Abstract: The performance of CAB blind adaptive beamforming was analyzed under cyclic frequency error. The analysis shows cycle leakage effect of sinc window due to finite data samples. An improved algorithm was presented, called forgetting factor CAB(F CAB), in which cyclic frequency need not be estimated. The analysis and simulation results show that adopting forgetting factor in estimation cyclic correlation matrix would largely depress the sensitivity of CAB to cyclic frequency error and the algorithm has excellent performance under little cyclic frequency error.