Mosfet Driver issue

[mcdesca]

Dear everybody,

I am just trying to update my old H-Bridge design to a newer spec and (hopefully) better components. One of the newer components is the MIC4605 mosfet driver, that includes shoot-through protection.

I have been supplying the gate supply with 12V from an already proven converter and the main power bus from 30V. The PWM signal has a frequency of 30KHz (should range from 20 to 50KHz once the circuit is validated).

The MOSFETS are Infineon's bsc123n08ns3: 12.3mohm 6.3nC gate charge

I keep destroying the ic that drives the half bridge that has the low side activated at that moment (the motor is spinning in a particular direction in open loop).

It looks to me as if the driver cannot source/sink enough current (1A limit). The components around the driver and the mosfets themselves have not changed from my previous design (done with UCC27201 - 3A limit).

I took the current limitation as the inability to source/sink enough current, not as an maximum electrical operating limit. Is that correct?

Any thoughts?


Thanks for your time.

[mcdesca]

Hello,

That is what I wanted to update from my previous design, and so has this driver adaptive dead time.

Is that what you pointed me out or have I misunderstood you?


Best,

[bktemp]

Show us your layout.
It could be a spike from the half bridge destroying the ic by pulling the HS pin to a negative voltage.

[mcdesca]

I hope it is clear enough to see each component. I am just showing one half-bridge.


Best,

[bktemp]

The layout is not ideal, but should be ok.
R22 and R23 are a bit close and the the loop area formed by HS, HO and T1 could be a bit smaller. But both are probably due to space constraints.

[mcdesca]

It is not that easy to find a controller with adaptive dead-time protection and relative high current source/sink capability. Any proposals?

[nuno]

You have 22 Ohm gate resistors, so that's worst case 12V / 22 ~ 0.55A so a 1A driver output current limit shouldn't be a problem.

Looks like C14, the chip's decoupling cap, has one pin unconnected. Copy typo?

Never seen C10 / C11, what are they for?

22 Ohm gate resistor looks intuitively too low to me (what's your desired peak motor current?), that would be a very fast switching, and very fast switching -> very high voltage transients. That's around what I would use for FETs with 100-200nC of gate charge, and that gives very few hundred ns switching time.

You should scope your signals, of course... but you have to use a very short GND lead on the probe (the springy thing).

Also.... you have "beefy" caps from VIN to GND *very near* your FETs, right?

[mcdesca]

Thanks for replying.

C14 is connectd to the ground plane. I have not filled the polygon in eagle, so there is a missing connection on the layout. On the schematic it is connected to ground.

I just put the resistor that makes the step start to look like a typical capacitor curve (turn on/off time is around 500ns, scope probe without the alligator clip, just the tip spring). The controller should be able to switch up to 50 KHz, and that resistor gives me an extra margin for the switching to occur on fully on/off states.

C10/C11 are a dirty and cheap snubbers that significantly reduce the switching noise. That should become either an RC or RCD network.

Peak motor current is at 8 amps.

There are 100uf in total spreaded across ceramic capacitors very near to the mosfets (to be seen on the layout as the 5 pads near the two mosfets). I could not place the big electrolytic near enough.

The last ones have are 470uF x 2, reducing the ripple at 50mVp-p max when motor current is around 2 amps (did not go further).

I have replaced the mosfet driver for the lm5106 (100V rail 118V bootstrap 1.8/1.2 sink/source current). I will let you know when the new pcb arrives.

Thanks again.

[nuno]

A double pulse test of your bridge allows you to learn a lot about your bridge and is a test that avoids destruction:

http://www.athenaenergycorp.com/wp-content/uploads/2012/11/double-pulse-testing.pdf
http://www.scribd.com/doc/100812439/Double-Pulse-Test-Setup

1st pulse is to build up current in the coil up to the value you want to test (and to test the switch-off), so it's given by the famous formula Vbat = Lcoil x CurrentRaise / PulseLen . Some people will use only the charge stored at the capacitors for the test; charge capacitors, disconnect VIN from caps and double pulse it.

Then you scope nodes of interest (always with a the springy thing of the probe) such as Vgs and Vds while switching, and see what ringings/transients you have, how your switching looks like. This is an example from a bridge of mine:

[wraper]

In the datasheet example C1 is 22uF, your mosfets have about the same gate charge as in the example, so 100n seem to be rather small capacitance. And if you use some crappy ceramic, under working voltage there might be less than half of that capacitance left.

[helge]

Hi,

we're using the MIC4605 in one of our latest designs (to be validated) - any conclusions to this thread? Is there a fundamental issue with the MIC4605?