Sumitomo GaAs and GaN FETs

(Obligatory on-topic content: Sumitomo also lists some, e.g. the SGNE010MK, but nobody seems to stock them.)

Cheers

Phil Hobbs

We are just now updating our inventory database to qualify replacements for the eol NEC NE3508 and NE3509 smallish depletion-mode phemts. C is doing that, and she's ordered some samples to test, just to make sure. I'll post what we find, later this week.

These are small parts, for discharging ramps and such, Idss in the roughly 40 mA range. Some parts have low drain voltage ratings, like 4 volts, and we have some situations where we may go a bit above that. So we'll test them to see what the real numbers are. Some phemts are OK at 4x their rated drain voltage; RF people write weird specs, and seldom supply DC curves. Pump in some RF, and more comes out; that's all you need to know. Some fets have the gate named "RF IN" and the drain "RF OUT" and the part is called an "amplifier."

We use a couple of the Avago enhancement parts, which are usually easier to drive in switched apps, but they are not drop-ins for the older depletion fets.

The CEL web site is depressing. Most of the cool older RF parts are discontinued.

We are playing with the EPC GaN chip-scale parts, as pulse generator outputs in the 1-amp range, but they are a little hard to handle and are seriously sole-source. The GaN parts are mostly aimed at power apps. Are there any little ones?

Not that I know of. I'm looking at the Broadcom/Avago VMMK-1225 pHEMT, which looks really cool, even if it's $8: Ciss = 1 pF, g_M ~ 1 S. That ought to be quiet if they've managed to control the 1/f noise.

My fave e-pHEMT is the ATF55143, but I haven't tried many. That's one of two device candidates for that 100-ps $2 TDR.

For switching and normal FET-type stuff I really like the SKY65050 because it has a decently large drain impedance, not like the Avago/Broadcom ATF38143, which doesn't even work as a source follower because it's so limp. Cascoding cleans it right up, of course. Their flatband noise is awesome, about 0.3 nV for the Broadcom part and 0.4 nV for the Skyworks (1 Hz, natch).

Unfortunately the ATF38143 has a 1/f corner of about 10 MHz, and the SKY65050's is 30 MHz.

The ATF38143 also has that neat self-biasing feature--if you leave the gate open, it floats negative by a couple of hundred millivolts. I used that feature in an amp once (it was a special for a research lab, so it didn't need to work in mass production).

Cheers

Phil Hobbs

For the Samsung labs front end, I added a free-of-charge calibrator that made really triangular triangle waves. A tri wave plus a really small cap (>

We use that one. Poke the gate positive into conduction, and it really turns on hard. I conjecture some sort of bipolar effect.

Abs max specs aren't everything.

Phemts as source followers is a frightening concept. A cascoded follower is terrifying.

Weird gadgets.

Can you get that Skyworks part? Digikey is discouraging.

used

Mouser has like 9000 of them.

Invest in reels!

Cheers

Phil Hobbs

I was doing it back in the 1980's - fast beam-blanking electronics for elec tron beam testers. The emphasis was more on broad-band transistor - BFR92 a nd the like - than on specialised microwave transistors.

We did use one Hewlett-Packard RF power transistor to let us switch a usefu l amount of current quickly - I'd tried to make a Percival distributed ampl ifier to put a bunch of smaller transistors in parallel, but it hadn't been quick enough - 800psec versus 500psec.

OT: I read "arebody" as "arsebody". Oh Freud... :^)

Tim