chip inductor model ?

vcpd inductor

Dear All :
I want to do chip inductor , but I have no sense to do it , Does anyone have any comment ? How can I do th chip inductor for cadence simulation . Thanks

on chip inductor model

People typically use a Pi model, either a 9-element or 11-element. In cadence, there is a tool name VCPD (before named VPCM) which is integrated in MMSIM. It will generate the model as well as layout for you. It can be 2.5D or full 3D simulator. There are some other tools like HFSS, IE3D, Sonnet which are 3D simulators also. There are some free 2.5D tool like ASITIC

Dear Sir:
Thanks, Do u ever use it ? How about ADS ?
Do u ever compare the two tools ? Do u have any document for design chip inductor ? Thanks

You mean ADS momentum? It is essentially a 2.5D tool. I think it is good also for the first trial of inductor design.

There is no absolute judgement on which tool is better. It all depends on how good you provide your technology file to your simulator. I would say Full 3D solver always give you high accuracy even for the first time. However, if you have the money to fabraicate a lot of inductors (> 100) and calibrate the 2.5D tool, it might give you high accuracy also

Maybe you can ask helps from your foundry.

Check out www.helic.com/products/VeloceRF

mitgrace:

The planar EM solvers can handle thick metal pretty well now, which was not the case even a few years ago. For really high accuracy modeling of currents on chip inductors, I think you might get better performance for resolving currents for accurate Q calculation using one of these. Full 3D tools can be used, but if you are solving at frequencies around 1 GHz on metal that is lossy (like aluminum), then you need to model the penetration of the current into the metal (skin depth).

For aluminum conductors at 1 GHz, you have a skin depth of penetration of 2.6um. So your modeling method needs to be able to simulation this current penetration and not assume that all current simply flows on the outside skin of your conductors. I think it's actually more difficult for these solvers to do a good job matching measurements at lower frequencies than at high frequencies because of this issue. At higher frequencies, the skin depth is shallower, and you can approximate current flow more correctly with a hollow "tube" model for the conductor.

This affects both your Q, and also the effective coupling distance from trace to trace of the turns of your inductor. The turn-to-turn coupling depends on the effective distance between the electrons flowing in each conductor, which at lower frequencies is, on average, some distance below the surface of the metal.

This means that if you use a full 3D EM software for realistic metal on a real semiconductor process (aluminum or even copper conductors) you should make sure you also mesh inside your conductors. This becomes a rather big job in a hurry for full 3D volume meshers like HFSS or CST.

If you use a planar EM solver, make sure that it has a way to model the actual penetration of the current. The only commercial planar EM solver that I am aware of that will model the actual current penetration is Sonnet, which uses a multi-sheet model to approximate the penetration. I believe Sonnet also has an integrated interface to the Cadence Virtuoso platform, and they can give a broadband spice extraction model as well as S-parameters inside of Cadence.

Remember that to get correct results, you need something that correctly predicts where the current is, and where it isn't.

Good luck,

--Max