Spice Simulates Custom Random Distributions for Monte Carlo Analysis (.PDF Download)

Feb. 5, 2019
Spice Simulates Custom Random Distributions for Monte Carlo Analysis (.PDF Download)

Not all Spice versions perform Monte Carlo simulations. Even those that do may only have a small number of available distributions, much less custom ones. LTSpice, for example, has built-in random distributions limited to Uniform and Gaussian (defined by sigma, but no specified limit). Unfortunately, this excludes many real-world distributions such as Gaussian (with maximums of 2-sigma, 3-sigma, etc.), bimodal, triangular, and U-shaped for example. A review of web reference shows few solutions beyond the available functions.

However, using straightforward statistical theory, you can create a random process of any distribution listed above. Even histograms from manufacturers’ datasheets or your own lab measurements are easily simulated. What are the minimum requirements? A calculation tool having a uniform random-number generator and a lookup table function. We’ll apply the method here using Excel and Spice.

This statistical/simulation adventure was both fun and insightful. The nuts and bolts of creating a circuit simulation developed an intuitive understanding beyond the abstract statistical theories.

Comments

To join the conversation, and become an exclusive member of Electronic Design, create an account today!