Rqolqhduulghqwlilfdwlrqqdqggprgholqjjriiodujhhvfdohhdqdorjlqwhjudwhg Flufxlwulhvviruu'077edvhggdssolfdwlrqv

Starting from the Volterra approach and assuming fading memory, a very efficient method to identify weakly non-linear systems is presented. Making special assumptions on the Volterra kernels, one ends up in the Hammerstein model, which has less complexity than the Volterra model, thus reducing the numerical effort. Using FIR filters to realize the Hammerstein model, no starting values are necessary to solve the linear optimization problem. To deal with the ill-posedness of the inverse problem, special regularization techniques have been used. Applying these methods to a large scale analog integrated circuitry, namely a broadband SLIC (subscriber line interface circuitry) for a splitter-less ADSL-Lite central office application, results are given for the buffer, consisting of a few hundred components in BiCMOS technology.