Multi-rate polyphase DSP and LMS calibration schemes for oversampled data conversion systems

Abstract

Architectural schemes for low-power calibration of oversampled analog-to-digital (A/D) systems are presented. Conventional full-rate least-mean squares (LMS) calibration has two well-known limitations: slow convergence and increased computational complexity/power dissipation for higher adaptive filter orders and sampling frequencies. Half (f <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">s</sub> /2) and quarter-rate (f <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">s</sub> /4) LMS calibration for oversampled A/D decimators are used to reduce the computational complexity. Noble identities and polyphase decimation are used to implement these schemes to match digital noise-cancellation filters (NCF) to the corresponding transfer functions of an analog fourth-order cascade sigma-delta (ΣΔ) ADC. Energy savings up to 30% compared to conventional full-rate (f <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">s</sub> ) schemes are confirmed using an Altera Stratix II field programmable gate array (FPGA). The analog front-end comprises a switched-capacitor 2-2 cascade ΣΔ ADC implemented in 0.13 μm CMOS. Using differential-pair opamps with gains of only 22 db and an oversampling ratio OSR = 8, the ΣΔ ADC system achieves 11-bit accuracy over a 9.4 MHz bandwidth with SNR = 67 dB and SFDR = 75 dB.