Adaptive Semiblind Calibration of Bandwidth Mismatch for Two-Channel Time-Interleaved ADCs

Bandwidth mismatch between sample-and-hold (S/H) circuits in a time-interleaved analog-to-digital data converter (ADC) causes undesirable distortions in the output spectrum. To reduce these undesired spectral components, methods are needed to estimate and correct the mismatch. In this paper, we introduce a hybrid filter-bank model of a two-channel time-interleaved ADC. The model allows the development of digital domain estimation and correction techniques. The estimation method is a semiblind estimation technique where a test tone of small amplitude is injected just below the half sampling frequency and is used to estimate the bandwidth mismatch without affecting the normal ADC operation. The estimation procedure relies on an adaptive filtering structure with three fixed relatively short finite-impulse response (FIR) filters and a single adjustable tap representing the estimated relative bandwidth mismatch of the two S/Hs. The correction method removes the first-order effects of S/H mismatches. It only requires a first-order fixed IIR filter and two fixed FIR filters. Simulations demonstrate the effectiveness of the proposed estimation and correction techniques.