Source-linked AI summary
Low RF-Complexity Technologies to Enable Millimeter-Wave MIMO with Large Antenna Array for 5G Wireless Communications
Xinyu Gao, Linglong Dai, Akbar M. Sayeed
TL;DR
The paper addresses limitations of traditional low-RF-complexity approaches for mmWave MIMO and examines PAHP and LAHP. It reviews their designs and compares their sum-rate, channel-estimation robustness, and power-efficiency characteristics.
Problem
Traditional antenna selection suffers serious performance loss in highly correlated mmWave channels, while analog beamforming supports only single-stream transmission without multiplexing gains.
Method
The paper reviews low-RF-complexity technologies, analyzes PAHP and LAHP, proposes an adaptive LAHP selecting network, and systematically compares their performance.
Results
PAHP achieves higher sum-rate with perfect channel knowledge, whereas LAHP is more robust to channel-estimation errors and has higher power efficiency.
Takeaways & Limitations
PAHP and LAHP provide different trade-offs between hardware cost, power consumption, sum-rate, and robustness for low-RF-complexity mmWave MIMO.
Takeaways & Limitations
Existing low-RF-complexity technologies are mostly designed for narrowband and time-invariant channels, whereas outdoor mmWave channels are more likely broadband and time-varying.
Abstract
from arXiv · showhide
Millimeter-wave (mmWave) MIMO with large antenna array has attracted considerable interests from academic and industry communities, as it can provide larger bandwidth and higher spectrum efficiency. However, with hundreds of antennas, the number of radio frequency (RF) chains required by mmWave MIMO is also huge, leading to unaffordable hardware cost and power consumption in practice. In this paper, we investigate low RF-complexity technologies to solve this bottleneck. We first review the evolution of low RF-complexity technologies from microwave frequencies to mmWave frequencies. Then, we discuss two promising low RF-complexity technologies for mmWave MIMO systems in detail, i.e., phased array based hybrid precoding (PAHP) and lens array based hybrid precoding (LAHP), including their principles, advantages, challenges, and recent results. We compare the performance of these two technologies to draw some insights about how they can be deployed in practice. Finally, we conclude this paper and point out some future research directions in this area.
I. INTRODUCTION
mmWave MIMO combines wide bandwidth and compact large arrays with a practical bottleneck: hundreds of costly, power-hungry RF chains. The paper reviews low-RF-complexity approaches and focuses on PAHP and LAHP.
- mmWave offers nearly 2 GHz bandwidth, compared with 20 MHz in current 4G without carrier aggregation.
- Large mmWave antenna arrays fit into small physical sizes because short wavelengths permit dense antenna packing.
- Each antenna typically requires a dedicated RF chain, making 256-antenna systems costly and power-intensive.RF-chain consumption is cited as 250 mW at mmWave frequencies versus 30 mW at microwave frequencies.
- The paper reviews low-RF-complexity technologies and details phased array based hybrid precoding (PAHP) and lens array based hybrid precoding (LAHP).It also compares their performance for practical deployment.
- Antenna selection and analog beamforming provide foundations for PAHP and LAHP.
A. Antenna selection
Antenna selection chooses the best connected antennas from the array, whereas analog beamforming uses phase shifters to form a single-stream beam. Their benefits depend on channel conditions and transmission demands.
- Antenna selection: Antenna selection uses CSI to choose NRF best antennas from N antennas for data transmission and sum-rate maximization.
- Antenna selection: When NRF exceeds Ns, antenna selection causes negligible performance loss under IID Rayleigh fading channels.
- Analog beamforming: Analog beamforming uses one RF chain and a phase-shifter network to control signal phases for array gain and effective SNR.
- Analog beamforming: Phase-only control forces equal beamforming-vector amplitudes, which incurs performance loss.
- Analog beamforming: Analog beamforming supports only single-stream transmission, excluding multi-stream and multi-user scenarios.
C. Can traditional low RF-complexity technologies be used in mmWave MIMO systems?
Traditional low-RF-complexity methods do not fully match outdoor mmWave MIMO channels: limited scattering creates correlation, while analog beamforming cannot provide multi-stream multiplexing.
- Outdoor mmWave scattering is limited because short wavelengths produce poor diffraction around comparatively large obstacles.Scattering can also incur 5-20 dB attenuation, and 28 GHz outdoor channels average 2.4 paths in one New York study.
- Antenna selection is unsuitable for mmWave MIMO because highly correlated channels cause serious performance loss.
- Analog beamforming cannot fully exploit mmWave MIMO spectrum efficiency because it supports single-stream transmission without multiplexing gains.
III. PHASED ARRAY BASED HYBRID PRECODING
Hybrid precoding combines analog and digital processing to retain multiplexing gains while reducing RF-chain requirements. PAHP implements the analog stage with phase shifters, using full or sub-connected architectures.
- Principle: Fully digital precoding offers arbitrary amplitude and phase control but requires one dedicated RF chain per antenna.Hybrid precoding is presented as a compromise between design freedom and hardware cost or power consumption.
- Principle: PAHP realizes the analog beamformer with phase shifters and uses a digital precoder for Ns single-antenna users.
- Principle: Full-PAHP connects every RF chain to every antenna, while sub-PAHP uses subsets of antennas and a block-diagonal analog beamformer.
- Principle: Sub-PAHP reduces phase shifters from NNRF to N and avoids power combiners, but loses array gain by a factor of 1/NRF.
B. Advantages
PAHP reduces RF-chain requirements while preserving near-optimal performance when enough RF chains cover the channel rank, but its design and channel estimation remain challenging.
- B. Advantages: PAHP can reduce RF chains from N antennas to Ns streams, lowering power consumption while retaining near-optimal multiplexing performance when RF chains exceed channel rank.The example reduces N = 256 antennas to NRF = Ns = 16 RF chains.
- B. Advantages: PAHP hybrid-precoder design maximizes sum-rate under non-convex analog-beamformer constraints, including constant-modulus phase-shifter amplitudes.A convex approximation can produce a near-optimal hybrid precoder with low complexity.
- B. Advantages: Spatially sparse OMP and SIC-based schemes obtain near-optimal hybrid precoders for full-PAHP and sub-PAHP, respectively.The SIC-based method solves sub-phased-array problems sequentially.
- B. Advantages: PAHP’s maximum gain requires perfect CSI, but low estimation SNR and fewer RF chains than antennas prevent direct observation of the channel matrix.These constraints motivate two-step estimation and low-rank-based approaches.
A. Principle
PAHP uses phase-shifter-based analog processing, while LAHP uses a lens array and selecting network to transform signals into beamspace.
- A. Principle: Full-PAHP uses phase shifters connecting every RF chain to every antenna, whereas sub-PAHP connects each RF chain to a subarray.Full-PAHP uses a full analog beamformer; sub-PAHP reduces connectivity and array gains.
- A. Principle: LAHP realizes the analog beamformer with a lens array and selecting network.The lens array provides directional energy focusing with antennas located on the lens focal surface.
- A. Principle: The lens array acts as an N × N spatial DFT matrix whose columns are orthogonal beams covering predefined directions across the angular space.These beams form the beamspace representation used by LAHP.
- A. Principle: LAHP’s beamspace model uses received signal ỹ = H̃Ds + n, with D as the digital precoder and H̃ = HU as the beamspace channel.The passage defines ỹ, s, n, D, and H̃ dimensions and roles.
B. Advantages
LAHP reduces RF complexity by selecting dominant beams from a sparse mmWave beamspace channel while preserving array gains and near-optimal performance.
- B. Advantages: LAHP selects a small set of dominant beams from the sparse beamspace channel, reducing MIMO dimension and required RF chains without obvious performance loss.Limited scattering makes the mmWave beamspace channel sparse.
- B. Advantages: The low-cost lens array preserves array gains, allowing a simple selecting network to maintain satisfying performance.This distinguishes LAHP’s hardware-efficiency mechanism from approaches that sacrifice array gain.
C. Challenges and recent results
LAHP’s main challenges are selecting beams efficiently and estimating sparse beamspace channels with suitable hardware, motivating incremental, reduced-dimensional, and adaptive methods.
- C. Challenges and recent results: Exhaustive beam selection is optimal but prohibitively complex because its complexity grows exponentially with the number of selected beams.More efficient beam-selection schemes are therefore needed.
- C. Challenges and recent results: Magnitude maximization selects high-power beams simply but loses performance by ignoring interference, while incremental selection chooses beams sequentially more efficiently.The incremental method is developed from antenna selection.
- C. Challenges and recent results: LAHP channel estimation targets a sparse beamspace channel and differs from PAHP estimation because the hardware architectures differ.The estimation target is not the conventional low-rank spatial channel.
- C. Challenges and recent results: Two-step estimation first scans beams and selects strong ones, then applies least squares to the reduced channel.This reduces the dimension of the beamspace estimation problem.
- C. Challenges and recent results: An adaptive selecting network enables compressed sensing by producing a full-rank random 0/1 sensing matrix during channel estimation.The same network selects beams for data transmission but connects each RF chain to all antennas for estimation.
V. PERFORMANCE COMPARISON
The comparison evaluates PAHP, LAHP, and fully digital precoding in single-cell and multi-cell mmWave MIMO settings, focusing on sum-rate and power efficiency. With substantially fewer RF chains, LAHP and PAHP approach fully digital sum-rate performance, while LAHP offers stronger robustness and power efficiency under the reported conditions.
- Single-cell scenario: The study compares full-PAHP, LAHP with adaptive selecting, and fully digital precoding for a 256-antenna base station serving 16 users.The single-cell evaluation uses the Saleh–Valenzuela multipath channel model, with one line-of-sight path and two non-line-of-sight paths per user.
- Single-cell scenario: With 16 RF chains instead of 256, PAHP and LAHP achieve sum-rate performance close to fully digital precoding.PAHP exploits low-rank spatial channels, whereas LAHP benefits from sparse beamspace channels.
- Single-cell scenario: With perfect channel knowledge, PAHP outperforms LAHP by about 2 dB because phase shifters provide greater design freedom than switches.The comparison also reports that LAHP is more robust to channel-estimation error because only the dimension-reduced beamspace channel is effective for data transmission.
- Power-efficiency comparison: Both PAHP and LAHP provide much higher power efficiency than fully digital precoding when the number of users is not large, specifically Ns ≤8.Power efficiency is defined as achievable sum-rate divided by transmission power plus hardware power consumption.
- Power-efficiency comparison: When Ns > 12, LAHP retains high power efficiency, whereas PAHP performs worse than fully digital precoding because its phase-shifter count grows rapidly.The analysis uses conservative high hardware-power values, so the reported efficiencies are lower bounds for practical implementations.
- Multi-cell scenario: In the two-cell setting, each 256-antenna base station serves 16 users without base-station cooperation and has channel knowledge only for its own-cell users.The multi-cell comparison uses the same channel-generation model as the single-cell evaluation.
VI. CONCLUSIONS
The paper reviews PAHP and LAHP as low-RF-complexity approaches for large-array mmWave MIMO and proposes an adaptive selecting network for LAHP. Its comparison finds different trade-offs between ideal-channel sum rate, estimation robustness, and power efficiency, while broadband time-varying channels remain an open challenge.
- Conclusions: The paper reviews PAHP and LAHP, detailing their principles, advantages, challenges, and recent results for large-array mmWave MIMO.It also provides a systematic performance comparison of the two technologies.
- Conclusions: The proposed adaptive selecting network gives LAHP a low-cost, low-power architecture and formulates beamspace channel estimation as sparse signal recovery.The approach is described as considerably reducing pilot overhead.
- Conclusions: PAHP achieves higher sum rate with perfectly known channels, whereas LAHP is more robust to channel-estimation error and has higher power efficiency.LAHP replaces the phase-shifter network with a lens array and switches.
- Future research: Most existing low-RF-complexity technologies target narrowband, time-invariant channels, leaving broadband and time-varying outdoor mmWave channels as an urgent open problem.Broadband operation complicates frequency-independent analog beamforming, while time variation requires frequent channel re-estimation and hybrid-precoder recomputation.