Frequency Response Mismatch Calibration in Generalized Time-Interleaved Systems

• Published in: IEEE transactions on instrumentation and measurement 2022, Vol.71, p.1-17
• Main Authors: Pan, Zhixiang, Ye, Peng, Yang, Kuojun, Huang, Wuhuang, Zhao, Yu
• Format: Article
• Language: English
• Subjects: Algorithms; Calibration; Circuit design; Data acquisition; Dynamic range; Equivalence; Filtering theory; Frequency measurement; Frequency response; frequency response mismatch (FRM); Frequency-domain analysis; mismatch calibration; Performance degradation; periodic time-varying filter; Polynomials; System performance; time-interleaved (TI)
• ISSN: 0018-9456; 1557-9662
• DOI: 10.1109/TIM.2022.3170981

The presence of frequency response mismatches (FRMs) in time-interleaved (TI) systems can severely degrade system performance. In addition, it is difficult to represent FRMs in generalized TI systems using low-order polynomials or physical circuits, which complicates the design of calibration methods. In this study, first, we propose a digital time-interleaved (DTI) structure to establish a relationship between the TI system and a periodic time-varying filter. The DTI structure uses less resources in time-varying filter implementation than the method that utilizing filters and multipliers. We also develop a cascaded calibration model and its equivalent noncascaded form based on the DTI structure. The equivalent noncascaded form uses only one time-varying filter rather than cascades multiple time-varying filters, which is a significant advantage because it can improve the calibration performance without increasing hardware resources during implementation. Furthermore, a filtering process was included in the frequency domain, rather than in the time domain, to cope with the challenges of designing an arbitrary frequency response filter. Simulations based on a four-channel generalized frequency response TI system showed that the proposed method achieved more than 45-dB spur-free dynamic range (SFDR) improvements if measurements of frequency responses are unbiased. Also, more than 60-dB SFDR can be achieved even if a ±5% estimation bias in the measurement of frequency responses. Errors between calibration frames were also minimized by introducing an overlapping frame technique, which ensured that the algorithm could calibrate large data quantities with high precision. Calibration was also implemented in a real four-channel 10 gigabit sampling per second (GSPS) system. Results indicated that the spurious-free dynamic range increased from 47.27 to 62.21 dB at a 5-GHz bandwidth after calibration.