Massive MIMO world record

A research team has set a world record for 5G wireless spectrum efficiency using a laboratory massive MIMO testbed.

Massive multiple-input, multiple-output (MIMO) technology promises to provide huge capacity and energy efficiency gains for future 5G networks, which will have to accommodate increased data rates and the rapid proliferation of smart connected devices without consuming any more of the radio spectrum.

Using National Instruments (NI) platforms to develop a 128-antenna, real-time massive MIMO testbed, we were able to use just 20 MHz of spectrum (within the 3.5 GHz band) to simultaneously serve 12 client devices over-the-air, with an aggregate data rate of 1.59 GBps, and set a new world record for 5G wireless spectrum efficiency.

The Communication Systems & Networks (CSN) Group at the University of Bristol formed in 1985 to address the research demands of the fixed and wireless communications sectors. It combines fundamental academic research with a strong level of industrial application. The group has well-equipped laboratories with state-of-the-art test and measurement equipment and first-class computational facilities.

Figure 1: Researchers Paul Harris and Siming Zhang with the BIO 128-antenna massive MIMO system.

‘Bristol Is Open’ (BIO) is a joint venture between the University of Bristol and Bristol City Council, which supports initiatives that contribute to the development of a smart city and the IoT.

Lund University seeks to be a world-class university that works to understand, explain and improve our world and the human condition. The Electrical Engineering and Information Technology Department (EIT) covers a wide range of research areas in the fields of analog and digital, as well as communications system design.

It has been at the forefront of massive MIMO research including massive MIMO theory, channel measurements and accelerator design.

The journey to 5G

In addition to mobile phone subscribers, who are predicted to each consume 20 GB per month in North America by 2020, networks will also need to provide broadband internet access to rural areas.

Most prominently, future networks must also accommodate the Internet of Things (IoT) and the proliferation of smart telemetry devices.

Analysts predict that, by 2020, each person in the United Kingdom will own and use 27 internet-connected devices. This will contribute to the expected 50 billion connected devices worldwide.

Aside from connectivity, new industrial applications (smart factories and machine-to-machine communications) and consumer applications (4K video streaming and driverless vehicles) require high data rates, lower latencies and improved reliability.

This is challenging telecommunications engineers to innovate rapidly to ensure that the fifth generation of cellular networks (5G) can cope with these unprecedented demands.

A massive MIMO system can spatially multiplex more devices without consuming any more radio spectrum, which is an extremely valuable and scarce resource.

Coupled with its ability to average out the effects of fast fading in multipath propagation environments (most urban and industrial settings), it can also fundamentally improve latency at the physical layer by reducing the number of errors caused by sudden drops in signal level.

Benefits of massive MIMO

In a multi-user MIMO communication system, devices can simultaneously transmit on the same frequency band whilst the base station uses multiple antennas to unravel their respective data streams in the spatial domain.

For downlink transmission, the base station performs the reverse process, transmitting back to all users simultaneously using a technique known as beamforming.

If the spatial signatures of each device are uncorrelated enough, the result is a K-fold increase in system capacity, where K is the number of users present.

For signal processing reasons, the base station requires at least the same number of antennas as users. MIMO is currently found in both Wi-Fi and 4G cellular, operating with up to eight antennas.

Developing the testbed

Through a collaborative effort with the NI Advanced Wireless Research Group in Austin, Lund University in Sweden and the Bristol City Council, we successfully implemented a 128-antenna massive MIMO system that can serve 12 wireless devices on the same time-frequency resource.

Figure 2: A highly scalable massive MIMO system, combining PXI and USRP RIO.

The testbed is designed with the NI massive MIMO reference design, combining five NI PXIe-1085 chassis. The master chassis contains an NI PXIe-1085 controller, an NI PXIe-6674T timing module and four NI PXIe-7976R FlexRIO FPGA modules. Four slave chassis are linked via x8 MXI.

We connected 16 USRP-2943R software-defined radios (SDRs) via x4 MXI links to each slave chassis, collectively providing a total of 128 RF chains. The accurate 10 MHz OXCO reference from the NI PXIe-6674T along with a digital trigger was distributed to all USRP SDRs through eight OctoClock clock distribution modules, ensuring tight hardware synchronisation.

Finally, we used an additional six USRP-2953R SDRs with x1 MXI links to laptops that mimic user clients. We used LabVIEW software, the LabVIEW FPGA Module and NI-Sync to develop the massive MIMO reference design that powers the system.

Figure 3: Six USRP RIO clients in front of the BIO massive MIMO testbed.

In a large and complicated system such as this, there are many things that can go wrong. However, NI provided an unrivalled, ubiquitous level of integration between its software development tools and commercial hardware products, which helped us easily modularise this complex system. We wanted a flexible, powerful solution built on a single, well-supported platform.

The PXI Express platform is a solid foundation for any high-throughput, low-latency system, and it is well supported by many years of experience from NI. We were able to build upon this well-established standard and integrate nearly 100 different pieces of hardware, yet seamlessly develop our entire application within a single software framework.

This highly modular approach and tight software and hardware integration not only gives us the solution we need right now, but ensures that future changes to the hardware configuration are cost- and time-effective.

To spatially separate and distinguish the signals from all 12 wireless devices whilst meeting real-time constraints, we implemented parallel MIMO processing across the integrated FPGAs within the four FlexRIO modules. Each needed to perform 24 million 12x128-128x1 matrix multiplies per second for signal detection alone.

This leads to another major benefit of PXI Express — peer-to-peer (P2P) streaming — which enables the deterministic transfer of data between cards within the PXI Express chassis.

P2P was pivotal to the success of our application, enabling direct point-to-point transfer between our 68 FPGAs without having to send any of the data through the host processor or memory. This empowered us to achieve the optimal throughput and latency performance required for real-time operation.

The P2P functions in LabVIEW simplified the implementation of each stream so we could easily map source, destination and data type.

Record-breaking results

This was the world’s first real-time demonstration of a 128-antenna massive MIMO system simultaneously serving 12 devices over-the-air in the same frequency band.

With a sum rate of 1.59 GBps in only 20 MHz of bandwidth, we achieved 79.4 b/s/Hz — the highest recorded spectral efficiency in the world to date.

This technology can enable a 12-fold increase in spectrum efficiency compared with current LTE (4G) networks, whilst offering the connection reliability and decreased latency required for Industrial IoT and real-time control applications.

Figure 4: Uplink throughput in real time at the base station. Left: Individual stream rates for each user. Right: System sum rate.

As the BIO testbed has proven, massive MIMO can deliver a strengthened network capacity that will help network operators to reliably host an ever increasing number of wireless devices.

Furthermore, with a 100x increase in radiated energy efficiency, we can greatly reduce the power consumption and operating cost of a wireless network.

From a consumer perspective, wireless devices and mobile handsets will also experience improved battery life.

Next steps

The NI PXI Express platform provided a solid framework to build our system on, which is rugged, easily reconfigured and capable of meeting the demanding throughput requirements for 128-antenna MIMO operation.

We will soon deploy the massive MIMO testbed on a rooftop site within the city of Bristol and connect it to the BIO fibre-optic network to conduct further research in real-world deployment scenarios.

Figure 5: Base station host interface, displaying channel information.

Eventually, we will split the system into four 32-antenna subsystems and use the fibre network to implement a distributed massive MIMO configuration.

All of this work ultimately pushes forward the validation of this promising technology, allowing network operators to consider practical deployments for real networks.

Acknowledgements

We would like to thank the collaborating professors and PhD students in Bristol — Mark Beach, Andrew Nix, Paul Harris, Siming Zhang, Henry Brice, Wael Boukley Hasan and Benny Chitambira — and at Lund — Fredrik Tufvesson, Ove Edfors, Liang Liu, Steffen Malkowsky, Joao Vieira, Zachary Miers, Hemanth Prabhu, Erik Bengtsson, Xiang Gao and Dimitrios Viastaras — for contributing to the success of this project.

Figure 6: The University of Bristol team in front of the record-breaking massive MIMO testbed.

We would also like to thank the members of the NI Advanced Wireless Research Group for their extensive work in developing the FPGA architecture for the massive MIMO reference design and for continuing to work closely with us to bring the system online. We genuinely could not have achieved this fantastic result without them.