Chapter Summary

Chapter Summary

Key Points

• 1.

  The cell-edge problem is the fundamental performance bottleneck of cellular networks: users at cell boundaries experience weak desired signal and strong inter-cell interference simultaneously, resulting in 5th-percentile rates that are 10--50 times lower than the median.

• 2.

  Coordinated Multipoint (CoMP) Joint Transmission converts interferers into helpers, providing a coherent combining gain of $|\mathcal{B}_k|^2$ from a cluster of $|\mathcal{B}_k|$ cooperating BSs. However, CoMP merely moves the boundary from cell edges to cluster edges and requires high-capacity, low-latency backhaul that is rarely available in practice.

• 3.

  Cell-free massive MIMO distributes $L$ simple access points across the coverage area, with all APs jointly serving all $K$ users. There are no cells, no boundaries, and no cell-edge users. The system operates in TDD mode with local MMSE channel estimation at each AP.

• 4.

  Under conjugate beamforming with imperfect CSI, the achievable downlink SINR for user $k$ is $\text{SINR}_k^{\text{cf}} = (\sum_l \sqrt{\eta_{lk}} \gamma_{lk})^2 / (\text{interference} + \text{noise})$, where the coherent sum in the numerator captures the macro-diversity gain from distributed APs.

• 5.

  Channel hardening in cell-free systems arises from averaging over many independent fading channels from different APs. Unlike co-located arrays, this also averages out shadowing, providing an arguably stronger form of hardening.

• 6.

  Max-min fair power control, solvable via bisection over SOCPs, ensures that the weakest user in the network receives the highest possible rate. Under this criterion, cell-free massive MIMO provides 5--10 times higher 95%-likely per-user throughput than small cells with the same total antenna count.

• 7.

  The practical challenges include fronthaul capacity (Level 1 processing requires $O(L \cdot K)$ scalars per coherence interval), AP synchronization, and power supply. These are addressed in subsequent chapters on user-centric design (Chapter 12) and fronthaul-aware processing (Chapter 14).