Challenges for network planning and testing at mmWave frequencies

In a mmWave world, turning a corner or spinning in your desk chair might very well mean your device losing the signal. While high frequencies have been used for point-to-point network backhaul links, cellular carriers have never before made common use of these frequencies to serve individual user devices directly. But the dearth of unused spectrum available below 6 GHz, and the huge channel bandwidths available at mmWave, mean that they’re doing so now.

mmWave makes possible blazing 5G speeds that simply can’t be achieved in the single channels available at lower frequencies, but that performance comes with a hefty price in propagation. In an outdoor environment, mmWave calls for even further densifying the networks that carriers have been packing with small cells for the past few years. Complicating things further, traditional outdoor-in coverage is well-nigh impossible to achieve with mmWave, particularly when you’re trying to get a signal into modern, highly energy-efficient buildings.

When it comes to designing, planning and testing millimeter-wave networks,  the necessary planning and testing tools — much like the networks themselves — still have some evolving to do. Some of the challenging and evolving aspects here include:

mmWave networks themselves demand greater scale, but the resources to deploy them aren’t scaling at the same pace. Andy Asava, EVP of global networks for InfoVista, posited that for however many RF design engineers a company had putting together its 4G network, the number of sites for 5G could be 10x that number – but the number of engineers won’t be. Instead, he said, planning tools have to “scale horizontally” with processing power and better accuracy to adjust to the necessary density, must be able to take in a massive amount of data from sophisticated sources and have to automate as many steps in the planning and deployment process as possible.

Models for outdoor environments could be super-granular but have to balance the compute power that will be necessary to draw on either high-definition satellite or lidar-based maps. InfoVista, Asava said, began looking at three-dimensional automated cell planning (ACP) in preparation for 5G and considering, for instance, what level of granularity to draw on for geodata. He sees three areas of need in which planning and design tools have to scale their capabilities and be able to draw on resources such as virtual machines on-demand for planning: High-resolution geodata is crucial for mmWave, when line-of-sight is the basis of planning and something as simple as foliage can foil designs. Satellite imagery can get down to 1-meter resolution, if you have the means to process it; likewise, lidar data sources can now provide high-resolution data that can be combined to increase accuracy of planning, and opening up APIs to include data sources such as social media heatmaps can inform planning with usage insights.

However, while geodata-based mapping information can provide detailed insights around relative building heights and clutter, there is other information that can greatly impact mmWave propagation, such as what materials that surrounding buildings are made of, for which databases simply don’t exist.

Indoor planning is particularly challenging because there isn’t an analogous source of detailed data about every single building that will need an indoor system. “You need good indoor maps and knowledge of the building, and the challenge is how to get that cost-effectively,” said Paul Challoner, VP Network Product Solutions for Ericsson North America. A site survey – usually performed by walking the building – will provide the necessary level of detail, but it’s costly and time-consuming, and Challoner said that objective in indoor planning is to avoid site surveys as much as possible.

Ericsson found that even simple building floor plans – such as the type found on the inside of a hotel room door that show people the basic layout and exits – could be digitized and put into planning tools to give a “good, first approximation” for planning purposes. “If you have pictures and extrapolations from actual photography in the building, that’s the next step. But the simple floor-plan import is sort of a good way to start, and that gets you there in terms of as far as where the presence of the radios should be.”

MmWave networks are mostly planned on a line-of-sight basis, and getting good models that take into account non-line-of-sight reflections accurately will take time and evolution. Building that possibility into planning tools and procedures to take advantage of it consistently and appropriately in large-scale deployments is a different matter than engineers applying it as a creative solution in a tricky coverage zone.

New data sources can offer new types of visibility into the network. Arnd Sibila, technical marketing manager at test company Rohde &Schwarz offered that it may be possible to integrate sensor data from smartphones into network testing tools, along with QoE measurements from those devices, to put together a richer picture of the indoor physical and RF environment – but, he cautioned, such capabilities aren’t in a state where they are automated. Another testing aspect for 5G that R&S is working on is coming up with “use case classes” for test parameters that can be applied broadly to ensure that 5G QoE meets the needs of a particular application class, without having to test every app of that type. In LTE, user experience testing has often involved running certain popular apps like YouTube or Facebook to test performance. Use case classes would make testing more efficient and broader at the same time. He gives the example of mobile gaming: There are hundreds or thousands of games, so in order to avoid having to test multiple games or figure out which ones are representative of the QoE needs, it would be much more efficient to put together parameters that will satisfy the needs of a class of games (or remote meeting apps, or AR/VR apps) and test to those parameters.

Planning and design tools are often still more manual than they could be, and more automation will help in improving network deployment efficiency – if it’s accurate enough. What’s the next step for mmWave planning and design? “Closed-loop automation and bringing machine learning into the mix,” Asava says. That could mean automated tagging of information, for example, or the ability to automatically pull early drive- or walk-testing data into planning and design tools for faster integration. Challoner said that being able to automate the integration of even simple floor plans would be helpful for indoor planning.

“We as an industry are going to have to continuously look and learn as we deploy more and do more massive-scale deployments,” Asava said. “I think the key is making sure that we have the right platforms in place that can adapt.”

Comments are closed.