Radio Free America is a set of programs that takes television transmitter contour data generated by the FCC and overlays those contours onto a Blue Marble-derived basemap image of the United States. It then applies a population dataset in the form of 925 1x1 degree Census data block files.

Calculations are then performed to determine a) the estimated number of people covered by a given television frequency range and b) the efficiency of this channel allocation, as a percentage of the total population.

The image output uses the following colors to indicate area coverage by a corresponding number of transmitters:

1	2	3	4-5	6-7	8-10	11-14	15-25	26-53

Output Images

	Spectrum Utilization Map [1280x547] Spectrum Utilization Map [2560x1024] Spectrum Utilization Map [7860x3360]
	Population Map [2048x875]
	Composite Spectrum + Population Map [1280x527] Composite Spectrum + Population Map [1600x659] Composite Spectrum + Population Map [2560x1024]

Purpose

The point of coming up with visualizations like this is to benchmark how much per-person use is squeezed into the over-the-air television spectrum. In heavily populated areas, there are corresponding incentives for TV network buildout. But in sparsely populated areas, many television channels are left as guard bands for a 60-year old analog broadcast standard. Instead of allowing for more mixed use of the spectrum, possibly for wireless Internet, 3G/4G telephony, WiMAX, or any number of licensed and unlicensed wireless high-data-rate services, the spectrum sits unused and protected by laws that are not amenable to innovation. The point of data visualizations like these is to draw out patterns that help futher discussions on the role and efficiency of government in setting spectrum policy.

There is an implicit argument here for the idea that spectrum rights as assigned to the individual, within some demarcation of space, generates a measure of total utilization versus underutilization. e.g. For any given bandwidth that is unused, there is a lost opportunity to more fully connect that individual to a larger telecommunications infrastructure. In other words, national spectrum coordination is important to create market incentives for structured deployment to a point; however, as the Internet, Linux, Wikipedia, SourceForge, Blogger, etc., have adequately demonstrated, providing tools and a space for individuals to connect in an open fashion can generate profound and socially-valuable organic developments.

Summary Statistics

Channel      Covered        Total   Utilization      Non-
          Population   Population                Utilization
------------------------------------------------------------
    2      135645739    294833827      46.0%        54.0%
    3      110345749        "          37.4%        62.6% 
    4      146670279        "          49.7%        50.3% 
    5      147071290        "          49.9%        50.1% 
    6      101762390        "          34.5%        65.5% 
    7      121638961        "          41.3%        58.7% 
    8       89743845        "          30.4%        69.6% 
    9      128739381        "          43.7%        56.3% 
   10      123518548        "          41.9%        58.1% 
   11      108257685        "          36.7%        63.3% 
   12       95030922        "          32.2%        67.8% 
   13      136470745        "          46.3%        53.7% 
   14       55114286        "          18.7%        81.3% 
   15       44566351        "          15.1%        84.9% 
   16       40539673        "          13.8%        86.2% 
   17       94855829        "          32.2%        67.8% 
   18       68722995        "          23.3%        76.7% 
   19       91931877        "          31.2%        68.8% 
   20       91599128        "          31.1%        68.9% 
   21      101467478        "          34.4%        65.6% 
   22      102728116        "          34.8%        65.2% 
   23      108628601        "          36.8%        63.2% 
   24       84685218        "          28.7%        71.3% 
   25       95512042        "          32.4%        67.6% 
   26       89363816        "          30.3%        69.7% 
   27      100894159        "          34.2%        65.8% 
   28       91281652        "          31.0%        69.0% 
   29      115694199        "          39.2%        60.8% 
   30       88168472        "          29.9%        70.1% 
   31      132827743        "          45.1%        54.9% 
   32      119082598        "          40.4%        59.6% 
   33       71185656        "          24.1%        75.9% 
   34      118453601        "          40.2%        59.8% 
   35      118438494        "          40.2%        59.8% 
   36      102807197        "          34.9%        65.1% 
   37      ------- Reserved for Radio Astronomy  --------
   38       98163455        "          33.3%        66.7% 
   39      112703469        "          38.2%        61.8% 
   40       95470297        "          32.4%        67.6% 
   41      125561642        "          42.6%        57.4% 
   42       97061686        "          32.9%        67.1% 
   43      130154567        "          44.1%        55.9% 
   44      104843921        "          35.6%        64.4% 
   45       80667672        "          27.4%        72.6% 
   46       78893663        "          26.8%        73.2% 
   47       73812777        "          25.0%        75.0% 
   48      136729680        "          46.4%        53.6% 
   49       89802640        "          30.5%        69.5% 
   50      115522415        "          39.2%        60.8% 
   51       79903413        "          27.1%        72.9% 
   52      111816006        "          37.9%        62.1% 
   53       78407248        "          26.6%        73.4% 
   54       88734724        "          30.1%        69.9% 
   55       61465676        "          20.8%        79.2% 
   56       92324756        "          31.3%        68.7% 
   57       77979680        "          26.4%        73.6% 
   58       88400169        "          30.0%        70.0% 
   59       71071601        "          24.1%        75.9% 
   60       64572082        "          21.9%        78.1% 
   61       60274853        "          20.4%        79.6% 
   62       81255403        "          27.6%        72.4% 
   63       44100281        "          15.0%        85.0% 
   64       37022807        "          12.6%        87.4% 
   65       57100734        "          19.4%        80.6% 
   66       83545056        "          28.3%        71.7% 
   67       52129049        "          17.7%        82.3% 
   68       55016781        "          18.7%        81.3% 
   69       43624435        "          14.8%        85.2% 
------------------------------------------------------------
Totals                               2117.1%      4682.9% 

In other words, 21 channels of 6 MHz spectrum are completely utilized, if we look at spectrum on a purely per-Population-covered basis. Whereas 46 channels worth of 6 MHz spectrum are either, a) not utilized for economic reasons in a market area or b) not utilized because of guard band industrial policy, and therefore have no population coverage. While this measure conveniently ignores the fact that the underutilized spectrum by channel is non-contiguous across the continental United States, the digital television transition could be used to structurally adjust the organization of the television spectrum to pack broadcasters into a tighter contiguous frequency range, thus freeing up large amounts of propagation-valuable spectrum.

There is also the issue that some markets are already mostly packed with channel coverage, which this particular analysis discounts by averaging. But this channel coverage is usually due to the fact that broadcasters were granted an additional, free DTV license, so very likely in the major Metropolitan Statistical Areas, the actual spectrum necessary to transmit program content could be chopped in half by removing the analog simulcast. This was ostensibly the original intent of the DTV transition, rather than turning into an incumbent spectrum giveaway.

Finally, in examining the Per-Channel Coverage Plots below, the question should be asked: Is it still worth nationally guarding the underutilized spectrum used by a few lingering analog television stations on Channels 60 - 69? Or are there more valuable uses of the bandwidth that could be expedited, as originally intended, towards the end of 2006?

Per-Channel Coverage Plots

All plots cover a region extending from [52°N x 128°W – 24°N x 64°W] using 120 pixels / degree. [link]

Source

Source code only. (2006/03/26)
More, different source. (2006/04/02)

Future

The programs have a number of current limitations.

1. They assume 100% coverage, 100% of the time in the covered area, e.g. f(100,100), which is wrong.
2. A number of program parameters are hard coded and should be parameterized.
3. Analog and digital coverage are combined in this map. Also, all license classes are included.
4. The cache generation program is extremely fat. (That's what I get for using wx-devcpp.)
5. The coverage cache data generator and the image generators are separate programs right now.
6. They use television coverage data from 10/2005. The latest FCC data (03/2006) seems to cause problems with it. (It appears that this data also has some kind of corruption. Or maybe my program has problems with the data, I haven't checked.)

A number of features would be nice.

1. Signal coverage modeling with Longley-Rice or another propagation model using terrain data. (This data is available from the FCC.)
2. Statistics generation for coverage in km² by channel, by percentage of population, by city, etc.
3. Output by television service type, e.g. filtering by digital/analog, commercial/educational/low-power, translator/repeater, etc.
4. Optional state outlines and major city markings might be nice. Though it's actually more fun trying to guess which blobs are which cities.

I think that a lot of this type of analysis could be performed in something like ESRI's ArcView or another expensive GIS package, but then I haven't got that kind of cash, so this is a little hack that works.

Final Note

Some of the astute amongst you may notice an odd rectangle in the population coverage extending east of Philadelphia through to the New Jersey coast. This is a problem in the population dataset I don't feel like tracking down (digging through a million lines of binary data in a hex editor is only fun if you're doing something naughty). This leads to the odd situation that some 1.28 million people are packed instead into a 30-arcsecond area, or roughly 1km², if I remember right. Since the algorithm I use to read the data is constant across all input files, I think the error is data driven. I don't know if the FCC knows this, but I am using the same data files they use to process broadcast license paperwork, and I based my read code on their utterly ancient Fortran code (rd_pop_data_full.f). I shudder to think what machine their code compiles on though, as g95 doesn't like it. Just think, U.S. national spectrum policy is run on flat files and old Fortran!

Links

The Citizen's Guide to the Airwaves
New America Foundation Spectrum Policy Program
Free Press
Public Knowledge
Consumer's Union
Digital Promise / Digital Opportunity Investment Trust (DOIT)
CATO Institute Tech, Telecom, and Internet Studies
Heritage Foundation