3

---

Fast Monte Carlo Method

---

The Fast Monte Carlo (FMC) engine is implemented in the Spectre simulator. The FMC method lets you extract useful statistical information without running the complete set of Monte Carlo samples, especially for high-sigma analysis. This reduces the simulation time.

The FMC method identifies the tail samples out of total N Monte Carlo points without simulating all N points for the specified circuit measurement specification or set of measures. The number of sample points that is actually simulated is circuit dependent.

In FMC method, you must specify any two of the following parameters:

• Total Samples (numruns)
• Target Yield (fmcsigma)
• Tail Samples (fmcnumtailsamples)

For example, the following settings are equivalent, which indicates that for 1M samples, the sample at 4.2 Sigma is the 14th worst sample:

• Total Samples = 1M and Target Yield = 4.2
• Target Yield = 4.2 and Tail Samples = 14
• Total Samples = 1M and Tail Samples = 14

The following table lists the relation between the fields, Max Points, Target Yield, and Tail Samples for FMC method.

Total Samples	Target Yield	Reported Samples	Sigma Range
20K	3	28	(2.99, 4.06)
100K	4	4	(3.98, 4.42)
1M	4.2	14	(4.2, 4.89)
100M	5.4	4	(5.39, 5.73)
500M	5.8	3	(5.73, 6)
1B	5.8	4	(5.79, 6.11)

Outputs and Target Specifications

To reduce the run time, it is recommended that you enable only the minimum set of specifications of interest. If both min and max for the goal are not needed for a high-sigma analysis, you can use either of them based on what is the actual criterion for a failure of the specific measurement.

Related Topics

Workflow of Fast Monte Carlo Method

Running the Fast Monte Carlo Method

Spectre FMC Analysis

Return to top