RBF Based Performance Model for High Dimensional Analog Designs

Advancements in semiconductor technology in terms of size and complexity contribute to sophisticated manufacturing processes. Notably, the analogue part of designs is becoming increasingly complex. The design space is too large to be evaluated by accurate simulators to come up with the required design configuration. Linearising a large design space with several interacting parameters without compromising accuracy is becoming crucial, each with a wide range and covering potential design points, and is one way of problem-solving. Hence, a satisfactory model of the system can be used to evaluate to save the design time and can help in time to market in microelectronic manufacturing. These complex nonlinear systems make the designing of accurate linear models difficult. RBFN (Radial Basis Function networks) evolutionary approaches are used in this proposed model to show how to build non-linear models that make sure the design space is filled correctly. Achieved research direction with suggested design models are performing accurately when subjected to a smaller number of inputs for generating the results could produce an estimation of performance space precisely. This technique shows the possibility of approximation or designing performance space automation in different stages.