Matching confusion: usin RFSim99 vs Smith Chart software vs?

Hello

Please can someone plough through this explanation and show me where I have got hold of the wrong end of the stick i.e. got myself muddled up?

I have a low power RF transceiver IC whose datasheet says the output will match into a 50 ohm environment by use of a 10nH series L. I have confirmed that without/with this inductor I derive powers of -10dBm/-4dBm into a 50 measurement system (in my case a Marconi 6950 power meter). So far so good.

I then need to drive the signal into a single PHEMT power amplifier with a gain of approx 10dB - using the S2P dataset for the device shows I need a matching net looking into it because, at the frequency of interest, the VSWR is hopeless (35 or similar - very capacitive junction).

If I use Fritz Delssperger's excellent Smith Chart program, I see a series and shunt L of approx 40nH each, from the transceiver inductor to the gate of the phemt. Doing so gives me a match of much better than VSWR=2 - in fact at the spot frequency the match is a return loss of 30dB or better and for a few MHz either side, all points transform into the VSWR=2 circle on the Smith chart.

However If I use RFSim99 I see that, for the same input matching components, the match is comepletely wrong.

Confusion 1 - if my phemt is represented by an S2P file, what should I be doing with the output side of the S2p file? If I am looking at S11 for the phemt then I should presumably terminate S22 in a 50R load - and then when I want to next solve for S22 can I leave the previously-determined network on the input port ?

Confusion 2 - if the transceiver needs 10nH on its output to appear as 50R, then I model sourcing from a 50R source into an L network with a series and a shunt L, can't I absorb the transceiver's 10nH with the series element of 40nH in the matching network?

Confusion 3 - if the phemt output is transformed to look like 50R looking back into it, then does that Z get transformed through the phemt such that the load now seen looking into the phemt from the other side has changed?

I don't know which to model first - S11 or S22 and do I do it with or without 50R loads terminating the "unwanted" port?

Thanks for any clarifications anyone can offer - going round in circles here (if you'll excuse the pun)

RF

http://www.fritz.dellsperger.net/Dow...th%20V3.10.zip
http://electroschematics.com/wp-cont.../rf-sim-99.zip