A Discussion into Massive MIMO

With the evolution of wireless communication technology, high-speed data services and ubiquitous access are undergoing explosive growth. It is estimated that by 2020, service traffic will grow 1,000 times compared with current service traffic. To meet users’ service requirements, we need to improve the capacity of wireless broadband access networks.
To target the needs of wireless broadband access, the European Union, China, Japan and United States have already started research into the requirements and key technologies of 5G mobile communication systems. The upgrade from 2G and 3G to 4G always comes with the emergence of new technologies and aims to solve the primary needs at the moment. In the post-4G era, cells will become denser and demands for capacity, energy consumption, and services will increase. The following ways can be used to improve network throughput: increase the transmission rate of point-to-point links; expand spectrum resources; deploy high-density heterogeneous networks.
As data traffic develops, multi-antenna technologies, such as 8-port MU-MIMO and CoMP, in existing 4G cellular networks are insufficient to meet user demands for bandwidth. The latest research shows that large-scale antenna arrays comprising hundreds of antennas or more can be used on a base station side to significantly improve its performance. This technology is called large-scale antenna systems (LSAS), or Massive MIMO.

Applications
Major applications for Massive MIMO include urban coverage, wireless backhaul, suburban coverage and local hotspots. Urban coverage involves macro-coverage and micro-coverage (within high-rise office buildings). Wireless backhaul involves dealing with data transmission between base stations, especially between macro base stations and small cells. Suburban coverage focuses on wireless transmission in remote areas. Local hotspots are used for high-density areas, such as major sports events, concerts, shopping malls, open-air gatherings, or transportation hubs.
Considering the antenna size and installation, distributed antennas become pragmatic. The coordination mechanisms and signaling transmission between antennas are big considerations. In the future, the main scenarios for large-scale antennas are outdoor macro coverage, high-rise building coverage, and indoor coverage.

Research Directions
Massive MIMO is also called large-scale MIMO. With Massive MIMO, hundreds of antennas (128, 256 or more) are installed on a base station to transmit data simultaneously.
Under MAC+PHY structure of the existing LTE systems, main research directions for the physical layer of Massive MIMO include base station antenna structure design, base station precoding, base station signal detection and channel estimation, and control channel performance improvement (Table 1).

As the number of antenna elements increases sharply, these elements need to be expanded to two-dimensional planes/surfaces or three-dimensional arrays. Should an omni-directional (spherical) antenna array, or a planar (plane-shaped) antenna array, or any irregular-shaped antenna array be chosen? This needs to be studied carefully. Meanwhile, although there are a large number of antennas, only larger antenna arrays can match the isolation. Consequently, the usage of higher frequency bands (> 5 GHz) is one of the research topics.
An increasing number of antennas leads to larger antenna dimensions, which causes larger near-field deviation in case of channel modeling using the traditional plane wave method. Therefore, appropriate channel modeling needs to be chosen.
In addition, the transceiver mode for active antennas to be used by MIMO has already been applied in LTE systems.

The performance of massive MIMO will remain stable as the number of antennas is increased. In this case, multi-user MIMO (MU-MIMO) can be adopted. The key to MU-MIMO technology is precoding. Currently, major precoding technologies include MRT, ZT and DPC, among which DPC is considered to be the best and MRT the worst. It is important to find appropriate precoding algorithms; for instance, the ZF algorithm is generally used in engineering. Whether we can find new precoding algorithms with high complexity and performance is a crucial issue at the physical layer.
Compared with traditional MIMO, the channels of Massive MIMO tend to be orthogonal when many antennas are involved. Most system performance is related only to the large rather than the small scale. The pilot design of hundreds of antennas in a base station costs lots of time frequency resources, so pilot-based channel estimation is not advisable. Specific solutions are implemented in TDD and FDD modes, where TDD has intrinsic advantages and is preferred. As the number of antennas is increased, the consumption of CSI-RS is also increased, In TDD mode, channel estimation can be conducted based on channel reciprocity without using pilots. In FDD mode, with wide coverage and high popularity, codebooks with lower consumption can be used for channel coefficient estimation and feedback, and compressive sensing (CS) algorithm can be used for channel feedback. Therefore, channel detection, estimation and feedback cannot be neglected.
After the number of antennas is increased, service channel coverage can meet user requirements, but the capacity of control channel is not enhanced. Accordingly, control channel coverage will become a bottleneck to system performance.
The main MAC research directions for Massive MIMO include MU-MIMO matching algorithms, user scheduling, and resource allocation strategies.

Massive MIMO involves numerous antennas and the channels between multiple users tend to be orthogonal, so the same time frequency resources can be used for user data transmission. When MU-MIMO is used, the base station simultaneously sends data to multiple users, the actual transmission power obtained by each user decreases sequentially. Lower power causes performance loss. Then it is crucial for Massive MIMO to determine which users are suitable for matching and what kind of matching optimizes system performance.
The spacing is bigger when the antennas are far apart. In order to improve system capacity by fully utilizing the antennas, some antennas are allocated to user A and others to user B. This is the allocation strategy for user antennas.
Transmit antennas can simultaneously send data to one or more users, which is RB resource allocation strategy. At present, the traditional resource allocation method can be used if it matches the user retransmission strategy.
What’s more, Massive MIMO can be used together with other networking technologies, such as CoMP technology in LTE systems.