Abstract: 3D MIMO (Three Dimension Multiple Input Multiple Output) is one of the key technologies in next communication system (5G). In order to better assess the performance of 3D MIMO communication system, it is necessary to build an accurate 3D channel model. Among many of the existing 3D MIMO models, 3D Kronecker model due to its simplicity and low computation, has been widely used at present. However, 3D Kronecker model, which requires the separation of spatial transmit and receive correlation, does not consistent with the real characteristics of wireless channel. In order to circumvent this restriction, a 2D MIMO channel model, which is introduced elevation angle, is extended to 3D on the basis of WINNER II. And the correlation matrices of two link ends respectively are derived in this paper. A 3D Weichselberger model is constructed through the eigenvalues decomposition of correlation matrices at two link ends respectively. This paper focuses on the analysis of the spatial coupling matrices of six typical scenarios. And the formulas are derived about the six scenarios to analyze the characteristics of 3D MIMO channel. Furthermore, the upper bound of mutual information for 3D Weichselberger model is derived. Simulation results show that, with respect to 3D Kronecker model, 3D Weichselberger is more close to actual 3D MIMO channel characteristics. The reason is that the spatial coupling matrix of the Weichselberger 3D model, which has a certain correlation between the transmitter and the receiver, can truly reflect the real channel characteristic.