Temporal Oversampling in One-Bit Quantized Massive MIMO Systems With Correlated Noise

One-bit analog-to-digital converters (ADCs) reduce cost and power consumption in massive MIMO systems but introduce signal distortion that complicates data detection. Temporal oversampling at the receiver can mitigate performance loss from coarse quantization. In this paper, we analyze the performance of temporally oversampled massive MIMO systems with one-bit quantization in the presence of correlated noise, where the noise correlation arises from both filtered thermal noise and inter-symbol interference (ISI). Specifically, we propose a linear symbol-wise MMSE receiver and reveal how noise correlation affects its resulting MSE. Two approximations of the MSE are developed to offer valuable insights. The first approximation suggests that the effects of different noise correlation pairs on the MSE can be treated as independent. The second approximation demonstrates that the MSE variations induced by noise correlation are jointly determined by the pulse-shaping and receive filters. For a system employing a raised cosine pulse, noise correlation degrades the receiver performance. Moreover, the trade-off between noise correlation and the oversampling ratio is analyzed. The case of imperfect channel state information (CSI) is also considered, where the analysis developed for data detection is extended to a linear MMSE channel estimator. Finally, numerical results show that for a system employing a raised cosine pulse, an oversampling ratio of 2 is beneficial, offering practical guidelines for system design.