Iterative WLS Design of SAW Bandpass Filters

It has been demonstrated by several authors that the well-known weighted least squares (WLS) approximation can be equiripple if a suitable weighting function is applied. In the present paper, the WLS algorithm is generalized to SAW filter synthesis with prescribed magnitude and phase specifications. Several weighting techniques producing quasi-equiripple designs are presented. The frequency sampling technique is used for SAW filter frequency response approximation to reduce the number of optimized variables. The WLS algorithm rapidly converges both for linear and nonlinear phase SAW filters. Typically, no more than 5-10 iterations are required to obtain the WLS solution to accuracy better than 0.5-1 dB in the stopband when compared with the optimum Chebyshev approximation. Moreover, it is shown that the WLS technique can be effectively applied for second-order effect compensation.