Application of neural networks to efficient design of wireless and RF circuits and systems

Complexity of RF designs in a wireless system continues to increase significantly, to support multiple standards, multiple frequency bands, need for higher bandwidth and stringent adjacent channel specifications. Using existing design tools, time required to carry out a virtual prototyping of such complex circuits and their trade- off analysis with the baseband circuitry, is unacceptably long, because both the circuit simulation and optimization procedures can be very time-consuming. In this paper we show how Xpedion Design Systems, Inc. can considerably speed up, the modeling of RF circuits by using the neural network approach. Also, these neural net based C models are in industry standard format, and hence compatible with circuit level and system level simulators for accurate trade-off analysis of a wireless system, early in the design process