Building High Compression/Low Latency 5G Fronthaul Networks That Will Power the Internet of Things

Keysight

We are on the cusp of the next mobile revolution. 5G cellular mobile communications are coming, and with them a host of new applications. But there are many technological challenges as well. The higher bandwidth and lower latency of fifth generation wireless technology will allow us to push more data faster, but force us to rethink the way we engineer and deploy our transport networks.


To support massive 5G traffic within the licensed spectrum bands, 5G radio foot-print needs to be much smaller than in 4G.  Part of the 5G spectrum is offloading to the millimeter-wave band. But, millimeter-wave has a shorter reach and this also shrinks the cell size.  Ever-increasing number of  5G cell-sites requires a shift in the way mobile transport networks operate. We need to look at the fronthaul: the way that a baseband processor interacts with antenna signals.

5G networks will allow us to push more data faster, but will force us to rethink the way we engineer and deploy our transport networks.


The closer proximity and higher bandwidth of 5G cells will alter the way we use near-field communications and the Internet of Things (IoT). Imagine several small cells in a store or shopping mall providing real-time cloud-based augmented reality (AR) to customers. Another major innovation will be the way we deploy home internet services. Instead of cabling an entire building, we can run fiber to rooftop cells that will provide high-speed low-latency access to all its occupants. While all of this is already possible with existing wireless networks, 5G will greatly enhance the delivery of such services.

From Atomic Clocks to Cyber Security 

We know a few things about standards here at the National Institute of Information and Communications Technology (NICT). The NICT promotes a full spectrum of research and development activities in this country. Our aim is to make Japan a world leader in information communications technology through close ties with academic and business communities here and abroad. The findings of our R&D activities impact society in a broad range of fields beyond academia and commerce.


NICT serves as a technical advisory body that helps the Japanese government with matters pertaining to radio spectrum planning and bandwidth allocation. We also research advanced fiber optic and photonic network technology. Fiber is reaching the limits of its capacity, and we are trying to push those limits. Recently, we developed a photonics and electronics convergence device that will convert data transmitted as light across fiber-optic networks to highly efficient millimeter-wave radio signals for wireless transmission. 

Optimizing Compression and Reducing Latency

In collaboration with The Graduate School for the Creation of New Photonics Industries (GPI), we are working on the optimization of low-latency data compression for the fronthaul of 5G  transport networks. 


In cellular communications, digital data is converted into analog radio frequency signals that are transmitted from a base station to user equipment, and vice versa. In the past, the entire modulation/demodulation process is done at each cell-site. This impacts on the cost of the cell-site. With 5G networks, part of this process is moved from the cell-site to the cloud. As a result, data has already been converted into RF signals when it reaches the cell-site. Splitting of base-station functionality drastically simplifies the cell-site, but requires broader bandwidth in the fiber link between the cell-site and the cloud, namely fronthaul.  


For efficient fronthauling, we need high-throughput, low-latency, space-and-time compression techniques for 5G radio frequency signals. In other words, we need to reduce the amount of data being transported from the multiple antennas at the cell-site to the cloud, and the reduction needs to be quickly done before the wireless channel environment changes. 


One of the ways we are doing this is by compressing merged radio channels. Say you have 100 antenna signals, but only 10 or 20 users are currently on those channels. We can go in, extract only the data in channels being used and compress them. We save a lot of overhead by not compressing the data streams of antennas that are not actively transmitting.


We extract this data blindly and adaptively with algorithms called subspace tracking. We then use low latency audio compression techniques—similar those used to master CDs and MP3s, but at far higher frequencies—to reduce the size of these signals. Using these tools, we were able to route 256-antenna LTE signals, which require a 260-gigabaud optical interface in the conventional fronthauling technique like CPRI,  over a 10-gigabaud optical interface. It was one of our biggest engineering achievements of 2018.

Testing Network Latency in Real Time

The compression side of the equation is easy. Once you have figured out how to extract the data and how to reduce it in size, you keep repeating what you are doing. The process never changes. However, latency is another matter. It can be affected by many factors. The number of channels and antennas; the distance between cells; the type of DSP circuits, DACs and ADCs being used; and the network infrastructure all play a part in determining latency. You can also simulate compression using a computer model, but the same is not true for latency.


We needed to test different configurations of our fronthaul architecture and our compression solution for latency. The only way to do this is in real time. We could have purchased FPGA chips from Xilinx and peripherals from other vendors, hooked them up to a test environment by integrating with high-speed DAC and ADC interface boards, and then asked a technician to program them, but this would have taken too much time, and cost far too much money. This would have amounted to $100,000 per trial, and hard to catch up with rapid 5G/IoT R&D.  

Accelerating R&D with Keysight

We needed a partner who provided a simpler hardware implementation solution to support us during the testing phase, and we found it in Keysight Technologies.


Keysight's FPGA-base multiple-channel arbitrary waveform generator (AWG) and digitizer combination solution (M3300A) was the best fit for analyzing our space-time fronthaul compression solution in real time. With their ready-to-go FPGA programming environment and prompt support by Keysight's engineers, we had attained almost the same level of expertise as a dedicated programmer, and was able to use the Keysight M3300A test platform to verify and  then demonstrate various iterations of our design, within a year.  It saved us months of extra effort and also saved approximately $50,000 by doing the FPGA programming ourselves, with guidance from Keysight’s engineers. This helped us focus on the R&D process and freed us to direct all our efforts/resource into redesigning, upgrading, and updating our algorithms as we moved toward a definitive version of our fronthaul technology.