Terrible bondwire effect. How could I reduce its' effect?

Dear all,

I design a LNA. Firstly, I don't added any bondwire effect (1nH in series with 1ohm to pretend a bondwire) on each pad for the simulation. Afterwards, I added the bondwire model at each pad and do the simulation again. I found the NF is 3dB higher compared to the LNA without bondwire model. It is terrible. I have searched each bondwire model. I found the major contribution of the NF is the bondwire connect between bottom current source (core differential LNA) of the transistor's source and ground. Actually, how could I fix this problem? I have tried to use two pads to decrease the bondwire's inductance value. However, it is fail.
Another interesting thing is that the LNA is same in ADS and Cadence. I have added all the bondwire effect in both platform. However, the Cadence will give wrose result.

Thanks

wccheng

In your simulation you have to include the bond wire model when you run the Noise Figure optimization, and re-tune all the input match components.
Seems that the inductance (and capacitance to the ground) of the bond wire is changing the Gamma optimum point where you get the lowest Noise Figure.

Yep, Noise Figure is very dependent on ground bond wire inductance! It has to do with positive feedback of ground reactace to the input terminal. Keep it small--lots of bond wires that are short. A flip chip might even make sense!