new GaN fet

It's usually on the data sheet. But V/I curves often aren't! That's crazy. So we have to measure that.

Does Spice have any

Well, the DC curves matter too, if you want to inject carriers.

I've done that with Spice. I just add some noise generators and run a time-domain model.

It would be more convenient if the noise models worked in time domain, but we can deal with that. We understand it.

But they handle a very narrow group of cases.

If you are saying that Spice is sometimes incomplete, we understand that and can adapt. But I still want to know if this is a depletion fet, and what Idss is.

Most depletion phemts turn out to enhance usefully, or a lot. 2x Idss or 1/2 Rds-on can be useful. Nobody mentions that! Actually, I've never seen an Rds-on spec for a phemt... it's an RF part!

But they can be dynamite RF switches, except that the RF boys don't seem to think about parts switching.

Here's a model used by ADS:

Spice models only include things that Spice understands, though. Carrier lifetime, diffusion behaviour, doping profiles and so on aren't in there. (The diode model does have the voltage-variable capacitance, but AIUI it's basically behavioural--it sure doesn't model forward recovery.)

Assuming they're reasonably accurate. The s-parameters are probably correct at one bias point at least. ;)

Cheers

Phil Hobbs

Have you actually done any of that stuff yourself?

Cheers

Phil Hobbs

Sure. And we understand that.

I don't know much about the big RF suites. Can they model pin diodes and srds and such? Can they model the time-domain switching behavior of a pin diode switch?

We're using THS4302s, which do have s-params. But they also have DC specs and capacitances and pulse shots and slew rates. Best of both worlds.

No, I have not. But I have worked with people who have and still do such modeling and have seen their analyses.

Joe Gwinn

PIN Diodes, yes - I see that all the time.

SRDs - not very well.

The big practical problem with SRDs is jitter inherent to the device physics. The jitter causes lots of phase noise. For low phase noise comb generators (used for frequency multiplication) people have gone to nonlinear transmission lines, such as those made by MACom, despite the component cost.

Joe Gwinn

Which is to say, it fills in parameters for a built-in SPICE model, designated NJF, that is documented somewhere... older SPICE variants had J2N3819-named models for JFETs. I wouldn't designate that text 'a model', rather I'd call it a 'model statement' for the particular SPICE variant in use.

If you are designing Doherties then it is very relevant.

Skittles

So are you the "beer & skittles" kind or the kid candy kind?

Cheers

Phil Hobbs

Definitely beer and skittles.

Tally Ho!

Skittles