Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fronthaul Compression for Uplink Massive MIMO using Matrix Decomposition

Aswathylakshmi P, Radha Krishna Ganti

TL;DR

This work tackles the fronthaul bottleneck in uplink massive MIMO by introducing a blind, iterative matrix-decomposition approach that exploits the convolution structure of OFDM signals. By representing the frequency-domain received data as $\mathbf{Y_f} \approx \hat{\mathbf{X}}\mathbf{F_L}\hat{\mathbf{H}}$ (SU) or its multi-user block form (MU), and solving via alternating minimisation, the method achieves compression ratios far exceeding PCA-based techniques while preserving target SER in link-level simulations. The derived compression formulas $CR_{SU}=\frac{NN_r}{N+LN_r}$ and $CR_{MU}=\frac{NN_r}{N_u(N+LN_r)}$ show that, for large $N$, the proposed method scales with a factor of approximately $L$ relative to PCA, enabling near an order of magnitude higher fronthaul efficiency. Recovery at the BBU uses the reconstructed $\hat{\mathbf{Y_f}}$ with conventional MRC/ZF equalization, validating performance parity with uncompressed systems at practical SNRs. Overall, the approach offers a scalable path to deploying uplink massive MIMO with much reduced fronthaul bandwidth requirements.

Abstract

Massive MIMO opens up attractive possibilities for next generation wireless systems with its large number of antennas offering spatial diversity and multiplexing gain. However, the fronthaul link that connects a massive MIMO Remote Radio Head (RRH) and carries IQ samples to the Baseband Unit (BBU) of the base station can throttle the network capacity/speed if appropriate data compression techniques are not applied. In this paper, we propose an iterative technique for fronthaul load reduction in the uplink for massive MIMO systems that utilizes the convolution structure of the received signals. We use an alternating minimisation algorithm for blind deconvolution of the received data matrix that provides compression ratios of 30-50. In addition, the technique presented here can be used for blind decoding of OFDM signals in massive MIMO systems.

Fronthaul Compression for Uplink Massive MIMO using Matrix Decomposition

TL;DR

This work tackles the fronthaul bottleneck in uplink massive MIMO by introducing a blind, iterative matrix-decomposition approach that exploits the convolution structure of OFDM signals. By representing the frequency-domain received data as (SU) or its multi-user block form (MU), and solving via alternating minimisation, the method achieves compression ratios far exceeding PCA-based techniques while preserving target SER in link-level simulations. The derived compression formulas and show that, for large , the proposed method scales with a factor of approximately relative to PCA, enabling near an order of magnitude higher fronthaul efficiency. Recovery at the BBU uses the reconstructed with conventional MRC/ZF equalization, validating performance parity with uncompressed systems at practical SNRs. Overall, the approach offers a scalable path to deploying uplink massive MIMO with much reduced fronthaul bandwidth requirements.

Abstract

Massive MIMO opens up attractive possibilities for next generation wireless systems with its large number of antennas offering spatial diversity and multiplexing gain. However, the fronthaul link that connects a massive MIMO Remote Radio Head (RRH) and carries IQ samples to the Baseband Unit (BBU) of the base station can throttle the network capacity/speed if appropriate data compression techniques are not applied. In this paper, we propose an iterative technique for fronthaul load reduction in the uplink for massive MIMO systems that utilizes the convolution structure of the received signals. We use an alternating minimisation algorithm for blind deconvolution of the received data matrix that provides compression ratios of 30-50. In addition, the technique presented here can be used for blind decoding of OFDM signals in massive MIMO systems.

Paper Structure

This paper contains 11 sections, 1 theorem, 27 equations, 3 figures, 3 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. System Model
  3. Compression using Matrix Decomposition
  4. SU-MIMO System
  5. MU-MIMO System
  6. Results and Discussion
  7. Comparison of achieved Compression Ratios
  8. Symbol Error Rate Performance
  9. Algorithm Complexity
  10. Conclusion
  11. Acknowledgements

Key Result

Lemma 1

If {$\mathbf{\hat{X}, \hat{H}}$} and {$\mathbf{\tilde{X}, \tilde{H}}$} are two solutions to (4), then they are related to each other by the scalar transform

Figures (3)

  • Figure 1: Proposed Fronthaul Compression Scheme
  • Figure 2: Uncoded SERs of the proposed method (after 10 iterations of Algorithm 1) and PCA compression for single user in the system with (a) all $N_r=64$ antennas used during MRC (b) only $N_{r}/2=32$ antennas used during MRC. SER for the uncompressed system for all cases are also plotted as baseline. Proposed method matches the SER of uncompressed system in both cases while providing a CR above 50, whereas SER for PCA compression degrades in case (b) in spite of offering a CR of only about 5.
  • Figure 3: Uncoded SERs of the proposed method (after 10 iterations of Algorithm 2), PCA compression and no compression cases for 4 users in the system with (a) all $N_r=64$ antennas used during ZF (b) only $LN_{u}=48$ antennas used during ZF. In both cases, the proposed method offers CR of nearly 14 while matching the SER of PCA compression and uncompressed system at low SNRs. The SER of the proposed method degrades at higher SNRs due to loss of diversity.

Theorems & Definitions (2)

  • Lemma 1
  • proof