next up previous contents
Next: 3.4.2 Comparison Up: 3.4 Practical Use Previous: 3.4 Practical Use

3.4.1 Analysis Using Response Surface Methodology

The procedure for analyzing a complex model with several possible control variables is listed in points 1 - 9.

1. Define one or several responses which should give information of the suitability of the model characteristics.

2. Choose out of the possible input parameters those which have the highest impact on the response. Assign ranges for these input parameters.

3. Select an experimental design. Criteria for the type of experimental design are given in Section 3.1.18.

4. Additional transformations can be chosen both for the input and output parameters. This is discussed in Section 3.2.4.

5. Generate from the parameters, their ranges and transformations with the selected experimental design, a set of sample points.

6. The model, usually a sequence of several simulations, has to be evaluated on the sample points generated. For the scheduling of the simulations a simulation environment, like the VISTA or the SIESTA framework is useful.

7. The experimental points and there results of the simulations are collected in an experiment table.

8. The factors of the RSM-model -- the factors of the polynomial -- are calculated from the data from the experiment table. Here again the transformations of the controls and the responses are used.

9. The fitted model is analyzed.

In Section 5.1 the optimization of a DMOS transistor is demonstrated. In this example the influence of the epi-layer on the on-resistance using Design of Experiments and Response Surface Methodology is given.


next up previous contents
Next: 3.4.2 Comparison Up: 3.4 Practical Use Previous: 3.4 Practical Use

R. Plasun