Massive MIMO for 5G networks
Recent technology developments have made Massive MIMO the preferred option for large-scale deployments in 5G mobile networks. Massive MIMO enables state-of-the-art beamforming and MIMO techniques that are powerful tools for improving end-user experience, capacity, and coverage. As a result, Massive MIMO significantly enhances network performance in both uplink and downlink. Finding the most suitable Massive MIMO variants to achieve the potential performance gains and cost efficiency in a specific network deployment requires an understanding of the characteristics of Massive MIMO solutions.
Introduction
End-user requirements continue to increase1, putting high demands on the radio access network (RAN) to deliver increased coverage, capacity, and end-user throughput. Since data usage is currently increasing at a much faster rate than corresponding revenue, communications service providers (CSPs) must evolve the RAN in a way that enables a reduced cost per bit while meeting new demands for end-user performance. Even as a relatively new technology, Massive MIMO solutions have already proven to be instrumental in today’s 5G mid-band deployments to meet the network requirements. Most Massive MIMO solutions have already undergone several hardware and software generations, making them highly competitive in terms of size, weight, cost, performance, energy efficiency, and ease-of-deployment.
What is a Massive MIMO solution?
A Massive MIMO solution or simply Massive MIMO (formerly called advanced antenna system, or AAS) is a combination of a Massive MIMO radio and a set of Massive MIMO features. A Massive MIMO radio consists of an antenna array tightly integrated with the hardware and software required for the transmission and reception of radio signals, and signal processing algorithms to support the execution of the Massive MIMO features. Compared to conventional systems, this solution provides much greater adaptivity and steerability, in terms of adapting the antenna radiation patterns to rapidly time-varying traffic and multi-path radio propagation conditions. In addition, multiple signals may be simultaneously received or transmitted with different radiation patterns.
Multi-antenna techniques
Multi-antenna techniques (here referred to as Massive MIMO features) include all variants of beamforming, null-forming, and MIMO. Applying Massive MIMO features to a Massive MIMO radio results in significant performance gains because of the higher degrees of freedom provided by a large number of radio chains.
Beamforming
During transmission, beamforming is the ability to direct radio power through the radio channel toward a specific receiver, as shown in the top left quadrant of Figure 1. By adjusting the phase and amplitude of the transmitted signals, constructive addition of the corresponding signals at the user equipment receiver can be achieved which increases the received signal strength and, thus, the end-user throughput. Similarly during reception, beamforming is the ability to collect the signal power from a specific transmitter. The beams formed are constantly adapted to the surroundings to give high performance in both uplink (UL) and downlink (DL).
Figure 1: Beamforming and MIMO with the different colors of the filled beams that represent different data streams
Although often very effective, transmitting power in only one direction does not always provide an optimum solution. In multi-path scenarios, where the radio channel comprises multiple propagation paths from the transmitter to receiver through diffraction around corners and reflections against buildings or other objects, it is beneficial to send the same data stream in several different paths (direction and/or polarization) with phases and amplitudes controlled in a way that they add constructively at the receiver2. This is referred to as generalized beamforming, as shown in the upper right quadrant of Figure 1. As part of generalized beamforming, it is also possible to reduce interference to other UEs, which is known as null-forming. This is achieved by controlling the transmitted signals in a way that they cancel each other out at UEs that would otherwise be interfered.
Note that the concept of generalized beamforming can be considerably more complex than illustrated in Figure 1, see for example [2, Ch. 6].
MIMO (Multiple Input, Multiple Output) techniques
Spatial multiplexing, here referred to as MIMO, is the ability to transmit multiple data streams, using the same time and frequency resource, where each data stream can be beamformed differently. The purpose of MIMO is to increase throughput. MIMO builds on the basic principle that when the received signal quality is high, it is better to receive multiple streams of data with reduced power per stream, than one stream with full power.
The potential is large when the received signal quality is high, and the beams carrying the data streams are designed not to interfere with each other. The potential diminishes when the mutual interference between streams increases. MIMO works in both UL and DL, but for simplicity, the description below will be based on the DL. The details of how these work are explained in detail in [2, Ch. 4,6,13].
Single-user MIMO (SU-MIMO) is the ability to transmit one or multiple data streams, also called layers, from one transmitting array to a single user. SU-MIMO can thereby increase the throughput for that user and increase the capacity of the network. The number of layers that can be supported, called the rank, depends on the radio channel and the minimum number of antennas on each side. To distinguish between DL layers, a UE must have at least as many receiver antennas as there are layers.
SU-MIMO can be achieved by sending different layers on different polarizations in the same direction. SU-MIMO can also be achieved in a multi-path environment, where there are many radio propagation paths of similar strength between the Massive MIMO radio and the UE, by sending different layers on different propagation paths, as shown in the bottom left quadrant of Figure 1.
In multi-user MIMO (MU-MIMO), which is shown in the bottom right quadrant of Figure 1, different layers in separate beams are transmitted to different users using the same time and frequency resource, thereby increasing the network capacity. To use MU-MIMO, the system needs to find two or more users that need to transmit or receive data at the very same time. Also, for efficient MU-MIMO, the interference between the users should be kept low. This can be achieved by using generalized beamforming with null forming such that when a layer is sent to one user, nulls are formed in the directions of the other simultaneous users.
The achievable capacity gains from MU-MIMO depend on receiving each layer with good signal-to-interference-and-noise-ratio (SINR). As with SU-MIMO, the total DL power is shared between the different layers, and therefore the power (and thus SINR) for each user is reduced as the number of simultaneous MU-MIMO users increases. Also, as the number of users grows, the SINR will further deteriorate due to mutual interference between the users. Therefore, the network capacity typically improves as the number of MIMO layers increases to a point at which power sharing and interference between users result in diminishing gains and eventually also losses.
It should be noted that the practical benefits of many layers in MU-MIMO are limited by the fact that, in today's real networks, even with a high number of simultaneously connected users, there tend not to be many users who want to receive data simultaneously. This is due to the bursty (chatty) nature of data transmission to most users. Since the Massive MIMO and the transport network must be dimensioned for the maximum number of layers, the CSP needs to consider how many layers are required in their networks. In typical mobile broadband (MBB) deployments with the current 64T64R Massive MIMO variants, the vast majority of the DL and UL capacity gains can be achieved with up to 8 layers. For other services than MBB, e.g. fixed wireless access (FWA), there is use for more layers compared to MBB. Eight layers are however usually sufficient also for FWA.
Acquiring channel knowledge for Massive MIMO
Knowledge of the radio channels between the antennas of the user and those of the base station is a key enabler for beamforming and MIMO, both for UL reception and DL transmission. This allows the Massive MIMO to adapt the number of layers and determine how to beamform them.
For UL reception of data signals, channel estimates can be determined from known signals received on the UL transmissions. Channel estimates can be used to determine how to combine the signals received to improve the desired signal power and mitigate interfering signals, either from other cells or within the same cell.
DL transmission, on the other hand, is typically more challenging than UL reception because channel knowledge needs to be available before transmission. Whereas basic beamforming has relatively low requirements on the necessary channel knowledge, generalized beamforming has higher requirements as more details about the multi-path propagation are needed. Furthermore, mitigating interference by using null-forming for MU-MIMO is even more challenging, since more details of the channels typically need to be characterized with high granularity and accuracy. There are two basic ways of acquiring DL channel knowledge: UE feedback and UL channel estimation.
To acquire DL channel knowledge based on UE feedback, the base station transmits known signals in the DL that UEs can use for channel estimation. Relevant channel information is then extracted from the channel estimates and fed back to the base station.
What type of DL channel knowledge can be acquired based on UL channel estimation, also referred to as UL sounding, depend on whether time division duplex (TDD) or frequency division duplex (FDD) is used. For TDD, the same frequency is used for both UL and DL transmission. Since the radio channel is reciprocal (the same in UL and DL), detailed short- term channel estimates from UL transmission of known signals can be used to determine the DL transmission beams. This is referred to as reciprocity-based beamforming. For full channel estimation, signals should be sent from each UE antenna and across all frequencies. For FDD, where different frequencies are used for UL and DL, the channel is not fully reciprocal. Longer-term channel knowledge (such as dominant directions) can, however, be obtained by suitable averaging of UL channel estimate statistics.
The suitable channel knowledge scheme to use depends on UL coverage and UE capabilities. In cases where UL coverage is limiting, UE feedback offers a more robust operation, whereas full UL channel estimation is applicable in scenarios with good coverage. In short, both reciprocity and UE feedback-based beamforming are needed.
Antenna array structure
The purpose of using a rectangular antenna array, as shown in section A of Figure 2, is to enable high-gain beams and make it possible to steer those beams over a range of angles in horizontal and vertical directions. The gain is achieved, in both UL and DL, by constructively combining signals from several antenna elements. Typically, the more antenna elements there are, the higher the gain. Steerability is achieved by individually controlling the amplitude and phase of smaller parts of the antenna array. This is usually done by dividing the antenna array into so called sub-arrays (groups of non-overlapping elements pairs), as shown in section C of Figure 2 and by applying two dedicated radio chains per sub-array (one per polarization) to enable control, as shown in section D. In this way, it is possible to control the direction and other properties of the created beam.
Figure 2: A typical antenna array (A) is made up of rows and columns of individual dual-polarized antenna element pairs (B). Antenna arrays can be divided into sub-arrays (C), with each sub-array (D) connected to two radio chains, normally one per polarization.
To see how an antenna array creates steerable high-gain beams, we start with an antenna array of a specific size, which is then divided into sub-arrays of different sizes. For illustrative purposes, we describe only one dimension. The same principles do, however, apply to both vertical and horizontal dimensions.
The array gain is referred to as the gain achieved when all sub array signals are added constructively (in phase). The size of the array gain, relative to the gain of one sub-array, depends on the number of sub-arrays – for example, two sub-arrays give an array gain of 2 (i.e. 3 dB). By changing the phases of the sub-array signals in a certain way, this gain can be achieved in any direction, as shown in section A of Figure 3.
Each sub-array has a certain radiation pattern describing the gain in different directions. The gain and beam width depend on the size of the sub-array and the properties of the individual antenna elements. There is a trade-off between sub-array gain and beam width – the larger the sub-array, the higher the gain and the narrower the beam width, as illustrated in section B of Figure 3.
The total antenna gain is the product of the array gain and the sub-array gain, as shown in section C of Figure 3. The total number of elements determines the maximum gain and the sub-array partitioning allows the steering of high-gain beams over the range of angles. Moreover, the sub-array radiation pattern determines the envelope of the narrow beams (the dashed shape in section C of Figure 3). This has an implication on how to choose an antenna array structure in a real deployment scenario with specific coverage requirements.
Since each sub-array is normally connected to two radio chains and each radio chain is associated with a cost in terms of additional components, it is important to consider the performance benefits of additional steerability when choosing a cost-efficient array structure.
Figure 3: An array of sub-arrays supporting high total antenna gain and steerability
Deployment scenarios
Determining what kind of Massive MIMO configuration is most appropriate and cost- effective for a particular deployment scenario requires a mix of knowledge about the scenario, possible site constraints, and available Massive MIMO features, particularly the need for vertical steerability of beams, the applicability of reciprocity-based beamforming and the gain from MU-MIMO. It should be noted that horizontal beamforming is a very effective feature that provides large gains in all scenarios since the users are generally spread in the horizontal dimension. Therefore, a large number of columns is beneficial in all scenarios.
We have chosen three typical use cases to illustrate different aspects of Massive MIMO deployment: rural/suburban, urban low-rise, and dense urban high-rise. More comprehensive and practically useful recommendations can be found in3. The scenarios, including relevant characteristics, suitable Massive MIMO configurations, and performance potential are depicted in Figure 4. More elaborate evaluations of the performance achievable with Massive MIMO are available in reference2 and3.
Figure 4: Suitable Massive MIMO configurations, schematic MU-MIMO and SU-MIMO usage ranges, and typical capacity gains in different deployment scenarios
Deployment scenario #1: Dense urban high-rise
As depicted in section A of Figure 4, the dense urban high-rise scenario is characterized by high-rise buildings, short inter-site-distances (ISDs) of 200-500m, large traffic volume, and high subscriber density with significant user spread in the vertical dimension. The main network evolution driver has increased capacity or equivalently high end-user throughput for a given traffic load.
For conventional non-beamformed systems such as 2T2R, the vertical spread of users in combination with the small ISD creates a situation where many users are outside the vertical main beam of the nearest base station. Together with the high site density, this leads to a situation where the signals from interfering base stations are strong, and severe interference problems may occur.
Desired Massive MIMO characteristics in the dense urban high-rise scenario include an antenna area large enough to ensure sufficient coverage (UL cell-edge data rate). Further, the vertical coverage range needs to be large enough to cover the vertical spread of users. This calls for small sub-arrays, which have a wide beam in the vertical direction. Partitioning the antenna into small vertical sub-arrays results in high-gain beams that can be steered over a large range of angles and effectively addresses the interference problems seen with conventional systems. The Massive MIMO radio needs to have a sufficient number of radio chains to support the relatively large number of sub-arrays. The good coverage and large spread of users mean that the potential for reciprocity-based beamforming and MU-MIMO with a relatively large number of multiplexed users is high, and the Massive MIMO radio should support these techniques. A good trade-off between complexity and performance could be achieved with 64 radio chains controlling small sub-arrays.
Deployment scenario #2: Urban low-rise
The urban low-rise scenario illustrated in section B of Figure 4 represents many of the larger cities around the world, including the outskirts of many high-rise cities. Base stations are typically deployed on rooftops, with inter-site distances of a few hundred meters. Compared to the dense urban high-rise scenario, traffic per area unit is lower. There is generally a mix of building types, which creates multipath propagation between the Massive MIMO radio and the UE. Maximizing the antenna area is important for improving the UL cell-edge data rates, especially for higher frequency bands employing TDD. Due to larger ISDs and decreased vertical spread of users (lower buildings), the vertical coverage range can be decreased compared to dense urban high-rises; hence, larger vertical sub-arrays can be used and there is less gain from vertical beamforming. Using larger sub-arrays for a given antenna area means that fewer radio chains are required. Reciprocity-based beamforming schemes will work for most users, but there will be users with poor coverage that need to rely on techniques such as feedback-based beamforming. MU-MIMO is also appropriate at high loads due to the multi-path propagation environment, good link qualities, and UE pairing opportunities. A good trade-off between complexity and performance is a Massive MIMO radio with 16 to 32 radio chains.
Deployment scenario #3: Rural/suburban
Rural or suburban macro scenarios, as depicted in section C of Figure 4, are characterized by rooftop or tower-mounted base stations with inter-site distances ranging from one to several kilometers, low or medium population density and very small vertical user distribution. This scenario calls for a Massive MIMO radio with a large antenna area and the ability to support horizontal beamforming. Vertical beamforming, however, does not provide any significant gains as the vertical user spread is low. Therefore, large vertical sub-arrays with small vertical coverage areas are possible. Reciprocity-based beamforming is supported for a smaller fraction of users than in the other scenarios, and MU-MIMO gains are more limited. A good trade-off between complexity and performance is a Massive MIMO radio with 8 to 16 radio chains.
Evolution of Massive MIMO
The brief explanation of Massive MIMO above reflects the solutions in use to date (2022- Q4). The evolution of Massive MIMO is very rapid, and several tracks are being investigated to achieve higher performance. A few examples include the use of higher numbers of radio chains, larger array panels, the use of new and higher frequencies, and the use of multiple transmission points (multi-TRP). In addition to advancements in technologies specific to Massive MIMO, the use of interworking between Massive MIMO and conventional radios on other frequency bands add additional capacity beyond the sum of the two, respectively. Other developing technologies, e.g. artificial intelligence and machine learning (AI/ML) will also be applied in Massive MIMO to improve performance. Yet other technology developments, relating to for example energy performance, cost efficiency, and site deployment, are coming into use to make Massive MIMO a highly competitive and commercially viable option for mass deployment in a large variety of scenarios.
Massive MIMO is also used to support a growing number of services in addition to MBB. Today Massive MIMO is already used for FWA, IoT and new industries and in the near future also XR services. With the development of private networks, the number of services supported is expected to grow very fast.
Conclusion
Recent technology developments have made Massive MIMO (advanced antenna systems) a preferred option for large-scale deployments in 4G and 5G mobile networks. Massive MIMO enables state-of-the-art beamforming and MIMO techniques that are powerful tools for improving end-user experience, capacity, and coverage. As a result, Massive MIMO significantly enhances network performance in both uplink and downlink.
The Massive-MIMO solution toolbox is versatile and selecting a suitable Massive MIMO (2x) solution depends on aspects such as deployment environment, traffic load variations and ease-of-deployment. Massive MIMO products provide significant benefits across a very wide range of deployment scenarios, making it possible for mobile network operators to enjoy the benefits of cost-efficient Massive MIMO across their networks. Massive MIMO solutions have already proven invaluable in many 5G deployments, and their importance will likely to increase even further in future network deployments.
Glossary
Contributors






