Multi-channel in azimuth is a technique to achieve high-resolution as well as wide-swath in Synthetic Aperture Radar (SAR) systems. Channel error is inevitable in multi-channel systems and it induces blurring in subsequent SAR imagery. Existing compensation approaches are sensitive to system parameters as well as the imaging scenes. Uncertainty of the parameters impacts the validation of these algorithms. In this paper, an improved approach is presented to remove the channel error. Based on the error form, this approach models channel error as three parts: the range gain error, the pulse sampling clock error, and the transmission phase error. The range gain error and the pulse sampling clock error are removed alternately. Then the proposed approach uses cost function to estimate the transmission phase error so that it is independent from the imaging scene. Point target simulations are carried out to investigate the performance, and real data comparison experiments are carried out to verify this approach.