When 4G was young, all the talk was of a spectrum famine, which loomed if mobile data traffic rose as exponentially as expected. Every vendor included a slide showing the widening gulf between traffic levels and available licensed spectrum (not to mention new revenue).
In fact, the story was over-simplified. The growth in cellular data traffic, while rapid, was slower than many had expected, outside of a few very mobile-centric markets like South Korea. Very high-bandwidth applications such as enhanced gaming were still largely used with fixed connections, and the stress on wireless networks was mitigated by the increasing reliance on WiFi, and the ability to move flexibly between cellular and WiF from a single handset (often encouraged by operators to reduce their need to expand network capacity).
The best evidence that spectrum shortage fears were overblown lies in the 2.5 GHz and 2.6 GHz bands, which were widely allocated for 5G, and in many markets – especially in Europe – remain chronically underused. Those bands also highlight two other important features of the spectrum landscape of a decade ago. One, the main approach to meeting high capacity demands was for MNOs to chase wide, contiguous slices of virgin spectrum, usually paired (FDD); but two, the business case for building LTE in midband spectrum remained challenging – the expense of building the relatively large numbers of base stations required to achieve broad coverage in 2.6 GHz was rarely justified by the increased revenue potential in competitive consumer markets.
The 5G landscape is very different, partly because of the technology itself, and partly because of changes in regulatory policy and operator business models.
Cellular traffic will continue to rise, though not as dramatically as envisaged 10 years ago. However, there will be far higher expectations of the way mobile data is handled, as 5G reaches out to new user experience and new enterprise markets, the only way to generate sufficient new revenue to make the business case work. These new services will require various combinations of very high data rates, assured quality of service, very low latency and critical availability.
To support all these requirements, the dream of a single chunk of unused spectrum will be abandoned. Yes, midband spectrum, mainly in the 3.4-4.2 GHz range, will come into its own. In some markets this is unoccupied or easily vacated; it has plentiful capacity; and 5G-era technologies like Massive MIMO mean it can achieve similar or better coverage than 2.5 GHz, improving the cost:revenue ratio.
But it remains a relatively short-range band, suited to dense capacity zones such as cities and in some markets, notably the USA, it has incumbent users, forcing the regulator to think of creative schemes to make it usable by MNOs – the multi-tiered CBRS sharing system in 3550-3700 MHz; the incentive auction for satellite spectrum in the 3700-3980 MHz portion of the C-band. There are even discussions about allocating lower C-band spectrum in 3100-3550 MHz for wireless broadband in future.
Even where 3.5 GHz is clean, it will not achieve full coverage, and in most cases, operators will need to use a far more complex combination of bands than they did in 4G. For LTE, most deployed in a sub-GHz band for broad coverage and a slightly higher band, often 1.8 GHz, to add capacity. Carrier aggregation enabled spectrum in several bands, including non-contiguous, to be bound together, and the technique was extended to unlicensed spectrum with LTE Licensed Assisted Access.
In 5G, to deliver the full variety of use cases and network behaviors demanded, many operators will need to stitch together a richer patchwork of airwaves and use different spectrum in a flexible, even dynamic way to maximize utilization, efficiency and performance. T-Mobile USA’s ‘layer cake’ approach will become increasingly common as operators gather together as many pieces of spectrum as possible, and as tools evolve to make it feasible to build networks using what would formerly have been regarded as a mishmash of sub-optimal airwaves.
Those with 3.5 GHz licences are starting conventionally enough, building out capacity in that spectrum, while using 4G for coverage; then typically acquiring a new sub-GHz licence to extend 5G coverage and migrate to 5G Standalone. But some MNOs, notably in the USA, but also other markets will not have these options readily available. And if operators want to support enterprise, industrial and low latency applications as well as mobile broadband, they will need a wider variety. Millimeter wave will start to enter the picture for fixed wireless (in the USA, Australia, South Korea and elsewhere) and for indoor or hotspot use cases; deep indoor penetration or private networks and slices may need dedicated airwaves; as 2G and 3G networks are switched off, those bands will be in play.
The standards support a new approach to spectrum. 5G New Radio is the first to be developed first for TDD (unpaired) spectrum, which provides greater flexibility in allocating capacity to different tasks (notably the uplink). And it supports aggregation across far larger numbers of bands, though unlicensed will not be included until the forthcoming standards release.
These techniques – together with dynamic spectrum sharing (DSS) and emerging methods to share airwaves using spectrum access systems (SAS), as in CBRS – will all help operators to adopt more nuanced approaches to buying and deploying spectrum, than in the largely unnecessary capacity race of early 4G. But greater flexibility, and ability to customize spectrum to specific use cases, can work against the established MNOs too. They make it easier for a challenger with a non-ideal assortment of assets to build a viable network, or for a new entrant or private network operator to harness shared spectrum to support 5G services. In both cases, the spectrum assets enshrined in long term, exclusive licences are likely to be devalued, while the real power and differentiation in a network model will move up the orchestration, slicing or dynamic resource management layers.