For DOA estimation of coherent signals in switch antenna array (SAA), a new fast algorithm is proposed. Instead of conventional sub-space algorithm’s covariance matrix, a Toeplitz matrix is constructed with a single cycle of sampled data. It is proved theoretically that the ranks of the Toeplitz matrix is equal to the number of signal sources and has no relations with the coherency of the signal source. Through eigenvalue decomposition, signal and noise subspace are obtained respectively, then DOA estimation can be done by one-dimensional spectral peak searching according to the MUSIC algorithm. The theoretical analysis and simulation results demonstrate validity and superiority of the novel algorithm.