Learning Forum

[Jesus Guzman]

When plotting the plane impedance of a PDN should the renormalization impedance of the port be changed to 0.1ohms vs the default 50ohms?

[Dan Dvorscak]

Typically yes. By using the defualt of 50 Ohms, you have a huge mismatch between the low impedance of the PDN (at lower frequencies where it is close to lossless) and the port impedance. One of the issues here is that doing so causes a loss of numerical precision in the resultant S-parameters So instead of magnitudes varying betwen values of 0 to 1, they will only go between 0 to ~0.001 (potentially). This is especailly traumatic when converting the S parameters to Z. Small errors in S can be made much worse by the lack of precisions. Then at low frequencies where the stucture is very low loss, those small errors in S can produce a huge variation in Z.

[Jesus Guzman]

What would be the best approach then when doing the decoupling analysis for a board where the sweep is done up to 1GHz to see the effect of the package inductance and the lower frequencies in the KHz to check the bulk capacintance?

[Dan Dvorscak]

I would still re-normalize any ports on the PDN to 0.1 or 1 Ohm. The problems mentioned above are predominant at lower frequencies where the magnitudes of the Z-parameters are very small and highly sensitive to any issues in S. I don't think I've ever seen a case where renormalizing those ports to 50 Ohms was the best option, regardless of how high in frequency you are sweeping. 

[Jesus Guzman]

Good to know. Thank you very much for the reponse.