A new approach to sensitivity computation of microwave circuits

The authors introduce an approach to first and second-order sensitivity computation of microwave circuits, based on the adjoint network concept applied to an all-lumped equivalent circuit. A general sensitivity analysis program has been written for microwave circuits. It computes the absolute and relative first and second-order sensitivity of any selected response (gain, reflection coefficient, VSWR, etc.), with respect to all parameters. The adopted procedure totally avoids matrix inversions or multiplications; it is therefore accurate and fast. It is concluded that the results are useful in the tolerance characterization of microwave circuits or in optimization by efficient gradient methods.<>