Multi-user massive multiple-input-multiple-output (massive MIMO) will play a key role in future wireless communication networks. Since spatial channel diversity is the fundamental merit of this technique, high channel correlation may significantly restrict its abilities. This study investigates the impact of channel correlation on a prototyped massive MIMO network with the objective to identify an antenna array geometry which has reduced mutual coupling and channel correlation. To this end, a highly-efficient directional wideband single antenna element was designed for the antenna arrays and the user equipments (UEs). The designed array geometry is tested in an experimental indoor wideband massive MIMO setup. Important system parameters, such as channel correlation, power delay profile, and average received power from the UEs, are studied by analyzing the measured channel data. Furthermore, system-level simulations and network capacity calculations are performed based on the measured channel data to evaluate the performance of the prototyped antenna arrays. A regular array was also fabricated and used for benchmarking comparison. Moreover, a power control algorithm is introduced for the uplink, which was shown to improve the network capacity by up to 3 dB. The results demonstrate that the introduced antenna array outperforms the uniform antenna array in terms of mutual coupling and channel capacity. Massive MIMO, antenna array, mutual coupling, channel correlation, power control algorithm * This work was presented in part at 2019 IEEE International Symposium on Antennas and Propagation and USNC-URSI Radio Science Meeting [1].† This study has been accepted to be published at IEEE Transactions on Antennas and Propagation, this is a preprint and is not the final version of the study which will be published by the journal.