Improving the accuracy IDC with thick metal by IE3D

what should be done for the best accuracy of simulating IDC (width of slot: 3um; thickness of metal: 1.3um) by IE3D? And why were some element values in the lumped-element equiavlent circuit automatically generated by IE3D after the s-parameter simulation negative?

Hi, Burton:

What is IDC? If it is a slot structure with width of 3 um and thickness of 1.3 um, the thickness is no longer very small compared to width. It is possible the modeling of true thickness is critical. You can send me your model and I can have it checked for you (jian@zeland.com).

Quite some RFIC and packaging designers want to know the equivalent circuit's values. However, given a structure, what is its equivalent circuit? It is not fixed. The users have to pick their favourite model. IE3D only tries to match the values of the equivalent circuit model picked with the s-parameters. The fitted values may not have physical meaning. You have to pick the right model for it in order to get meaningful equivalent circuit parameters for you.

Best regards,

Hi, Jan

It is an interdigital capacitor, the width of all of the gap is 3.5 microns, the thickness of metal is 1.5 microns, and the metalic finger width is 12 microns.

When the metalic loss is accurately calculated, I should use ARTT and AEC, and in this case the actual metalic thickness 1.5 microns can be replaced by 0.01 microns. But if I do not want to calculate the metalic loss, should I still use ARTT (AEC should be always used) ? And should the metal thickness be 1.5 microns? Can only AEC guarantee that accurate S-paramter can be obtained in this case?

Regards.

Hi, burton:

For IDC structure, if the width of the finger is 12 um, the thickness of 1.5 um may or may not need to use thickness model. Certainly, thickness of 1.5 um is not too small compared to the gap of 3.5 um. You may give it a try with or without thickness (either ARTT or manually created thickness model) and see how much difference you will get. I suspect the non-thickness model may just slightly under estimate the capacitance between the fingers and you may see little difference in the s-parameters as long as you use AEC.

Using ARTT should further improve the accuracy. If you want to have even higher accuracy, you should try to use the CONTEMPORARY MESHING and APPLY EDGE CELLS TO THICKNESS POLYGONS. It should yield even higher accuracy.

Best regards,

thanks for your suggestions, Jian.

About meshing, I always use CONTEMPORARY MESHING which is more effective than classical meshing.

I used the thickness model without AEC and obtained the results which is similar to that obtained HFSS at the cost of a long time. And I compared the results using thin model with AEC with the HFSS results, there is a slight difference. So I will use thickness model without AEC for my further design because it seems that the meshing is more effective and resonable for my special design by using thickness model than by thin model.

Best Regards.

Hi, Burton:

Thickness model without AEC is already very good. In fact, even you select thickness model with AEC, we will not create edge cells on the thickness traces normally. You will get the same results as thickness model without AEC. Only you choose thickness model, AEC and the option fo applying AEC TO THICKNESSS TRACES TOO will make the difference. If you choose all of the above, it will yield even more accurate results. However, it may be an over-kill because you should be able to obtain very accurate resutls using thickness model without AEC (or with AEC and the default option of APPLYING AEC TO OPEN EDGES only. Regards.