Due to limited accuracy of the modern fabrication technologies, cascaded sigma-delta modulators could suffer from finite amplifier gain and capacitor mismatch. Unfortunately, these non-ideal analog components cause inevitable mismatch of the analog circuits and digital noise cancellation filters; consequently it leads to a poor signal-to-noise ratio (SNR). In this paper, the SNR performance enhancement problem is formulated as a robust H-infinity filtering problem in finite frequency domain. Specifically, we propose a heuristic finite impulse response (FIR) filter design that deals with the problem by treating the non-ideality as parametric uncertainty in the integrators' pole and zero of the modulators. In contrast with the existing approaches, the proposed design is a direct approach which directly adjusts the gain responses of the noise transfer function to yield high-order noise shaping by searching the coefficients of FIR filters via solving linear matrix inequality (LMI) feasibility problems. We have confirmed the validity of the proposed method via simulations of a cascaded 2-1 modulator with 2% deviation in the analog integrators. Simulation results show that the modulator matched by the proposed filter achieves better SNR performance than that matched by the conventional one.
2010 International Conference on System Science and Engineering (ICSSE),p.p401 - 406