Problem with Hbridge circuit?

Hi Jon,

I'm not at all an expert on this sort of thing, but I have a few suggestions:

Here's what I'd do:

1) Add a capacitor across your output pins (P3.1/P3.2). The idea is that the catch diodes (those intrinsic to U1-U4 or separate ones you've added) take a finite amount of time to turn on (usually in the ns range... they act like little capacitors), and during that time you can get little spikes applied do your upper FET sources and lower FET drains. This can exceed Vgs of the upper FETs (likely) or Vds of the lower FETs (less likely) and cause failures. If you want to get fancy, you can ues an R-C network instead (a proper "snubber") -- this can waste less power than a capacitor alone does, but you probably don't care for a low-power circuit. 2) Add a ~10V Zener diode to each upper FET between gate and source (cathode at gate, anode at source). The idea here is that rapidly rising/falling drain-source voltages will capacitively couple back to your gate, potentially exceeding Vgs and breaking it down. See section 3 here: 3) Add some small (~22 ohm) resistors in series between your driver outputs and the FETs' gates. Although this will slightly slow down your switching, it also damps any oscillations on the gate of the FET and (somewhat) protects the Max5062's outputs. (Such oscillations are generally bad news to the health of your circuit...) Note that you might not *have* any such oscillations (sometimes they're high enough in frequency that they're difficult to see, though), but the idea is to add the resistors, make sure everything works, then remove them, and make sure everything still works rather than risking something blowing up while you're troubleshooting other problems.

At a few Hz, the high side drive voltage storage capacitors (C2 & C1) might be losing their charge -- have you looked at the gate drive voltage on your upper FETs while driving the LED?

Do the FETs fry as well? I'm guessing that the most likely cause is one of your upper FETs getting punched through by inductive spikes from the load, and those high-voltage spikes next see the gate drivers and kill them as well.

---Joel

a) Use smaller bootstrap caps. 10-20x gate charge is plenty. Bigger caps are harder to charge quickly, and force high currents through the drivers' bootstrap diodes.

b) You can't leave the outputs high--you have to switch them. Otherwise the gate drive bleeds down, the top FETs go linear then fry.

Could be.

HTH, James Arthur

Are you pulling the low side Fets (both) low in the off duty cycle with the PIC ? I can see right off what would happen if you only trigger lets say U1 for the HIgh side and U4 for the low side. In this config, the driver on that side is discharging the Boost cap for U3 how ever, you're not using U3 your using U1 which needs it's boost cap properly discharged so that it can create a boost for the next on duty cycle. This simply gives you no boost for U1.

The same would apply if you were to reverse your direction on the other side.

Now what would work maybe depending on what you're driving here is to have both low side FETS come on during the off duty cycle. Other wise, all I see here is a set of high side fets (N channels) getting no boost due to the cap not getting discharged in the correct order.

If you looked at the data sheet for these drivers, they used PWM into a half bridge using both N-channels. Those examples force the cap to be pulled (discharged) on the off cycle of the PWM so the next ON cycle for the high side will have a boost.

Maybe you are already pulling both low sides down at the same time now during the off cycle ? I don't remember you saying that you were? If could cause an issue depending on what you have for a load on the circuit for example a DC motor? the Regen could get loaded in the low side of the bridge when doing this since both leads would then be shorted at that time.

Some food for thought.

"

The boost capacitor is required to supply ~9uA of quiescent current to the driver circuit. With a 0.1uF boost capacitor, you'll lose 5V from the voltage headroom on this capacitor in 55mSec. (~20Hz)

So if your pwm control frequency is low enough to produce visible blinking in the led load, you'll have to use a larger boost capacitor.

A static 'ON' state is not possible in either direction - ensure that the switching function is faithfully maintained at either control extreme.

Rapid charging of the severely depleted boost capacitor may produce unpredictable supply voltage and ground trace perturbations.

RL

Thanks guys... I'll try some of the suggestions but remember that the circuit was taken directly from the datasheet applications section... of course I doubled it so I could have a full H-Bridge but that shouldn't make any difference.

The application note unlikely included the two types of IC's, as you originally hooked them together, nor did it suggest adopting audible or sub-audible operating frequencies.

Even a well-proven schematic requires considerable discipline in it's physical rendering.

Drawings from app notes should be considered as suggestions or starting points to a practical end-use application. There is no reasurance that they accurately reflect a working circuit, typos aside.

We have only your assertion that the schematic presented for examination reflects the breadboard that you have assembled, or your description accurately relates real physical circumstances, as they actually occured.

You really should use a scope to troubleshoot a circuit like this.

RL

Joel, RL, and I all agree: you can't drive this thing at a few Hz. It won't work.

I saw the datasheet recommended the 100nF boost caps, which is probably excessive (you didn't specify which FETs you're driving), but that's not the problem now.

You have to drive it faster.

Cheers, James Arthur

I have a friend who's an FAE, and he tells me that a huge percentage of the circuits he sees at customers are directly copied from the data sheet example

Can you explain to me for my own benefit why it matters how fast I drive it? If I pulse it once a minute or 10000 times a second it shouldn't matter? (assuming transitions are fast enough and its not too fast)

Basically your telling me if I turned on one of the inputs then it would blow up cause its at 0hz. This makes no sense to me. Maybe it has to do with that boost cap thing though? (I'm not sure how it works but maybe that cap is to smooth out the pulses to read the average value so it knows how much to boost or something? too slow and it gets the wrong reading?)

You don't always get what you think you're paying for.

Power supplies have always been considered as simple or an after-thought. It breaks my heart, sometimes.

RL

the boost cap is the power supply for the upper fet gate drive

inside the driver chip is a diode from Vcc to the boost cap, so everytime the lower fet is on the boost cap is charged to Vcc.

when the lower fet is off, the driver will slowly discharge the boost cap and eventually the voltage is too low to fully turn on the fet.

_unless_ you at regular intervals turn off the upper fet and turn on the lower fet to recharge the boost cap

-Lasse

As Lasse mentioned, the problem is that the "flying" capacitor that drives the gate of the upper FET will slowly discharge, so if you switch too slowly that capacitor is no longer a decent ("solid") voltage source and on every (assumed-to-be-too-slow) switching cycle, your upper FETs first turn on solidly, and then (over time, as the capacitor charge decreases so Vgs drops) go through their linear regions (where they look like resistors) before turning off. If your currents are high enough, this can case a lot of power dissipation in those FETs... *possibly* enough to kill them, although this probably isn't the cause of *your* FET failures, since the time constants involved are typically going to be on the order of many seconds and there's a good chance you'll notice the FETs getting quite hot before they die. (And they might not die at all, depending on the particular circumstances.)

But it does appear to be the reason that driving your LEDs at a couple Hz wasn't working.

---Joel

Exactly right. You have to switch the driver to keep its boost cap charged.

There could be other problems--you still haven't told us which parts fried, or did I miss that?

"I took my pwm circuit and ran it one of the mosfets and it took relatively significant amount of current to get good transitions. (I was driving and LED and it was acting funky with low current.. I assume essentially the mosfet was probably always conducting because it didn't blink like it was suppose to(I'm only driving it at a few hz))"

Ahh, here it is:

"But I seem to be having issues when my load is large and inductive (a large motor). The driver chips burn up for some reason that I can't figure out why;/ I've added diodes across the mosfets which are not shown in the circuit but still same problem."

Those better be fast recovery diodes you've added across the FETs...are they?

And are the FETs still good after the driver pops?

What does the driver do after it's fried? Stuck high? or low? or open? You could be blowing the driver's boost diodes.

We need those clues too.

Joel mentioned inductive loads can backdrive the FET gates negative, exceeding their Vgs rating.

Maybe, depending on the FET, but more likely they'll zap the driver first--the voltage undershoot goes with the ratio of Crss / (Crss + Ciss), typically about 1:10. With a 125volt supply, you'd not usually get more than -12v at the gate, not enough to kill it.

But the driver might need protection against such undershoot.

These drivers switch a 5nF gate in 33nS--a loose layout, inductive sense resistor, oscillating FET, or stray drain inductance can screw the thing easily. Adding series gate resistors will ease the situation.

HTH,

James Arthur

Finally some one understands the real problem!

And getting back to the author of this fine design ;)

It aint going to work unless you turn on the lower FET that is part of the half bridge of that side!. of course, that's not the side that you are turning on in the low cycle!. so there you go.

Oh well, such as life.

Kind of reminds me of the mini DC drive boards inside of our electric reel movers at work.. It uses a double Half bridge using N channels and when the battery gets low, the high side don't fully switch on thus, taking out a couple of fets if they hammer it. When I say take out, I mean split the cases open!

Lack of understanding can cost you a lot in R&D

"

strange ;/ why the hell would they generate the voltage like this? ;) Seems like there are other ways that are just as good? Essentially it is a charge pump that depends on the lower?

(so essentially these chips are not really made for H-bridge driving?

I wonder if there is a way to modify it, say by takign the boost cap off the hbridge and somehow tricking the chip? (Just wondering... not like I will do it)

Yes, this isn't occuring. I'm leaving one fet on and one fet off for large periods of time while the other side will have one off and the other toggling. Hence if your explination is correct about the boost cap(which I'm not saying it isn't) then this means I am not operating the chips properly.

Although they did work when driving an LED so I guess because it was such a low current that it was working? But eventually it should stop working because the cap would be drained?

One time a mosfet did go bad(didn't fry or anything... not sure what happened) but its mainly been the IC's.

Yes but the fets themselfs have built in diodes and I was using those at first with same problem.

I think so(at least in one occasion I replaced the chip and it worked).

One time it was stuck low. I don't remember the last time and I already removed the ICs.

The voltage is only 12V for the drain on the high side and 12V input to the drivers.

Well, it seems its the way I'm using htem as they don't seem to be made for hbridges. That is, if the both high and low side fets on one side need to be toggled to actually charge up the boost cap. In that case it never happens in my application. (But it worked with an LED so I'm not sure.. maybe the small amount of current was the reason?)

Well, I'm glad your such a god in everything.... Just wish I was as perfect as you... maybe I should go should myself for being such an idiot?

The fets never got hot. The drivers did though... sometimes for no apparent reason.

It was working. Thats why I thought the circuit was working fine.

It seems that I just wasn't using the chip as it was designed. I didn't realize that the high side voltage was generated in such a way. I figured they generated it independently of the driven circuit and the boost cap was sort of a sensor to determine how much to drive the high side gate(or something like that).

I guess that method they use is a little trick to easily generate the high side voltage in that application(when top and bottom are toggled quickly)?

Right. The chip *is* made to drive an H-bridge, but you've got to switch it at some minimum rate to keep the boost cap filled.

How often? Worst case it'll draw 40uA (Ibst) from the boost cap, which will discharge a 100nF boost cap by 1 volt in 2.5 milliseconds. It looks like the undervoltage lockout kicks in around 7v, so with the boost cap starting at 11v you'd have to switch the driver at about 100Hz, minimum (worst case), or about 25Hz (typical).

If you want the top FET to stay ON all the time, well, the FET driver won't do that. You usually just bang the H-output low for a blip occasionally and settle for the top FET being ON 95% of the time. The motor won't notice.

You really should put gate resistors in series with the driver--motors have large interwinding capacitances; you don't want to blast them with 12v in 30nS. 1-5uS switching time would be fine, and avoid many problems.

HTH, James Arthur