Noise and sensitivity analysis of periodically switched linear circuits in frequency domain

Abstract

This paper presents new theories and efficient computational methods for noise and sensitivity analysis of multiphase periodically switched linear (PSL) circuits in frequency domain. Tellegen's theorem for PSL circuits in the phasor domain, frequency reversal theorem, and transfer function theorem, are introduced. The adjoint network of PSL circuits is developed using a frequency-domain approach. An adjoint network-based noise analysis algorithm for PSL circuits is proposed. It is shown that the computational overhead associated with multiple noise sources is eliminated by using the transfer function theorem. It is also shown that the excessive cost of computation due to aliasing effects is significantly reduced when the frequency reversal theorem is employed. In sensitivity analysis, the incremental form of Tellegen's theorem for PSL circuits in the phasor domain is introduced and frequency-domain sensitivity of PSL circuits Is obtained. It is shown that frequency-domain sensitivity of PSL circuits is a series summation of the network variables. Both the baseband and sideband frequency components of the network variables contribute to baseband sensitivity. The method yields sensitivities of one output with respect to all circuit elements in one frequency analysis. Sensitivity networks of PSL circuits are introduced. It is demonstrated that both the adjoint and sensitivity network approaches give the same sensitivity. Numerical results computed using the proposed methods are compared with measurement data and those from other CAD tools.