Online secondary path modeling with higher modeling accuracy and faster convergence is desirable in active noise control systems. This paper proposes a method based on variable step size LMS algorithm for modelling filters and Multiple Error Filtered-X LMS (MeFxLMS) algorithm for noise control filter in 1×2×2 multichannel active noise control systems. The proposed idea uses a novel three stage approach for step size selection of modeling filters in order to improve the performance of secondary paths modeling. Consequently, a faster initial convergence and less steady state error are achieved in comparison with existing online secondary path modeling methods for multichannel ANC systems. The number of computations per iteration of the proposed method is comparable to the existing computationally efficient method for multichannel ANC with online secondary path modeling. The performance of the proposed method is verified through simulation results.