Debate continues to rage about the 5G air interface – will the new generation of mobile networks require an entirely new air interface, or several; or will it rely on evolutions of current OFDM technologies; or be primarily an umbrella framework tying different options together seamlessly and flexibly according to use case? The choices, and the companies supporting them, will become clearer as the standards process gets into gear, but it is a brave company which tries to propose such a fundamental element of 5G, if it is not one of the inner circle of giants.
That is what start-up Cohere is aiming to do, however, and is likely to find itself up against Qualcomm (the most public about its air interface ideas so far), as well as the major OEMs, Nokia, Ericsson and Huawei.
Many believe that an evolution of OFDM – which underpins LTE and WiFi – will be retained for sub-6 GHz bands and RAN technologies targeted at coverage and throughput. This will not be a new start but will improve on current implementations in areas like energy consumption, latency, spectral efficiency – in other words, continuing the work which is already ongoing in the 4G era.
In two other areas, a new air interface will be required, many argue – for sub-6 GHz Internet of Things applications, and for high frequency bands. As Alan Carlson, European head of InterDigital, a major contributor to 3GPP standards, wrote recently: “Two licensed band cellular definitions are expected below 6 GHz. The second will be especially tailored for IoT support. It is less likely in my opinion that this will be based on OFDM. OFDM is great for video, but in the IoT world of trillions of randomly occurring access events, its strict synchronization needs render a severe handicap.”
Qualcomm agrees with this assessment, and it pitching its non-orthogonal RSMA (resource spread multiple access) technology for IoT transmissions with low data rate and signalling, but high requirement for reliability – while sticking to OFDM elsewhere. A new multiplexing technique would allow traffic requiring very low latency to take priority automatically and to use RSMA. This technology uses time and frequency spreading and overlaps users in a way that aims to improve network efficiency and power consumption. It can support mobility and downlink meshing, as well as network-assisted mesh on the uplink.
John Smee – a senior director of engineering at Qualcomm Research and head of the firm’s 5G technical work – told EETimes: “We are designing a network for a future we don’t know. It’s like our work on 4G, 10 years ago when people carried flip phones to the meetings and thought about a network capable of downloading video.”
He said: “It’s not a new waveform but an expansion of OFDM for wider use cases with a family of numerologies with scaled tone spacing. No one numerology fits all use cases, but a family of three or four provides a checkerboard of design parameters.” This then allows systems be optimized for different cell sizes and frequency bands, a step towards the infinite flexibility which most envisage for a 5G network which will have to adapt to many different applications and business models, some as-yet undefined.
Qualcomm summarizes this approach as “a unified air interface with optimized OFDM-based waveforms and multiple access, with a flexible framework that can scale from low spectrum bands to mmWave, from macro deployments to local hotspots, and will support licensed, unlicensed, and shared licensed spectrum from the beginning … For targeted use cases such as sporadic uplink traffic from battery-powered IoT sensors, the use of non-orthogonal RSMA helps further reduce device complexity.”
While this suggests that the chip giant will favor OFDM for mmWave too, others are less sure. Carlson says: “The third high capacity solution will be provided in the so-called mmWave bands that start around 30 GHz. This is probably where the most industry debate is going on at this time. This radio may or may not be based on OFDM. A single carrier-based approach is more likely to be selected. Simply stated, in moving to higher frequencies and narrower beams, many of the benefits of OFDM such as MIMO integration and higher order QAM schemes diminish and further PAPR (peak to average power ratio) becomes a handicap. So, it is fair to expect something new here as well.”
Meanwhile, Huawei’s proposal focuses on another non-orthogonal access technology based on Sparse Code Multiple Access (SCMA), and also on a Filtered-OFDM solution. And Nokia is promoting a universal filter band approach which uses guard bands to protect adjacent channels and so make better use of them.
Cohere is also moving away from OFDM and proposing OTFS (orthogonal time frequency and space), which can nevertheless coexist with OFDM to support smooth migration from current networks. The company is led by CEO Shlomo Rakib and CTO Ronny Hadani, the latter a leader in OTFS research, formerly at University of Texas at Austin. Rakib has co-founded a string of companies – Terayon Communication Systems, which went public in 1998 and was later acquired by Motorola, as well as Gainspeed and Novafora. He was also a key inventor of S-CDMA technology.
Cohere says OTFS creates a two-dimensional view of the delay and signal fading of a wireless channel. It spreads information across time and frequency, which allows signals to benefit from diversity in the channel while also penetrating through concrete and glass, and to achieve high levels of immunity to fading, multipath and other signal impairments. In its early tests it trialled its technology among tall buildings and mountains and in moving vehicles, and said that the OTFS radios never faded, even over distances above four kilometres, as OFDM ones would have done.
A minimum of 4bps/Hz was achieved in 10 MHz of spectrum in the 3.5 GHz band, driving speeds of 120Mbps to 320Mbps. The company’s initial trials have used currently available 2×2 or 4×4 MIMO antenna arrays but its roadmap is to produce Massive MIMO radios with 64×64 arrays, and to move up to the millimeter wave bands.
“We solved some very fundamental problems around imperfect channel information and timing references that the wireless community has been struggling with forever,” Rakib told EETimes. “The good news is this is a thin layer above OFDM-based cellular silicon today, requiring small transforms in the modulation process that should require less than 10% in addition to today’s silicon.”
Phil Marshall, chief research officer at Tolaga Research, said OTFS is “not an entirely new modulation scheme as it sits on top of OFDM. It’s a scheme for diversity that adds to MIMO with an equalization technique.”
After a kick-off workshop, at which some of the companies outlined their ideas, the 3GPP 5G working group will hold its first meeting this month, to decide on details of use case requirements, and then will go on to define channel models for each use case. After that, it will take proposals for air interfaces, and presumably the usual horse trading process will ensue – as recently seen in NB-IoT – to try to come to a consensus solution. The best opportunity for smaller innovators like Cohere is likely to be to secure a big name supporter to fight its case for inclusion, though like many mmWave pioneers, it is developing its technology anyway, for particular applications, such as fiber extension, backhaul and campus small cell networks, which do not need to wait for mobile standards.
Rakib says the company has almost finished the architectural definition of its ASIC, probably to be made in 2018 to support gigabit speed wireless connections via a low cost home or office gateway. He added: “Today we are a systems company, but if we become part of the 5G standard, we will have to add on licensing models, and eventually we might have to build silicon.”