Sharing Government Spectrum
This week I’ve had the pleasure of taking part in two off-the-record discussions about spectrum policy as well as two media inteviews that were very much on the record. While I can’t write about the comments I heard at the off-the-record briefings, I can repeat what I said and why I said it across the board, so let’s get to it.
Sharing Government Spectrum
One of the burning questions is what to do about government spectrum that’s being used in a sub-optimal way. If government agencies were operated like businesses, they would be able to deal with this problem in a rational way: They would upgrade their equipment to reduce their spectrum footprint, and they would sell or lease their remaining spectrum allocation to cover the cost of the upgrades and make capital available for investments or dividends.
If a business with more licensed spectrum available than it needed had some vague ideas for possible future uses, it would most likely lease its excess on a short-term basis so as not to lose the opportunity to deploy future systems when and if their plans took concrete shape.
But government agencies don’t operate like businesses and aren’t free to make deals with commercial firms that trade spectrum for upgrades. That’s been tried and it’s simply not lawful. So we have convoluted spectrum sharing schemes such as those laid out in the 2012 PCAST report on spectrum sharing, “Realizing the Full Potential of Government-Held Spectrum to Spur Economic Growth.” The PCAST report proposed novel spectrum sharing models that had not been used before the report was published and still haven’t been used four years later.
Lazy Spectrum Databases
The PCAST report relies on Spectrum Access System (SAS) administrators and Environmental Sensing Capability (ESC) operators that were introduced in the discussion about TV White Spaces ten years ago. While several firms have applied to the FCC to operate these services, there’s a stark shortage of network services that want to use the databases and comply with the limitations this approach entails.
SAS and ESC databases are updated on a daily basis, and perhaps on an hourly basis, but they don’t reflect network conditions in real time. So they can only be useful in the rare circumstance where the primary system – the government agency – is either fully on or fully off for hours at a time. Normal digital radio systems are fully on or fully off for fractions of a second at a time, so this SAS “lazy monitoring” is about as efficient as the DMV.
The databases could be useful for sharing long-lasting information between spectrum users that share reliance in the digital domain; this means systems that can coexist by sharing technical parameters that allow them to transmit at the same time without interfering with each other. CDMA, for example, uses logic and math to convert a desired sequence of bits to a sequence of codes that can be transmitted at the same time as other systems are transmitting without causing interference.
Network Parameter Sharing
But CDMA only works when all the transmitters are coding their transmissions differently than the others. Wikipedia likens CDMA to people speaking different languages at the same time. If I’m talking to you in English while someone close by is speaking Russian to someone else, you can understand me and the Russian speakers can understand each other because we tune into our native languages and easily ignore the others.
So we could design an SAS database that authorized one network to use codes 0 – 127 and the other 128-256, for example. This would ensure that each network can use half the total bandwidth and only half as long as their coding systems were compatible. If one of the networks only operates for an hour one day a week, it could grab the codes it needed when it needed them and leave them to the other network (or networks) the rest of the time.
The downside of this form of sharing, as with all other forms of spectrum sharing, is that upgrades would need to be coordinated. Since government isn’t good at upgrading its systems, this brings us back to the original problem. The way out of that dilemma is to make on of the spectrum systems the network operator and the other a guest, like a Mobile Virtual Network Operator (MVNO) is on today’s large mobile networks.
The downside of that approach is that it assumes the MVNO – the government system – has needs that can be satisfied by the technology used by mobile networks. In some cases they do, and in others they don’t. But it would certainly work when government is using spectrum to transmit and receive bits, which is at least a large share of government spectrum systems.
Every spectrum system includes a function that coordinates spectrum access with other potential and actual users. In some networks spectrum coordination is centralized, and in others it’s distributed. Cellular networks centralize the function in each cell and between overlapping cells. Wi-Fi distributes spectrum coordination across each potential Wi-Fi user.
There are some upsides to the distributed approach: there’s no need to “ask permission” before transmitting, which pleases those with idealistic views of network organization. It’s also resilient, because no single equipment failure renders the network unusable (except the condition where a network fails in a mode that features non-stop transmission.) But it has a huge downside: because Wi-Fi nodes don’t share the intention to transmit with others most of the time (they can say how long they expect each transmission to last, but that’s a last-instant thing with limited value.)
Soviet Spectrum Sharing
So distributed coordinators have an information deficit, not unlike the central planners of the Soviet economy, who had very poor knowledge about what producers could create and what consumers wanted. Centralized spectrum coordinators take advantage of their location in the functional organization of a network to allocate spectrum fairly and efficiently.
They can’t make long-term decisions any better than the Soviet planners could, but they can make the trains run on time. And that’s the job that needs to be done where government and industry are sharing resources. A centralized spectrum coordinator has better information on which to make its decisions than does the distributed planner who only knows what it wants and who is using the network at any given instant.
Sharing Government Spectrum the Right Way
So network-resident coordinators are both more fair and more efficient than the distributed coordinators. So one might wonder why the PCAST report and its fans favor an inefficient and unfair system. I don’t think they mean to, but they’re swayed in that direction by stakeholders with fuzzy ideas about how wireless networks actually work.
The government actually does know how to share certain kinds of resources with the private sector. The best model is the General Services Administration (GSA), the government’s property and purchasing coordinator. GSA rents office space for government agencies and procures technology products and services. If government spectrum were sold to the private sector and then leased back by GSA when needed, government’s spectrum use would be more rational and less costly to the overall economy. If GSA finds that it can combine government systems on common bands with a private contractor managing the band – and upgrading equipment to use less spectrum when practical – government spectrum use would be more rational and less costly to the general economy.
So why don’t we do that? A one-time sale would pay for government use for the next twenty years quite easily, and by the time the money runs out we’ll have new technology that allows for much less expensive and non-rivalrous operation. Sensitive applications needed by law enforcement and the military could be combined and managed by competent operators at less expense to the taxpayer.
Isn’t that a win-win?