package modeling using Q3D Extractor

Did anyone ever try to model a package (e.g. BGA package) using Q3D Extractor ? There are some examples given by Ansoft for this purpose. I am a bit confused with the example. In the model, a ground plane which is supposed to model the pcb ground plane is included. This ground plane is usually a few milimeter below the package. Is there any consideration in defining the distance? Why the pcb ground in modeled in a such way that covers an area almost as big as the packge. Shouldn't it be a narrow track only?

The consideration for pcb ground comes from what is usually done in RF application boards: typical a gnd plane is put under the RF IC in order to have a very good ground. Typical distances are in the range of 0.1 to 0.5 mm. I do not expect to have it modeled as a narrow track because there will be a nonzero RF voltage in the ground if you'll use small track.
Bye
Mazz

Hi Mazz,

Thanks a lot for the reply. It's really helpfull. If you have access to IEEE paper, could you please have a look at the paper below:

SI issues associated with high speed packages
Jain, N.; Silvestro, J.; Cendes, Z.; Potluri, S.;
Electronic Packaging Technology Conference, 1997.
Proceedings of the 1997 1st , 8-10 Oct. 1997 Pages:310 - 312

In that paper, the ground plane is modeled 4mm below the package. Can you comment on this? What thickness should I make for the ground plane. Does it have to be a lot thicker than the skin depth? One more question, before I export the circuit model, I should reduce the matrix (ground net), shouldn't I ? What about the "Float at Infinity" matrix reduction. Should I also do that after the "Ground Net" reduction?

Thanks once again.

I'm sorry, I don't have IEEE access and I have not used Q3D but "only" HFSS in which I cannot generate any netlist. I think I cannot help you any more, to be honest.
But regarding the ground plane I can add that a good way to model it (in HFSS) is to define it as a perfect E plane (so no thickness defined). I hope it can help.
Bye
Mazz

Thank you mazz. Actually I want to compare the results obtained from Q3D and HFSS. Anyway, thanks.

In my experience both the simulators have good and let's say the same accuracy, if of course you use Q3D in the allowed frequency range (if I well remember the longest dimention of your package should be smaller than a fraction (1/8?) of the lambda). The main problem that I discovered is using the generated netlist, that sometimens oscillates!, and there is a long discussion between Ansoft and CST on the stability of their netlist extractors.
Bye
Mazz