S Band LNA Design, but conditional stable at 16 GHz

Hello

I recently am designing an S Band LNA using ATF-541M4, an Avago Technologies' p-HEMT.

But the stable factor shows that the amplifier is stable at any frequency except for somewhere around 16 GHz. This causes me a lot of problem. I've tried to put a series resistor at gate, but still can't kill the instability.

I've found that the reason why it is conditional stable is that the connection between the source of the FET and the ground is imperfect. The via hole really act as inductors, maybe resistors also, at 16 GHz. How to deal with that?

But I still remember that my goal is to design an S band LNA, should I really need to care about what is happening at 16 GHz? Can I ignore it? Or is there any possibility that what happens in simulation, the conditional stability at 16 GHz of cause, will never take place in real world?

Yes, if add all the parasitics of components may be possible to do not get the 16GHz oscillation.
Do you see an oscillation, or the simulation give a potential unstable amplifier at that frequency (K<1)? Meantime potential unstable doesn't mean it will oscillate.

No, I'm still doing my simulation work, and it shows potential unstable.

If the RF Source and Load impedance remain 50 Ohm at that frequency, I don't think it will oscillate. However, it is not my working frequency. In my opinion, the input and output filters will make equivalent source and load impedance at somewhere near the edge of the Smith Chart, so I'm afraid it would oscillate.

And I found that if I change my bias point by increasing current from drain to source, the stability is getting better. But the current is actually close to the maximum current it can bare, and the model can not kill itself if the current is larger than the maximum.

If your LNA has an "potential instability" ( because even the stability factor is less than unity, it doesn't meant that the amplifier will absolutely oscillate ) point around 16GHz, you may cutt-off by using a low pass filter at the output.But this oscillation amplitude is too high, it drives the LNA saturation and the LNA gain may drop.
The best way is to cut-off the oscillation by applying stability practices.