University of Texas at Dallas

University of Texas at Dallas Introduction to Wireless Communications Systems EE6390 Initiation and Performance Analysis of Handoff in CDMA Cellular Systems By Gandhar Dighe Hemanth Srinivasaraghavan Mukul Bhatnagar Dec 6,1999

ABSTRACT This project presents an overview of soft handoff, an idea which is becoming quite important because of its use in the IS-95 code division multiple access (CDMA) cellular standard. It involved a brief study of the various parameters affecting the handoff procedure. It also includes a Matlab implementation of the received power(with gaussian noise) versus the distance in the soft handoff case for a mobile moving between two base stations separated by distance d, along a straight line with fixed velocity. Furthermore a theoretical study on the performance of handover due to combined effect of channel availability and initiation was also done. Performance criteria of interest here were the forced termination probability and blocking probability. 1

INTRODUCTION The code division multiple access (CDMA) scheme has been considered as one of the possible choice of future standards in cellular networks because of its various advantages. In CDMA systems, the narrowband message signal is multiplied by a very large bandwidth signal called the spreading signal (a pseudo noise code sequence that has a chip rate in order of magnitude greater than the data rate of message signal). The process of transferring a mobile station from one channel or base station to another is called a handoff, which is an essential element of cellular communication. Since CDMA uses only one frequency, it uses a special handoff scheme with diversity, so called soft handoff. Soft handoff is a process in which a mobile unit can commence communication with a target station without interrupting the communication with the current serving base station (make before break). The traditional handoff scheme which requires the mobile to break communication with the current base station before establishing a new communication with other base station is called hard handoff(break before make). Efficient handoff algorithm can enhance system capacity and service quality cost efficiently. Soft handoff is a very intriguing technology and has some well established benefits over conventional handoff schemes, such as fade margin improvement, higher uplink capacity, reduction/elimination of ping-pong effect [Appendix B] and hysteresis margin and imposing fewer time constraints on the network. It also has certain disadvantages such as increase in downlink interference, requirement of additional network resources and its complex implementation. 2

But due to its promising scopes in further improving the cellular system scenario, a great deal of research and study is continuing for the different tradeoffs involved in the handoff parameter settings. Successful and reliable handoffs are major issues for system performance and thus we have concentrated our research, analysis and implementation in studying some of the factors, which project the advantages of using soft handoff and factors which enable successful and reliable handoffs. PROJECT DESCRIPTION: The analysis is done for a model with two base stations separated by a distance d and a mobile moving from one base station to another along a straight line. The analysis and implementation is based upon various factors affecting the handoff procedure. The decision to initiate a handoff can be made by measuring several quantities such as received signal level from the communicating and neighboring base stations, received signal strength to interference ratio, and the bit error rate. The simplest and the most commonly used method is based on received signal strength. The received signal strength in land mobile communication has three kinds of variations i.e. path loss, shadowing and Rayleigh fading [Appendix B]. Here we are considering the effects of the received signal strength due to path loss only. For path loss in a macrocell environment, we have used the Hata-Okumura model (which considers the propagation loss between isotropic antennas, quasi smooth terrain and using the (standard ) empirical formula for urban area propagation loss) due to simplicity of the formula in relating the distance with path loss. This model is however applicable only for a flat urban environment and to make the model applicable to suburban and rural area a 3

ground cover factor [Appendix B] has to be introduced. A limitation of this model is that it does not consider the structure of buildings and roads. Furthermore, the project involves a theoretical study on performance of handover due to combined effect of initiation and channel availability. Performance criteria of interest here are average number of handoff attempts per call, forced termination probability and blocking probability [Appendix B]. Figure 1 Figure 1 [1] shows the relationships between forced termination probability, hysteresis level and offered traffic. The forced termination probability increases as the hysteresis level or offered traffic increases. At high traffic loads, the channel availability becomes increasingly significant in determining the forced termination probability. It can be seen that the forced termination probability decreases and tends to a minimum level as the offered traffic decreases. When the offered traffic is low, the forced terminations are 4

mainly due to handoff initiations. It is observed that unrealistically low forced termination probability is obtained if channel availability is not considered. IMPLEMENTATION: The standard formula for Hata-Okumura path loss model is given by L(urban)(dB) = 69.55 + 26.16logfc - 13.82 log hte - a(hre) + (44.9-6.55log hte)log(d) where, fc is the frequency in MHz from 150MHz to 1500MHz, hte is the effective transmitter antenna height(meters) ranging from 30m to 200m, hre is the effective receiver antenna height(meters) ranging from 1m to 10, d is the T-R separation distance(km) and a(hre) is the correction factor for effective mobile antenna height which is a function of the size of the coverage area. For a small or medium sized city the correction factor is given by a(hre) = [(1.1log fc 0.7)hre (1.56log fc-0.8)]db We also compared the Free Space path loss model with the Hata-Okumura model (Figure 2). The Free Space Path loss equation is given by L (free) = 32.4 + 20logfc + 20log (d) The average received power at the mobile station is calculated as Pr(d) = EIRP(dBm) - L(urban)(dB) + Gr(dB) Where EIRP is the Effective Isotropic Radiated Power (1 kw) and Gr is the gain of the receiving antenna. Figure 3b shows variation of received signal strength of the mobile from both base stations when no noise was present in the system. 5

To analyze a slightly more realistic cellular environment two Gaussian noise sequences were introduced. By introducing a hysteresis margin, it is clearly seen that number of handoffs decrease (Figure 3a has a hysteresis margin of 10 db, we see that instead of handoffs taking place at A, B, C & D we see handoffs taking place only at X&Y). However an excess hysteresis margin will cause an initiation delay. If the delay persists for longer intervals of time the call will be dropped due to deteriorating signal conditions. Clearly, there exists a tradeoff between hysteresis and initiation delay. Figure 2 Figure 3a & 3b 6

CONCLUSION The simulation observations do show the decrease in the number of unnecessary handoffs on incorporation of an optimum hysteresis margin and also gives an insight of the tradeoffs involved in introducing an optimum hysteresis margin and the associated initiation delays. The detailed study of [1] shows that handoff initiation as well as channel availability has significant impact on the forced termination probability. If either one of them is not considered, one may obtain unrealistically low forced termination probability. The average number of handoff attempts per call and the forced termination probability cannot be minimized simultaneously. Tradeoffs have to be made between the forced termination probability and average number of handoff attempts per call. Soft handoff promises a better performance than hard handoff, through the exploitation of macroscopic diversity and minimizing the hysteresis margin. It is a complex technology and various studies indicate that system performance may be very sensitive to the settings of some parameters. FUTURE DIRECTIONS The quantitative tradeoffs between various advantages and disadvantages of soft handoff need to be further investigated, as do the parameter settings. As in [1], there is a need to consider the combined effects of various parameters on the overall performance of handoff procedure rather then considering their individual effects only. 7

Scope within the project To make the communication environment more realistic there is a need to consider the effects of fading and shadowing in the propagation model. It is also possible to extend the study by comparing other Propagation Loss Prediction Models like the Walfisch-Bertoni Model to investigate microcellular environments, effects of structure of the buildings etc. to get a better understanding on the factors effecting system design. ACKNOWLEDGEMENT We wish to thank Dr. Murat Torlak for his continuous guidance and suggestions during the project. We also, wish to thank Ramakrishna Vedantham our teaching assistant, for his guidance in implementation of our experiment. REFERENCES: [1] J.L.Pan, Djuric.P.M and Rappaport, A Simulation Model of Combined Handoff Initiation and Channel Availability In Cellular Communications, VTC, Mobile Telephony for the Human Race, IEEE 46 th Vol 3, pp 1515-1519, 1996. [2] M.Hata, Empirical Formula for propagation loss in land mobile radio services, IEEE trans. On Vehicular Technology, vol-29, pp317-325, Aug 1980. [3] M Hata, Propagation Loss Prediction Models for Land Mobile Communications, Microwave and Millimeter Wave Technology Proceedings, pp15-18, ICMMT 1998. [4] Daniel Wong, Teng Joon Lim Soft Handoffs in CDMA Mobile Systems, IEEE Personal Communications Dec 1997. 8

[5] R.Peterson, K.Cutts, and J.Huang, System Performance prediction for Personal Communication Systems, IEEE VTC, pp 749-753, 1995. [6] Rajiv Vijayan and J.M.Holtzman, A Model for Analyzing Handoff Algorithms, IEEE trans. On Vehicular Technology,vol-42,pp351-356, Aug1993. [7] Tripathi.N.D, Reed J.H, and Vanlandingham, Handoff in Cellular Systems, IEEE Personal Communications Dec 1998. [8] Wireless Communication: Principal and Practice, Theodore S. Rappaport. [9]Mobile Cellular Telecommunications Systems, William C.Y.Lee. [10] Murase A, Symington.I and Green E, Handover Criterion for Macro and Microcellular Systems, VTC 1991, Gateway to the Future Technology in Motion, 41 st IEEE, pp524-530, 1991. [11] http://www.cdg.org [12] http://iel.ihs.com 9

APPENDIX A: MATLAB CODE % Study of a propagation loss using Okumura-Hata Model, and comparing it with Free space model % Making analysis for various frequencies and also varying antenna heights % Velocity is being kept constant, analysis has been done for various distance cells % The program calculates the path loss and received power for the uplink, downlink and urban environment frequencies clc clear all; hte=150; %height of transmitting base station antenna in meters hre=10; %height of receving antenna of mobile station in meters sda=3; %standard deviation of noise for Base station A sdb=5; %standard deviation of noise for Base station B noisea=sda*randn(1,50); noiseb=sdb*randn(1,50); disp('uplink freq=835 Mhz') disp('downlink freq=880 Mhz') disp('urban environment =900 Mhz') fc=input('do You want path loss for uplink, downlink or urban environment frequency?=') for d=1:50 % path loss calculation Between Mobile & Base station A LA(d)=(69.55+26.6*log10(fc))-(13.82*log10(hte))-((1.11*log10(fc)-0.7)*(10)+(1.56*log10(fc)- 0.8))+((44.9-6.55*log10(hte))*log10(d)); %path loss calculation Between Mobile & Base station B LB(d)=(69.55+26.6*log10(fc))-(13.82*log10(hte))-((1.11*log10(fc)-0.7)*(10)+(1.56*log10(fc)- 0.8))+((44.9-6.55*log10(hte))*log10(51-(d))); % path loss calculation for free space model LF(d)=32.4+20*log10(fc)+20*log10(d); % Received power at A without noise SrA(d)=60-LA(d); % Received power at B without noise SrB(d)=60-LB(d); % Received power at A with Gaussian noise sd=3 PrA(d)=60-LA(d)+noiseA(d) ; % Received power at B with Gaussian noise sd=5 PrB(d)=60-LB(d)+noiseB(d); end figure(1) subplot(2,1,1);plot (PrA); hold on plot (PrB,'m'); axis([0 50-90 -50]); xlabel('distance');ylabel('signal strength') grid subplot(2,1,2);plot(sra); hold on plot(srb,'m'); axis([0 50-90 -50]); grid figure(2) plot(la); hold on plot(lf); 10

APPENDIX B : Ping-pong effects in Hard Handoffs: The handing off back and forth between two base stations in a relatively short period of time is known as ping-pong effect. This problem is analogous to the movement of a ping-pong ball between the two ends of the table in a ping-pong game. The problem is that each time a handoff is executed there is some overhead in the network resulting in undesirable use of network resources. In order to reduce the ping-pong effect a standard feature of hard handoff algorithms is the incorporation of hysteresis. Performance Measures: Channel Conditions Signal Conditions 1 2 3 4 W S S S B X S S B B W-Both signals received from A&B acceptable. X-Signal received from base station A is acceptable but that received from B is not. Y-signal received from B is acceptable but that received from A is not. Z-Both signals received from A & B are not acceptable 1-Channel at both A & B are available 2-Channel at A is available but at B is not 3- Channel at B is available but at A is not 4-No channels at A & B are available Y S B S B Z B B B B Table 1:Blocking Conditions B: Blocked, S: Successful Table 1 summarizes all the cases including those of blocked new call attempts. The number of blocked new call attempts NB is calculated as follows: NB = NW4 + NX3 + NX4 + NY2 + NY4 + NZ1 + NZ2 + NZ3 + NZ4 The blocking probability PB is defined by: PB = NB/NG Where NG represents the total number of generated new call attempts. Initiation & Channel Conditions Signal Conditions 1 2 3 E C C C F C D D G D D D Table 2 :Dropping Conditions D: Drop,C: Continue E- signal quality of the current link is acceptable F-signal quality of the current link is not acceptable but that of the alternative link is acceptable G-Both the current and alternative signal qualities are not acceptable 1-Handoff is initiated and there is an available channel in the alternative base station 2- Handoff is initiated and there is no available channel in the alternative base station. 3-Handoff is not initiated 11

Table 2 shows all the cases for which an ongoing call is dropped.dropping of a call at any time during its intended duration time results in forced termination of the call. Any call which enters service and is subsequently dropped encounters one of the events F2, F3, G1, G2 or G3. The forced termination probability, PFT is defined by PFT = ND/(NG NB) Where ND is the number of observed call dropping events. NAV = NH/NG Where NH represents total number of generated handoff attempts. Variations in received signal strengths: Figure 1 This figure shows the variation in the signal strength as the mobile moves from BS1 to BS2. The variation in the signal strength is due to path loss, shadowing and fast fading. The fast fading is usually modeled as a Rician distribution with K (Rice factor) varying with distance. When K=0 the variation is Rayleigh fading. Ground Cover Factor: The ground cover factor is a function of the percentage of the area covered by buildings. To calculate this a database of buildings and roads is required. α = [(Σ Sj)/area]*100 %, where Sj = Total number of structures in the given area. 12