de-embedding s-parameters: nonphysical results/positive s11 & s22

Hello,

I'm measuring a transistor device with the goal of obtaining transit frequency from the H-21 curve. I have open and short de-embedding structures to deembed using two methods outlined in the paper below:
AN IMPROVED DEEMBEDDING TECHNIQUE FOR ON-WAFER HIGH-FREQUENCY CHARACTERIZATION

the first one is the simple
Y_transistor = Y_dut - Y_open

when i use this method, at low frequencies Y_transistor = Y_dut. however at higher frequencies, I observe positive S11 and S22 for the de-embedded results. I have calculated k, stability factor, and it is indeed unstable for some frequencies, but I am not sure if this invalidates the transit frequency calculation from H21 (which, incidentally, the ft from the deembedded results is slightly smaller than the ft of the DUT itself).

the second method is:
Y_transistor = ((Y_dut - Y_open)^-1 - (Y_short - Y_open)^-1)^-1

when I use this method, i obtain positive s22 for the de-embedded results, though not positive s11. additionally the low frequency behavior of the s21 and s22 does not at all coincide with the behavior of the DUT. additionally, the ft is 10x higher than the ft of the DUT by itself.


If you have any comments or experience in this please help!

Hello, kput,

Presumably, you are referencing Koolen, et al from 1991. I cannot access it on my APS account, but probably wouldn't read it carefully enough anyway.

There's plenty of stuff on de-embedding on the internet, but it might be best to start first with your fixture. What can you say about it?

Mark