10.06.2016 Views

eldo_user

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Monte Carlo Analysis<br />

Introduction to Uncertainty Analysis<br />

If the mean is zero, the relative standard deviation scaled by<br />

is used instead:<br />

The coefficient of variation is defined as a multiple of the RSEM. Refer to “Mean and<br />

Standard Deviation” on page 478 for more information.<br />

Uncertainty as an Interval<br />

Under the assumptions that the probability distribution of the output measure can be assumed to<br />

be normal (Gaussian). The uncertainty is often defined as an interval about the measurement<br />

result that may be expected to cover a given fraction of the distribution of values. This interval<br />

can be associated to a number u(H) obtained by multiplying the standard deviation by a<br />

coverage factor κ p :<br />

The number p is the coverage probability or level of confidence of the interval. A measurement<br />

is then conveniently expressed as:<br />

Then for example, one may take κ p = 2 for having an interval with confidence interval of<br />

approximately 95 percent. By taking κ p = 3, a confidence level of approximately 99 percent<br />

would be obtained.<br />

It is important to understand that the assumption of normality cannot be used in any cases, and<br />

this motivates the introduction of Monte Carlo simulation methods as general and flexible tools<br />

for addressing difficult estimation problems. The Monte Carlo method provides a numerical<br />

approximation of the output distribution, which may be non-Gaussian. This distribution is in<br />

general not symmetric, and a coverage interval is not necessarily centered on the average value.<br />

Yield and Failure Probability Estimation<br />

You may want to compute the probability values associated with lower and upper bounds<br />

(y L , y U ) for a circuit performance Y. Conversely, you may want to compute the bound or<br />

threshold that corresponds to a target probability value π L, U .<br />

Eldo® User's Manual, 15.3 413

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!