new GaN fet

It does specify the threshold as -1.5 to -3.5V, but doesn't give the corresponding I_D.

Nice package--wonder if it comes in DIP? ;)

(Almost certainly way too rich for my blood, if it's intended for cell tower use, but an interesting part, for sure.)

Cheers

Phil Hobbs

BTW: they're specifying the output power at *3 dB compression*. Who ever heard of that?

Cheers

Phil Hobbs

BTW EPC has a very interesting report on reliability vs. gate drive for their GaN FET parts. You really really don't want to exceed 8V.

Cheers

Phil Hobbs

I took some data on EPC gate degradation vs time, when biased above +5 volts. In my tests, high gate voltage caused increased gate leakage over hours or days, but the parts still worked.

High drain voltages have a soft leakage that's not destructive if the current is limited.

It's very common in RF power amplifiers, often the final amplifier in a radio transmitter, striking a balance between linearity and output power capability. One also sees 1 dB compression.

Joe Gwinn

I like that package. The EPC BGAs are a pain and can't get heat out.

I'll try to get pricing and samples.

Cheers

Phil Hobbs

P1db is more common, but less relevant in a pulse application. The data sheet introduction does say they can sustain linear operation, but these devices seem to be specified mainly for pulse applications like radar, c.f. the quiescent drain current of 15mA. No way you're gonna build a linear amp with such a low drain current.

CH

Those loadpull charts on page 5 are a mess, unreadable, If this is what its like, then No wonder a lot of RF guys retire unexpectedly.

RF lives in the dark ages. We need Spice models.

Am 03.03.22 um 05:54 schrieb snipped-for-privacy@highlandsniptechnology.com:

Don't mke me laugh so hard. Spice IS the dark ages. What we need is AWR or ADS design kits.

Gerhard

On a sunny day (Wed, 02 Mar 2022 13:47:42 -0800) it happened John Larkin <jlarkin@highland_atwork_technology.com> wrote in snipped-for-privacy@4ax.com:

Yes I sort of like the datasheet Notation like 2.2-j2.3 shows capacitance effect etc. ?

On a sunny day (Wed, 02 Mar 2022 20:54:28 -0800) it happened snipped-for-privacy@highlandsniptechnology.com wrote in snipped-for-privacy@4ax.com:

Those are just illusions, BUILD the circuit and test!

"Stop, you're both right." ;)

Cheers

Phil Hobbs

(former RF guy)

What's wrong with knowing all the voltages and currents as a function of time?

If you know that, you know all the RF stuff. That doesn't work in reverse.

And what's wrong with knowing drain current as a function of gate voltage? RF data sheets usually say "turn the trimpot until the RF comes out."

I wonder how people generate those "design kits" if they don't know the basic electrical properties of the part. Maybe it's like "load pull" engineering.

I like to Spice things to prepare my instincts, and get close to a working circuit. Less soldering. That blows up fewer expensive parts too.

I don't! The part is supposedly untuned, but the few impedances are specified from 3400 to 3800 MHz. Strange.

At a single frequency and bias condition, in a power amp. A Spice model would include the nonlinear capacitances and the inductances, which might matter for RF. They sure matter in wideband time domain.

They might sell more parts if they considered a wider range of use than a power amp over one small band.

They probably make about eight of them over the lifetime of the part, so statistics are hard to come by. ;)

Cheers

Phil Hobbs

Rf is a strange world. I think its traditions pre-date computers, so they emphasize analytical, necessarily linear, ideas like s-params and Smith charts. Things that could be sort-of handled with pencils and slide rules.

The nonlinear stuff became load-pull, basically documenting a few bench tests.

The EPC GaN fets have Spice models and LT Spice examples, probably because the intended market is switching power supplies.

The Cree SiC fets have Spice models, for the same reason.