Meeting the need for spectrum

<b style="color:#900;">commentary</b> Efficiency and new towers alone can't satisfy growing demands placed on mobile networks. Inventor Richard Bennett says spectrum must be freed up from legacy uses such as broadcast TV and obsolete government systems, and soon.

Richard Bennett Inventor and network engineer
Richard Bennett is one of the inventors of Ethernet switching and the Wi-Fi system architecture. A long-time contributor to Internet and Open Systems Interconnection standards and protocols, he currently works at the Information Technology and Innovation Foundation as a senior research fellow.
Richard Bennett
4 min read

Editors' note: This is a guest column. See Richard Bennett's bio below.

The National Broadband Plan delivered to Congress by the Federal Communications Commission last year recommended the licensing of 500MHz of new spectrum to mobile broadband providers, including 120MHz currently held by local TV broadcasters. The nature of the incentive auction that would enable willing broadcasters to transfer this spectrum to mobile operators (for a portion of auction proceeds) is under debate in Washington, as are means of transferring additional spectrum currently held by government agencies.

Wheels move slowly in Washington, and mobile providers can't wait for the outcome of policy debates that may take years to reach successful resolution. AT&T proposes to meet its short-term spectrum needs by purchasing 12MHz from Qualcomm and to purchase T-Mobile's spectrum assets, along with the rest of the company.

Both of these moves are controversial, although the issues with the Qualcomm deal are much less serious than critics maintain (PDF). Those who oppose AT&T's efforts to increase its spectrum holdings generally insist that the demands for spectrum can be satisfied by increased technical efficiency and the construction of new radio towers. However, as we assess the virtues or dangers of the merger of AT&T and T-Mobile, it's unrealistic to suggest that efficiency and tower construction can meet the demand for spectrum alone.

Who's hogging the airwaves?
It's disingenuous to label any of the mobile operators a slothful spectrum hog. Unlike other wireless networks, mobile networks are the products of ongoing deep investment in technology, deployment, and spectrum: The mobile networks have advanced through four generations of technology in 10 years, while the broadcast TV and radio networks are only one generation removed from their original form.

It's certainly not the case that any one of the mobile operators today is wasting the spectrum they've purchased through auction. A recent report by Bernstein Research shows that the larger the network, the greater its efficiency. The greatest oversupply on a per-user basis is currently found in the combined Sprint-Clearwire network that floundered on an unfortunate experiment with WiMax.

Current spectrum holdings
Source: Bernstein Research, used by permission from Craig Moffett

Forecasting spectrum demand
The linchpin of the efficiency argument is the simple observation that spectrum demand depends on two factors: one, user demand for data capacity; and two, network efficiency in terms of data capacity per unit of spectrum. These two factors don't progress at the same rate. The rule of thumb for spectrum efficiency is Cooper's Law, the prediction that spectral efficiency doubles every 30 months. Demand for capacity moves several times faster than this, driven by the replacement of dumb cell phones with smartphones and by the eventual desire of smartphone users to do more things from more places.

The radical increase in spectrum use since the advent of the iPhone has been driven by replacement, but we can reasonably predict that demand will continue to increase even after all the legacy phones are decommissioned. We've seen this dynamic on the residential broadband networks that are now pushing 100Mbps per user, a dizzying increase from the megabit-per-user capacity they had when deployed only a decade ago. How far the demand ultimately goes is anyone's guess: several countries are building Gigabit residential networks on the assumption that future application bundles will require cavernous capacity.

Past growth in network utilization is a poor guide to future demand, because utilization is always capped by capacity, so most exercises in demand forecasting are ultimately backward-looking exercises in measuring network investment. There's good reason to believe that mobile app developers can find ways to use dozens of megabits per smartphone user, at least for short periods of time, as we become more accustomed to mobile video, augmented reality, and mobile conferencing. Mobile network connections aren't shared to the same extent as residential broadband, but the investment required to meet mobile demand is greater on a per-megabit basis than it is for wireline.

Meeting the demand without breaking the bank
The efficiency argument tends to overlook the fact that all means of increasing mobile network capacity aren't equally costly. Because Cooper's Law moves slower than Moore's Law, technology alone can't solve the mobile capacity problem. Critics recognize this, for the most part, and argue that tower construction can relieve the remaining deficit. Construction is a bricks-and-mortar issue that doesn't leverage technology at all. A cell tower costs roughly $150,000 to build, and splitting a cell into a group of smaller cells generally requires three new ones. Splitting cells across the nation could easily cost large network operators $40 billion apiece, roughly twice as much as the cost of their current networks. This money can only come from consumers.

Expense isn't the only issue with new tower construction: Each new tower requires approval from local zoning boards that move even more slowly than Washington, flouting the FCC's tower siting "shot clock" rule. Thousands of tower siting applications currently await approval.

Freeing up spectrum for use by existing towers is a much more economically efficient path. As more spectrum is available per tower, network operators can bond channels and focus their beams on smaller sectors of the geography around each tower. These means enable users to transition to leading-edge applications without breaking the piggybank. Efficiency is important, but there's no substitute for spectrum.

The solution
Maintaining adequate competition among mobile network operators depends on freeing up spectrum currently assigned to legacy uses such as broadcast TV and obsolete government systems and making it available for high-demand applications. The appropriate mechanisms for doing this are spectrum auctions, secondary markets, secondary uses, and flexible policies.

The U.S. lags the rest of the world in LTE spectrum allocations because we haven't moved rapidly enough at reassigning spectrum. This is a problem that sound policy and decisive government action can address.