No one likes “dead zones” There are various ways of locating them. Plotting the distribution of “femtocells” may be one, because cell phone users in areas of poor reception can work around the problem of poor reception using either (1) a repeater; or (2) a femtocell.
A repeater comprises an outside antenna, a bi-directional amplifier and an interior antenna. The weak “down link” signal received by the outside antenna is amplified and rebroadcast inside. Conversely, the “up link” signal from the cell phone is picked up by the interior antenna and amplified before being sent back by the external antenna. The outside antenna is directional and needs to be aimed towards cell phone towers. So the approximate location of “nearby” cell phone towers must be known. Applications like OpenSignalMaps can be very helpful in finding them.
The other approach is to set up a femtocell, which is basically a very low power “cell phone tower” that depends on a good internet connection (cable or DSL) for the “back haul”. A femtocell covers a tiny area — about the size of a house or office, and tends to be easier to set up than a repeater, particularly for those less technically inclined, and is the approach encouraged by some cell phone carriers in areas of poor signal.
Generally urban areas tend to be well provided with cell phone base stations, while rural areas often are not. On the other hand, man-made obstacles in urban areas can block cell phone signals and contribute to poor reception there as well. Many users would like to know where the “dead zones” are and what areas have a higher share of weak signal spots than others. “Heat maps” generated through crowd sourcing, such as those provided on the OpenSignalMaps web site can help with that.
Another way to locate areas of weak signal would be to plot where there are a lot of repeaters — or a lot of femtocells. It is difficult to determine where repeaters are installed, but femtocells are equipped with GPS receivers so they can report their location (This is critical in order to trace the origin of an emergency call). Femtocells also have unique “IDs” just like regular cell phone base stations. These IDs are composed of four components, much like internet addresses (In the case of CDMA, the unique ID consist of MCC (Mobile Country Code), SID (System ID), NID (Network ID), and BID (Base Station ID). For GSM the unique ID consists of MCC (Mobile Country Code), MNC (Mobile Network Code), LAC (Location Area Code), and CID (Cell ID)).
As, an example, consider the distribution of Verizon Wireless base stations and femtocells in the Eastern New England area. Presently there appear to be roughly 17,000 femtocells. Compare that to the distribution of cell phone base stations (“macrocells”) on about 2,400 towers and buildings — for a total of roughly 7,000 antennas (most carriers have three sector antennas per tower pointing off in different directions). So there are a lot more femtocells than cell phone company towers! Of course, a femtocell covers a tiny area, nevertheless, this indicates that areas of weak cell phone reception are not uncommon.
The distribution of femtocells is not quite what one might expect. For example, quite a few femtocells can be found in urban areas where one might think service must be good. These femtocells appear to be mostly used were there is a good outside signal, but a steel and concrete building severly attenuates it. Also, there are few femtocells per square mile in very remote areas, but then again the population density there is very low. So normalizing the raw statistics with respect to population (rather than say area) seems appropriate. Now the population density (persons per square mile) varies over three or four orders of magnitude in the New England area:
While the number of base stations per square mile also varies over several orders of magnitude, the density of base stations (per thousand people) does not — mostly hovering between 0.5 and 2 (on the left — blue). Similarly, the number of femtocells per square mile varies over several orders of magnitude, while the density of femtocells (per thousand people), varies only between 0.5 to 6 (on the right — red):
Note that the areas of high base station density (bright blue) are not the same as the areas of high femtocell density (bright red).
A scattergram showing femtocell density (vertical) versus base station density (horizontal) shows that high femtocell concentration (red dots — top of figure) occurs mostly at relatively low base station density. Conversely high base station density (bright blue dots — right of figure) goes with low femtocell density. Most counties fall somewhere in between — or are unfortunate in having both low base station density and low femtocell density.
We can combine the plots of base station density and of femtocell density into a single figure where each combination of base station density and femtocell density corresponds to a particular color, as indicated by the inset in the top left of the figure:
The areas of relatively high femtocell density (four red counties in New Hampshire) are in areas of relatively low base station density. Conversely, areas of relatively low femtocell density are in areas of relatively high base station density (three bright blue counties in Maine).
The above is based on averages over counties. The raw data reveals additional clustering of femtocells. In some deep valleys, for example, cell phone reception is inherently iffy and so more people will resort to repeaters and femtocells. But beyond that, there appears to be a “word of mouth” effect (an archaic form of social networking), where, if someone installs a femtocells, the neighbors soon find out and want one too.
Clicking on an image (JPEG) above links to the corresponding full resolution PDF file.
Base station coordinates are not the coordinates of the cell phone towers, but the centers of areas in which contacts where reported by cell phones (hence not perfectly accurate - in rural areas may be as much as a few kilometer off).
Femto cell coordinates are often “spot on” (within 10 meters) if reports come only from the cell phone of the owner — which is what happens quite often in rural areas.
A few “base stations” may be bogus combinations of SID:NID:BID produced by a bug in Android software which updates SID:NID before updating BID — leading to incorrect combinations of new values of SID:NID with old value of BID.
A few “femtocells” may be duplicates where the same femtocell was assigned different BIDs at different times (although most of these duplicates were eliminated based on proximity, where the reported locations were closer than 100 meters).
A small percentage of base stations were not assigned to any county because (1) their “average footprint” coordinates ended up in the ocean (common where coastlines have large concavities such as in Rhode Island's Newport area, near Boston Harbor and along Down East coast of Maine) or because (2) the county outlines were “generalized” and so do not depict the county outlines with perfect fidelity. Overall, between 1% and 2% of the raw data was lost in this fashion.
The county data is provided in the form of a CSV file. The first item in each line is the GEO_ID that matches that found in the Google “Fusiontable” county outlines. The county populations are from the 2010 Census.
In a few places there is some “spill over” into adjacent counties in Vermont and Connecticut, as well as Western Massachusetts, that are not in the area covered by the survey. Those counties are included in the table as well, but the computed base station and femto cell densities are not useful (no color in the figures).
The underlying raw base station data is available as a KML file.
The underlying raw femtocell data is available as a KML file. (This file is near the size limits of what Google Maps can handle).
To use either of the KML files in Google Earth, just launch the file (assuming you have Google Earth installed).
To use either of the KML files in Google Maps, type the URL into the “Search” box after launching Google Maps from your web browser.
Alternate versions of the figures for OpenSignalMaps blog, using Google's Fusion Tables can be found here.