5G: Just an Expensive New Interface?
At MWC this year, 5G was inevitably the focal point of many discussions, and there are a growing number of conferences on the topic: Indeed, I spoke at one last week. That said, most wireless experts recognize that 5G remains some way off and, if we're to get excited about it, we first ought to have some idea of what it may look like.
But I'm increasingly concerned that 5G could shape up to be a stunningly expensive fiasco.
5G will be more subtle, strategic and complex than previous cellular technology generations in a myriad of ways. The "More of the same!" commentary misses the opportunity to make 5G genuinely interesting and worthwhile.
While there is a flurry of debate over what exactly 5G will be, and how we achieve it, one of the few things everyone agrees on is the schedule. The key milestone dates are designed to coincide with the South Korea Winter Olympics 2018 and Japan Summer Olympics 2020.
These dates seem feasible given the traditional eight-to-ten year cycle of major cellular releases, even more so given that the first 3rd Generation Partnership Project (3GPP) 5G oriented study on New Services and Markets Technology Enablers was approved last month. With the final stage of this study due to be completed by March 2016, 5G looks set to be ready in time for these major events -- but we're still not quite sure what it will look like,
In terms of what people hope 5G will deliver, the headline aspirations include:
- 1,000-fold gains in capacity density
- Scalability to connect at least 100 billion devices
- 10 Gb/s per individual user
- 1ms latency and response times
- Incredible device density: 1 device/m^2 across huge areas
There's also significant consensus on what users need from 5G:
- Perceived improvement in user performance, particularly in becoming more consistent and ubiquitous
- Increased support for new services like Internet of Things
- Lower latency and cost, increased robustness
- Energy efficient services for devices
- The ability to launch new services effectively and efficiently
But the reality is that 5G cannot possibly deliver all of these things at the same time. In fact, there is no way that a single technology could ever meet such a disparate set of needs on its own, for two key reasons.
First of all, there's the small matter of the laws of physics. You simply can't have long range, high bandwidth and low power. Rather, you must select two of the three –- and appreciate that each permutation would require a different air interface, or at least a much more flexible, dynamically reconfigurable air interface.
Secondly, we have almost hit the Shannon Limit, meaning that there are few more gains we can make on the radio link itself. Wider channels may give us more speed, and more MIMO will give us greater efficiency, but these are not new generations of technology.
Ideas like better FEC and clever approaches to modulation, massive MIMO interference cancellation, power efficiency improvements and (my favorite) full duplex are great -- but those are not dramatic, revolutionary upgrades. Rather, they could be introduced as incremental upgrades to any standard, just as GSM evolved from GMSK to QAM and Release 7 added MIMO to UMTS.
There are plenty of areas for improvement, but none that necessitate a "rip up and replace" new generation.
In fact, most of the benefits will come from better densification, virtualization, coordination and cooperation. More cells; better, smarter networks; distributed MIMO; coordinated scheduling; macro-diversity -- as Paul Jacobs said a few years ago, "It's topology, not technology."
Right now, we have a long list of desired features, and the proposals for a new air interface will do all of them poorly. It is rather like going to the hardware store to buy a screwdriver, a hammer and a saw -- and returning with a very expensive Swiss Army penknife "because it can do all of those things." Well, sort of... but not as well as the individual tools would do them, and the fact that you now have a doohickey for removing stones from horses' hooves will not make for a better cellular network.
There is also the problem of money -- and at the end of the day, it is money that matters.
By 2020 the global investment in LTE will amount to something like $1 trillion, with physical infrastructure, spectrum licenses, software and services. LTE will have had many significant updates, and be on Release 16. Why would the CFO of an MNO want to obsolete that? What is the financial case for an upgrade? Why can't we sweat that asset, rather than scrapping it?
I fear we are in danger of creating a very expensive white elephant: a compromise that does not actually deliver any compelling advantages or business case.
Instead, I would argue we need to think of 5G as an architecture and not just a new, slightly shinier air interface.
Just as the Internet is not really one uber-network but is instead a "network of networks" that allows many different connections to work together, we should think of 5G as a framework: a structure that supports several different requirements in an elegant, efficient, secure, scalable way. One that allows different air interfaces for different needs, but managed in a coherent, cost-effective way. Different protocols, some optimized for resilience, some for ultra-low-power, others for high data rate -- but provisioned and managed coherently.
This could leverage the investment in LTE while still enabling new services and new performance improvements -- and crucially, a scalable mechanism for simply and flexibly taking advantage of the huge variety of spectrum out there in an efficient and scalable way.
That is much more complex, but much more important, than simply a "new modulation method." And a lot more commercially attractive than the world's most expensive Swiss Army penknife that can do everything but does nothing well.
— Rupert Baines, CMO, Technology Strategy & Marketing, Real Wireless