Realistic worst-case parameters for circuit simulation

The spread in device performance due to random variations in the manufacturing process is usually characterised for the designer by a set of parameters representing typical transistors along with the best and worst cases. The success of a design is obviously dependent on the accuracy of these parameters and it is shown that obtaining a worst and best case set by combining the worst and best values of individual parameters gives unrealistic results. An alternative technique is proposed which results in an accurate parameter set using a single measurement to characterise transistor performance. Both methods have been used to obtain parameter sets for n- and p-channel devices.