Monolithic means one rock. Silicon is one rock, GaN is one rock, so that's... yep, two rocks.
Marketing might like to call it monolithic, though. It's hard to push 'duolithic' as a virtue. Pentalithic would be the best, because pantathletes could be spokespersons...
There are papers and patents for monolithic (one rock) GaN-on-silicon cascodes, but I don't know if any are commercial yet.
Actual circuits are layers of material, and as long as the layers are conti nuous and coherent, it seems fair to call them monolithic, even if it is a layer of GaAs on silicon dioxide on a silicon substrate, and there's an adj acent layer of doped silicon on more silicon dioxide on the same substrate.
If you plonk a lump of something else into a package, it's called a hybrid.
You can grow GaN on the 111 surface of silicon, using a couple of intermediate layers. Thus it's possible to make monolithic GaN-Si chips using wafer-bonded SOI with 100 Si on the bottom and 111 Si on top.
See e.g.
Cheers
Phil Hobbs
The manufacturers of available si-GaN or si-SiC cascodes are evasive, but it looks like they are mostly 2-chip devices, or 3-chip for the ones that include a gate driver.
GaN seems to be stuck at about 600 volts lately. I could use more. SiC is hard to drive. Blasting a gate from -10 to +20 for 10 ns is a chore.
I had good results with -3 to +16 volts drive.
My apparent gate swings for the 1200v Cree are -10 and +18, but I don't know what the gate actually sees. There is PCB inductance and the internal gate inductance and resistance and 1200 volts of Miller feedback. A little added inductance in series with the gate actually helps, probably by twanging the rise/fall a little.
"Series peaking" from the video amplifier days.
SiC fets have largish gate resistances; some are so high that fast switching is impossible.
The Crees don't like to be driven much harder. Trust me on that one.
Have something to add? Share your thoughts — no account required.
Ask the community — no account required