Alex Lidow: Inventor of the Power MOSFET

Haven’t watched it but heard him speaking (probably the same stuff) on the Amp Hour podcast - what really jarred me was he completely ignored Siliconix V mos which for me marked the start of power mos devices before his.

The EPC parts are great. They claimed that they would never make packaged (non-BGA) parts, but they are now.

GaN ICs will be very cool.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Dunno. There are no P-channel GaN FETs, because the hole mobility in III-V systems is the pits--much worse than in silicon. That's why, as they used to say, "GaAs is the material of the future, and always will be." ;)

IIRC somebody posted some months back about a new material that had really amazing hole mobility.

Cheers

Phil Hobbs

Yes, something like a p-channel GaN fet would be wonderful. I'm making half bridges with GaN fets and the high side driver is a true pain.

There are integrated drivers but they are all slow.

Simon is getting a bunch of EPC2037 and 2038 breakout boards made. I want to try them out as front ends and class-H series pass devices.

They’re super high-g_M pHEMTs with interelectrode capacitances similar to a CPH3910 and very low gate resistance.

The noise of an ideal FET is

e_N = sqrt ((4kT/e)*2/(3 g_M)),

i.e. a bit below the Johnson noise of 1/g_M.

Since g_M is huge, and the actual gate resistance is less than an ohm, the total is only a bit more than an ohm.

So maybe they’re down in the 150 pV range. Or maybe not.

Cheers

Phil Hobbs

I did breakouts for them, and some other EPC parts.

I found the 2037 and 2038 to be very fragile. Touch them, and they may shatter or just jump off the board. I put them on a little mouse-bite pcb; we test them and then glop the GaN fets. If a fet fails, we can replace the entire little board.

Even the glopping is tricky.

Wonderful parts.

Similar but smaller—0.3 inch square, single sided, no castellated holes. I really want them to be good in front ends!

Cheers

Phil Hobbs

I know someone who is designing a pretty big, very dense PCB using 32 of the EPC2037 BGA parts to make eight full-bridges. There's not much room around the fets, so it looks like it would be really hard to replace one. Nearly impossible. Should be interesting.

It might be worthwhile to "package" the EPCs onto tiny PCBs just to make them replacable.

Jeroen Belleman

It would be interesting to know the noise level of these EPC gadgets.

One issue is that they have gate leakage current that increases radically with gate voltage. I've mostly measured them above 5 volts on the gate, where they leak and the gates are gradually damaged.

I'd expect gate current to have full shot noise, or worse.

GaN seems to have slow damage effects. They have steady-state max drain voltages, and higher voltage ratings for short-term use.

Ideally a FET’s noise is sqrt(2/3) times the calculated Johnson noise of g_M, a bit more than a BJT, which is sqrt(1/2).

But these things are like 1A/V, and the specified gate resistance is under an ohm, so the spherical-cow noise is about 0.15 nV in 1 Hz.

Cheers

Phil Hobbs

Wouldn’t surprise me. OTOH, I’m hoping not to have to pour on the coal too much in a front end. ;)

Cheers

Phil Hobbs

It's not the right dimension. It should be E_n = sqrt (4kT*2/(3 g_M)).

Jeroen Belleman

Right. You didn’t like my way of putting it?

Cheers

Phil Hobbs

I just removed the factor 1/e, which I believe shouldn't be there.

Jeroen Belleman

I’ve just got shot noise on the brain. ;)

Cheers

Phil o

I get that a lot, too...

Jeroen Belleman

The gate current shot noise may be the thing that wrecks the idea. Unless you are amplifying the sig from a current shunt or something low impedance.