|
|
Nathaniel Egwu
| Speaker: |
Nathaniel Egwu
|
| Date: |
Friday August 13, 2010
|
| Title: |
Accelaration
Techniques for Extraction within Semi-automatic Design Optimization
(Master's
thesis presentation)
|
Abstract
|
|
The simulation of the
physical layout of parameterized cells (pcells) with parameters
as their defining properties is vital in the circuit design process.
The steps involve generation of the layout by the optimization loop,
extraction and netlist creation for simulation. The usual optimization
of the physical layout is done several times and extraction has to be
repeated at every iteration. This greatly increases the time required
to extract the physical layout. We present a method that can be used to
speed up extraction. The formulation is based on investigating the
relationship between parameters of pcells and the parasitic
capacitances obtained in the netlist.
In contrast to full extraction of the entire layout, we implement
partial extraction of the layout of the individual pcells and compare
the time obtained in both cases. The effects of parameters will also be
classified as discrete or continuous based on the structure of the
netlists created. The procedure is implemented to some pcells which
form part of the low noise amplifier.
Some models that can be used to predict the netlist will be proposed
and comparison made between the extracted values and the predicted
model. This enables us to obtain the relative error between the
predicted and extracted netlists which is used to determine the
accuracy of the models. For models with high accuracy, time can be
saved by performing partial extraction where there exist a
significant difference between the relative error between the predicted
and extracted values. This can speed up the whole process since
extraction has to be performed at specific steps during optimization.
|
|