Article ID: 14.20170851
Bandpass sampling (BPS) is a widely used technique where one samples a bandpass-filtered signal at a sampling rate below its Nyquist rate (twice the upper cutoff frequency). For a BPS time-interleaved ADC (TIADC), its dynamic range is subject to the frequency-dependent mismatches generally introduced by the differences of process, supply voltage and temperature between constituent ADCs. This paper proposes a calibration method for frequency-dependent mismatches in BPS TIADCs. The frequency-dependent mismatches is analyzed in both time and frequency domain with the aid of Taylor's series and binary Hadamard transform (BHT). The calibration is composed of coefficients estimation, equalization and mismatches compensation. The coefficients are estimated by solving a linear equations in the foreground based on least square (LS) algorithm. Equalization is to cancel out the linear distortion ahead of compensation to provide a better approximation of the input signal. The principle of mismatches compensation is to reconstruct the mismatches-induced spurious components and subtract them from the original TIADC's output. Simulation results are provided to show the improvement of both effective resolution and dynamic range, and investigate the proper length of training signal, the proper order of Taylor approximation, differentiator and Hilbert filter.