1: \begin{abstract}

2:     This paper proposes a channel estimation method for hybrid wideband \acf{MIMO} systems in high-frequency bands, including \acf{mmWave} and \acf{sub-THz}, in the presence of beam squint effects and array errors arising from hardware impairments and environmental time fluctuations such as thermal effects and dynamic motion of the array.

3:     % 

4:     Although conventional channel estimation methods calibrate array errors through offline operation with large training pilots, the calibration errors remain due to time-varying array errors.

5:     % 

6:     Therefore, the proposed channel estimation method calibrates array errors online with small pilot overhead.

7:     % 

8:     In the proposed method, array response matrices are explicitly decomposed into a small number of physical parameters including path gains, angles and array errors, which are iteratively estimated by alternating optimization based on a \acf{ML} criterion.

9:     % 

10:     To enhance the convergence performance, we introduce a switching mechanism from an on-grid algorithm to an off-gird algorithm depending on the estimation accuracy of the array error during algorithmic iterations.

11:     % 

12:     Furthermore, we introduce an approximate mutual coupling model to reduce the number of parameters.

13:     % 

14:     The reduction of parameters not only lowers computational complexity but also mitigates overfitting to noisy observations.

15:     % 

16:     Numerical simulations demonstrate that the proposed method effectively works online even with small pilot overhead in the presence of array errors.

17: \end{abstract}

18: