saturating cascode

Fun.

The falling edge would be vaguely fine with BJTs, but the rise would be gross, on account of charge storage.

Cheers

Phil Hobbs

I'll be making a short negative pulse into a 50 ohm load, so the load itself provides my pullup current.

Strangely, in my single-fet prototype, and in the simulated cascode, the fall time is slower.

I intend to apply rigid scientific principles and analysis, namely lay out a lot of proto boards and see if any work. Much below about 200 ps, I don't entirely trust Spice.

"Below 1 ns, things get hard. Below 100 ps, they get really hard."

I stuck it into LTspice with that hacked-together SAV-551+ model, and the spherical-cow showed about 80 ps fall time and 200 ps rise. That was considerably faster than a plain common-source stage.

Cheers

Phil Hobbs

I stuck it into LTspice with that hacked-together SAV-551+ model, and the spherical-cow showed about 80 ps fall time and 200 ps rise. That was considerably faster than a plain common-source stage.

(Reposted to add model inline instead of a library.)

Cheers

Phil Hobbs

Cool. I thought this was a text group that didn't allow attachments.

In my version, I used different fets, 4 of them, and drove them harder; ~1 volt gate drive from low impedance, and higher cascode bias. My duty cycle will be very low (envision firing largish lasers) so I'm optimistic I won't blow out the gates. I asssume that gross gate over-drive makes them conduct harder; we'll see. [1]

I'm seeing pretty fast edges, but my fet model is crudely hacked from yours, and doesn't include wirebond inductances.

(I am pretty heavily NDA'd on this one, so the concept is general. I could go private if you are interested.)

[1] I have long suspected a bipolar gain mechanism in mesfets and phemts, namely huge drain currents when the gate begins to conduct serious current. Maybe that's just some exponential and some imagination.

Nice. Small inductances matter a lot--I recently talked about that fancy vehicular lidar TX board that I got to blow up on purpose--the same EPC GaNFETs you like, with a few-nanofarad 0402 cap and a triple-stacked Osram diode laser, producing ~30A pulses of 5 ns FWHM. I backed out the strays from my measurements, and found a really good fit with 200 pH in the drain and 180 pH in the source. Not a bad layout! (The source inductance was a worse problem.)

Cheers

Phil Hobbs

Does this need four FETs, or just two? Did you add the extra pair for stronger drive? Forgive me if I'm missing something...

Clifford Heath