Efficient estimation and correction of mismatch errors in time-interleaved ADCs

Abstract

We present a novel method for the estimation and correction of mismatch errors in time-interleaved analog-todigital converters. The estimation of the mismatch errors requires a training signal, but it is efficient, accurate, and effective while keeping low the complexity of the associated algorithm, when compared with other works in the literature. The compensation strategy uses the estimated parameters for offset and gain mismatch error correction, and implements a low-complexity (similar to that of a finite-impulse response filter) Lagrange interpolation filter to correct errors due to timing mismatch, which depends on the dynamics of the input signal. The proposed compensation strategy achieves almost ideal behavior over a wide bandwidth, i.e., only 12% of oversampling is required for an almost complete cancelation of distortion. The method has been tested under severe mismatch conditions (up to 10% timing mismatch and 5% gain and offset mismatch), showing an improvement of over 6 bits in the effective number of bits and 50 dB in the spurious-free dynamic range. In addition, when compared with the available techniques, no limitation on the number of channel converters is introduced since the compensator effectively cancels the distortion even in the presence of large mismatches.