Exponential spatial correlation with large-scale fading variations in massive MIMO channel estimation

To provide the vast exploitation of the large number of antennas on massive multiple-input–multiple-output (M-MIMO), it is crucial to know as accurately as possible the channel state information in the base station. This knowledge is canonically acquired through channel estimation procedures conducted after a pilot signaling phase, which adopts the widely accepted time-division duplex scheme. However, the quality of channel estimation is very impacted either by pilot contamination or by spatial correlation of the channels. There are several models that strive to match the spatial correlation in M-MIMO channels, the exponential correlation model being one of these. To observe how the channel estimation and pilot contamination are affected by this correlated fading model, this work proposes to investigate an M-MIMO scenario applying the standard minimum mean square error channel estimation approach over uniform linear arrays and uniform planar arrays (ULAs and UPAs, respectively) of antennas. Moreover, the elements of the array are considered to contribute unequally on the communication, owing to large-scale fading variations over the array. Thus, it was perceived that the spatially correlated channels generated by this combined model offer a reduction of pilot contamination, consequently the estimation quality is improved. The UPA acquired better results regarding pilot contamination since it has been demonstrated that this type of array generates stronger levels of spatial correlation than the ULA. In contrast to the favorable results in channel estimation, the channel hardening effect was impaired by the spatially correlated channels, where the UPA imposes the worst performance of this effect for the discussed model.