Having decided on the density of cells, it is then important to select the best cell sites. Broadly speaking, if the system is coverage constrained,
then cell sites on high buildings or high areas will be important. If the system is capacity constrained, then lower buildings may be more appro- priate because they will allow other buildings to act as shields, dividing cells and preventing the signal from spreading too far.
As a first pass, the coverage area should be divided into a regular grid (normally a hexagonal grid), with the radius of the hexagons being the calculated cell radius as required for coverage or capacity. Base station sites then should be sought as near as possible to the centers of the nominal hexagons.
The next stage is to make use of an appropriate planning tool, as described in Section 6.6. With the use of a planning tool, base station sites can be tried near the center of the nominal hexagon until one is found that approximates as closely as possible to the desired coverage. The process is repeated, building up the entire grid. As one base station is selected, neighboring cells may need to move increasingly from the nominal grid, such that they provide a minimal overlap with the first cells.
Once the sites have been selected on the computer, they should be checked visually to make sure that the LOS paths are as predicted. Suitable mounting points for antennas should be investigated, as should equipment housing and power supplies. If all these look promising, it is time to enter into negotiation with the owner of the building for the rental of space to
Traffic (E)
1 3 5 7 9 11 13 15 17 19
Linesrequired
0 5 10 15 20 25 30 35
P = 0.1%B P = 1%B P = 2%B
Figure 17.3 Channels versus Erlangs for a range of blocking probabilities.
Rolling Out the Network 259
site the equipment. There is little that can be said about this, other than a good negotiator may be required and it might be helpful to find out the going rate for rental from other operators.
One approach that can avoid the need for site negotiation for every cell is to make a deal with an organization that owns numerous buildings, such as a bank. If that is done, the radio planning becomes quite different.
Instead of working from a grid, base stations are now simulated on all the banks (or whatever), and sites then are sought to fill in the resulting gaps.
The number of cells required by this approach should be compared with the number required if the buildings to be used are unconstrained. If additional cells are required when using the banks, the extra cost associ- ated with the additional cells can be compared with the savings made for site rental and time spent negotiating to determine whether it is cost effective to use sites on the banks.
Radio planning is a resource-intensive procedure. Trained operators will need to spend some time using planning tools, trying different configurations and cell sites to achieve the best layout. Depending on the complexity of the environment, the number of cells planned per person per day can vary dramatically. Operators may consider using contract staff to perform the planning; as once the network has been rolled out, there will be little need for such staff.
There are some subtleties to the cell site planning. The first is the use of sectorization. Instead of using an omnidirectional antenna providing coverage over a circle, directional antennas are deployed, splitting the coverage into a number of wedge-shaped slices. Nonsectorized and sectorized cells are shown diagrammatically in Figures 17.4 and 17.5, respectively.
The benefits of sectorization differ dramatically for CDMA and TDMA systems. For CDMA, sectorizing a cell results in an increase in capacity, broadly in line with the number of sectors deployed. That is because adjacent cells already use the same frequency. By splitting a cell into a number of sectors, each sector can use the same frequency. The result is that the interference to adjacent cells is increased, because there are now more users in the cell that has been sectorized. That reduces the capacity in the adjacent cell, but the reduction in capacity is much less than the gain in capacity in the cell that has been sectorized. To a first approximation, the capacity is increased in line with the increase in the
f1, f2, f3
f1, f2, f3 f1, f2, f3
f1, f2, f3
Figure 17.4 Omnidirectional cell arrangement.
f1 f2
f2 f1
f1 f2
f2 f1
f1 f2
f2 f1
f1 f2
f2 f1
f1 f2
f2 f1
f1 f2
f2 f1
Figure 17.5 4-sector cell arrangement.
Rolling Out the Network 261
number of sectors (e.g., when moving from a nonsectored to a three-sec- tored cell, the capacity increases by a factor of 3).
For TDMA, the gains are much more limited. Because the base station still transmits with the same power after sectorization, neighboring cells are unable to use the same frequencies. And because additional frequen- cies are required in the sectorized cell (since each sector must use a different frequency), the overall number of frequencies required and, hence, the cluster size, rise. That has the effect of offsetting the gains in capacity associated with making the sectors smaller than the original cell.
Careful planning using the directionality of the antennas in the sectors can result in some gains. For example, moving from a nonsectored system with a cluster size of 3, then a three-sectored system would require a cluster size of around 7 or 8. The overall gain from sectorization is then around between (3 × 3)/8 = 1.1 to (3 ×3)/7=1.3. With dynamic channel allocation, it may be possible to increase those gains slightly more.
Each sector requires a separate base station unit, resulting in addi- tional cost, but the site rental typically is unchanged, making the use of sectors on a single cell less costly than installing multiple small cells.
Given the limited gains of sectorization for TDMA systems, it is reasonable to ask why most GSM operators use a sectorized cell system in city areas. The reason is that sectorization allows for a greater path loss than a nonsectored approach because the gain of the sectored antennas can be included in the link budget. That has the effect of increasing the range. Given that a number of base station transceivers would be required to handle the traffic generated in city areas anyway, connecting those transceivers to different antennas results in little increase in cost. Further, in city areas, sectorization prevents some multipath reflections that might occur if an omnidirectional antenna was used, increasing signal strength and finally a small capacity gain is achieved. Overall, for a small cost increase, improvements in capacity, range, and signal quality can be achieved by sectorization in this area.
In suburban and rural areas, however, sectorization is used less. That is because in those areas it may be that only one carrier is required to provide sufficient capacity. To use additional carriers to feed each sector would engender a significant cost increase.
Another approach is the use of hierarchical cell structures. In a hierarchical structure, an oversailing macrocell provides coverage to a large area, which typically generates more traffic than the capacity provided by the cell. Smaller microcells or minicells then are inserted in areas of particularly high traffic density, taking some load off the macro- cell. Such minicells need to use a different radio frequency to avoid interference. That can be a sensible approach in a large city where the average traffic levels are low, but there is a small area in the center with high traffic densities. If the nonhierarchical approach is adopted, then one cell would be required in the center and perhaps six cells around it to cover the entire city. Each of the six cells would be operating at well below capacity. With a macrocell-minicell approach, only two cells provide coverage in the same area.