hlep!! Microwave amplifier design in ADS

Does anybody try to design a microwave amplifier in ADS with a transistor nonlinear modeling? How accurate is this simulation prediction? Why measured efficiency is 15% more than the simulation? Seems like too much different. Is it a common problem of ADS? What could be the possibilities the simulation did not give good agreements?
Regards

The differences should not be owed to software, it's your resposibility to understand how the program operates on your device model and how much is the reliability of the model provided by foundry.

And for an RF engineer, the intuition to parasitics is a must.

Actually you can go less than 15% accuracy if you would use precise model for that transistor and tweak a few things in balancing in circuit.

Remember the golden rules in stability (stern stability factor etc) in high frequencies. Amplifier is very stable and oscillator is very unstable. Testing it alone would seem okay but putting the amplifier in whole design propability is your IC's will act erratically only because your amplifier is a little unstable.

Actually I am using the component modelling with the parasitic effects. The non-linear modelling of the transistor includes the package effect as well. It could be the difference made by fabrication. Thanks guys for your suggestions.
Regards

We obtain approx. %2.5 accuracy with HB simulation inADS compare to measurements.
All depends on modelization of the components..

@BigBoss

how do you get your models? I mean, how do you model lumped elements? Do you measure response of each and take the S parameters in the simulation?

What about active devices, do you directly use nonlinear models (including package) from manufacturer? Or do you modify them somehow?

Where do you think the major errors are made in the RF simulation (nonlinear)?

Thanks,
rfmw

hi,
please attach below pdf file,
thanks.

Maybe BigBoss is referring to integrated circuits.
Usually in companies that has foundry there are dedicated groups who develops models for their own technologies, and can be very complex one, for both actives and passives devices. I'm speaking about thousands of man-hours of very skilled people.
In discrete designs the job is somewhat more simple (S parameters are given by manufacturer, linearity can be directly measured in test boards) but less accurate, sometime.
In my experience I've seen very good performin circuits designed with discrete parts, not well optimized in term of power consumption, compared to less performing IC much more efficient.
I hope it can help.
Mazz

Absolutely you're right...

My company has own models that are very accurate ( maybe best ) and other passive components are measured very precisely and the library group make many measurements "on the wafer" so, we have very accurate models and very consistant statistics about them.

So, simulations and measurements are in a harmony within %2.5 accuracy...

It should be so , otherwise we loss money.

For commercial available models, I can say only that are statistically proven within some tolerances and the models are rarely diverge from mean values that is very normal.
The best way to design very accurate circuits to measure a certain lot of the component that will be used in the circuit. It's not so easy but should be...

Who comes from a certain world speak a certain language...
Is the same company or we are in competition... but not in elektroda!
Auf Wiedersehen
Mazz