Eliminating Phase Misalignments in Cell-Free Massive MIMO via Differential Transmission

TL;DR

This work tackles phase misalignment in the downlink of cell-free massive MIMO by proposing two differential strategies that avoid reliance on receiver CSI: differential space-time block coding (DSTBC) and differential PSK (DPSK). The authors adapt these differential schemes to a CF-mMIMO setting, detailing how to distribute DSTBC code matrices across APs and how differential encoding enables non-coherent detection at the UE, supported by a mathematical outline and ML-based decoding. Through simulations under realistic CF-mMIMO settings, the methods are shown to restore BER and SE that are degraded by AP phase misalignments, with DSTBC offering improved BER due to diversity at the cost of pre-log factors, and DPSK providing a robust alternative with PSK limitations. The results demonstrate practical robustness against oscillator-induced phase errors and reduced CSI requirements at the UE, highlighting potential benefits for dense AP deployments and challenging fronthaul/channel conditions.

Abstract

This paper proposes two approaches for overcoming access points' phase misalignment effects in the downlink of cell-free massive MIMO (CF-mMIMO) systems. The first approach is based on the differential space-time block coding technique, while the second one is based on the use of differential modulation schemes. Both approaches are shown to perform exceptionally well and to restore system performance in CF-mMIMO systems where phase alignment at the access points for downlink joint coherent transmission cannot be achieved.

Figures (3)

• Figure 1: An illustration of how the information signal $\mathbf{C}^{t}_{k}$, designed at the CPU, is row-wise split among the AP serving UE $k$. Each row of matrix $\mathbf{C}^{t}_{k}$ is assigned to the corresponding AP according to the mapping function $m(\cdot,k)$. In this example, the AP transmit the elements of the assigned row over $L_k\!=\!4$ consecutive symbol intervals. As an example of the meaning of the mapping $m(l,k)$, note that, since the fourth row of the matrix $\mathbf{C}^{t}_{k}$ is sent to AP$_7$, we have that $m(7,k)=4$.
• Figure 2: CDF of the average BER and SE for each setup. Parameters setting: $L = 40$, $K = 20$, $N = 4$, and $L_k = 4$.
• Figure 3: CDF of the SE for the proposed approaches and synchronized CF-mMIMO system. Here, $L \!=\! 40$, $K \!=\! 20$, $N \!=\! 10$, and $L_k \!=\! 2$.