Generating variation-aware library data with efficient device mismatch characterization

• US 8,204,730 B2
• Filed: 06/06/2008
• Issued: 06/19/2012
• Est. Priority Date: 06/06/2008
• Status: Active Grant

• Current Assignee: Synopsys Incorporated
• Original Assignee: Synopsys Incorporated
• Inventors: Liu, Jinfeng, Taraporevala, Feroze P.
• Primary Examiner(s): Thangavelu, Kandasamy
• Assistant Examiner(s): GUILL, RUSSELL L
• Application Number:
  US12/135,033
• Publication Number:
  US 20090306953A1
• Time in Patent Office: 1,474 Days
• Field of Search: 703/13
• US Class Current: 703/13
• CPC Class Codes: G06F 30/367 Design verification, e.g. using simulation, simulation program with integrated circuit emphasis [SPICE], direct methods or relaxation methods

Abstract

In a method of generating variation-aware library data for statistical static timing analysis (SSTA), a “synthetic” Gaussian variable can be used to represent all instances of one or more mismatch variations in all devices (e.g. transistors), thereby capturing the effect on at least one timing property (e.g. delay or slew). By modeling device mismatch with synthetic random variables, the variation behavior (in terms of the distribution of delay, slew, constraint, etc.) can be interpreted as the outcomes of process variations instead of modeling the variation sources (e.g. transistor shape variations, variations in dopant atom density, and irregularity of edges).

Claims

1. A method of generating variation-aware library data, the method comprising:

receiving a circuit simulation model on at least one device mismatch parameter;

performing a sensitivity analysis on all device mismatch parameters, the sensitivity analysis including computing a rate of change of a timing property compared to a rate of change of a device mismatch parameter;

deriving a distribution of each timing property using the sensitivity analysis;

determining a mapping of one or more synthetic parameters to the distribution, each synthetic parameter being a random variable representing an impact on a timing parameter of all device mismatch parameters grouped to the synthetic parameter;

selecting an assignment of device mismatch parameter values; and

performing a deterministic circuit simulation using the assignment of device mismatch parameter values to generate the variation-aware library data corresponding to a 1 sigma of the synthetic parameter.

2. The method of claim 1, wherein the circuit simulation model is a commercial SPICE model.

3. The method of claim 1, wherein at least one device mismatch parameter is a threshold voltage or a gate length.

4. The method of claim 1, wherein performing the sensitivity analysis includes computing a rate of change of the timing property compared to a rate of change of a device mismatch parameter.

5. The method of claim 4, wherein sensitivity values generated by the sensitivity analysis are expressed in one of linear terms and non-linear terms.

6. The method of claim 1, wherein deriving the distribution of each timing property using the sensitivity analysis includes:

performing a Monte Carlo sampling for all device mismatch parameters; and

computing each sample value of the distribution by performing a linear sum of a product of each sample value of Vi_j and Si_j,

wherein Vi_j is a specific device parameter and Si_j is a sensitivity value of the specific device parameter.

7. The method of claim 6, wherein deriving the distribution of each timing property using the sensitivity analysis includes computing the 1 sigma from the distribution.

8. The method of claim 6, wherein deriving the distribution of each timing property using the sensitivity analysis includes computing the 1 sigma, wherein computing the 1 sigma includes squaring each sensitivity, summing the squared sensitivities, and taking a square root of the sum, when all device mismatch parameter are independent Gaussians with linear timing responses.

9. The method of claim 1, wherein performing the sensitivity analysis includes characterizing only active devices.

10. The method of claim 9, wherein characterizing only active devices includes evaluating a predetermined number of points in a two-dimensional library table space to determine the active devices.

11. The method of claim 10, wherein the predetermined number of points is 3-7.

12. The method of claim 1, wherein performing the sensitivity analysis includes reducing library table points for sensitivity extraction based on nominal timing responses or using predetermined library table points.

13. The method of claim 12, wherein a predetermined number of points is 3-7.

14. The method of claim 1, wherein performing the sensitivity analysis includes:

running full sensitivity extraction at a first point in a two-dimensional library table space to determine dominant parameters;

extracting sensitivity values for most dominant parameters at a second point in the two-dimensional library table space;

extracting sensitivity values for selected, less-dominant parameters at the second point; and

interpolating sensitivity values for any remaining, less-dominant parameters at the second point using the sensitivity values for the selected, less-dominant parameters at the second point and the full sensitivity extraction at the first point.

15. The method of claim 14, wherein the two-dimensional library table space is based on input slew and output load.

16. The method of claim 14, wherein the two-dimensional library table space is based on clock slew and data slew.

17. The method of claim 1, wherein determining a mapping of a synthetic parameter to the distribution includes replacing arbitrary quantile values that correspond to the synthetic parameter when the distribution is not Gaussian.

18. The method of claim 1, wherein multiple synthetic parameters can be determined by at least one of an arbitrary grouping of device mismatch parameters, and performing principal component analysis to distributions of timing attributes.

19. A non-transient, computer-readable medium storing computer-executable instructions for generating variation-aware library data, which when executed by a computer performs steps of:

receiving a circuit simulation model on at least one device mismatch parameter;

performing a sensitivity analysis on all device mismatch parameters, the sensitivity analysis including computing a rate of change of a timing property compared to a rate of change of a device mismatch parameter;

deriving a distribution of each timing property using the sensitivity analysis;

determining a mapping of one or more synthetic parameters to the distribution, each synthetic parameter being a random variable representing an impact on a timing parameter of all device mismatch parameters grouped to the synthetic parameter;

selecting an assignment of device mismatch parameter values; and

performing a deterministic circuit simulation using the assignment of device mismatch parameter values to generate the variation-aware library data corresponding to a 1 sigma of the synthetic parameter.

20. The computer-readable medium of claim 19, wherein the circuit simulation model is a commercial SPICE model.

21. The computer-readable medium of claim 19, wherein at least one device mismatch parameter is a threshold voltage or a gate length.

22. The computer-readable medium of claim 19, wherein performing the sensitivity analysis includes computing a rate of change of the timing property compared to a rate of change of a device mismatch parameter.

23. The computer-readable medium of claim 22, wherein sensitivity values generated by the sensitivity analysis are expressed in one of linear terms and non-linear terms.

24. The computer-readable medium of claim 19, wherein deriving the distribution of each timing property using the sensitivity analysis includes:

performing a Monte Carlo sampling for all device mismatch parameters; and

computing each sample value of the distribution by performing a linear sum of a product of each sample value of Vi_j and Si_j,

wherein Vi_j is a specific device parameter and Si_j is a sensitivity value of the specific device parameter.

25. The computer-readable medium of claim 24, wherein deriving the distribution of each timing property using the sensitivity analysis includes computing the 1 sigma from the distribution.

26. The computer-readable medium of claim 24, wherein deriving the distribution of each timing property using the sensitivity analysis includes computing the 1 sigma, wherein computing the 1 sigma includes squaring each sensitivity, summing the squared sensitivities, and taking a square root of the sum, when all device mismatch parameter are independent Gaussians with linear timing responses.

27. The computer-readable medium of claim 19, wherein performing the sensitivity analysis includes characterizing only active devices.

28. The computer-readable medium of claim 27, wherein characterizing only active devices includes evaluating a predetermined number of points in a two-dimensional library table space to determine the active devices.

29. The computer-readable medium of claim 28, wherein the predetermined number of points is 3-7.

30. The computer-readable medium of claim 19, wherein performing the sensitivity analysis includes reducing library table points for sensitivity extraction based on nominal timing responses or using predetermined library table points.

31. The computer-readable medium of claim 30, wherein a predetermined number of points is 3-7.

32. The computer-readable medium of claim 19, wherein performing the sensitivity analysis includes:

running full sensitivity extraction at a first point in a two-dimensional library table space to determine dominant parameters;

extracting sensitivity values for most dominant parameters at a second point in the two-dimensional library table space;

extracting sensitivity values for selected, less-dominant parameters at the second point; and

interpolating sensitivity values for any remaining, less-dominant parameters at the second point using the sensitivity values for the selected, less-dominant parameters at the second point and the full sensitivity extraction at the first point.

33. The computer-readable medium of claim 32, wherein the two-dimensional library table space is based on input slew and output load.

34. The computer-readable medium of claim 32, wherein the two-dimensional library table space is based on clock slew and data slew.

35. The computer-readable medium of claim 19, wherein determining a mapping of a synthetic parameter to the distribution includes replacing arbitrary quantile values that correspond to the synthetic parameter when the distribution is not Gaussian.

36. The computer-readable medium of claim 19, wherein multiple synthetic parameters can be determined by at least one of an arbitrary grouping of device mismatch parameters, and performing principal component analysis to distributions of timing attributes.