For the past two years, I’ve been writing on a book about Massive MIMO networks, together with my co-authors Jakob Hoydis and Luca Sanguinetti. It has been a lot of hard work, but also a wonderful experience since we’ve learned a lot in the writing process. We try to connect all dots and provide answers to many basic questions that were previously unanswered.
The book has now been published:
Emil Björnson, Jakob Hoydis and Luca Sanguinetti (2017), “Massive MIMO Networks: Spectral, Energy, and Hardware Efficiency”, Foundations and Trends® in Signal Processing: Vol. 11, No. 3-4, pp 154–655. DOI: 10.1561/2000000093.
What is new with this book?
Marzetta et al. published Fundamentals of Massive MIMO last year. It provides an excellent, accessible introduction to the topic. By considering spatially uncorrelated channels and two particular processing schemes (MR and ZF), the authors derive closed-form capacity bounds, which convey many practical insights and also allow for closed-form power control.
In the new book, we consider spatially correlated channels and demonstrate how such correlation (which always appears in practice) affects Massive MIMO networks. This modeling uncovers new fundamental behaviors that are important for practical system design. We go deep into the signal processing aspects by covering several types of channel estimators and deriving advanced receive combining and transmit precoding schemes.
In later chapters of the book, we cover the basics of energy efficiency, transceiver hardware impairments, and various practical aspects; for example, spatial resource allocation, channel modeling, and antenna array deployment.
The book is self-contained and written for graduate students, PhD students, and senior researchers that would like to learn Massive MIMO, either in depth or at an overview level. All the analytical proofs, and the basic results on which they build, are provided in the appendices.
On the website massivemimobook.com, you will find Matlab code that reproduces all the simulation figures in the book. You can also download exercises and other supplementary material.
MAMMOET (Massive MIMO for Efficient Transmission) was the first major research project on Massive MIMO that was funded by the European Union. The project took place 2014-2016 and you might have heard about its outcomes in terms of the first demonstrations of real-time Massive MIMO that was carried out by the LuMaMi testbed at Lund University. The other partners in the project were Ericsson, Imec, Infineon, KU Leuven, Linköping University, Technikon, and Telefonica. MAMMOET was an excellent example of a collaborative project, where the telecom industry defined the system requirements and the other partners designed and evaluated new algorithms and hardware implementations to reach the requirements.
In 2012, when you began to draft the project proposal, Massive MIMO was not a popular topic. Why did you initiate the work?
– Theoretically and conceptually it seemed so interesting that it would be a pity not to work on it. The main goal of the MAMMOET project was to make conceptual progress towards a spectrally and energy efficient system and to raise the confidence level by demonstrating a practical hardware implementation. We also wanted to make channel measurements to see if they would confirm what has been seen in theory.
It seems the project partners had a clear vision from the beginning?
– It was actually very easy to write this proposal because everyone was on the same wavelength and knew what we wanted to achieve. We were all eager to start the project and learn from each other. This is quite unique and explains why the project delivered much more than promised. The fact that the team got along very well has also laid the fundament for further research collaborations.
What were the main outcomes of the project?
– We learned a lot on how things change when going from small to large arrays. New channel models are required to capture the new behaviors. We are used to that high-precision hardware is needed, but all the sudden this is not true when drastically increasing the number of antennas. You can then use low-resolution hardware and simple processing, which is very different from conventional MIMO implementation.
Some of the big conceptual differences in massive MIMO turned out to be easier to solve than expected, while some things were more problematic than foreseen. For example, it is difficult to connect all the signals together. You need to do part of the processing distributive to avoid this problem. Synchronization also turned out to be a bottleneck. If we would have known that from the start, we could have designed the testbed differently, but we thought that the channel estimation and MIMO processing would be the challenging part.
What was the most rewarding aspect of leading this project?
– The cross-fertilization of people was unique. We brought people with different background and expertise together in a room to identify the crucial problems in massive MIMO and find new solutions. For example, we realized early that interference will be a main problem and that zero-forcing processing is needed, although matched filtering was popular at the time. By carefully analyzing the zero-forcing complexity, we could show that it was almost negligible compared to other necessary processing and we later demonstrated zero-forcing in real-time at the testbed. This was surprising for many people who thought that massive MIMO would be impossible to implement since 8×8 MIMO systems are terribly complex, but many things can be simplified in massive MIMO. Looking back, it might seem that the outcomes were obvious, but these are things you don’t know until you have gone through the process.
What are the big challenges that remains?
– An important challenge is how to integrate massive MIMO into a network. We assumed that there are many users and we can all give them the same time-frequency resources, but the channels and traffic are not always suitable for that. How should we decide which users to put together? We used an LTE-like frame structure, but it is important to design a frame structure that is well-suited for massive MIMO and real traffic.
There are many tradeoffs and degrees-of-freedom when designing massive MIMO systems. Would you use the technology to provide very good cell coverage or to boost small-cell capacity? Instead of delivering fiber to homes, we could use massive MIMO with very many antennas for spatial multiplexing of fixed wireless connections. Alternatively, in a mobile situation, we might not multiplex so many users. Optimizing massive MIMO for different scenarios is something that remains.
We made a lot of progress on the digital processing side in MAMMOET, while on the analog side we mainly came up with the specifications. We also did not work on the antenna design since, theoretically, it does not matter which antennas you use, but in practice it does.
The deliverables contain a lot information related to use cases, requirements, channel modeling, signal processing algorithms, algorithmic implementation, and hardware implementation. Some of the results can found in the research literature, but far from everything.
Note: The author of this article worked in the MAMMOET project, but did not take part in the drafting of the proposal.
I have been wondering for years if “MIMO” will always be a term exclusively used by engineers and a few well-informed consumers, or if it eventually becomes a word that most people are using. Will you ever hear kids saying: “I want a MIMO tablet for Christmas”?
I have been think that it can go either way – it is in the hands of marketing people. Advanced Wifi routers have been marketed with MIMO functionality for some years, but the impact is limited since most people get their routers as part of their internet subscriptions instead of buying them separately. Hence, the main question is: will handset manufactures and telecom operators start using the MIMO term when marketing products to end customers?
Maybe we have the answer because Sprint, an American telecom operator, is currently marketing their 2018 deployment of new LTE technology by talking publicly about “Massive MIMO”. As I wrote back in March, Sprint and Ericsson were to conduct field tests in the second half of 2017. Results from the tests conducted in Seattle, Washington and Plano, Texas, have now been described in a press release. The tests were carried at a carrier frequency in the 2.5 GHz band using TDD mode and an Ericsson base station with 64 transmit/receive antennas. It is fair to call this Massive MIMO, although 64 antennas is in the lower end of the interval that I would call “massive”.
The press release describes “peak speeds of more than 300 Mbps using a single 20 MHz channel”, which corresponds to a spectral efficiency of 15 bit/s/Hz. That is certainly higher than you can get in legacy LTE networks, but it is less than some previous field tests.
Hence, when the Sprint COO of Technology, Guenther Ottendorfer, describes their Massive MIMO deployment with the words “You ain’t seen nothing yet”, I hope that this means that we will see network deployments with substantially higher spectral efficiencies than 15 bit/s/Hz in the years to come.
Several videos about the field test in Seattle have recently appeared. The first one demonstrates that 100 people can simultaneously download a video, which is not possible in legacy networks. Since the base station has 64 antennas, the 100 users are probably served by a combination of spatial multiplexing and conventional orthogonal time-frequency multiplexing.
The second video provides some more technical details about the setup used in the field test.
IEEE ComSoc is continuing to deliver webinars on 5G topics and Massive MIMO is a key part of several of them. The format is a 40 minute presentation followed by a 20 minuter Q/A session. Hence, if you attend the webinars “live”, you have the opportunity to ask questions to the presenters. Otherwise, you can also watch each webinar afterwards. For example, 5G Massive MIMO: Achieving Spectrum Efficiency, which was given in August by Liesbet Van der Perre (KU Leuven), can still be watched.
In November, the upcoming Massive MIMO webinars are:
The concept of superimposed pilots is (at least 15 years) old, but clever and intriguing. The idea is to add pilot and data samples together, instead of separating them in time and/or frequency, before modulating with waveforms. More recently, the authors of this paper argued that in massive MIMO, based on certain simulations supported by asymptotic analysis, superimposed pilots provide superior performance and that there are strong reasons for superimposed pilots to make their way to practical use.
Until recently, a more rigorous analysis was unavailable. Some weeks ago the authors of this paper argued, that all things considered, the use of superimposed pilots does not offer any appreciable gains for practically interesting use cases. The analysis was based on a capacity-bounding approach for finite numbers of antennas and finite channel coherence, but it assumed the most basic form of signal processing for detection and decoding.
There still remains some hope of seeing improvements, by implementing more advanced signal processing, like zero-forcing, multicell MMSE decoding, or iterative decoding algorithms, perhaps involving “turbo” information exchange between the demodulator, channel estimation, and detector. It will be interesting to follow future work by these two groups of authors to understand how large improvements (if any) superimposed pilots eventually can give.
There are, at least, two general lessons to learn here. First, that performance predictions based on asymptotics can be misleading in practically relevant cases. (I have discussed this issue before.) The best way to perform analysis is to use rigorous capacity lower bounds, or possibly, in isolated cases of interest, link-level simulations with channel coding (for which, as it turns out, capacity bounds are a very good proxy). Second, more concretely, that while it may be tempting, to superimpose-squeeze multiple symbols into the same time-frequency-space resource, once all sources of impairments (channel estimation errors, interference) are accurately accounted for, the gains tend to evaporate. (It is for the same reason that NOMA offers no substantial gains in MIMO systems – a topic that I may return to at a later time.)
IEEE ComSoc provides new online material every month and in August the focus is on Massive MIMO.
First, four carefully selected articles are offered free of charge, see the screenshot below and click here for details.
More precisely, IEEE offers free access to the published versions of these articles, while the accepted versions were already openly available: Paper 1, Paper 2, Paper 3, and Paper 4.
Second, a live webinar entitled “5G Massive MIMO: Achieving Spectrum Efficiency” is organized by IEEE ComSoc on August 24. The speaker is Professor Liesbet Van der Perre from KU Leuven. She was the scientific leader of the MAMMOET project, which is famous for demonstrating that Massive MIMO works in practice. You can expect a unique mix of theoretical concepts and practical implementation insights from this webinar.
Colleagues at Lund University presented last month a working circuit that performs, in real time, zero-forcing decoding and precoding of 8 simultaneous terminals with 128 base station antennas, over a 20 MHz bandwidth at a power consumption of about 50 milliWatt.
Impressive, and important.
Granted, this number does not include the complexity of FFTs, sampling rate conversions, and several other (non-insignificant) tasks; however, it does include the bulk of the “Massive-MIMO”-specific digital processing. The design exploits a number of tricks and Massive-MIMO specific properties: diagonal dominance of the channel Gramian, in particular, in sufficiently favorable propagation.
When I started work on Massive MIMO in 2009, the common view held was that the technology would be infeasible because of computational complexity. Particularly, the sheer idea of performing zero-forcing processing in real time was met with, if not ridicule, extreme skepticism. We quickly realized, however, that a reasonable DSP implementation would require no more than some ten Watt. While that is a small number in itself, it turned out to be an overestimate by orders of magnitude!
I spoke with some of the lead inventors of the chip, to learn more about its design. First, the architectures for decoding and for precoding differ a bit. While there is no fundamental reason for why this has to be so, one motivation is the possible use of nonlinear detectors on uplink. (The need for such detectors, for most “typical” cellular Massive MIMO deployments, is not clear – but that is another story.)
Second, and more importantly, the scalability of the design is not clear. While the complexity of the matrix operations themselves scale fast with the dimension, the precision in the arithmetics may have to be increased as well – resulting in a much-faster-than-cubically overall complexity scaling. Since Massive MIMO operates at its best when multiplexing to many tens of terminals (or even thousands, in some applications), significant challenges remain for the future. That is good news for circuit engineers, algorithm designers, and communications theoreticians alike. The next ten years will be exciting.