System-level optimization of baseband filters for communication applications

Abstract

In this paper, a design approach for the high-level synthesis of programmable continuous-time baseband filters able to achieve optimum trade-off among dynamic range, distortion behavior, mismatch tolerance and power/area consumptions is presented. The proposed approach relies on building programming circuit elements as arrays of switchable unit cells and defines the synthesis as a constrained optimization problem with both continuous and discrete variables, this last representing the number of enabled cells of the arrays at each configuration. The cost function under optimization is, then, defined as a weighted combination of performance indices which are estimated from macromodels of the circuit elements. The methodology has been implemented in MATLAB TM and C++, and covers all the classical approximation techniques for filters, most common circuit topologies (namely, ladder simulation and cascaded biquad realizations) and both transconductance-C (G m -C) and active-RC implementation approaches. The proposed synthesis strategy is illustrated with a programmable equal-ripple ladder G m -C filter for a multi-band power-line communication modem.

A 0.18μm CMOS Low-Noise Elliptic Low-Pass Continuous-Time Filter

“…The active structure of the filter must be scaled in order to achieve the smallest possible area and power consumptions for a given set of specifications [6]. The optimization algorithm used in the design of the proposed filter has been described in [7] and takes into account design aspects, such as limited quality factor of integrators, internal peak amplitudes, noise (they are evaluated by simple matrix methods), linearity (using Volterra series), as well as area and power estimations. The optimizer uses simulated annealing techniques and explores the design space of all those filter configurations de- †.…”