How to set up the port in HFSS when using a thick substrate

Hi! I tried to model a planar interdigital capacitor above a grounded silicon substrate in HFSS. The substrate thickness is 280um, and the size of the port extension of the capacitor is 1.24um. I used 2 lumped ports for simulation of the capacitance, both port sizes are 1.24um*280um. I found that the arrow of the mode at the port is distributed only on top of the port, instead of pointing to the ground. For comparision, I also used IE3D with the advanced extension ports to simulate. It turned out that the result in HFSS is much bigger than that in IE3D ( 100 times). While I increased the mesh density of the fingers in HFSS, it seemed little help to explain the result difference with IE3D. I wonder if something is wrong with the port setup in HFSS, maybe the port is too long in such a thick substrate. Could somebody give me some advice? Thanks!

You should try SonnetLite (www.sonnetsoftware.com, I work for Sonnet). Thick substrates are typically not a problem for Sonnet port calibration. When you get it installed and registered (takes about 5 minutes total), start it up, click on Help->Examples->Resistors, Capacitors, Inductors. Then click on the CAP example (top right hand corner of the example page). Then click on "Load into Project Editor". Next save the project somewhere (so you can modify it). The example is for a capacitor on GaAs. Change the substrate to your Si substrate (Circuit->Dielectric Layers), set your frequencies (Analysis->Setup...) and click Analyze ("em" icon on the right end of the tool bar at the top).

Sonnet is the only EM tool with perfect port calibration. Approximate port calibration is OK if you do not need high accuracy and you do not push limits. Thick conductive substrates push limits. Sometimes in these cases people even think that transmission line theory is no longer valid, but in reality it is just port calibration error. Transmission line theory is 100% valid, as is easily seen when you use perfect port calibration. Sonnet's prefect port calibration will have no problem with thick substrates...unless your line is overmoded. If that is the case, then you will also have problem when you build it. To check, just re-analyze with longer connecting lines, but still calibrated up to the capacitor. If this is not clear, I will be glad to provide more detail.

Thanks so much for your advice! I will do it now!

You are most welcome, Yanyan. Please let me know how it goes.

Hi, rautio,

Could you please explain a little bit in detail what a 'perfect port calibration' is as compared to an 'apporoximate port calibration'?

Thanks!

Hi Zhang -- By perfect calibration, I mean perfect to within numerical precision provided that there are no over-moded port connecting lines. For example, radiation (a second mode of propagation), or box resonances (another mode of propagation), or microstrip lines 1/2 wavelength wide are problems to be avoided.

To demonstrate that port calibration is perfect, I usually take a 6-resonator hairpin filter. This filter is in the Sonnet examples directory (Help->Examples). I split it into two halves by cutting it horizontally across all the hairpins. Then I analyze each half separately, then connect the two halves back together with circuit theory. The result in identical to an em analysis of the entire filter. This can only be done with perfect port calibration, i.e., it can only be done with Sonnet.

Another way to see this is described in my paper:

EM-Component-Based Design of Planar Circuits - Aug 2007 IEEE Microwave Magazine, pp. 79-90.

You can download it by going to www.sonnetsoftware.com and clicking on the paper title on the right side if the home page. The article describes how we can take lots of little pieces out of a circuit on silicon, put in perfectly calibrated ports, then analyze each little piece and put them back in with circuit theory. The result is exactly the same as an em analysis of the entire circuit. However, things can go wrong. So be sure to read the article if this is important to you.

You can tell approximate port calibration in several ways. Some of them require a separate analysis of the Zo of a port connecting line. Sometimes they do not even allow two ports close together because that would require coupled port connecting lines. Sometimes they require that you specify a "line integral" to determine line voltage. All these approaches are approximate. They are OK at low frequency, with thin substrates, and with low loss substrates. They start to fail otherwise. Perfect port calibration does not fail in these cases.

It is easy to check if you are using perfect or approximate port calibration. Simply run the tests in the paper I described above.