Why Spectrum Swaps Serve the Public Interest

A reporter asked me to explain what’s up with the spectrum swap that AT&T and T-Mobile have asked the FCC to let them do. The benefits are pretty simple and straightforward, but practically nothing happens with networks these days that somebody doesn’t complain about. The merger of cable companies Charter, Time Warner Cable, and Brighthouse seems pretty straightforward, for example. The combined firm, “New Charter”, would be about the same size as Comcast, which would seem to bode well for competition and economies of scale.

The merger isn’t nearly as dubious as the Comcast/Time Warner Cable would have been, and Charter has offered to abide by a number of conditions that net neutrality advocates have called for in the past, such as flat-rate pricing for unlimited data volume. Despite these proposed concessions, the FCC has received complaints from the National Association of Broadcasters, Dish Network, and populist crusaders Free Press and Public Knowledge. The complaints appear to be coordinated, which is par for the course in commercial disputes that can be spun as public interest issues.

Similarly, the transition from copper wire telecom networks to fiber-optic networks is an undeniably good thing that engineers have been pushing since the late 1970s. Nevertheless, there are still critics who insist that copper wire is just fine and there’s no need to replace it with fiber. This is nonsense, of course, but it’s an amusing story that can still get some people riled up, like the fantasy that vaccines cause autism (they don’t). Just this week, there were news reports claiming the transition is just one gigantic scam to raise telephone bills. Companies have to ignore this sort of thing and stay the course with their transition plans.

Hysterical claims move people, and when the subject is technology – something that few people understand in depth – it’s not difficult to mobilize campaigns across the Internet. Remember the Internet uprising against the 2012 anti-piracy bills that were alleged to “break the Internet”? The bills were withdrawn, but other countries that have mandated the blocking of piracy-oriented sites have seen a decline in piracy with no loss of utility in the Internet. IFPI, a UK-based pro-intellectual property organization, reports that blocking major piracy sites reduces piracy traffic from blocked and unblocked sites as well:

Since the Pirate Bay and numerous other sites have been blocked in the UK there has been a 45 per cent decline (from 20.4m in April 2012 to 11.2m in April 2014) in visitors from the UK to all BitTorrent sites, whether blocked by ISPs or not.

UK’s Internet is every bit as free as the US Internet is where lawful sites are concerned. So the predictions of Internet breakage were, shall we say, a bit premature.

So returning to the reporter’s question, why would AT&T and T-Mobile want to swap spectrum and how does this affect the public? The short answer is it will make our mobile broadband connections run two to 10 times faster in the markets where the swap takes place. Mobile broadband spectrum has typically been licensed in narrow channels, 5 to 10 megahertz (MHz) wide. When we used mobile phones for making phone calls, these allocations looked very large because each phone call only consumed 50 kilohertz (KHz) or so of bandwidth. Hence, a tower with a 5 MHz channel could support 100 simultaneous calls.

But a 5 MHz channel is very limited when it comes to data. With LTE, the target for downlink spectral efficiency is 5 bps per hertz, so 5 MHz x 5 bps/Hz means the 5 MHz channel can support 25 Mbps at a time. If it’s lightly loaded – fewer than 20 active users – each user sees performance close to the peak download rate of 25 Mbps, but beyond that it begins to degrade and people start to experience high rates of packet loss, delay, and dropped web connections. So the carrier can improve performance right away from moving from 5 MHz channels to 10 MHz and even 25 MHz channels. This is what the spectrum swap allows AT&T and T-Mobile to do.

You may wonder whether it’s better to have a single 25 MHz channel than to have 5 channels that are each 5 MHz wide. The answer is yes, for two reasons: channel boundaries are dead spots for spectrum, in the same way that easements are dead spots on real property. Wi-Fi channels, for example, are 25 MHz wide but only 22 MHz is usable because there needs to be a small buffer between channels to prevent cross-channel interference. Signals spread in transmission.

If we combine two adjacent Wi-Fi channels into one, as IEEE 802.11n allows, we end up with 47 MHz of usable spectrum where we only had 22 MHz before; this is because we’ve put the buffer zone between the two channels to work. So eliminating boundary losses is one good reason to “bond” channels.

There is another technique known as “carrier aggregation” that makes two non-adjacent channels behave as one, but it’s less efficient because it doesn’t overcome boundary losses. It’s a logical form of channel bonding that would net 22 + 22 or 44 MHz in the Wi-Fi example. That’s a best-case result, because in practice aggregated channel pairs perform no better than twice as well as the poorer of the two channels. If one channel sees a lot of interference, this drags down the partner channel as well.

We can also do more efficient statistical sharing with one fat channel than we can with several skinny ones. In practice, this is the difference between the Internet and plain old telephone service. POTS uses 64 Kbps channels, while broadband channels can be multiple Gbps. When a POTS channel is unused, its capacity goes to waste, but when a user of a shared broadband connection is inactive, everyone else’s transmissions are just a little bit faster.  Sharing 100 Mbps with 10 neighbors is much better than having 10 Mbps for yourself because most of the time you’ll get more than 10 Mbps when you want it on shared connection, but you’ll never get it on the private connection.

So there you have it, sports fans: sharing fat cellular channels is better for everyone than sitting in skinnytown channel slow lanes. Just don’t overshare.

You might believe that adding more spectrum to the cellular network isn’t necessary because you’ve heard something about small cells being a better way to increase performance than adding spectrum to big cells. That’s not really true in the general case because 5 MHz imposes a 25 Mbps limit on small and large cells alike, and we’d all benefit from peak speeds much higher than that. Once we have adequate spectrum to support the fastest phones running the most demanding apps in the most places, we can optimize the network by adding smaller cells in congested areas.

But small cells are not a substitute for adequate spectrum any more than lots of skinny channels are anywhere near as good as a few fat ones.