Partitioning and Net Grouping

The electrical behavior of a signal net depends on its proximity to other objects, such as power and ground planes or other signal nets (e.g., characteristic impedance is controlled by nearby supply planes, while crosstalk is determined by the locations of both planes and other traces. For meaningful analysis of any particular net, you need to determine what other planes and nets are nearby and include them in the simulation).

When you perform an analysis in CPA, the software automatically identifies nearby signal nets and power/ground planes. During full-package analysis, the signal nets are divided into several partitions, each of which consists of a number of nets that are all close to one another. This is done in order to efficiently and accurately analyze the entire package.

The following are some characteristics of partitioning:

• Different nets in each partition are treated differently. Some nets are considered to be valid members of the partition because they are sufficiently well surrounded by other nearby nets (in the same partition) to ensure that all neighborhood effects on them have been adequately considered. They are called valid when the inductance and capacitance values for those nets are an accurate representation of their true values.
• If a net is not valid for a particular partition, we call it nonvalid within that partition. This means that the net was included within the partition to model the environment of another (valid) net, but its own environment has not been modeled sufficiently to consider its self and mutual inductances to be correct.
• A net is always valid in at least one of the generated partitions. That same net may appear as a nonvalid net in a other partitions because it was needed there to model environmental effects on other nets.
• After the analysis completes, SIwave collects valid parts of the inductance and capacitance matrices for different partitions, discards the nonvalid parts, and combines the results together into a single accurate solution matrix.