Granted Patent B2

US 7,369,974 · App. 11/217,577 · Granted May 6, 2008

Abstract

Claims (38)

1. A method for generating a polynomial for use in circuit modeling comprising:

a. iteratively curve fitting a characteristic map having a set of domain regions characterized by a current domain count without using data points in a domain region having a greater current domain count;

b. determining, for each of the domain regions, an error area in the characteristic map;

repeatedly performing the steps of:

c. generating a patch region enclosing the error area; and

d. curve fitting, using an nth order polynomial, the patch region, until there is no remaining error area in the patch region, wherein the polynomial is associated with the domain count, which is incremented for each patch region.

2. The method of claim 1 further comprising generating the nth order polynomial using data points generated by a circuit simulator.

3. The method of claim 2 further comprising providing extra data points by bilinear interpolation of the data points from the circuit simulator.

4. The method of claim 1 further comprising outputting the polynomials in order of decreasing the domain count.

5. A system for generating a polynomial based mode of a circuit comprising:

a circuit simulator operable to generate a data set comprising a plurality of data points;

a curve fitter operable to:

a. generate a characteristic map from the data set, wherein the characteristic map has a set of domain regions characterized by a current domain count;

b. determine an error area in the characteristic map;

c. determine a patch region that will contain the error area; and

d. iteratively curve fit, using an nth order polynomial, the patch region until there is no remaining error area in the patch region, without using data points in a domain region having a greater current domain count, and by associating the polynomial with the domain count, which is incremented for each patch region.

6. The system of claim 5 wherein the curve fitter is further operable to determine extra data points by bilinear interpolation of the data points from the circuit simulator.

7. The system of claim 5 wherein the curve fitter is further operable to:

increment a patch region count each time step c and d are executed; and

output the polynomials for each patch region in order of decreasing patch count.

8. The system of claim 5 wherein the characteristic map is a delay map.

9. The system of claim 8 wherein the delay map is generated by the nth degree polynomial from input slew data and output capacitance data.

10. A method for determining polynomials to model circuit delay comprising:

a. determining an error area in a characteristic map;

b. setting a current domain count of a domain region to zero;

c. determining a patch region that will contain the error area;

d. curve fitting the patch region and increasing the current domain count by one;

e. repeating the steps c-d until there are no error area within the patch region;

f. determining a previous domain region having the largest domain count and an error area and curve fitting the previous domain region without using data points in any of the domain regions having a greater current domain count than the previous domain region;

g. if, after curve fitting the previous domain, the previous domain region contains at least one error area, repeating steps c-e for the at least one error area;

h. repeating the steps f-g for all domain regions less than the previous domain region that has at least one error areas, until all error areas are removed from all domain regions.

11. The method of claim 10 wherein the step of determining an error area of a characteristic map comprises generating the characteristic map using data points generated by a circuit simulator and a nth order polynomial.

12. The method of claim 11 further comprising providing extra data points by bilinear interpolation of the data points from the circuit simulator.

13. The method of claim 10 wherein the curve fitting of steps c and f further comprises:

using a nth order polynomial for curve fitting; and

associating the polynomial with the domain region.

14. The method of claim 10 further comprising outputing the polynomials for each domain region in order of decreasing domain count.

15. The method of claim 14 wherein the step of determining an error area in a characteristic map further comprises determining an error area in a delay map generated by a nth degree polynomial from input slew data and output capacitance data.

Assignments (18)