Multicell Downlink Capacity with Coordinated Processing

Sheng Jing, David N. C. Tse, Joseph B. Soriaga, Jilei Hou, John E. Smee, Roberto PadovaniEURASIP J. Wirel. Commun. Netw 2008 · DOI

Abstract: This research investigates the impact of base station (BS) cooperation on downlink capacity in multicell networks, with a focus on mitigating inter-cell interference for cell-edge users. By analyzing linear precoding techniques within a modified Wyner-type model, the authors identify a "singularity problem" in symmetric networks that can severely limit performance. The study demonstrates that practical approaches, including multiuser scheduling and the exploitation of natural channel fading, can effectively overcome these limitations, allowing simple linear precoders to achieve performance levels approaching the theoretical optimum of Dirty Paper Coding (DPC).

Introduction

In conventional cellular architectures, base stations operate independently, leading to significant inter-cell interference. This problem is most acute for "cell-edge" users situated at the boundaries of multiple cells, where they receive competing signals of comparable strength. While non-linear techniques like Dirty Paper Coding (DPC) offer a theoretical solution for interference cancellation, their implementation complexity is prohibitive. This paper explores whether simpler linear precoders—such as Zero-Forcing (ZF), Minimum Mean Square Error (MMSE), and Cophasing—can provide comparable benefits when base stations coordinate their transmissions.

Network Models and Precoding Strategies

The authors employ a modified Wyner-type model, where base stations are arranged in a ring to facilitate analytical tractability. Two primary network configurations are examined:

  1. Single-class model: Focuses exclusively on cell-edge users who experience interference from adjacent base stations.
  2. Double-class model: Incorporates both cell-interior users (located near the BS) and cell-edge users.
The Single-class model focuses on coordination for users at the cell boundaries.
The Double-class model accounts for both interior and edge users, reflecting more realistic network deployments.

To evaluate these models, the study derives sum-rate expressions for several precoding schemes. DPC serves as the performance upper bound. ZF aims to nullify all inter-user interference, while MMSE seeks a balance between interference cancellation and noise suppression. Cophasing is used as a baseline for coherent beamforming to a single user. The analytical findings are further validated through Monte-Carlo simulations under Rayleigh fading conditions.

Overcoming the Singularity Problem

A major challenge identified in large-scale coordinated networks is the "singularity problem." In a perfectly symmetric, non-fading network where a user receives equal signal strength from two base stations, the channel matrix for ZF precoding becomes nearly singular. As the network size grows, the performance of ZF collapses toward zero.

However, the research highlights two key factors that mitigate this issue:

  • Natural Fading: Random channel fading introduces inherent asymmetry in path gains. This prevents the channel matrix from becoming singular, significantly improving the performance of linear precoders.
  • Multiuser Scheduling: By implementing scheduling algorithms that select users based on "path asymmetry"—prioritizing those with naturally stronger connections to one specific BS—the network can break the problematic symmetry. This allows simple ZF precoders to achieve rates nearly identical to DPC.

Key Results and Tradeoffs

The analysis reveals several critical insights into multicell coordination:

  • Performance Gaps: In networks with asymmetric paths ($\alpha < 1$), the gap between DPC and ZF is a constant $\log_2(1+\alpha^2)$, regardless of the number of cells.
  • MMSE Versatility: The MMSE precoder demonstrates dual behavior, performing like Cophasing at low Signal-to-Noise Ratios (SNR) and converging to ZF performance at high SNR.
  • Efficiency of TDM: The study shows that Time Division Multiplexing (TDM) within a cell—serving one user at a time—incurs minimal capacity loss compared to simultaneous multi-user transmission, offering a path toward simplified hardware.
  • Double-Class Optimization: For networks with mixed user types, "Cell-breathing with Superposition Coding" was found to recover approximately 50% of the capacity gap between non-cooperative baselines and the theoretical limit.

Conclusion and Future Directions

Coordinated base station processing offers a powerful means to enhance multicell downlink capacity. While the "singularity problem" poses a theoretical threat to linear precoding in symmetric networks, practical scheduling and the presence of natural fading make these schemes highly effective. Future work should extend these findings from the 1D Wyner model to more complex 2D hexagonal layouts and investigate the impact of imperfect Channel State Information (CSIT) and network asynchrony on coordination gains.

Summary updated

BibTeX
@article{jing2008multicell,
  author = {Sheng Jing and David N. C. Tse and Joseph B. Soriaga and Jilei Hou and John E. Smee and Roberto Padovani},
  title = {{Multicell Downlink Capacity with Coordinated Processing}},
  journal = {EURASIP J. Wirel. Commun. Netw.},
  year = {2008},
  doi = {10.1155/2008/586878},
}
Figure