Influences of oversampling and analog imperfections on Hybrid Filter Bank A/D converters

Abstract : Hybrid Filter Banks (HFB) are considered as good candidate for implementing wide-band, high-frequency A/D parallel converters. The performance of HFB (original and twostage) structures is analyzed for different values of oversampling ratios leading to the choice of the optimal value. The electronic elements of the analysis filter bank of HFB structure are considered with real constraints (realization errors and drifts). Real analysis filters of HFB are considered with the analog elements for which the real values are randomly Gaussian-distributed. Sensitivity to the deviations from real values is simulated and the performance of two HFB (original and two-stage) structures are compared. The possible methods for decreasing the sensitivity to realization errors such as Total Least Squares (TLS) optimization are discussed. Original structure is found to be a better candidate rather than the two-stage one in practical cases. However, it is shown that a blind estimation method would be necessary to compensate the realization errors for practically implementing the HFB-based A/D converters.