Macrocells and Their Limitations

Because a cell is the basic constituent unit of a point-to-multipoint network, the distribution of
cells determines the coverage and capacity of the network and of course has a major bearing on
capital costs. Thus, cell size matters very much.
Network engineers tend to think of radio cells in terms of the areas they encompass. A
large cell with a radius of miles is known as a macrocell, while cells of a kilometer or two in
radius constitute microcells. Cells with radii measured in the hundreds of meters are picocells.
The usual pattern of broadband wireless operators, as indicated in Chapter 2, has been to
launch the network with a single base station defining a single cell and with sufficient transmitting
power to reach all or most of the potential customers within the metropolitan market
addressed by the network operator. Both low-frequency microwave and millimeter microwave
operators have used this type of single-cell architecture.
Lower-frequency microwave operators have been able build macrocells with radii exceeding
15 miles, and if they utilize sectoral antennas—actually clusters of high-gain antennas
distributed around a central mounting pole—they can reuse the available spectrum as much
as four times over in theory, eight being the maximum number of sectors that is practical with
conventional nonadaptive antennas and four being the maximum reuse factor within a single
cell. It would appear to follow, then, that if one starts with a generous frequency allocation—
say, the nearly 200 megahertz (MHz) available in the Multichannel Multipoint Distribution
Service (MMDS) bands—and multiplies that by four and assumes a three-bit-per-hertz ratio,
then one is looking at minimally a couple of gigabits per second total system throughput, a rate
that translates into quite a respectable system capacity.
None of these theoretical maxima can be realized in practice, however, and they cannot
even be approached. The narrow beams that define each sector will spread out over distance
and begin to interfere with one another at the outer edge of the macrocell so that all eight
sectors cannot be extended the full width of the cell. Then too, bit rates drop off with distance
as the signal attenuates, the fade margin declines, and the error rate ascends. In fact, real throughput rates are apt to be half or less the theoretical maxima at the farthest transmission
distances.
In practice, only about a gigabit of capacity is likely to be reliably available, even assuming
that generous 200MHz spectrum allocation. If you settle on an oversubscription rate of four to
one, a fairly conservative figure, then the network can accommodate 400 subscribers at
a 10Mbps throughput rate for each albeit with no service guarantees. If one can charge appropriately—
say, $500 to $1,000 per month—that may be the basis of a sustainable business, at
least in the short run, though an annual revenue in the $2 million to $4 million range may not
permit the network operator to capitalize the expansion of the network.
Furthermore, in such a macrocellular architecture—absent the newer NLOS technologies—
obstructions will put many potential subscribers entirely out of reach. What percentage
of potential subscribers cannot be served is a matter of dispute among authorities, and in
fact the particular topography in which the network is being deployed can give rise to enormous
variations in this regard. By most accounts, at least 40 percent—and perhaps as many
as 80 percent—of potential subscribers are unlikely to be served because of the presence of
obstructions.
Faced with such constraints, network operators can resort to a number of ploys. They can
use repeaters to reach certain subscribers, a sort of partial cell-splitting approach. They can
employ sectoral antennas to achieve frequency use within the cell. They can employ dual
polarization to reuse spectrum. They can resort to full-on cell splitting and begin to create a
microcellular network. Or they can use NLOS technologies to access out-of-reach subscribers,
though only in the immediate vicinity of the central base station and not at the outer periphery
of the macrocell. They can also do all of these. But what they must do in all cases is to increase
spectral efficiency, that is, overall carrying capacity of the networks based on available spectral
resources.