Happy New Year! With the AWS-3 auction in the US reaching round 247 as I write, and $44bn already committed, new spectrum is still in the headlines in 2015. The NTIA continues in its mission to release US Federal spectrum holdings , reading a bit like the UK regulator Ofcom in a recent blog about UK spectrum. Yes, we’re looking to free up some spectrum. No we haven’t done it yet, but we promise we will real soon. The new bit, I think, is this “spectrum sharing” angle, which ironically too has a retro feel – harking back to the 1800MHz (DECT Guard Band) shared spectrum auction in the UK in 2006 the difference being, of course, the scale of the spectrum under discussion. Where low power GSM shared out 3.3MHz FDD, the NTIA and Ofcom are talking about 100MHz TDD. Definitely a game changer. The question will be, how will the regulators pass it on? Fully unlicensed (which is the default position, I expect, and should increase the appeal of LTE-U), or a shared license auction (like LP-GSM, and enabler for local private networks such as our customers PMN, Sirran and so on) or a fully exclusive license (like the AWS-3 auction that’s going now). Given that the AWS-3 auction is raising $40+bn for the US treasury, it must be very tempting to go the full license route. The higher frequency and the TDD nature of the 3.5GHz band makes it quite a bit less valuable than the AWS band, and certainly much less valuable than the 600MHz band that’s going to get licensed in 2016. Even so, spectrum is spectrum, and if Google can convince chip vendors to create silicon for the new band and enable it in the software, it’s going to attract some value, as the Register hooks into here.
But spectrum sharing? I hear this a lot in the more academic end of our industry. Of course, unlicensed spectrum is intrinsically shared, and works ok if it’s not congested. What people are starting to talk about is sharing by time-of-day, by geography, by usage, between multiple licensees of the same block. I’m pretty sceptical about this model, I must say. The gains depend on highly uncorrelated usage patterns between the user groups, and the reality is, most spectrum usage correlates with human activity, and human activity is a fairly regular diurnal pattern with peaks at the same times and troughs at the same times. Just like the bank who’s ready to loan you an umbrella when it’s not raining, spectrum sharing on this model offers you MHz when you need kHz, and limits you to kHz when you need MHz.
There are small examples of time decorrelated use – when the phone or the network wants to do something independent of the human user – for instance App and OS updates can go on happily overnight with no human intervention required. But this isn’t really going to move the needle. What may bring supply into line with demand in spectrum usage is surge pricing. Just as Uber charges people more during busy periods, so mobile tariffs may increase to encourage social media storms to happen more slowly, with lower peaks. But you can imagine the outrage – as Uber noticed when the algorithm spotted huge demand in Sydney one morning and applied a massive surge premium, not realising the underlying cause was a hostage situation in a business district cafe. If people are getting gouged for tweeting politically significant events, it’s not going to win any PR awards.
No matter which way you look at it, the only decorrelation of significance is one of space. Data usage happens in clusters. Clusters are best served by small cells that can re-use the spectrum over and over again from cluster to cluster. If the economics of licensed radio are going to be able to serve this clustered growth, then spectrum needs to be freed in a way that lets technology providers take advantage of it economically. The unspoken advantage of WiFi is that it only has to serve two bands worldwide. If we continue to license spectrum band-by-band piecemeal across the globe, we’re going to incrementally and in the end, fatally burden the equipment costs of licensed radios pointlessly. At the most basic level, we’re making new spectrum too expensive to access.
But as the days start to lengthen this January, thoughts turn to the coming spring. I’m starting to wonder if the 3.5GHz band is the nearest thing we have in the licensed world to a global band. It certainly looks like the US and Europe are lining up behind it. If Asia comes too, then 3.5GHz could become the licensed radio counterweight to 5GHz in unlicensed operation. If a licensing regime can be established worldwide that allows free, yet harmonised access to this spectrum, then we have the best of all worlds – freely licensed radio access without the overhead of market fragmentation, or the technology overhead of listen-before-talk, that can truly compete with its unlicensed cousin. It looks like a while before the handset guys will be ready, but spectrum is spectrum. Where it goes, people follow.