The ability of neural netw orks to predict the electromagnetic field of one dimensional sources as well as of two- and three dimensional sources in a wide dynamic range with small error is an imperative condition to use neural networks in the analysis of complex and fast printed circuit boards. As presented in this paper, Sigma-Pi networks are able to achieve this condition. Because of the high dynamic of the input and the output data within the region of investigation, a very fast back propagation learning algorithm is used to train the neural network and also presented in this paper. Finally, a short outlook is given to the use of neural networks in analysis of printed circuit boards.