Problem driving an H-Bridge

[dteck]

Hi everyone,

So I'm trying to drive an h-bridge at a fairly high speed 2.4mhz. I want to use it to switch a piezo element at 24v.
To get the timing right I set up an LTC6902. It gives me 4 outputs with 2phases so I thought I could use it to drive the gates on an n-mos H-bridge.
I laid it out, got the parts, and when I first powered it up things seemed good. I didnt hook up the piezo I just put scope probes on it to try and set the output frequency.
I started with low voltages about 6v and I thought the square waves looked good. I had identical waves at 2.4mhz but the output voltage was lower than 6v it was only 2 or 4 volts. So I turned it up to 12v and the output went to about 8. So I ramped it up again to about 18 volts. Around there the waveform changed I started getting nothing but noise on my probes and things were getting pretty hot. I showed a couple hundred mA on my benchtop supply. I turned it all off let it sit then ramped it back up to 5v Now the first channel is overshooting and oscillating quite badly. The second channel seems fine. But I also noticed that now the wave forms seem to be overlapping rather than out of phase. I honestly cant remember if they were out of phase in the beginning now. I want to say they were but maybe I was just dumping power and thats what blew my bridge?

I was hoping someone could look over my design and sanity check it for me. I feel like it should be ok, but maybe I've made a stupid mistake and I'm not seeing it. 

I'll attach my schematic, my board layout, and my latest scope reading (powered at 6v).

Thanks

[dteck]

So I assembled a second board to get better test results.

On the second board I ran it at 7volts, put a 5v regulator on the oscillator, bumped the cap after the regulator to 1uf instead of 0.1uf, left the spread spectrum pin floating (unintentional mistake), and put it back on the scope.

The outputs are out of phase like I wanted but the voltage on the outputs is only 3.6v when the supply voltage is 7v. I'm not sure why I'm dropping so much across the mosfets. This was as you might imagine accompanied by a good amount of heat.

Does anyone with some H-Bridge experience have some protips for me?

[bktemp]

LTC6902 is not designed to be used as a mosfet driver. 100ohms output impedance is way to high for driving power mosfets at anything above a couple of 10kHz.
But that's not the only problem: You don't have any level shifters in your circuit. If your input voltage is higher by than the LTC6902 supply voltage, the P-channel mosfet will always be switched on (at around +2V) and short the supply voltage!

[rob77]

you got it wrong - you need some mosfet drivers to drive the high side mosfets, otherwise it won't work (you need higher voltage than the supply voltage to drive the upper/highside  N-chan mosftets).
i would feed the ltc6902 oscillator's outputs to half bridge drivers and drive the mosfets with those (e.g. use IR2110 mosfet drivers).

[rob77]

another option would the P-channel high side with level shifting as bktemp suggested above.
but the IR2110 (or other) half bridge driver is adding some extra safety - it makes sure you can't drive both high and low side at the same time.

[bktemp]

IR2110 is probably a bit too slow for 2.4MHz. It has a deadtime in 100ns range, because it is designed for operation at many 100V.
One cheap+dirty solution is using a fast low side driver and level shift the high side signal using a capacitor + diode+resistor for dc level restoration.
Getting the deadtime right at 2.4MHz is probably not easy. The best solution is  using a dedicated highspeed mosfet driver. There are many nice parts designed for synchronous CPU core voltage stepdown regulators.

[dteck]

HI Rob, Bk,

Thanks for your advice. I looked into bootstrapping mosfets. It showed me where I was going wrong. (good video about it here https://youtu.be/ZZDdlAgZfvI)

Given that and the issue with output impedance I thought I would switch over to using a gate drive as was suggested.
I was thinking the Fairchild FAN3278 30V PMOS-NMOS Bridge Driver might work and then switching the h-bridge to the Diodes Inc DMHC6070LSD.

I'm hoping to get some feedback on if these might work before I jump the gun again and build it.

Thanks

[bktemp]

FAN3278 looks ok. Depending on the supply voltage it will probably get quite warm if there is no large copper area to dissipate the heat.