Paired-Interval Mapping Calibration for Spectral Leakage in Non-Uniform DBI-ADCs

In digital bandwidth-interleaved analog-to-digital converters, analog bandpass filter non-idealities leak out-of-band signals into the processing chain during sub-band decomposition. Frequency conversion shifts these leaked signals non-trivially, generating conversion-parameter-dependent aliasing errors at the output. Consequently, conventional calibration methods like synthesis filter banks, designed for architectures such as hybrid filter bank analog-to-digital converters without frequency conversion, suffer performance degradation. This paper reveals the mapping relationship between the aliasing errors and output signals, based on which an output-guided error correction method is proposed. By acquiring error mapping information, this method reconstructs and eliminates the error spectrum from the contaminated output signals. The proposed method effectively addresses aliasing errors in uniform scenarios and is naturally applicable to arbitrary non-uniform channel partitioning. Experimental results demonstrate the efficacy of the proposed method, with the SNDR within the operating band improved from 36.58 dB to 64.20 dB.