My main mistake was to imagine that the FCC was trying to protect TV reception against cellular; it’s more likely the other way around. It goes back to the “Channel 51 Problem”; here’s how Coleman Bazelon explains it in his 2009 IEP paper, "Too many goals: Problems with the 700 MHz auction":
“Television channel 51 broadcasts are adjacent to the Lower 700 MHz A Block and can cause interference to the A Block licensees, making the affected licenses less valuable. These channel 51 television broadcasts will continue in 41 markets, including important markets such as Chicago. The effect of those broadcasts is to render the A Block spectrum around the broadcast tower largely unusable for two-way mobile communications. Near the TV transmitter, the TV signal is sufficiently powerful that use of the A Block is unlikely to significantly interfere with TV reception. Rather, the strong TV signal would likely drown out the devices using the A Block. ”About a year ago, the FCC froze all applications for TV channel 51 by both applicants for full-power and low power facilities. This was seen as a victory for cellular interests like the CTIA and Rural Cellular Association (coverage here and here; see also Charlie Rhodes’s technical discussion here). In effect, the FCC conceded that cellular licensees were entitled to a 6 MHz guard band between their allocation and TV broadcasts.
That’s bad policy. Licensees should internalize guard bands, to use the economic jargon. That is, they (rather than society at large) should directly bear the cost protecting themselves from interference by paying for such channels at auction; the guard bands they need should be part of, i.e. internal to, the frequencies they obtain at auction. Reserving guard bands such as these, as well as the building in the quiet duplex gap that cellular operators require between their uplink and downlink bands, amounts to providing cellular operators with significant value at no cost.
Rather than inserting guard bands in the incentive auction, cellular license frequencies should go all the way up to the boundary with TV channels, but licenses should include an explicit statement of the interference protection TV receivers are entitled to. Cellular licensees could then decide how close to the TV channels they want to operate. To prevent cellular service degradation, they might have to back off the frequency boundary by (say) 6 MHz given today’s technology, but in future they could move closer, e.g. because they paid the TV stations to reduce power or because they improved the interference tolerance of cellular receivers. And at the same time, TV licensees would have more clarity about the protection they would be entitled to from interference from cellular transmissions; see e.g. my post “The FCC's TV/cellular guard bands don't compute”. In other words: rather than giving away guard bands, the FCC should state the rights and responsibilities for dealing with harmful interference on either side of the boundary between television and cellular, and let industry figure out the details.
Downsides? There are few, of course, including impacts on auctions and unlicensed. Cellular licenses close the TV boundary won’t raise as much revenue for the Treasury as less encumbered ones, as happened in the 700 MHz auction (see Bazelon’s paper, cited above). However, at the moment the income from them is zero anyway, since they’re guard bands; and revenue should in any case not be the main goal of an auction (yeah, right).
Commons supporters will object that society will lose the value of guard bands being allocated to unlicensed. I’d respond, first, that unlicensed operation in the guard bands as proposed is impractical if not impossible: the maximum allowed power will be minimal (40 mW, if the FCC follows the approach it used for similar channels in the whitespace rulemaking); and I don’t believe that unlicensed receivers will be able to reject the adjacent channel interference from TV and cellular base station transmitters. Second, if commons proponents are so confident of unlicensed’s value, they should put that value on the table, either by stating deployment targets that, if not met ten years, would trigger withdrawal of the unlicensed allocation in ten years (see my post "Time limiting unlicensed authorizations"), by bidding for those guard band blocks against cellular in an auction a la Bykowsky, Olson and Sharkey (Efficiency gains from using a market approach to spectrum management, Information Economics and Policy, 22(1):73-90).
If guard bands are bad, it’s going to get worse. They will only get wider as the proceeding progresses, as all the parties to the proceeding (TV, cellular and unlicensed) stand to gain by having them as wide as possible, and their cost, in terms of the more intensive use that is foregone, is borne by society at large.
Wide or narrow or none, we need a quantitative way to think about the guard bands. The harm claim threshold approach (the FCC defining the interfering signal levels that have to be exceeded before a receiving system can claim harmful interference) can help here. The FCC, in consultation with the affected parties, should determine the threshold for interfering signal levels that would be considered harmful interference – mainly from cellular into TV receivers, but also vice versa. It can then use these profiles of energy over frequency to determine the combination of guard band width and interference protection, notably the responsibility that cellular licensees will have to mitigate degradation of TV reception by nearby cellular base stations, e.g. by providing home owners with receiver filters.
Oh, and why doesn’t the FCC worry about interference from cellular into TV? Probably because of the assumption that no-one watches over-the-air TV any more (well, fewer than 15% of households, though I've heard that the number's been increasing recently), and that actual cases of harmful interference into TV will in practice be rare enough to ignore. Ah, practical politics…