GaN FETs and overvoltages (eg mains transients)

[cupoftea]

Page 22 of the innoswitch3-ep datasheet shows that 150V of margin is recommended between VDS peak and the rated voltage of the GaN FET.
150V margin is never requested for Si FETs. Does this show a weakness to overvoltage of GaNFETs?
Si FETs can reasonably well handle overvoltage ti an extent by just avalanching.

Innoswitch3-ep datasheet

https://www.power.com/sites/default/files/documents/innoswitch3-ep_family_datasheet.pdf

 

[Easy peasy]

depending on the dv/dt - you may not see the true peak volts with a normal scope setup and probes

 

[cupoftea]

Hi,
Just tried to find a 25+ Amp rated GAN FET for a 120W , 37vin to 6vout sync Buck at 150kHz.
Vds of 60 to 150v

There is nothing on digikey other than BGA or "ball strip array" parts......these are totally unsuitable for a powerFET, not to mention expensive and time consuming to do the footprint for.

There doesnt seem to be a heatpad, like you get with the SMD Si FETs.

The parts require solder mask defined pads, and those solder mask openings are not easy shapes, they are oval.

Why are the footprints so bad for GAN in this voltage range?

They also look like a nightmare to solder with a hotplate and hot air gun.

 

[mtwieg]

I'm assuming you're referring to parts from EPC.

You've made it clear that you don't like their packaging. Do you need a "head pad"? Have you estimated power dissipation and temperature rise and found it to be too high?

The footprints are all doable, though will require better processes from the fab than other power devices. EPC makes a few different kinds of footprints, including some with wider pitch (though this makes their thermal performance worse than devices with finer pitch and more balls/bars).

Why are you looking at GaN anyways? 150kHz isn't very high.

 

[cupoftea]

I'm assuming you're referring to parts from EPC.

thanks, yes, thats all digikey gives you for 60-150v gan fet from 25A+

Why are you looking at GaN anyways? 150kHz isn't very high.

..thanks...synch buck..dont want rev rec problem when upper fet turns on

 

[mtwieg]

thanks, yes, thats all digikey gives you for 60-150v gan fet from 25A+

EPC definitely has the most selection for lower voltage ranges. But you should also look at GaN Systems, they have a few 100V devices which might work for you. And they don't have BGA/LGA packages.

..thanks...synch buck..dont want rev rec problem when upper fet turns on

Does this actually contribute enough to power dissipation at 150kHz to make GaN worth it though?

If this is for the very low duty cycle buck converter project mentioned in other threads then yes GaN is appropriate for that. With proper gate drivers, GaN half bridges can work well with pulse widths down to 10ns.

 

[cupoftea]

Thanks, we are using EPC2204A which needs up to 6Vmax drive voltage, and gets it nicely from the LT8391.
The GaN ones are also with very low rdson for the price i am seeing.
Going to be a nightmare putting thermal vias in though, as those bars are only 0.5mm pitch....thermal via restrings will go below 150um, and pcb price will shoot up.

 

[mtwieg]

Thanks, we are using EPC2204A which needs up to 6Vmax drive voltage, and gets it nicely from the LT8391.

The gate drive levels should work fine for eGaN FETs. Though the timing specs on the drivers won't take full advantage of their switching speed.

Going to be a nightmare putting thermal vias in though, as those bars are only 0.5mm pitch....thermal via restrings will go below 150um, and pcb price will shoot up.

I wouldn't assume thermal vias are necessary. EPC has an app note where they compare thermal performance with thermal via-in-pad vs thermal vias just adjacent to the pads. The difference is quite small.

I recall several years ago when all their devices came in packages with 0.4mm pitch, and their demo boards had several vias in every bar, with 2oz copper to boot. I tried sending the gerbers for a demo board, unmodified, to a few fabhouses, and couldn't even get quotes. They wouldn't let it past their DFM review. I honestly think EPC did themselves harm by convincing some engineers that the devices require HDI processes to work properly. But they definitely do not.