A neural network model for propagation delays in systems with high speed VLSI interconnect networks

Models used for simulating propagation delays in networks containing high speed VLSI interconnects for the purpose of iterative design optimization must be accurate and yet computationally simple enough to allow for convergence of the optimization algorithm in a realistic time frame. Existing models capable of accurately simulating the relationship between the signal delays, the terminations, the network topology, and the interconnects with distributed parameters, are highly CPU intensive and are not suitable for computer automated optimization. In our paper we present a propagation delay model employing feedforward neural networks which, being simple and computationally inexpensive, can be used in efficient optimization routines.