You are on page 1of 3

Wireless Networks (teacher: G. Gelli, a.y.

2008/09)

Matlab laboratory # 1 Trunking in cellular systems

Introduction
For this lab you will need: (a) to review Lecture 9, with particular attention to the part regarding trunking and Erlang B model (BCC); (b) to review your Matlab skills; (c) to download from the course website some Matlab functions for performing calculations with Erlang B and Erlang C models. Please try to work cleanly, by writing one Matlab M-function for each exercise.

Exercises
(1) In a cellular network, the available channels (100 channels per cell) are equally divided between two operators. (a) Compute the number of users served in each cell by each operator if every user generates 0.1 Erlangs of trafc. (b) Compare with the situation where one monopolistic operator uses all the 100 channels. (Assume a BCC model and GOS = 5% in your calculations). (2) For a N = 7 system with P (blocking) = 1% and an average call length of 2 minutes, nd the loss in trunking efciency when going from omni-directional antennas to 60o sectored antennas. (Assume a BCC model and that the average number of calls made by each user is 1 per hour). (3) A cellular operator decides to use a digital TDMA scheme which can tolerate a C/I ratio of 15 dB in the worst case. Find the optimal value of N (cluster size) for: (a) Omni-directional antennas (b) 120o sectoring (c) 60o sectoring Should sectoring be used? If so, which case (60o or 120o ) should be used? (Perform your analysis for a path loss exponent of = 3 and = 4 and consider trunking efciency, assuming GOS = 1%).

Wireless Networks (teacher: G. Gelli, a.y. 2008/09)

(4) A cellular system can support a given trafc A with C = 100 channels per cell and GOS = 2%. Verify what happens to GOS if the trafc increases in some periods of the year by a factor 2, 5, or 10. How many channels would be needed to keep the GOS unchanged in these situations? (5) This exercise explores some issues in non-uniform xed channel allocation (FCA). Consider a cellular system with cluster size N and Nc channels available in each cluster. For each cell of the cluster, let the triple (Ai , Ci , GOSi ) represent the cell trafc Ai , the number of channels Ci of the cell, and the grade of service GOSi (blocking probability) of the cell (i = 1, 2, . . . , N ). In the BCC model, these quantities are related by the Erlang B formula:
i AC i Ci !

GOSi = f (Ai , Ci ) =

Ci

k=0 N

Ak i k!

Note that
i=1

Ci Nc (where the equality holds when all the channels are used).

In uniform FCA, each cell is assigned the same number of channels, i.e., Ci = C = Nc /N (note that oor rounding is needed since C must be an integer number). This is clearly optimal if Ai = Aj , i = j (uniform trafc). In non-uniform FCA, instead, each cell is assigned a different number of channels, i.e., more channels are assigned to cells with higher trafc. The problem considered here is how to allocate the channels to different cells according to some performance measure. Since the GOS is not necessarily the same in all cells, we can dene an overall GOS at least in two different ways: (a) Average GOS: a weighted sums of the GOSs
N

GOS =
i=1

wi GOSi ,

wi =

Ai N i=1

Ai

where the coefcients wi weight the GOS in each cell according to relative trafc intensity. (b) Worst-case GOS: the maximum value among all the GOSs GOS =
i=1,2,...,N

max GOSi

For a given distribution of trafc (A1 , A2 , . . . , AN ) among the cells, the optimal channel assignment can be found as the solution of the following optimization problem:
N C1 ,C2 ,...,CN

min

GOS

subject to
i=1

Ci Nc

Wireless Networks (teacher: G. Gelli, a.y. 2008/09)

This is a constrained nonlinear optimization problem, where moreover the optimization variables belong to a discrete set. Note that solving the problem by a brute-force approach is not advisable (the number of congurations of channels to be tested increases exponentially with N ). However, the case N = 2 can be dealt with rather simple math (although the solution cannot be written in closed-form, mainly due to the fact that the Erlang B formula cannot be easily inverted). Provide a simple mathematical formulation for N = 2, and try to solve it with the help of Matlab, by using the two previous denitions of GOS (average GOS or worst-case GOS). Compare the obtained solutions for different non-uniform trafc distributions between the two cells, and compare them with the uniform assignment (you might assume Nc = 120 channels and an overall trafc in the two cells equal to Atot = A1 + A2 = 100 Erlangs, but feel free to experiment with other values, provided that the GOS of the system is reasonably below 5%).

You might also like