Background Calibration for Mixed Mismatches in TIADC Using Taylor-Volterra-Series-Based Model

The linear and nonlinear mismatches in time-interleaved analog-to-digital converter (TIADC) both severely degrade the system’s performance. Calibration techniques play a crucial role in restoring TIADC’s optimal performance. However, existing techniques have application limitations in nonlinear modeling and exhibit one-sidedness in error detection, leading to overfitting. This paper proposes a new background calibration for mixed linear and nonlinear errors. A unified model with wide applicability is presented to characterize mixed error generation based on the Taylor and the Volterra series. Then the errors are calibrated by a high-precision algorithm which addresses the issues of high dimensional optimization and overfitting problems simultaneously. The proposed technique is verified with simulations in MATLAB and experiments in a 16-bit 4-channel 260MS/s board-level prototype, improving the SFDR from 35.5 dB to 73.3 dB with a dual-tone input at 67.6 MHz and 119.6 MHz.