Hi all

First of all, thanks and congratulations to all the people in MW101. I think this is in the top 3 (#1) of the MW resources out there!!

I've designed and fabricated my forst version of a 1 to 40GHz DA MMIC (GaN on Si), with 20dB gain and Pout >1W P1dB for the whole bandwidth (hopefuly). Input and output match is good, gain is decent (some 17-18dB, with Vgs-s still to be tuned due to differences in Vth and Idss from EM simulation to measurements) and I still have to measure Pout.

BUT, There's a naughty absorption in my S21 measurement around 12GHz (the "A" dip in the first picture). My amplifier consists of 2 series connected cascode DA stages, one for gain and the second for power. I've checked in EM simulation that this effect is due to my first stage being loaded by the DC bias external circuitry (drain bias is applied to both sides of the drain line of stage #1, second picture).

When I apply bias to the second stage I also see this loading, if I apply it through the drain termination side. Only applying bias voltage directly through the output pads with a bias tee can I avoid this effect... And only in the power stage... And here comes the question...

Thinking about it, I would expect this effect to be present in every wideband distributed amplifier, as bias bondwires and the capacitance of the PCB areas where they land will resonate in the GHz range (be it higher or lower). But I have seen several papers of amplifiers with equal or higher bandwidths (though with lower power and gain) for similar and even higher frequencies that don't have such absorptions in their S21 response. What is more, some apply bias with only a line section with a capacitor (lambda/4 for several decades???), and this puzzles me completely, as I would expect those designs to behave like mine... What am I missing? Is it a matter of output line impedance (I guess most of us work with 50 ohm)? Do they use 4mm long bondwires or what? Because I guess they will have to encapsulate and somehow bias their MMICs for their final applications also...

And regarding my biasing problem... Any suggestions? I'm reading about using transistors as active sources but to be honest I've used some 85% of my MMIC area and don't think I'm going to be able to include anything substantially "big" in my design, so I need to think "out of the MMIC"... I'm trying to take resonances down below 1GHz but it doesn't seem to be easy...

Anyway, sorry for my dissertation and thanks again to all of you for reading this :smile:

Iban