Performance Estimation of DS-CDMA Cellular Systems

Get Complete Project Material File(s) Now! »

Long-Range Dependent Traffic

A number of studies have shown that certain types of traffic exhibit a characteristic known as long-range dependence e.g. LAN [63], FTP [64], and video [65] traffic. Traffic that exhibits this characteristic has structural similarities on a wide range of time scales and this means that the generated traffic stream has no natural burst length and instead appears bursty over a wide range of time scales, for example, from milliseconds to minutes [63]. This phenomenon is illustrated in Figure 4.3 where a long-range dependent Pareto traffic stream (see Section 4.4.4) is compared with a non-long-range dependent Poisson traffic stream (see Section 4.4.2) over a range of time scales. The figure shows that the long-range dependent traffic stream maintains its burstiness over a range of time scales while the non-long-range dependent traffic stream does not.

Traffic Model Development

This thesis investigates packet based services and the VBR traffic streams generated by individual users must be captured in a traffic model. Elements on the transmission path between the traffic source and traffic destination can alter the characteristics of the traffic and in many cases these elements will not be under the control of the system operator e.g. traffic originating from the public Internet could be affected by routers and firewalls that are external to the system. From a system operator’s point of view, it is best to develop traffic models that describe user traffic at a point within the system that is under the operator’s control. The traffic model reference point used in this thesis is shown in Figure 4.4 and this point was chosen because the radio component of a cellular system is the primary focus of this thesis i.e. the traffic offered to the BTS is modelled.

Poisson Model

A Poisson process represents a sequence of independent events that occur randomly in time [66]. This process is assumed in many engineering applications and, for example, is the key assumption in the well known Erlang traffic formulae that are used to estimate the probability that a service request is blocked in a telecommunications network [12]. Intuitively, a Poisson process may be an appropriate model for voice call arrivals or web browsing sessions (see Section 4.3) as these could be assumed to be independent events [67]. It is arguable whether such a model is appropriate for packet arrivals but it is still an interesting model to consider because of its simplicity and wide use in many traffic related studies [64].

Pareto Model

The Pareto model assumes that the inter-arrival time between packets has a Pareto distribution and with an appropriate parameter selection the resultant traffic stream exhibits longrange dependence and appears bursty over a wide range of time scales (see Section 4.3.1) [69]. This characteristic has been observed in a number of different types of traffic [63, 64, 65]. The probability density function of the Pareto distribution is given by fX(x) = a b 1 + x b −(a+1) b > 0 (4.7) where a is the shape parameter, b is the scale parameter, and x represents the inter-arrival time between fixed size packets offered to the BTS scheduling queue.

READ Generic Model for Power Electronic Systems

Traffic Scheduling

In DS-CDMA cellular systems there is a soft capacity limit on the radio channel and additional spreading codes can be allocated to users at the cost of increased system interference and reduced propagation dependent performance. With dynamic resource allocation, a BTS can vary the number and length of spreading codes allocated to a user on a frame-by-frame basis. This form of dynamic resource allocation is essential for VBR traffic streams as it would be inefficient to allocate a fixed size radio bearer for the duration of a user’s service, as there would be periods of time where the allocated bearer capacity would go unused and could be better allocated to another user.

Contents :

1 Introduction
- 1.1 The Evolution of Wireless Cellular Systems
- 1.2 Packet Based Services and Variable-Bit-Rate Traffic
- 1.3 Context and Objectives of this Thesis
- 1.4 Contributions and Structure of this Thesis
2 Performance Estimation of DS-CDMA Cellular Systems
- 2.1 Introduction
- 2.2 Direct-Sequence Spread-Spectrum
- 2.3 Direct-Sequence Code Division Multiple Access
- 2.4 DS-CDMA Cellular Systems
  - 2.4.1 Power Control
- 2.5 DS-CDMA Performance Estimation
  - 2.5.1 OVSF Spreading Codes
- 2.6 Receiver Diversity
  - 2.6.1 RAKE Receiver
  - 2.6.2 Antenna Space Diversity
- 2.7 Summary
3 Outdoor and Indoor Propagation Environments
- 3.1 Introduction
- 3.2 Radio Propagation Mechanisms
- 3.3 Large-Scale Path Loss
  - 3.3.1 Distance Dependent Path Loss
  - 3.3.2 Lognormal Shadowing
- 3.4 Small-Scale Fading
  - 3.4.1 Rayleigh and Ricean Fading
- 3.5 Propagation Models used in this Thesis
  - 3.5.1 Outdoor Macro-Cellular Model
  - 3.5.2 Indoor Engineering School Tower Model
- 3.6 Summary
4 Teletraffic and Quality of Service
- 4.1 Introduction
- 4.2 User Services
  - 4.2.1 Quality of Service
- 4.3 Traffic Characterisation
  - 4.3.1 Long-Range Dependent Traffic
- 4.4 Traffic Model Development
  - 4.4.1 CBR Model
  - 4.4.2 Poisson Model
  - 4.4.3 Negative Binomial Model
  - 4.4.4 Pareto Model
  - 4.4.5 Discrete Auto-Regressive (DAR) Model
- 4.5 Traffic Scheduling
- 4.6 Summary
5 Simulation Model Development
- 5.1 Introduction
- 5.2 Downlink Simulation Model
- 5.3 Traffic Generation
- 5.4 Traffic Scheduling
- 5.5 Cellular System Deployment
  - 5.5.1 Outdoor Macro-Cellular System
  - 5.5.2 Indoor Pico-Cellular System
- 5.6 Received Power Estimation
- 5.7 Performance Estimation
- 5.8 Summary
6 Downlink Performance in Outdoor Macro-Cellular Environments
7 Downlink Performance in an Indoor Pico-Cellular Environment
8 Downlink Performance Sensitivity to Traffic Type
9 Traffic Shaping and Quality of Service
10 Implications of Variable-Bit-Rate Traffic on System Design