User Guide

8 Understandin

Optimization Principles and Options

•

Increase simulator accuracy (e.g., reduce RELTOL).

The first proposed technique is preferred because it does not

affect the time required to run each simulation (usually the

determining factor in how long an optimization run takes).

Global and Local Minima

The curve in Figure 4-2 shows a 1-dimensional function with 2

minima. Point M1 is a local minimum; it satisfies the conditions

for a minimum, but there is another minimum which is smaller.

Point M2 is the global minimum for the function. There are no

points within the range of the function which are smaller.

All practical optimization techniques find local minima,

including the algorithms used by the PSpice Optimizer. This

may or may not present a problem. The application may not

have any local minima within the domain of interest. If local

minima do exist, the global minimum may be the nearest

solution to the starting point. This is discussed further in Starting

Points on page 4-8.

Startin

Points

It is important to begin with a good estimate of the starting point.

There are two reasons for this:

• The process may converge to the wrong solution (a local

minimum) rather than to the right solution (the global

minimum).

Example: Consider Figure 4-2. If point A is chosen, the

PSpice Optimizer will most likely find local minimum M1.

If point B is chosen, the optimizer will most likely find the

global minimum M2.

• The PSpice Optimizer may require a large number of

simulations to find a region close to a solution. It is usually

ure 4-2

Global and Local

Minima of a Function