A fast Fourier transform based method to estimate frequency response mismatches in time interleaved systems.

Abstract

Time interleaving (TI) technique is widely used in acquisition systems to improve the sampling rate. However, as a parallel sampling structure, it inevitably brings in channel mismatches, such as the offset mismatch, gain mismatch, and time mismatch. Moreover, the gain mismatch and time mismatch are frequency-dependent, which means that the gain mismatch and time mismatch will be different when the signal frequency changes. This is because the gain and time mismatches are the special circumstances of frequency response mismatches. In this paper, a frequency response mismatch estimation method in TI acquisition systems is proposed and analyzed. First, the two kinds of mismatch models are compared to find the relationship between frequency response mismatches and the gain and time mismatches. Then, a mismatch estimation method that can estimate mismatches when the analog bandwidth exceeds the sampling rate of a single channel in a time interleaving analog to digital converter is proposed. Sinusoids with different frequencies are utilized to convert the question of estimating frequency response mismatches to acquiring the mismatch coefficients at a series of frequency points. Furthermore, a semi-automatic estimation technique is proposed to reduce the operation time, and a 10 GSPS eight-channel TI system is introduced. Simulations show the accuracy and effectiveness of the proposed method. At last, the frequency response mismatches are calibrated by using the estimated parameters. The calibration improves the performance of an eight-channel TI system from the original spurious free dynamic range of 27.6639 to 56.7029 dB and signal-to-noise ratio of 25.5468 to 32.7667 dB. The calibration result demonstrates the usefulness of the proposed method.