Brooklyn 5G Summit
It’s important for industry and academia to meet and shape a common understanding of the fundamental technology basis for 5G. At the recent Brooklyn 5G summit, we had the opportunity to present and discuss several key technologies. Let us share our insights and reflections specifically on massive MIMO, channel models and RFIC technology.
Several speakers emphasized what we in Ericsson have been talking about since several years: that 5G should be designed as a platform to support new business opportunities. This has fundamental implications on the architecture of the network and its application interfaces. The wide array of potential use cases accentuates the need for flexibility and adaptability of the network. The technologies that are dear to our hearts need to come together to provide the underlying enabler for these opportunities.
Digging into the details of the 5G platform, looking at its components, we want to put the spotlight on three specific technologies.
Mikael was a speaker and panelist in the session on “Massive MIMO technology for 5G and LTE-A below 6GHz.” Antennas are a very powerful tool to improve the performance of wireless networks and massive-MIMO will be a key technology for 5G, with both the LTE-evolution, as well as new technologies. Massive-MIMO is also an area where there is some confusion on terminology. You have probably heard Massive-MIMO, MU-MIMO, FD-MIMO, VL-MIMO, etc… the list is long and clear definitions of these terms don’t exist. So let’s start by defining our interpretation: By Massive-MIMO we mean “a large number of dynamically steerable antennas.” This massive number of antennas can be used in different ways: SU-MIMO, where you use the antennas for beamforming and multiplexing to one device (per time/frequency resource) at a time; and MU-MIMO, where you schedule more than one user at a time per time/frequency resource.
There is no doubt that massive-MIMO has a very large potential. However, there are also a number of challenges that need to be considered in order to realize this potential. Of course, with challenges there are always opportunities for innovation!
Here is a highlight of a few areas that we think deserve attention going forward:
- Traffic modelling: Especially for realistic MU-MIMO gains it is important to look at realistic traffic patterns and user distributions. We need to get beyond the full-buffer assumption that’s quite often used.
- Reciprocity is often mentioned as a pre-requisite for massive-MIMO concepts. This is however an over-simplified view. Interference is not reciprocal, meaning that in practice we need closed-loop feedback in order to do link adaptation and other link optimizations. Channel state information will cost overhead in terms of pilots etc. It is a delicate optimization problem to design the most efficient massive MIMO system taking into account realistic traffic/user assumptions, number of users vs. number of antennas, propagation properties and also the fact that a large part of the spectrum below 6GHz is designated to FDD. Massive MIMO will also be of significant importance for FDD.
- Characterization and measurements: It is of key importance that massive-MIMO systems can be characterized both for performance and for co-existence with neighboring carriers and systems. The current regulatory framework for verification of requirements is centered on measurement points at the antenna connectors, which is not sustainable for large numbers of highly integrated massive MIMO solutions. For this, a new framework is needed that is agnostic to the number of antennas and specific configurations, while still capturing all relevant aspects and being feasible from measurement complexity perspectives.
In the session on “Channel Models: Key to 5G Air-Interface Technology,” Mikael was also on the discussion panel.
We feel that the most important role of channel models is to capture the essential characteristics of the wireless environment. With the trend towards large numbers of antennas, it is particularly crucial to capture the correlation among channels properly, in order to tone down the over-optimistic promise of multi-antenna techniques. Too often, the very high potential, indicated by early results with simplistic channel models, gets smaller and smaller as more realistic models are introduced. The end result would be very encouraging in isolation, but it may feel like a disappointment given the starting point, especially to those not directly involved with the technology. This is not just a perception issue, as it can slow the momentum of this important technology.
The channel modeling specialists had many strong opinions; in particular about whether channel models have to make physical sense. Some felt the model equations are just a way to fit the data from measurements, and only meaningful in the range of frequencies, distance etc. of the actual measurements. Others felt that the models need to be consistent with the underlying physics. For instance, the path loss would be with respect to a 1-meter reference, and the decay parameter would directly correspond to energy decay. Also, the models need to make sense at the boundary conditions of cmW and visible light.
There was much more harmony of opinions regarding the need for channel models which are as unified and common as possible, and cover as many key use cases as possible. This is important from an operational perspective, as it enables evaluations and comparisons in standards activities. It’s also very useful beyond standards, for deployment planning, evaluation of new technologies etc. There was also agreement about the urgency of additional measurement campaigns in various use cases.
One interesting perspective came from audience questions: Isn’t it time to move from a modeling approach to a big data approach? That is, the terabytes of measurements could be used directly in evaluations, rather than through a model proxy.
The session on “RFIC Technology for Massive MIMO and Beamforming” was of particular interest to us. As a group, the speakers were very optimistic about the ability of the ability of this key technology to address the challenges of 5G in mmW. This agrees with the opinion of Prof. Gabriel Rebeiz from UCSD, who visited us and also participated in our recent 5G workshop.
The collective message from the talks and the panel is that mmW transceiver technology is maturing fast, driven in particular by the experience with WiGig. We caution that the requirements of 5G are likely to be much tougher than those of WiGig. One reason is spectrum: We expect 5G to operate in licensed mmW spectrum, where requirements such as out of band emissions and blocking would be much tighter than in unlicensed spectrum. Bearing in mind that such requirements are driven by regulators, the mobile industry should take a hint and aim for looser requirements. But if mmW spectrum licenses are as expensive as in current spectrum, operators will want to make sure requirements stay nice and tight.
Another reason is transmit power: Even if the mmW range of 5G is limited to 100’s of meters, it will still require a couple of orders of magnitude higher power than WiGig, whose range is in meters. This will present a big challenge in terms of transceiver size and cost, especially in the terminal. The degree of difficulty will be compounded by tight requirements, such as out of band emissions. We should keep an eye on research progress in higher power mmW transmitters.
Another area of interest is testing and calibration of antenna arrays. The consensus appears to be that more progress is needed, and that integrated self-calibration schemes would be very useful.
About 200 people were present, and the event was also broadcast via video streaming by IEEE. New York University and Nokia Networks were the organizers. The agenda can be found here.
As we also attended the initial summit a year ago, it was instructive to get a perspective on the progress made on many fronts on the road to 5G. We sponsor the NYU Wireless Research Center, so we were happy to see them in the spotlight!
All in all, this was a great event with interesting talks, posters and also a number of demos. It will continue to be important for industry and academia to meet and shape a common understanding on fundamental technology basis for 5G. We look forward to a continuation of this event next year, as well as other initiatives shaping a common 5G platform – next in line is the Johannesberg Summit, a co-arrangement between Ericsson Research and the Royal Institute of Technology.
Mikael Höök Research Director, Radio Access Technology
Ali Khayrallah, Engineering Director Ericsson Research Silicon Valley