Faster than 2n7002

Especially if the source is not grounded. Bipolar emitter followers love to oscillate, too. I tend to blame the oscillation on wire bond inductances, but that's conjecture on my part. A series gate or base resistor is the usual fix.

They oscillate because the impedance viewed into the base is negative under certain levels of emitter load capacitance. ...Jim Thompson

With the two lags (from C_GS and C_DG), resistance in the source turns into negative resistance at the gate. You need a big enough series resistor to make the net resistance seen by the gate positive.

I'm currently building a 200-400 MHz low noise TIA using one of those nice Avago ATF55143 e-pHEMTs cascoded by a BFP640 40 GHz SiGe.

The transconductance is about 3k, not bad for a single stage at that frequency. (Well, two stages technically--there's also a BFT92A emitter follower driving the coax.)

Interestingly, its stability is limited by the capacitance of the 10-ohm bead in the cascode's base. It works great at bias currents below 6 mA, but much above there, it wants to oscillate at 6 GHz. Interestingly this is just what LTspice predicts if the capacitance across the bead is

Still not bad for a hand-wired protoboard.

Cheers

Phil Hobbs

We love those Avago e-phemts. We can drive them on/off from ECL. The bigger ones make nice laser drivers.

It might work better as a real surface-mount circuit on a multilayer board. It will certainly work different!

You got that right. I didn't have time to send it out--the electronics for this gizmo have to be working by Monday.

Cheers

Phil Hobbs

Thankfully--I've still got two reels in the workshop for hobby-stuff.

Nice fusible ESD-detectors, too.

Cheers, James

I don't think I've ever zapped one. I certainly aren't careful handling them. Must be the humidity.

They do switch fast. Most "modern" SOT23s have a low Rds-on and a lot of capacitance, so 2N7002 looks pretty good there.

FDC2612 is 200 volts and very fast.

Would the FDC2512 be faster? It's rated at 150V and has similar capacitance, but half the Rds(on). And it can switch with a considerably-lower Vgs, conducts 2A at Vgs = 3.7V compared to 5.4V. Both show a spec'd required Qg of 8nC at Vgs = 10V, but the '2512 can be fully switched at 6V, taking 5nC.

Or is lower Rds(on) not helpful?

Too much theory for me. We needed to make 100 volts into 50 ohms, through a transmission-line transformer, and we tried a bunch of fets. A few FDC2612s in parallel worked great, with ca 1 ns edges.