Massive MIMO: Top 10 Research Topics in Need of Prototyping

Overview

Multiple input, multiple output (MIMO) has emerged as a promising technology that uses multiple antennas to address the increasing demands for data throughput and capacity in next-generation wireless networks. Today’s MIMO deployments, however, use only relatively small numbers of antennas and frequently underperform simulated results. New research into higher channel count MIMO systems, also known as Massive MIMO systems1 indicates that increasing the number of antennas can significantly improve gains in spectral efficiency and capacity and can reduce energy consumption at the base station.2 Significant research is still needed to uncover and solve the technical challenges remaining between the promises of Massive MIMO and its deployment in commercial systems.

 

This white paper features key, yet-to-be published Massive MIMO topics and how a testbed can be used to accelerate wireless research by enabling the validation of simulated results through rapid prototyping. The paper also examines how the new NI MIMO Prototyping System hardware and software are being customized to explore novel facets of Massive MIMO that can help researchers find the technical solutions they need to mature the technology.

 

Scientists are already seeing significant progress in a few areas of Massive MIMO research. Teams from University of Bristol in the UK and Lund University in Sweden recently demonstrated the feasibility of Massive MIMO technology utilizing 128 base station antennas, shown in Figure 1, by using beamforming to communicate with 22 users simultaneously on the same time and frequency resource. The demonstration provided foundational evidence that large-scale MIMO arrays can provide significant gains in 5G. Bristol and Lund claimed a record-setting spectral efficiency of 145 b/s/Hz over a 20 MHz bandwidth, which built on their record-setting 79.4 b/s/Hz with 12 users on a fully bidirectional, real-time, over-the-air link at 3.5 GHz.3 That level of performance represents around a 10X improvement in capacity over current LTE standards,4 which is a vital step in the race to 5G.

 

Figure 1. 128-Channel Massive MIMO Base Station at Bristol University, UK

 

The foundational work at Bristol and Lund is just the beginning. Massive MIMO presents a wide variety of interesting technical obstacles in need of deeper exploration and publication by researchers. Without greater knowledge and real-world prototyping, Massive MIMO may not reach its full potential.

 

Contents

10 Massive MIMO Research Topics in Need of Further Study

 

      1. Distributed Massive MIMO

Current research has focused on testbeds with colocated radios connected to a single antenna array. However, Massive MIMO systems with noncentralized elements offer significant advantages. By distributing antennas or remote radio heads, greater spatial diversity is achieved, potentially enhancing the benefits of MIMO. There are also geometric benefits. Especially at lower frequencies, fitting over 100 antennas into a single array is difficult. Distributed Massive MIMO enables more convenient antenna placement, such as spreading them among multiple buildings. Additionally, distributed baseband units could be used to increase the number of potential use cases.

 

      2. High Mobility

Mobility is more difficult to achieve in Massive MIMO setups than in conventional wireless systems. Since Massive MIMO relies on up-to-date channel state information to pinpoint each user via beamforming, the shorter channel coherence time of a mobile user may prevent accurate beamforming. The channel coherence time is dependent on the speed of the user device, and the faster the device movies, the more often the channel state information must be updated. In other words, a user equipment (UE) could move faster than the beam could update or track. This is especially challenging in time division duplex setups. Work is needed to determine the physical layer modifications required to ensure Massive MIMO robustly supports high-mobility applications such as vehicular-based UEs.

 

      3. Antenna Patterns

Most Massive MIMO work has been conducted using patch antennas or omnidirectional dipoles arranged in either uniform linear arrays or 2D planar arrays with consistent spacing. Though this may be ideal for traditional beamforming applications, it’s not clear which antenna patterns provide the best Massive MIMO performance. Antenna geometry, such as curved and distributed arrays, or the greater use of directional antennas could yield significant improvements in performance.

 

      4. Hybrid Beamforming

Digital beamforming enables the full potential of Multi-User MIMO, but separate transceiver chains for each antenna in the system can be expensive and energy intensive. Hybrid beamforming aims to not only preserve the benefits of digital beamforming by using multiple digital transceivers but also improve cost and power consumption by using an analog front-end antenna array with multiple elements to perform beam steering. Further research will help develop algorithms, characterize performance, and optimize energy use of a hybrid beamforming approach and compare those results with existing methods.

 

      5. Energy Efficiency

Massive MIMO should limit radiated power significantly, but using hundreds of transceivers may present a power problem at a real-world base station. This is an especially complex issue with the faster analog-to-digital converters and digital-to-analog converters required for higher bandwidth deployments. New algorithms are needed to reduce energy by selectively impairing antennas, reducing sample rates, or using other strategies for power reduction. Prototyping will determine the extent to which these algorithms can be used.

 

      6. Improvements in Channel State Information Feedback

Massive MIMO base stations must regularly receive channel state information (CSI) from each UE. The transfer of CSI feedback, however, occupies valuable time that could be used to transmit data. Furthermore, until the CSI is received by the base station, no beamforming can occur, which decreases operational efficiency. New methods, such as beamforming CSI reference signals to the base station or using CSI quantization strategies, could solve these problems.

 

      7. MAC Layer Control

The potential of Massive MIMO to add more users is clear, but scheduling algorithms that unlock the full capacity benefits of the technology need to be developed. Smart scheduling approaches and user-tracking strategies featuring variable beamwidths, long-term channel statistics based beamforming, or the clustering of UEs based on direction offer the promise of greater efficiency in the MAC layer.

 

      8. Network MIMO

Several fundamental system level research problems arise when looking beyond a single MIMO base station and into multiple base-station setups. Pilot contamination, during which a pilot signal from one cell interferes with a pilot signal in an adjacent cell, remains a problem. Further research is needed to determine the best way to hand off a user from one base station to another. Another option is multipoint MIMO, which connects two or more base stations to the same UE. This requires additional synchronization and communication between base stations. Coordination between base stations could also limit interference between cells, which would enable denser networks.

 

      9. FDD Massive MIMO

Most Massive MIMO research has used time division duplex (TDD) operation. TDD is useful because it reduces the overhead of acquiring CSI by exploiting channel reciprocity. With frequency division duplex (FDD), on the other hand, CSI overhead increases as the number of base station antennas increases. However, FDD operation offers several useful benefits. Existing wireless networks in particular are largely implemented with FDD. Efficiently adapting Massive MIMO to use FDD would overcome a large hurdle in getting the technology to market.

 

      10. Spectrum Sharing

As new bands such as the 3.5 GHz band open up for commercial wireless applications, new rules include provisions for primary and secondary users. Additionally, the 5 GHz Wi-Fi bands include rules for carrier sense multiple access. To date, the viability of Massive MIMO systems in shared spectrum scenarios has not been evaluated, but it poses some of the greatest opportunities for near-term adoption. The higher frequencies and thus shorter wavelengths of both 3.5 GHz and 5 GHz decrease the overall size of the antenna array, which simplifies deployment.

 

These Massive MIMO research questions are barriers to 5G technology adoption, but they are also opportunities for groundbreaking research. To solve these problems, NI’s approach is to overcome the limitations of theory by enabling real-world, real-time, rapid prototyping. The new NI MIMO Prototyping System is designed to deliver this game-changing approach to MU-MIMO and Massive MIMO development.

 

NI's MIMO Prototyping System

 

The MIMO Prototyping System, shown in Figure 2, combines multiple hardware components to form a powerful yet flexible research platform. Up to 64 dual-channel USRP RIO devices are used as radio heads, offering 128 state-of-the-art RF transceivers combined with a high performance Kintex-7 FPGA for high performance signal processing. Data is routed into a PXI chassis via a low-latency PCI Express connection and then processed with multiple FlexRIO Kintex 7 based FPGA coprocessors. A PXI controller with an Intel i7 processor handles CPU-based host-side processing, and a highly accurate OXCO is used to generate a reference clock that synchronizes all 128 radio heads offer coherence clock phase for all of the channels—up to 128. The result is a powerful platform that is tightly synchronized yet completely open for design and development.

Figure 2. NI Massive MIMO Prototyping System in the 16-Channel Configuration

 

The MIMO Prototyping System combines this integrated hardware with ready-to-run yet customizable software: the MIMO Application Framework. This framework is a set of reference code that unlocks the potential of the MIMO Prototyping System by providing an open and fully reconfigurable reference design that can be customized with IP developed in LabVIEW or integrated with existing HDL-based IP. Out of the box, the MIMO Application Framework supports real-time communication using MU-MIMO and Massive MIMO, scaling from 4 to 128 channels (depending on research needs) at the base station and up to 12 simultaneous users in the same time and frequency block. Advanced features such as MMSE, ZF, and MRC precoding algorithms, channel reciprocity calibration, over-the-air synchronization, and channel visualization are included and fully open for user modification. As a result, the MIMO Prototyping System greatly reduces time to results and empowers researchers to conduct groundbreaking 5G research. Massive MIMO is an exciting candidate technology with a lot of work left to be done. The research topics listed in this white paper only scratch the surface as the number of base station antennas grows. Early results are very promising. NI is focused on providing an open and reconfigurable starting point for a wide variety of MU-MIMO research. Using a common hardware and software platform is allowing a community of researchers to come together, share code, and validate results providing tremendous gains in productivity.

 

References

1. T. L. Marzetta, “Noncooperative cellular wireless with unlimited numbers of base station antennas,” IEEE Trans. Wireless Commun., vol. 9, no. 11, pp. 3590-3600, Nov. 2010.

2. Vieira, J., Malkowsky, S., Nieman, K., Miers, Z., Kundargi, N., Liu, L., ... & Tufvesson, F. (2014, December). A flexible 100 antenna testbed for massive MIMO. In Globecom Workshops (GC Wkshps), 2014 (pp. 287-293).

3. “Bristol and Lund once again set new world record in 5G wireless spectrum efficiency”, Bristol.ac.uk, 2016. [Online]. Available http://www.bristol.ac.uk/news/2016/may/5g-wireless-spectrum-efficiency.html. [Accessed: 26- Jul- 2016].

4. Mobile and Wireless Communications for IMT-Advanced and Beyond, First Edition. Edited by Afif Osseiran, Jose F. Monserrat and Werner Mohr., 2011