Statistical design and optimisation for high-yield BiCMOS analog circuits
The use of the modified flexible statistical models for MOS and BJT allows the designer to incorporate the parameter variations in any nominal model deck. Therefore, the transistor parameters need not be re-extracted if a different nominal model is used. In addition, the important MOS transistor and BJT parameters for statistical simulation are identified. The strength of circuit simulation based on the modified flexible statistical model is its ability to estimate the performance variance of a given circuit without prior knowledge of the circuit performance. Therefore, the simulation model and methodology are not performance or circuit topology limited. This feature is important when a new design is simulated statistically . However, the simulations provided by the model do not yield information as to which transistor contributes to the performance variance. In order to enhance the capability of the modified and the original flexible statistical models, a sensitivity analysis (derived from the two factorial design for experiment technique) for the effect of parameter mismatch on circuit performance is developed. The sensitivity analysis provides an efficient way to detect the effect of each transistor mismatch on the circuit performance. Together with the flexible statistical model, it is proven useful in the analysis and design of precision analog and mixed-mode integrated circuit.
Statistical design and optimisation for high-yield BiCMOS analog circuits, Page 1 of 2
< Previous page Next page > /docserver/preview/fulltext/books/cs/pbcs008e/PBCS008E_ch17-1.gif /docserver/preview/fulltext/books/cs/pbcs008e/PBCS008E_ch17-2.gif