ELECTRONICALLY TUNABLE FILTERING NETWORKS - THEORY, ANALYSIS AND SIMULATION

A parallel four-port tunable conductance-susceptance network model is proposed for use in tunable filtering networks, with emphasis on its m odelling, sensitivity minimising and computer simulation. In contrast with the design of tunable transfer functions in the complex domain, w e synthesise a four-port linear and controllable conductance network i n the real domain, which is easier to implement with CMOS transconduct ance elements. The basic building blocks and general forms of the cond uctance matrix and structure for biquad filtering networks (BFN) are g iven. The redefined port parameters, equivalent port short-circuit res istance, equivalent port open-circuit resistance and equivalent transf er coefficients are given and used for the analysis of network sensiti vity. Theoretical analysis and preliminary simulation results show tha t the minimum sensitivity condition can be obtained for BFN when the t wo time constants at two energy-storage ports are equal. It also shows that the omega and Q parameters in biquad filtering networks and the weights in some neural networks are tunable by changing the gate signa ls, and the input impedance of the network is very large as there is n o backward-feed connection to input port. Because the construction is based on CMOS elements, the realisation can be readily integrated. Thi s method can also be used to design a tunable-weight network in some n eural networks, since a transconductance element can be easily used as a weight between any two neurons.