The angles-of-arrival (AOA) estimation and beam- forming by an array antenna plays important roles in MIMO communication systems. In multipath environment, almost conventional subspace approaches with or without the eigendecomposition adopt the spatial smoothing (SS) using multiple subarrays. This paper proposes a new iterative AOA estimation of partially or fully coherent signals without adopting spatial smoothing using subarrays. The algorithm is derived by eliminating the coherent components of the reconstructed signal covariance matrix. One of the notable advantages of the proposed method is to improve the angle resolution due to larger aperture compared to ordinary SS-based methods using subarrays with small size. Furthermore, the proposed algorithm can directly be applied to non-uniformly spaced linear array and circular array without using the array transformation matrix. The effectiveness is validated in numerical simulation compared to the FBSS-based MUSIC.