Massive MIMO (mMIMO) is crucial for meeting 5G’s use cases and promise for both performance and energy efficiency, so it is not surprising to witness rapid gains in shipments and deployments as 5G roll-out proceeds.
The driving factors are clear enough – the potential to serve multiple users and devices simultaneously while sustaining high data rates and consistent performance efficiently, which makes it a strong fit for 5G, particularly the eMBB (enhanced mobile broadband) high speed use cases.
Yet mMIMO was not originally envisioned to become such a major driver for 5G, having only broken into the cellular industry around 2006 on the back of a few research projects. It had already made its mark in WiFi and has roots in antenna diversity going back to the early days of television, the idea being that a receiving station has a far higher chance of recovering a transmitted signal accurately if it can make more than one observation of it. This proved especially valuable for urban and indoor reception by catering for multipath interference and fading effects, which are just as relevant today for 5G in those environments.
Antenna diversity was first proposed in its modern compact form for higher frequency transmission in wireless LAN systems, now mainly WiFi, during the early noughties. Multiuser MIMO is now playing an increasingly important role in the latest WiFi 6/6E as one of the two technologies boosting performance, the other being OFDMA modulation. Both of these are focused on improving multiuser operation, which is especially important for WiFi’s predominantly indoor use cases where there is often with contention for access.
MU-MIMO allows routers to communicate with up to eight devices simultaneously with WiFi 6, compared with four for the preceding WiFi 5 and only one previously when each had to wait its turn to communicate. For WiFi, this multiuser performance improvement is more important than headline one-to-one data rates.
In the case of 5G, other properties of mMIMO are more relevant, such as the ability to boost signal strength for reliable communication indoors and to deliver on some of the performance promises. Larger numbers of antennas will be deployed, for example with Ericsson’s AIR 6468, celebrated as the company first 5G NR radio, having 64 transmit and receive antennas. Huawei and ZTE have demonstrated 128 antennas each way. There is no theoretical limit on the number of antennas a mMIMO base station or router could have, just practical implementation constraints.
The two most common mMIMO configurations at present are 64 and 32 antenna pairs, with the former having higher capacity and range but at the cost of more power consumption. Clearly the trend is upwards, noting that most mMIMO systems with just 16 antenna pairs were withdrawn due to collapsing demand in 2019. At the user end though, mMIMO is not needed and indeed a single simple antenna will serve for a smartphone, even if most have two to four these days.
mMIMO does the same for a frequency band as DWDM (dense wave division multiplexing) does for a fiber optic cable, by increasing the capacity of a given physical medium. The total data volume, that is bit-rate per user multiplied by number of users, that a given network or cell can serve is greatly increased, potentially by up to several hundred times with technology currently on the horizon.
In the case of mMIMO, there is a second clear benefit in more even coverage, with a more uniform experience across the whole coverage area, rather than deteriorating reliability and performance towards the edge furthest from the base station.
mMIMO achieves these benefits by combining two related but distinct techniques, sometimes confused – beamforming and spatial multiplexing. Beamforming is sometimes called the ‘choreographer of the signal’, in the case of the mid and low bands, because it coordinates propagation and reception between paired antennas.
By directing signals at individual user devices, the energy required to sustain a given bit-rate reliably is greatly reduced. By contrast, mobile networks until now have simply apportioned their pool of spectrum to be shared between all devices requiring access at the time, without any reuse.
In the case of mmWave, beamforming operates slightly differently because there is usually line of sight (LOS) or near-LOS. Signals are subject to more rapid attenuation or fading at the higher frequencies and beamforming mitigates this loss along the path by focusing the energy towards the target. It helps users receive sufficiently strong signals without interference. In so doing it increases the effective capacity of each antenna.
Spatial multiplexing, as the name suggests, exploits 3D space to transmit multiple signals which can be separated on the basis of their different properties such as phase at the receiving ends, reflecting their varying trajectories. Unlike beamforming the principle can be used in any physical medium that supports different spatial modes of transmission, including multimode optical fiber.
Spatial multiplexing enables MU-MIMO to share base station spectrum among multiple users by exploiting all the space, that is the full range of angles in 3D. In principle this can achieve huge gains in capacity, but in practice those are limited given present traffic profiles by the relatively small amount of simultaneous data transmission typically taking place within a cell. For MU-MIMO to make any difference, there have to be at least two users transmitting or receiving data at the same time in a given cell. Even when the number of connected users is high, the bursty nature of data transmissions mean not many users want to receive data simultaneously.
In practice therefore there is no point at present configuring MU-MIMO for that many layers, that is separate spatial channels, each carrying different data over the same frequencies. As used so far in deployments for MBB, eight layers at most deliver all the achievable capacity gains from MU-MIMO.
That will change with greater densities and also with more use of streaming video, which requires continuous transmission, so over time more MU-MIMO layers will come into play, exploiting more of the scope. Indeed, the expectation of sustained increases in performance and capacity is propelling excitement around the technology and inspiring ever more bullish predictions of growth.
Certainly, growth so far in transceivers and the underlying chips has been faster than expected. This has happened as the underlying business case for mMIMO has strengthened since about 2017 as the technology came to be seen as a foundation for midband 5G NR in the first instance, followed by mmWave. This led to a surge in shipments, so given rate of 5G deployments and realistic expectations generated by spectrum auctions, admittedly tempted in some cases by overpayment that may curtail investment, quite bullish projections are justified.
Market sizing has been problematical because of the confusion around the boundaries between MIMO and mMIMO, the latter usually now in practice being taken to mean transceivers with 16 or more antennas, although nearly all sold have 32 or more. But early on eight-antenna systems might have been included and are after all sufficient for MU-MIMO at present.
As a consensus though we can take it that macro and small cell transceiver shipments for mMIMO configured systems totaled around 50m in 2019 at a value between $1bn and $1.5bn. However, it is clear that CAGR in 2020 has exceeded many predictions of around 50% growth because of rapid deployment in China, as well as South Korea, so revenues are likely to come in between $1.8bn and $3bn. Whatever the case, it looks as though mMIMO will be one of the bright spots for 5G chip and hardware makers for some years to come.