FDD Massive MIMO Based on Efficient Downlink Channel Reconstruction

Yu Han, Qi Liu, Chao-Kai Wen, Shi Jin, Kai‐Kit Wong

Introduction

Massive multiple-input multiple-output (MIMO) systems are a cornerstone of modern wireless communication, significantly boosting spectral efficiency by serving multiple users simultaneously. However, the performance of these systems hinges on the Base Station (BS) having accurate Downlink Channel State Information (CSI). In Frequency-Division Duplex (FDD) systems, where uplink and downlink operate on different frequencies, obtaining this CSI is a major challenge. Traditional methods require training and feedback overhead that scales linearly with the number of antennas, which becomes prohibitive as antenna arrays grow larger.

System model showing multipath propagation.

The Proposed Framework

To overcome the CSI bottleneck, this research proposes a full transceiver framework that exploits the spatial reciprocity of the wireless channel. While the complex gains of signal paths vary with frequency, the physical geometry—specifically the angles of arrival/departure and time delays—remains consistent across the uplink and downlink frequencies.

The framework operates in four distinct phases:

  1. Uplink Parameter Extraction: The BS uses an enhanced Newtonized Orthogonal Matching Pursuit (eNOMP) algorithm to analyze uplink sounding reference signals (RSs). It extracts frequency-independent parameters: downtilts, azimuths, and delays for each propagation path.
  2. Downlink Training: Leveraging the extracted geometry, the BS designs a greedy pilot scheduling scheme. Instead of exhaustive training, it identifies a minimal set of spatial beams to estimate the complex path gains for multiple users.
  3. CSI Feedback and Reconstruction: Users estimate their respective downlink gains and feed them back to the BS. The BS then reconstructs the full high-dimensional downlink channel matrix by combining these gains with the previously identified 'skeleton' of physical paths.
  4. Data Transmission: With the reconstructed CSI, the BS designs precoders to transmit data to multiple users with minimal interference.
The proposed four-phase transceiver framework.

Key Innovations

  • eNOMP Algorithm: Standard sparse recovery algorithms often suffer from 'grid mismatch' when signal paths do not align perfectly with discrete codebook points. The eNOMP algorithm incorporates a Newton-step refinement to precisely locate 'off-grid' parameters, significantly improving reconstruction accuracy.
  • Greedy Pilot Scheduling: By recognizing that users in a cell often share similar spatial paths, the BS can prune redundant training beams. This multi-user synergy drastically reduces the number of required downlink pilot symbols.
  • Exploiting the Channel 'Skeleton': The core insight is that the massive channel matrix is governed by a few physical paths. Shifting the estimation burden from the full matrix to these few path gains slashes both training and feedback overhead.

Performance and Results

Numerical simulations of a system with 128 antennas and 10 users demonstrate the framework's efficiency. The eNOMP algorithm achieves a much lower Normalized Mean Square Error (NMSE) compared to traditional linear estimation methods. Most importantly, the system's total data throughput (sum-rate) using the reconstructed CSI nearly matches the performance of a system with perfect, instantaneous CSI.

The overhead reduction is substantial: feedback was reduced from thousands of complex numbers to just 60, and training symbols were cut from 128 to as few as 12 in some scenarios.

Conclusion

This framework provides a practical path for FDD systems to realize the full potential of Massive MIMO. While it assumes a degree of channel sparsity, its ability to accurately reconstruct channels with minimal overhead addresses one of the most significant hurdles in modern wireless network design. Future work may focus on optimizing the computational complexity of the eNOMP algorithm for real-time hardware and adapting the framework for high-mobility environments.

Summary updated

Abstract

Massive multiple-input multiple-output (MIMO) systems deploying a large number of antennas at the base station considerably increase the spectrum efficiency by serving multiple users simultaneously without causing severe interference. However, the advantage relies on the availability of the downlink channel state information (CSI) of multiple users, which is still a challenge in frequency-division-duplex transmission systems. This paper aims to solve this problem by developing a full transceiver framework that includes downlink channel training (or estimation), CSI feedback, and channel reconstruction schemes. Our framework provides accurate reconstruction results for multiple users with small amounts of training and feedback overhead. Specifically, we first develop an enhanced Newtonized orthogonal matching pursuit (eNOMP) algorithm to extract the frequency-independent parameters (i.e., downtilts, azimuths, and delays) from the uplink. Then, by leveraging the information from these frequency-independent parameters, we develop an efficient downlink training scheme to estimate the downlink channel gains for multiple users. This training scheme offers an acceptable estimation error rate of the gains with a limited pilot amount. Numerical results verify the precision of the eNOMP algorithm and demonstrate that the sum-rate performance of the system using the reconstructed downlink channel can approach that of the system using perfect CSI.
BibTeX
@misc{han2019fdd,
  author = {Yu Han and Qi Liu and Chao-Kai Wen and Shi Jin and Kai‐Kit Wong},
  title = {{FDD Massive MIMO Based on Efficient Downlink Channel Reconstruction}},
  year = {2019},
  eprint = {1902.06174},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1902.06174},
}
Figure