In this article, stochastic channel models for massive multiple‐input multiple‐output (M‐MIMO) and extreme large MIMO (XL‐MIMO) system applications are described, evaluated, and systematically compared. This work aims to cover new aspects of M‐MIMO stochastic channel models in a comprehensive and systematic way. For that, we compare different models, presenting graphically and intuitively the behavior of each model. Each M‐MIMO channel model emulates the environment using different methodologies and properties. Using metrics such as capacity, SINR, singular values decomposition, and condition number, one can understand the influence of each characteristic on the modeling and how it differentiates from other models. Moreover, in new XL‐MIMO scenarios, where the near‐field and visibility region effects arise, our finding demonstrate that for the two assumed schemes of clusters distribution, the clusters location influences the performance of the conjugate beamforming and zero‐forcing precoding due to the correlation effect, which have been analyzed from the geometric M‐MIMO channel models.

In extra-large scale massive MIMO (XL-MIMO) systems, due to the large antenna array physical dimensions, the received signal at the base-station (BS) results in spatial non-stationarities and visibility regions (VRs), limited regions of the array which are visible to the users equipments (UEs). In crowded XL-MIMO scenarios, the reuse of pilot sequences is a path to provide massive connectivity, principally when the radio resources are scarce. However, such reuse increases the intra-cell interference, since UEs sharing the same pilot sequences may have overlapped VRs. Thus, despite the minimum mean-squared error (MMSE) combiner having a high computational complexity, its capability in mitigating the pilot contamination, and even in eliminating such effect in asymptotic antenna regime, justify its use to provide massive connectivity with high spectral efficiency (SE). In this work, we evaluate the performance of the MMSE combiner in XL-MIMO systems in terms of the SE, energy efficiency (EE), resource efficiency (RE), and capability of eliminating the intra-cell pilot contamination. The MMSE channel estimator is deployed to estimate the channel vectors. Indeed, the performance of the zero-forcing combiner and the randomized Kaczmarz regularized zero-forcing combiner is evaluated and compared with the MMSE scheme. For modeling the XL-MIMO propagation channel, we use the double-scattering channel model, which considers clusters of scatterers on both the BS-side and the UE-side to model the spatial channel correlation. In parallel, we develop the analysis considering the single-scattering channel model, a useful simplification of the double-scattering channel model.The numerical results reveal that, at the cost of a high computational complexity, the MMSE combiner achieves remarkable SE and EE results. However, such superior performance is limited by the number of antennas, being recommended for cases where the size of the VR (number of active antennas) is reduced.