Transient analysis of adaptive filters – Part II: The error nonlinearity case