In analog circuit design, we often use variation analysis to understand what causes performance to vary. We will use a tool like Monte Carlo simulation where multiple parameters for each transistor are taken from a distribution and we simulate the circuit to get a distribution of the relevant measurements. The measurement distribution can be compared to a target spec and we can project the probability of failure.
What I am looking for is a way to understand how the variations of the individual transistors influence the measured result. Because these circuits take a long time to simulate, it is difficult to get more than a few hundred to a thousand Monte Carlo trials. And because of the number of transistors, the number of independent variables is also in the hundreds. Intuitively we know that the circuit will be most sensitive to combinations of parameters. For example, the worst case may be when transistor A and B have the largest difference in parameter X, but only if one of the two has a large parameter Y.
I am looking for guidance on how to analyze my Monte Carlo results in JMP to find the underlying sensitivities that have the greatest influence on the measured results.
