next up previous contents
Next: 9.3.6 Simulation Results Up: 9.3 Inverse Modeling of Previous: 9.3.4 Handling Multiple Devices

9.3.5 Running the Inverse Modeling Procedure

Having this comprehensive model (Figure 9.7) which delivers match metrics associated to a set of doping profile parameters for each of our measured operating points, we are able to perform the actual inverse modeling procedure using SIESTA's optimization environment. Therefore, we define the optimization experiment based on LMMIN as shown in Example 9.6.
\begin{Example}
% latex2html id marker 9577
\caption{
The {TCAD}{} optimization
...
...g} \> \texttt{\dq{}demo.log\dq{}} )\\
)\end{tabbing}\end{minipage}\end{Example}
In this experiment description we define n3, ysigma3, n4, and ysigma4 as free parameters of the optimization. Therefore, the optimization tool will search for the shape and the peak concentration of the threshold adjust implant and for the well doping of the measured devices.

Figure 9.8 shows the inverse modeling procedure at work. Plots display the match vector for each iteration, and for the evolution of doping profile parameters. Figure 9.9 displays the job farming system and as one can see from this screen shot, computation hosts are utilized as far as possible. The user of this system is able to produce Postscript files of all plots, which can be displayed in the GUI, as it might be necessary for documentation purposes.


\begin{Figure}
% latex2html id marker 9504\centering
\includegraphics[width=0....
...ables a user to track the progress
of the inverse modeling system.}
\end{Figure}


\begin{Figure}
% latex2html id marker 9509\centering
\includegraphics[width=0....
...caption{
The job farming is on heavy duty
during inverse modeling.}
\end{Figure}


next up previous contents
Next: 9.3.6 Simulation Results Up: 9.3 Inverse Modeling of Previous: 9.3.4 Handling Multiple Devices
Rudi Strasser
1999-05-27