[Paper Review] Achievable Rates of FDD Massive MIMO Systems with Spatial Channel Correlation

Zhiyuan Jiang, Andreas F. Molisch+2 more|arXiv (Cornell University)|Jun 29, 2014

Advanced MIMO Systems Optimization25 references4 citations

TL;DR

This paper proposes eigenspace-based training sequences and feedback codebooks for FDD massive MIMO systems with spatially correlated channels, leveraging channel correlation to reduce CSIT acquisition overhead. It achieves a significant rate gain over i.i.d. FDD systems and narrows the performance gap with TDD massive MIMO, especially under strong correlation and moderate antenna counts, outperforming JSDM in high-coherence and high-angular-spread scenarios.

ABSTRACT

It is well known that the performance of frequency-division-duplex (FDD) massive MIMO systems with i.i.d. channels is disappointing compared with that of time-division-duplex (TDD) systems, due to the prohibitively large overhead for acquiring channel state information at the transmitter (CSIT). In this paper, we investigate the achievable rates of FDD massive MIMO systems with spatially correlated channels, considering the CSIT acquisition dimensionality loss, the imperfection of CSIT and the regularized-zero-forcing linear precoder. The achievable rates are optimized by judiciously designing the downlink channel training sequences and user CSIT feedback codebooks, exploiting the multiuser spatial channel correlation. We compare our achievable rates with TDD massive MIMO systems, i.i.d. FDD systems, and the joint spatial division and multiplexing (JSDM) scheme, by deriving the deterministic equivalents of the achievable rates, based on popular channel models. It is shown that, based on the proposed eigenspace channel estimation schemes, the rate-gap between FDD systems and TDD systems is significantly narrowed, even approached under moderate number of base station antennas. Compared to the JSDM scheme, our proposal achieves dimensionality-reduction channel estimation without channel pre-projection, and higher throughput for moderate number of antennas and moderate to large channel coherence time, though at higher computational complexity.

Motivation & Objective

• Address the high pilot overhead in FDD massive MIMO systems due to CSIT acquisition, which severely limits performance compared to TDD systems.
• Exploit spatial channel correlation to reduce the dimensionality of CSIT feedback and training, minimizing resource overhead.
• Design optimal downlink training sequences and user feedback codebooks that leverage multiuser spatial correlation for improved rate performance.
• Demonstrate that FDD massive MIMO can achieve performance close to TDD systems when spatial correlation is properly exploited.
• Provide a performance comparison with JSDM and i.i.d. FDD systems, showing advantages in high-coherence and high-angular-spread environments.

Proposed method

• Proposes eigenspace channel estimation schemes that exploit the spatial correlation structure of the channel, using the dominant eigenvectors of the channel correlation matrix.
• Designs downlink training sequences tailored to the spatial correlation structure to minimize estimation error and reduce training overhead.
• Develops a skewed feedback codebook design that quantizes the effective channel in the eigenspace of the correlation matrix, reducing feedback dimensionality.
• Uses regularized zero-forcing (RZF) precoding to mitigate multiuser interference while accounting for imperfect CSIT.
• Derives deterministic equivalents of achievable rates under spatially correlated fading using random matrix theory, enabling tractable performance analysis.
• Introduces a dimensionality-reduction approach that avoids channel pre-projection, differing from JSDM and enabling direct feedback on the eigenspace.

Experimental results

Research questions

• RQ1 Can spatial channel correlation be leveraged to reduce CSIT feedback and training overhead in FDD massive MIMO systems?
• RQ2 To what extent can FDD massive MIMO achieve rates comparable to TDD massive MIMO when spatial correlation is exploited?
• RQ3 How does the proposed eigenspace-based training and feedback scheme compare with JSDM in terms of throughput and complexity?
• RQ4 What is the impact of channel coherence time and angular spread on the performance of the proposed scheme?
• RQ5 Does exploiting second-order statistics (channel correlation matrices) significantly improve achievable rates in FDD massive MIMO with limited feedback?

Key findings

• The proposed eigenspace-based training and feedback design significantly reduces the rate gap between FDD and TDD massive MIMO systems, especially under strong spatial correlation and moderate numbers of base station antennas.
• When channel correlation is strong and the number of base station antennas is not very large, the achievable sum rate of FDD systems can even exceed that of TDD systems.
• Compared to the JSDM scheme, the proposed method achieves higher spectral efficiency for moderate to large channel coherence times and large angular spreads, though at the cost of higher computational complexity.
• The scheme enables dimensionality reduction in channel estimation without requiring channel pre-projection, offering a simpler and more direct implementation than JSDM.
• Numerical results show substantial rate improvements over i.i.d. FDD systems, validating the effectiveness of exploiting spatial correlation in FDD massive MIMO.
• The performance gain is most pronounced when the channel correlation matrix is well-estimated and the number of users is small relative to the number of base station antennas.