1. Asad Parvez Advisor: Dr. Akl
Impact of Actual Interference on Capacity and Call Admission Control in a CDMA Network
Asad Parvez
Advisor: Dr. Akl

2. Outline Introduction to CDMA networks
Interference model impact on capacity
Global call admission control
Local call admission control
Global vs local
Conclusions

3. Code Division Multiple Access (CDMA) Overview
Multiple access schemes
Call 1
Call 3
Call 2
Call 4
Frequency
Time
FDMA
Call 5
Call 6
Call 7
Call 8
Call 9
Call 10
Call 11
Call 12
TDMA
CDMA
Code
For radio systems there are two resources:
Frequency
Time
FDMA: Division by frequency, so that each pair of communicators is allocated part of the spectrum for all of the time
TDMA: Division by time, so that each pair of communicators is allocated all (or at least a large part) of the spectrum for part of the time
CDMA: every communicator will be allocated the entire spectrum all of the time. CDMA uses codes to identify connections.
May be cocktail party example.

4. Spread Spectrum: Direct Spreading
1 chip period
Data Signal
1 bit period
Data Signal
Interference
PN-code
Spreading
Coded Data Signal
Code Division Multiple access scheme allows multiple users to share the same bandwidth.Each user is identified by a code that is orthogonal to all other users. Unlike Frequency Division Multiple Access (FDMA), CDMA has a soft capacity.This means that there is no hard limit to how many users we can allow on the system.Each time a user is added, the noise floor for the other users is increased by a little bit
CDMA is a form of Direct Sequence Spread Spectrum communications.
Spread Spectrum communications is distinguished by three key elements:
1. The signal occupies a bandwidth much greater than that which is necessary to send the information
2. The bandwidth is spread by means of a code which is independent of the data
There are few ways to spread the bandwidth of the signal
Direct sequence. The digital data is directly coded at a much higher frequency. The code is generated pseudo-randomly, the receiver knows how to generate the same code, and correlates the received signal with that code to extract the data.
A PN-code is a sequence of chips valued -1 and 1 (polar) or 0 and 1 (non-polar).
CDMA codes are not required to provide call security, but create a uniqueness to enable call identification. Codes should not correlate to other codes or time shifted version of itself.
PN-code
Recovered Data Signal
De-spreading
Decoded data Signal

5. Factors Affecting Capacity
Power Control
Pt1: Power transmitted from c1
Pt2: Power transmitted from c2
Pr1: Power received at base station from c1
Pr2: Power received at base station from c2
Pr1 = Pr2 c2
Pt2
Pr2
Base Station
Pt1
Pr1 c1
CDMA is interference limited multiple access system. Because all users transmit on the same frequency, internal interference generated by the system is the most significant factor in determining system capacity and call quality. The transmit power for each user must be reduced to limit interference, however, the power should be enough to maintain the required Eb/No (signal to noise ratio) for a satisfactory call quality.
A critical problem in CDMA is the near-far problem.This problem occurs due to the lack of effective power control:If all mobiles were to transmit at a fixed power the mobile closest to the base station, from communication point of view, will overpower all others signals.Another reason for power control is the battery life time: d2 d1
Distance

6. Factors Affecting Capacity (cont.)
Soft handover of calls
Handover occurs when a call has to be passed from one cell to another as the user moves between cells. In a traditional "hard" handover, the connection to the current cell is broken, and then the connection to the new cell is made. This is known as a "break-before-make" handover. Since all cells in CDMA use the same frequency, it is possible to make the connection to the new cell before leaving the current cell. This is known as a "make-before-break" or "soft" handover. Soft handovers require less power, which reduces interference and increases capacity.9
Time

7. Factors Affecting Capacity (cont.)
Universal frequency use
Reverse link vs forward link
Voice activity factor E A F D A A A A A A A A G C A A B A
TDMA or FDMA
CDMA
The purpose of the variable rate Vocoder is to allow the transmitter to limit its average transmitted power in proportion to the user's voice activity factor. By lowering the average transmitted power, and therefore its self-interference power, the CDMA system is able to increase its capacity
. The vocoder used by the IS-95 system is variable rate, which means that the output rate of the vocoder is adjusted according to a user’s actual speech pattern. For example, if the user is not speaking during part of the conversation, the output rate of the vocoder is lowered to prevent power from being transmitted unnecessarily. For a uniform population, this reduces the average signal power of all users and consequently the interference received by each user. The capacity is thus increased proportional to this overall rate reduction
Power control is needed for the forward link (base station to mobiles station) as well as for the reverse link (mobile station to base station).But, power control is more critical for the reverse link. There are more limitations imposed on the reverse link that makes the power control problem harder to deal with.Some of these limitations are: i) mobile station has less computation power than the base station; ii) non-coherent demodulation, transmission of a pilot signal is a luxury available only for the base station; and iii) stricter transmission power limitation, unlike base stations, mobiles stations can not transmit at a very high power.
Forward link
Reverse link

8. Relative Average Inter-cell Interference Model
Relative average interference at cell i caused by nj users in cell j A
Cell j
Power control attempts to equlaize users received signal power at a given cell’s base staiton, for all users controlled by that base station. But interference also arrives from users controlled by other cells base stations. Its arrives at the given base station with lower power levels.
The propagation loss is generally modeled as the product of the mth power of distance and a log normal component representing shadowing losses.
Shadow loss: depending on the environment and the surroundings, and the location of objects, the received signal strength for the same distance from the transmitter will be different – buildings, walls etc.
Rayleigh fading: fast fluctuations due to movement.
Since the user is communicating with nearest base station, it will also be power-controlled by that base station. The user’s transmitter power gain thus equals the progagation loss for that cell.
ζ: decibel attenuation due to shadowing, and is a guassian random variable with zero mean and standard deviation sigma
χ: rayleigh random variable that represents the fading on the path from this user to cell I.
γ: ln(10)/10
Denominator: progagation loss to the given base station
Numerator: gain adjustment through power control by the nearest base station. B
Back

9. Interference Matrix 11 12 13 … 1M 21 22 31 32 M1 M2 MM 11 12 13 … 1M 21 22 31 32 M1 M2 MM
Interference matrix: every column I is the interference exerted by cell j on cell I.
Hence, the total relative average inter-cell interference experienced by cell i is C

10. Relative Actual Inter-cell Interference Model
Interference matrix F cannot be calculated in advance
Instead, a new interference matrix U is computed as follows
For a user k in cell j, the relative actual interference offered by this user to cell i is D
Hence, the total relative actual inter-cell interference at cell i caused by every user in the network is
k users in cell j E

11. Actual Interference Matrix U
Example: for a new call in cell 2, compute row matrix U[2,i] for i = 1,…,M using equation D
Update 2nd row of interference matrix U by adding the above row matrix to it. 11 12 13 … 1M 21 22 31 32 M1 M2 MM

12. Capacity
The capacity of a CDMA network is determined by maintaining a lower bound on the bit energy to interference density ratio, given by
W = Spread signal bandwidth
R = bits/sec (information rate)
α = voice activity factor
ni = users in cell i
N0 = background noise spectral density F
The power control scheme used in cdma networks is signal level based, I.e., the power control equalized the received power from the mobiles at the base station. W assume such a power control scheme throughput this work.
Number of users in every cell must satisfy: eq G
The capacities of the CDMA cells in a network must be considered jointly.
Let τ be that threshold above which the bit error rate must be maintained, then by rewriting Eq. F G
Back

13. Capacity Cases Equal capacity: all cells have an equal number of users
Optimized Capacity: A set of users in each cell obtained by solving following optimization problem H

14. Simulations Network configuration
COST-231 propagation model
Carrier frequency = 1800 MHz
Average base station height = 30 meters
Average mobile height = 1.5 meters
Path loss coefficient, m = 4
Shadow fading standard deviation, σs = 6 dB
Processing gain, W/R = 21.1 dB
Bit energy to interference ratio threshold, τ = 9.2 dB
Interference to background noise ratio, I0/N0 = 10 dB
Voice activity factor, α = 0.375
These values in Eq. G give upper bound on the relative interference in every cell, c_eff =
European COST (Cooperation in the field of Scientific and Technical Research) wave propagation model

15. Simulations – Equal Capacity
Average interference
Users in each cell: 18
Actual interference
Users in each cell: 17

16. Simulations – Optimized Capacity Vs Actual Interference Capacity
Optimized Capacity using average interference = 559
Simulated Capacity using actual interference = 554

17. More Simulations – Actual Interference
Simulated Capacity = 564
Simulated Capacity = 568

18. Individual Cell Capacity Comparison
Comparison of cell capacity for 3 simulation trials.
Comparison of average cell capacity for 50 simulation trials.

19. Extreme Cases Using Actual Interference – Non-Uniform Distribution
Maximum network capacity of 1026 with best case non-uniform user distribution
Maximum network capacity of 108 with worst case non-uniform user distribution

20. Results Actual interference model is computationally intensive.
Capacity obtained using average interference is close to the capacity obtained using actual interference for uniform user distribution.
Average interference model cannot predict extreme variations in network capacity under non-uniform user distribution.

21. Global Call Admission Control (CAC)
A CAC algorithm decides whether or not a network shall accept a call.
Designing a CAC algorithm for CDMA is harder than designing for TDMA or FDMA.
Self interference.
Affects the entire network.
A global CAC algorithm takes the entire network in account for every call making decision.

22. Mobility Model Call arrival process is a Poisson process with rate: λ
Call dwell time is a random variable with exponential distribution having mean: 1/μ
Probability that a call in cell i goes to cell j after completing its dwell time: qij
Probability that a call in progress in cell i remains in cell i after completing its dwell time: qii
Probability that a call will leave the network after completing its dwell time: qi
Poisson process: The numbers of changes in nonoverlapping intervals are independent for all intervals.

23. Mobility Model – Handoff Calls
Handoff calls (vji): calls that have moved from cell j to an adjacent cell i.
Handoff calls
Handoff traffic from cell j to an adjacent cell I is the sum of the proportion of new calls accepted in cell j that go to cell I and the proportion of handoff calls accepted from cells adjacent to cell j that go to cell I.
Bj : Call blocking probability for cell j
Aj : Set of cells adjacent to cell i
ρj : Total offered traffic to cell j j i
New arriving calls

24. Global CAC Algorithm
A new call is accepted if the following set of equations still hold upon acceptance.
Actual Interference case:
Average Interference case:

25. Simulator – Call Arrival and Admission Module (Global CAC)
From call removal module
To call removal module

26. Simulator – Call Removal Module (Global CAC)
From call arrival and admission module
To call arrival and admission module

27. Performance Measurements
Network throughput: Number of calls per unit time that are admitted and stay in the network till termination.
Blocking probability: For a cell, it is the ratio of rejected calls to total offered traffic to that cell.

28. Simulations – Network Parameters
Non-uniform traffic distribution
Group A (cells 5, 13, 14, 23) : 14 calls/time
Group B (cells 2, 8, 9, 19) : 14 calls/time
Rest of the cells : 3 calls/time
Ceff = 38.25
No mobility probabilities
qij = 0
qii = 0.3
qi = 0.7
Low mobility probabilities
High mobility probabilities Ai
qij
qii qi 3
0.020
0.240
0.700 4
0.015 5
0.012 6
0.010 Ai
qij
qii qi 3
0.100
0.000
0.700 4
0.075 5
0.060 6
0.050

29. Maximum users admitted per cell for average and actual interference for the three mobility cases.

30. Network throughput for average and actual interference for the three mobility cases.

31. Blocking probability for average and actual interference for the three mobility cases.

32. Results
Network throughput is always a little higher for average interference in all the three mobility cases.
Blocking probabilities are a little higher for actual interference for all three mobility cases.
Blocking probability is around 10% in all the three mobility cases for the cells with high demand.
Throughput is highest and blocking probability is lowest for the high mobility case.

33. Local Call Admission Control
A local CAC algorithm considers only a single cell for making a call admittance decision even though its design may look at the network as a whole.
A simple approach: Find N, the maximum number of users that are allowed in a cell, which is the same for all the cells in the network.
Disadvantage: Inefficient

34. Why? Maximum (N): 18 5 5 18 18 18 18 18 18 20 (5) 18 (5) 18 (18) 18
Rejected: = 2
Could have admitted 20 since adjacent cells have less traffic

35. Traditional CAC Algorithm
Define network throughput
A traditional CAC algorithm is formulated that calculates N, the maximum number of calls allowed in each cell. The optimization problem is given by

36. Our Optimized Local CAC Algorithm
Solve a constrained optimization problem that maximizes the network throughput with signal-to-interference ratio constraints as lower bounds.

37. Simulator – Call Arrival and Admission Module (Local CAC)
From call removal module
To call removal module

38. Simulator – Call Removal Module (Local CAC)
From call arrival and admission module
To call arrival and admission module

39. Erlang traffic and maximum number of calls allowed to be admitted per cell for all three mobility cases.
High mobility has an equalizing effect on non-uniform traffic distribution.

40. Network throughput for our optimized local CAC for all three mobility cases.

41. Theoretical and simulated network throughput for our optimized local CAC and traditional CAC for all three mobility cases.

42. Theoretical and simulated blocking probability for our optimized local CAC and traditional CAC for all three mobility cases.

43. Results
Our optimized local CAC algorithm adapts in response to the traffic demand due to users’ mobility.
Our local CAC network throughput is higher than traditional CAC throughput by nearly 13%.
Our local CAC algorithm strikes a good balance between the blocking probabilities of the low and high traffic cells.

44. Global CAC vs Local CAC Global
Call admission based on all the calls present in the network.
Slower.
Inherently optimized.
Adaptable.
Complexity: O(ceffM).
Local
Call admission based on calls present in the cell under consideration only.
Faster
Optimized only for a given traffic distribution profile.
Cannot compensate for big fluctuation in traffic.
Complexity: O(M)

45. Network throughput for our optimized local and global CAC algorithms.

46. Blocking probability for our optimized local and global CAC algorithms.

47. Summary
Capacity obtained using average interference is a good approximation for uniform user distribution only.
High mobility results in highest throughput because it equalizes non-uniform traffic.
Our optimized local CAC algorithm performance is better than traditional CAC algorithm.
Our optimized local CAC algorithm performance is just as good as a global for a given traffic distribution.