Performance of Automatic Frequency Planning and Optimization Algorithm for Cellular Networks

Each cell of the 3 cells uses a different subset of the allocated channels. Cells using the same frequency are placed as far away from each other as possible to ensure that co-channel interference is weaker than the signal originating from the cell.

For K-cell frequency reuse, the distance between the cells is given as; (1) D=(√3K). R {\rm{D}} = \left( {\sqrt {3} K} \right).\;{\rm{R}}

Where R= Cell radius

The most significant interference comes from the 6 closest co-channel cells considering that the cell is considered hexagonal in this case. (2) C/I=(1/6)(D/R)γ {\rm{C}}/{\rm{I}} = \left( {1/6} \right)\left( {{\rm{D}}/{\rm{R}}} \right)\gamma

For free space, path loss (γ) = 2 but in most cases it ranges from 3–4

For K = 3 and γ = 3.7, we have (3) C/I=(1/6)(3R/R)3.7=9.7095=9.872dB {\rm{C}}/{\rm{I}} = \left( {1/6} \right)\left( {3{\rm{R}}/{\rm{R}}} \right)3.7 = 9.7095 = 9.872{\rm{dB}}

The C/I value calculated above is the quoted GSM value of 9 dB. By the central limit theorem, as the number of interferers gets large, the total interference tends towards a Gaussian distribution and will hence resemble Additive White Gaussian Noise (AWGN). Using this, capacity is then approximated as; (4) C=BTlog2[1+SINR] {\rm{C}} = {\rm{BTlog\;}}2\left[ {1 + {\rm{SINR}}} \right]

Where SINR ≈ C/I

The capacity per area for a single frequency band used throughout a given area (A) is given as; (5) C/A=C/KΠR2 {\rm{C}}/{\rm{A}} = {\rm{C}}/{\rm{K}}\Pi {\rm{R}}2

The minimum K that provides acceptable SINR is used to achieve satisfactory performance and maximize capacity. Decreasing the value of R, increases the capacity per unit area of the system, this is a good approach but very uneconomical because it means installation of more base stations which is very expensive.

Systems with fewer cells also experience higher interference levels because the path loss exponent, γ tends to be distance dependent. Decreasing the cell radius, R, makes γ to approach 2. This is because the unobstructed line of sight propagation is very likely in fewer or smaller cells. It can be seen that γ decreases, C/I and SINR decrease if the reuse factor, K is fixed. To have good interference levels, when cell radius is decreased, K must be increased to maintain the acceptable SINR.

Consider the geographical area as shown in Fig.2 below, each cell having a radius of R. The distribution of channels among the cells, based on the resource planning model with cluster of size 3 is assumed. Minimum reuse distance, Dmin, is 3R. Meaning that the neighboring cells in interference is 6. For example, our reference cell is center cell B3 for which interference will be considered from the following 6 neighbors; A3, C3, A5, C4, A2 and C1.

Figure 2.

If a channel is used in a cell, then none of the other 6 interference neighboring cells uses it. When an MS sets up a call, a request message is sent to the immediate BTS and if there is a free primary channel, the BTS selects one and ensures that there isn’t any co-channel interference hence maximizing channel utilization. Some of the channel selection schemes use resource planning model to get better channel reuse. Static resource planning is done in a way where a fixed set of resources is allocated to cells. However with the increase in variation of traffic, few cells starve for spectrum while others remain completely unused. When cells starve, there are very high chances of calls getting blocked due to insufficient resources. In a variable traffic scenario, static resource planning becomes inefficient. To deal with the unbalanced resource distribution, a more flexible planning is required which dynamically varies the resources according to the traffic. Resources allocation in orthogonal frequency division multiple access (OFDMA) ensures that two users are not assigned a common resource in a cell at any given time. This eliminates intra-cell interference (due to transmission within the cell).

The rules for using resource planning models are as follows:

When all the primary channels are completely exhausted, then a cell requests the secondary channels.

Cells are divided into 3 linear disjoint subsets. i.e. A, B and C. These subsets are disjointed to avoid adjacent channel interference. Each of the 3 cells uses a different subset of the available channel. Each cell has 6 neighbors. Consider a total of 21 channels each with 7 primary channels;

Subset A: 1, 4, 7, 10, 13, 16, 19

The following assumptions are made:

The neighbors of each cell are well defined. With the center cell as B3, its neighbors are A3, C3, A5, C4, A2 and C1.

When cell B3 receives a call request, it checks the availability of primary channels and if there are vacant primary channels, it connects the call. If there aren’t any available primary channels, it initiate a search for free primary cells in the neighboring cells (A3, C3, A5, C4, A2 and C1). When it receives a confirmation message from the neighboring cells, it selects a channel randomly from the first one to arrive. The call is connected ensuring that the quality of the call is maintained.

B3 takes confirmation of the selected channel and marks it as interference channel. The channel is not used again until it is returned back by the borrower. The scheme focuses on the assignment of channels in such a way that channel utilization is maximized at the same time maintaining the voice quality. The algorithm ensures maximum reuse of channels.

In channel acquisition, information on free channels is collected ensuring that two cells within minimum reuse distance do not share the same channels. Acquisition phase of the distributed Dynamic Channel Allocation (DCA) algorithm consist of the search and update phases.

In channel selection, when cell B3 requires a free channel, the channel selection scheme is used to pick one available channel and confirms with its interfering neighboring cells whether it can use the selected channel or not. After that, when a cell acquires or releases a channel at any time, it informs its interference neighbors so that every cell in the system model always knows the available channels of its interference neighboring cells.