-circuited their side of the bridge... I'm guessing my hand's capacitance was to blame. I've been very careful when handling the energized circuit after that.

 

No, of course it didn't literally blow up in my face... ... I used an eye-catching title to see if I could draw more attention to my thread... And it worked

Bit you're right ... I guess that'll teach me not to rely too much on hyperbole ...

OTH ... it did literally burn my fingers. Twice!
That happened when I accidentally touched one of the bridge's pFets and they were somehow triggered and short-circuited their side of the bridge... I'm guessing my hand's capacitance was to blame. I've been very careful when handling the energized circuit after that.

Wear safety glasses... and gloves. Maybe some Class III body armor would be a good idea, too.

 

There are many threads on this site about H bridges and all of the have the same issue. I don't know where people get these ideas to connect Gates together for P- and N- MOSFETS.. Internet is full of example circuits to do things correctly and they warn shoot-through problems.
Here are few.....
View attachment 248646

I forgot to add a requirement in my description of the circuit: and that is that it must draw zero-power when inactive.

So the list of requirements is:

• Minimum amount of parts
• Only two outputs needed to trigger it
• Zero-power draw when inactive.

 

Wear safety glasses... and gloves. Maybe some Class III body armor would be a good idea, too.

Need to check the back of my shed for that last one ... maybe I'll find an Apollo 11 surplus space suit that I could use.

 

Then You can use 2- High-Power-FET-Gate-Drivers.
That's right, just 2- TO-220 Packages,
and a single TVS, or 2-Zeners for protection,
and no other parts.

They are available as Inverting, or Non-Inverting,
zero problems with "shoot-through".

The one in the Data-Sheet below is good for ~8-Amps Continuous, 30-Amps Peak.
~16-Volts over ~4-Ohm Load = ~4-Amps.

Don't operate any Transistor without a Heat-Sink.
.
.
.

 

Then You can use 2- High-Power-FET-Gate-Drivers.
That's right, just 2- TO-220 Packages,
and a single TVS, or 2-Zeners for protection,
and no other parts.

They are available as Inverting, or Non-Inverting,
zero problems with "shoot-through".

The one in the Data-Sheet below is good for ~8-Amps Continuous, 30-Amps Peak.
~16-Volts over ~4-Ohm Load = ~4-Amps.

Don't operate any Transistor without a Heat-Sink.
.
.
.

Thanks for the suggestion, but that device's specs state that it consumes 0.75 mA while idle. And that's a big no-no in my requirements.

 

You have a ~4-Amp-Load,
how could 1.5ma make a difference ?
.
.
.

 

I forgot to add a requirement in my description of the circuit: and that is that it must draw zero-power when inactive.

Zero power.... Just switch off the circuit completely by external MOSFET and use a proper driver.

 

You have a ~4-Amp-Load,
how could 1.5ma make a difference ?
.
.
.

Because it's battery powered. That's why the circuit must draw zero power when Idle ...

 

Zero power.... Just switch off the circuit completely by external MOSFET and use a proper driver.
View attachment 248655

That almost meets my requirements ... the drawback is that an extra output (input, in this case) is needed for it to work ... perhaps if an or-gate were added to CW/CCW and its output used as the kill switch?