Blind Timing Skew Estimation Using Source Spectrum Sparsity in Time-Interleaved ADCs

Timing skews in time-interleaved analog-to-digital converter (ADC) (TIADC) greatly degrade the spurious-free dynamic range of the TIADC, which can be improved effectively by digital compensation algorithms. Literature review shows that many nonblind compensation algorithms required the accurate timing skew parameters measured before the compensation process. In this paper, the blind timing skew estimation (TSE) for the TIADC system is investigated. With the evaluation of the working rationale of the TIADC system and the undersampling nature in each sub-ADC, the relation between the output digital spectra of the sub-ADCs and the source analog spectrum is formulated. A novel subband partition approach has been proposed, and the corresponding expression of the output digital subband spectra of the sub-ADCs in terms of the output digital subband spectra of the TIADC is obtained. With the assumption of the source spectra sparsity, the nonoverlapping frequency points can be determined from the output digital subband spectra of the sub-ADCs. Making use of the determined nonoverlapping frequency points, the phase information of the output digital subband spectra of the sub-ADCs can be employed to provide a good metric to estimate the timing skews, and a closed-form expression of the TSE has been derived. As the result, the source spectrum sparsity-based blind TSE (SS-BLTSE) algorithm has been developed and evaluated. Computer simulation results show that the SS-BLTSE algorithm provides good TSE performance for arbitrarily narrow or wideband source signals. Moreover, it has many merits over the existing blind TSE algorithms, such as there is no prior requirement of the source signal except that it has certain spectrum sparsity, no limit to the number of the sub-ADCs, and the frequency of the input signal can be approaching the Nyquist frequency.