Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database
Quick Links

# Method and System for Reducing Parameter by Utilizing Parameter Equivalence

IP.com Disclosure Number: IPCOM000234594D
Publication Date: 2014-Jan-21
Document File: 6 page(s) / 183K

## Publishing Venue

The IP.com Prior Art Database

## Abstract

A method and system is disclosed for reducing parameter by utilizing parameter equivalence. The method and system utilizes parameter equivalence and reduces parameter values in Statistical Static Timing Analysis (SSTA).

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 48% of the total text.

Page 01 of 6

Method and System for Reducing Parameter by Utilizing Parameter Equivalence

Disclosed is a method and system for reducing parameter by utilizing parameter equivalence. The method and system utilizes the parameter equivalence, wherein

multiple physical parameters are modeled as single equivalent virtual parameter given as,

PV = PE1, PE2,…, PEN

where,

PV = Virtual Parameter. Here, the virtual parameters are propagated, statistically processed and reported.

PEi = Equivalent Parameters. Equivalent parameters are original parameters present in a timing model. All instances of equivalent parameters are stored or processed as

virtual parameters.

The method includes the step of reducing overhead back to original non-separable model, given as,

(2)

In eq. (2), PE is Equivalent Parameter and represents PC1 - PCN

When any of PC1 - PCN is encountered, the method uses PE. In eq (2), corner variables are non-statistical by definition.

The method and system may optionally subdivide the PE parameter sensitivity into components, wherein one component or fraction contributes to the virtual parameter by

using a fully correlated fraction that serves as a corner variable sensitivity. This approach is called fractional equivalence. The remaining fraction may be modeled as E/L split in mean. Modeling the remaining fraction is non statistical which ensures bounding result. In addition, it reduces runtime or memory overhead due to parameter reduction. The remaining fraction can also be modelled as uncorrelated data which is statistical independent parameter and less pessimistic than the E/L split. This uncorrelated model reduces runtime or memory overhead due to parameter reduction. Further, the remaining fraction is modelled as standard correlated source of variation overhead. For example, by propagating it as original physical parameter. This standard correlated source model is less pessimistic than E/L split and more pessimistic than uncorrelated model. The standard correlated source model drives minor additional runtime and memory overhead. Fractional equivalence enables modeling of physical mistrack between parameters that require coverage. For instance, VDDVI1-VDDVIN

share some common variability due to a common power supply, but have independent

1

(1)

Page 02 of 6

variation due to separate distribution networks. This independent variation may be modeled using fractional equivalence.

In addition, utilizing parameter equivalence provides some benefits. It is more efficient than unity correlation, and can replace existing applications. For parameter equivalence, there is no data mapping required and it processes single rather than multiple external variables. In addition, parameter equivalence is less constrained than unity correlation.

Further, parameter equivalence is used for voltage Islands, which may be better suited

for some applications than headered voltage island approach. Headering occurs at higher level wherein all processing occurs ident...