Creating SPICE models

I have LT Spice models of the MiniCircuits SAV541 and SAV551 phemts, sorta similar parts but enhancment mode. Based on Phil Hobbs model, tweaked a bit based on measurements. They seem to work for my applications, fast switching and some Colpitts oscillators. They don't include package parasitics.

You can have them and tweak for the Sky part.

It's always a hassle, trying to use RF parts in time domain, namely in Spice.

What are you doing?

The ones I have are based on a destroked version of an Avago model for a fancier simulator than I possess.

From my omnibus SPICE library: ===================

• Depletion pHEMT

• This is the datasheet model, except that LTspice doesn't understand a bunch of

• the parameters given by Avago--we'll see if it works

*.MODEL ATF38143_chip NMF( vto=-0.75, Beta=0.3, Lambda=0.07, Alpha=4,

*

• B=0.8, Pb=0.7, Cgs=0.997E-12,

*+ Cgd=0.176E-12, Rd=0.084, Rs=0.054, Kf=1e6, Af=1)

• PH 2/10/20: Turns out that Avago's flicker noise amplitude Kf is 18 orders of

• magnitude too high--they had 1E6 and it's more like 5E-12 for a 10 MHz

• 1/f corner frequency. Hacked it to be 1/f (AF=2) and adjusted KF for a

• 24-MHz 1/f corner, which should be roughly right.

.MODEL ATF38143_chip NMF( vto=-0.75, Beta=0.3, Lambda=0.07, Alpha=4,

• B=0.8, Pb=0.7, Cgs=0.997E-12,

• Cgd=0.176E-12, Rd=0.084, Rs=0.054, Kf=5e-11, Af=2)

• Fixed to be enhancement .MODEL SAV551_chip AKO:ATF38143_chip VTO=0.08 ==================

JL's models are likely to be better in the time domain, since he's put in actual measurement data.

The SKY65050 has about 1-2 dB higher flatband noise (0.35-0.4 nV/sqrt(Hz) than the old Avago parts. On the other hand, it has a considerably higher drain impedance (i.e. higher Early voltage or whatever it's called in a FET) than the Avago parts, which is a big win in time domain stuff.

Using a pHEMT as a follower is a depressing experience--some of them have gains of, like, 0.6 as followers.

Cascoding with a SiGe BJT such as a BFP640 makes them amazingly better, but requires, *ahem* a bit of care. ;)

If you do a babytalk groups search in SED for the string

"pHEMT" "SiGe" "cascode"

a whole bunch of stuff about that will come up.

Cheers

Phil Hobbs