The never-ending quest for bandwidth

5 mins read

You can tell that Shannon’s Law provides little further scope for improving cellular bandwidth: it’s forcing the architects of what will form the 6G cellular standard to consider all manner of options.

The possibilities range from AI-enabled base stations to airships and a spectrum range that may top out in the terahertz, as long as electronics can keep up with the pace of change.

At the 6GWorld Symposium, Paul Hart, NXP Semiconductors’ general manager of radio power electronics, described the problems that the cellular industry now faces as the inexorable rise of data traffic and the power that traffic consumes comes into conflict with the need to reduce energy consumption across the board. “The current network is consuming twice the power of LTE networks a decade ago,” he added.

This is not because post-LTE systems are less efficient. “The increase in efficiency has been offset by a faster rise in bandwidth,” Hart said. Though there are research avenues in coding improvements that could make the next-generation of protocols more energy efficient by removing the need for high-resolution, power-hungry A/D converters in the down conversion stages (New Electronics, 24 November 2020) and taking advantage of the relative silicon efficiency of high-frequency digital CMOS versus analogue circuitry. But that is not going to be enough. Bits need to do way fewer air miles in order to deliver video and other high-bandwidth data.

Hart argues the best mechanism is to take advantage of Moore's Law and pack more intelligence into local base stations and the computers around them because, “it is more efficient to do computations at the farthest edge of the network than to transmit data back and forth over longer distance. Taking a holistic approach will allow us to create a much more sustainable network”.

Local intelligence

There are several ways in which this local intelligence might be deployed. One is simply to take a leaf from the content delivery network (CDN) operators who already cache a great deal of web-delivered media using server farms deployed around the world. Pushing caches further out into the network would help reduce the amount of data that would otherwise need to be repeatedly pumped through backhaul links. Similarly, local processing close to base stations would help cut latency for applications that need real-time responses, such as the V2X installations that should be deployed by the end of the decade to support driver-assistance systems as well as shared virtual-reality applications and industrial-control systems.

Much closer to the level of cellular technology is the local decision-making that the radio-access network undertakes. 6G encompasses such a wide range of RF protocols that a key issue is determining in real time which channel to use for what data and for which users on a per-second basis.

One of the key lessons learned from the development of 5G is that plenty of available spectrum can be exploited above the microwave domain favoured for most cellular networks. Even better, the available bandwidths are significantly larger, which allows for less aggressive and silicon-intensive symbol encoding techniques.

Mike Eddy, vice president of corporate development at RF filter specialist Resonant, said it comes down to one of the remaining options open under Shannon’s Law. “The formula is: how many RF paths I can get times bandwidth times signal-to-noise ratio. Bandwidth is the big one you can play with,” he said, although the higher frequencies do support the ability to steer beams in specific directions.


That directionality is also the curse of the millimetre-wave spectrum. It is easy for a mobile device to obtain near-perfect reception one second but have to switch to lower-frequency bands the next as the user moves around. A late-summer 2019 study carried out by a team from the University of Minnesota on several networks operating around 30GHz found smartphones could achieve download rates as high as 2Gbit/s but could easily reduce to 500Mbit/s even if the link was held. In practice, the experiment showed handoffs from 5G to 4G occurring more than 30 times in fewer than 8 minutes.

Though researchers have found millimetre-wave signals benefit as much from reflections as lower frequencies and so do not absolutely need line-of-sight to work, they are easily blocked by the user's own body with no useful reflection paths to avoid the obstacles.

“The focus globally at the moment for 5G is in the 3GHz to 5GHz range. I'm a believer in millimetre-wave even though it's hard to do. But the bandwidths are there,” Eddy said. “We do believe millimetre-wave will happen in 5G but it's not easy.”

A move up the frequency range into the terahertz promises even higher peak data rates albeit with more of the problems encountered with millimetre-wave transmissions. They suffer more strongly from strong atmospheric absorption, especially when they approach the infrared range. The issues with forming directional beams that can stay in contact with a mobile user for more than a second at a time may shift the attention to backhaul: the fixed nature of the terminals will make it easier to organise line-of-sight transmission or reflectors that can ensure uninterrupted signalling.

Unfortunately, water absorption becomes a major problem over longer distances, so the terahertz links may only be useful for placing small cells along a street fed by a nearby fibre-enabled macro-base station.

Smart surfaces

In the small-cell environment, organising reflections to cope with situations where line-of-sight transmission is impossible may well be one of the identifying features of 6G.

Stephen Douglas, head of 5G strategy at Spirent, said there is now a great deal of interest in smart surfaces to amplify and redirect radio signals around the environment. "That could be a real game changer," he said. Potentially, buildings could be armed with arrays of reflectors, both inside and out that serve to redirect millimetre and terahertz waves around the

These boosted areas, as CEA-Leti 6G programme director Emilio Calvanese Strinati calls them, will likely involve the installation of passive arrays covered with meta-materials or active panels that have steerable reflectors. His R&D group is part of an EU-funded project called RISE-6G that will trial boosted areas in two scenarios that offer a "wireless environment as a service".

One is a public network in a rail station in Rennes, France; the other a factory owned by Fiat-Chrysler Automobiles.

The factory scenario provides the opportunity to gauge how well the positioning information needed to guide robots around the factory can help direct the reflectors and phased-array antennas used for transmission.

The plan is to experiment with communications at 26GHz and below 6GHz while trialling centimetre-accurate localisation using 70GHz and 130GHz carriers and gauging how few of the smart reflectors will be needed in practice.

In Rennes, the reflectors will be used to help direct 26GHz and 70GHz channels to improve coverage in shops around the building and to rest areas that will advertise quick video download and gaming services. One issue is potentially cost with these arrays if they need to be large to provide effective coverage. Although simulations by NTT Docomo suggest more than 60 per cent of mobile users in a shopping mall could obtain data rates better than 100Gbit/s, it may in practice be simpler to limit terahertz coverage to tightly defined spaces where people are encouraged to perform large downloads, with conventional microwave transmissions used for other traffic.

A similar delay to the one for millimetre-wave in taking place in 5G may face designs targeting 100GHz-plus frequencies as 6G standards emerge. “Millimetre wave is already hard. You are going to have to reinvent everything for terahertz,” said Eddy.

The higher-frequency transmissions will call for research into novel materials that can surpass the performance of conventional silicon and the various passive elements used today. Imec, for example, has kicked off a 6G-focused research programme to look at the combination of compound semiconductors such as indium phosphide, an area that some groups at Intel and Qinetiq have been looking at for well over a decade.

There have been some successes so far. The EU Taranto project, of which Imec is a member, has produced a compact 140GHz radio module that enables single-link data rates up to 80Gbit/s but further work will be needed to boost peak frequencies and cut manufacturing costs.

In August, LG and the Fraunhofer Institute demonstrated a 150GHz-plus power amplifier transmitting signals over a distance of a hundred metres.

Given the uncertainties over ultra-high frequencies, 6G will likely be characterised by a smorgasbord of transmission technologies that are coupled together, providing further impetus to the push for artificial intelligence (AI) in the network.

Systems will need to employ smart algorithms to support rapid switching between RF modes as users move around and machine learning may prove to be the best way of determining how and when to make each switch. Douglas says that will introduce new issues. “How do you test a network that’s starting to think and self-drive? There are also questions over the governance of AI.”

However, the industry has roughly a decade to work out how to achieve it.