Abstract: To solve the problem of slow convergence in the modified constant modulus algorithm (MCMA), a variable step and dual mode blind equalization algorithm is proposed, based on the MCMA algorithm. The proposed algorithm uses a variable step size in it's start-up to speed up its convergence, and switches it self to a decision directed (DD) mode after it has been converged to keep the error in a robust state and stay in a reliable bound. The performances of the proposed algorithm and the original one are compared by Monte-Carlo simulation, in which channel data measured on practice is used. Simulation demonstrates that the proposed algorithm converges faster and gives out a smaller error in its steady state.