CDMA2000 1X CAPACITY DECREASE BY POWER CONTROL ERROR IN HIGH SPEED TRAIN ENVIRONMENT

Simon Shin, Tae-Kyun Park, Byeung-Cheol Kim, and Yong-Ha Jeon Network R&D Center, SK Telecom, 9-1, Sunae-dong, Bundang-gu, Sungnam City, Gyunggi-do, South Korea

Dongwoo Kim School of Electrical Engineering & Computer Science, Hanyang Univ. 1271 Sa-dong, Ansan, Kyungki-do 425-791, South Korea

Keywords: CDMA2000 1X, Doppler shift, capacity, power control, Korea Train Express

Abstract: CDMA2000 1X capacity was analysed in the high speed train environment. We calculated the power control error by Doppler shift and simulated bit error rate (BER) at the base station. We made the interference model and calculated the BER from lower bound of power control error variance. The reverse link BER was increased by high velocity although there was no coverage reduction. Capacity decrease was negligible in the pedestrian (5 km/h), urban vehicular(40 km/h), highway and railroad(100 km/h) environment. However, capacity was severely reduced in high speed train condition(300 km/h and 350 km/h). Cell-planning considering capacity as well as coverage is essential for successful cellular service in high speed train.

1 INTRODUCTION train with 300 km/h velocity. Received power, transmitted power, and pilot chip energy to Cellular mobile telephone and data communication interference ratio (Ec/Io) of mobile station were not correlated with the mobile velocity. We could serve services are very popular. Cellular service is usable in anywhere, even though tunnel, sea, and successfully the CDMA2000 1X in the KTX by underground places. Railroads and highways are existing cellular network. Experimental results confirm that coverage is not main service area of cellular service because many people move through them. Many countries adopts reduced by speed mobility. However, a few base the high speed train for transportation capability and station and frequency assignment (FA) is established in rural area. It makes the capacity shortage. One convenience for example TGV in France, ICE in Deutschland. Korea also constructs the Korea Train KTX train is composed of 20 cars and one car seats Express (KTX) between Seoul and Pusan. KTX 64 persons. It means more than 2000 calls can connect one base station when two opposite trains travels by 300 km/h speed and will be upgraded by 350 km/h speed. are in coverage of same base station. In the dense High speed mobility can damage pilot urban area, heavy traffic cell is split for dividing traffic. This method cannot be used for KTX acquisition, steady-state demodulation, code tracking, and power control. We experimented the because traffic cannot be split in one train. We must CDMA2000 1X service quality using channel increase the FA for capacity improvement. Adjacent cell uses same FA for smooth handoff. If one cell simulator in Test-bed network. There was no quality degradation such as MOS(mean of score), data increases the FA number, adjacent cells also throughput, call drop, and call fail. We got the same increase the FA number. Therefore we have to estimate the capacity of one FA for efficient network results in the KTX. Call origination, response, data download, and data upload was successful in the investment.

109 Shin S., Park T., Kim B. and Kim D. (2004). CDMA2000 1X CAPACITY DECREASE BY POWER CONTROL ERROR IN HIGH SPEED TRAIN ENVIRONMENT. In Proceedings of the First International Conference on E-Business and Telecommunication Networks, pages 109-113 DOI: 10.5220/0001398401090113 Copyright c SciTePress ICETE 2004 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS

Velocity does not influence the coverage, but it variance of random variable measured by Watt- Si does not mean that capacity is not varied by velocity. unit are as follows. Network performance can be changed with user 1 number because CDMA is the system limited by βm + β 2σ 2 E[S ] = e i 2 i interference. CDMA performance is influenced by i 2.2 multi-user interference. CDMA capacity is limited 2 2 2 2 by Walsh code on the forward link and by cell 2βmi +β σ i β σ i 2.3 Var(Si ) = e (e −1) loading on the reverse link. Cell loading is the received power increase by mobile station at the base station receiver. High cell loading increases the β is (ln10)/10 in equation 2.2 and 2.3. Received noise level at the base station receiver and it signal strength of user o is as equation 2.4. 2.4 degrades the bit energy to noise ratio (Eb/Nt). Low So = ε o (dB) Eb/Nt increases the transmitted power of mobile station by power control and high transmitted power In equation 2.4, it is assumed has a log-normal ε o degrades Eb/Nt repeatedly. CDMA2000 1X uses the distribution with variance 2 . σ e (dB) fast power control for corresponding to variable Received power from other cell users must be radio channel. High velocity of mobile causes the considered to analyse multi-cell environment. power control error due to Doppler shift and fast Interference from other cell is represented by in-cell radio channel environment variation. This paper interference. If there are K users in the each cell and analysed the capacity decrease by power control average propagation loss ratio of in-cell and other error. Section 2 will derive the calculation of power cell is g, other cell interference is gK. Standard control error and Section 3 will show the simulation deviation of received signal is the control function of results. base station transmitted power because of fading. Total interference considering both in-cell and other cell interference is as equation 2.5

2 POWER CONTROL ERROR K gK 2.5 SIMULATION I = ∑ S ei + ∑ S oi i=1 i=1 2.1 Interference Model S ei is signal power of in-cell user i with standard deviation . is signal power of other cell user i σ ei Soi We starts making interference model by followed with standard deviation . Therefore, expected σ oi assumption. value and variance of I is as follows - User o who is analysis target has the appropriate received signal strength 0 dB. K gK - There are K+1 users including target user o. 2.6 E[I] = ∑ E[Sei ] + ∑ E[Soi ] - K+1 users are served in same cell. i=1 i=1 - K+1 users are riding the KTX. K gK 2.7 - There is no other user who is not riding KTX in Var(I) = ∑Var(Sei ) + ∑Var(Soi ) the analysed cell. That is to say, all of users in i=1 i=1 the analysed cell are riding KTX. - Other cell user K interference is assumed in-cell Probability density function (PDF) of I can be user gK. approximated by log-normal random variable ξ , - There is no error in the power control procedure because I is sum of independent log-normal random of the other cell user. variables. Expected value and variance of ξ is as follows Received signal in base station from interference 1 user i is as follows: m + σ 2 K 2 K E[ξ ] = e 2.8 + Si = mi ε i (dB) 2.1 2 2 Var(ξ) = e2mK +σ K (eσ K −1) 2.9 is the signal strength without power control error. mi ε i is the random variable due to power control error It is assumed power control error variance of in-cell and has log-normal distribution with 0 dB expected 2 2 users and other cell is σ e and σ o , respectively. value and 2 dB variance. Expected value and σ i Expected value and variance of I is defined with

110 CDMA2000 1X CAPACITY DECREASE BY POWER CONTROL ERROR IN HIGH SPEED TRAIN ENVIRONMENT

expected value and variance of ξ using Wilkinson’s ∞ Method. γ 2.19 Pe = Q(e )g(γ )dγ ∫0 1 2 2 1 2 2 1 2 βm+ β σ e βm+ β σ o mK + σ K Ke 2 + gKe 2 = e 2 2.10 Equation 2.19 is approximated to equation 2.20 using the expansion of central difference. 2 2 2 2 2 2 2 2 2 2 Ke 2 βm+ β σ e (e β σ e − 1)+ gKe 2 βm+ β σ o (e β σ o − 1)= e 2mK +σ K (eσ K − 1) 2 m 1 m + 3σ 1 m − 3σ 2.11 γ γ γ γ γ 2.20 Pe ≈ Q(e )+ Q(e )+ Q(e ) 3 6 6 We get mk and σ k from equation 2.10 and 2.11.

⎡ ⎤ 2.3 Variance of Power Control Error

⎢ 2 2 2 2 2 2 2 2 ⎥ β σ e β σ e β σ o β σ o 2.12 2 ⎢e ()e −1 + ge (e −1)⎥ σ K = ln⎢ 2 +1⎥ Each base station controls the received power from 1 2 2 1 2 2 ⎢ ⎛ β σ e β σ o ⎞ ⎥ mobile station in cellular CDMA network. There is K⎜e 2 + ge 2 ⎟ ⎢ ⎜ ⎟ ⎥ four main error factor when tracking the received ⎣ ⎝ ⎠ ⎦ power. They are the quantum(σ q), decoding(σ d), 1 1 β 2σ 2 β 2σ 2 ⎛ e o ⎞ 1 2 measurement( ), and propagation delay( ) error. m = ln K + ln⎜e 2 + ge 2 ⎟ − σ 2.13 σ m σ p K ⎜ ⎟ 2 K ⎝ ⎠ Generally, σ m and σ p is much larger than σ q and

σ d. If error factor is statistically independent, 2.2 Mean BER Calculation 2 2 2 2 2 2.21 σ e = σ m + σ p + σ q + σ d Bit energy to interference ratio of BPSK having bandwidth W is given by equation 2.14. ν is the maximum velocity of mobile. fc is the

carrier frequency. c is the light velocity. We assume E S / R 2.14 b = eo b the bandwidth is narrow enough to neglect N + I 1 ⎛ K gK ⎞ 0 0 N + ⎜ S + S ⎟ bandwidth. Maximum Doppler shift is 0 W ⎜∑ ei ∑ oi ⎟ 2.22 ⎝ i=1 i=1 ⎠ f d = fcv / c

No is the power spectral density of background noise Propagation delay is 2d/c for closed loop power and Rb is the bit rate. BER of BPSK is given by control. d is the distance between mobile and base ⎛ E ⎞ station. Therefore, processing delay is P = Q⎜ 2 b ⎟ 2.15 e ⎜ ⎟ ⎝ N 0 + I 0 ⎠ α 2d 2.23 The received signal is log-normal random variable Tp < − due to imperfect power control and BER is given by f d c We define the pm is measured received power in dB γ 2.16 scale. is the received power in linear scale and p Pe = Q(e ) p0 m 2 is 10log p0 . If σ m1 is variance of natural γ is the Gaussian random variable and its mean and logarithm( of ln p0 ) p0 variance is given by 2 2 2 2.24 σ m = (10loge) σ m1 1 ⎛ W ⎞ ⎜ ⎟ 2.17 mγ = ⎜ln 2 − mK ⎟ 2 ⎝ Rb ⎠ It is assumed that fluctuation of received power can be neglected during measurement time Tm. 1 Measurement period T is the main factor of 2 2 2 2 2.18 m σ γ = (β σ e + σ K ) 4 processing delay Tp. Power control error is due to multi-user interference and white Gaussian noise. We assume the power control makes received power With mean γ , mean BER is calculated by from any mobile to be almost same. It enables us to

111 ICETE 2004 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS

0 0 10 10 5km/h 5km/h 40km/h 40km/h -1 100km/h -1 100km/h 10 10 300km/h 300km/h 350km/h 350km/h

-2 -2 10 10 e e t ra r r rat -3 -3 10 10 rro e erro t t i i B B -4 -4 10 10

-5 -5 10 10

-6 -6 10 10 5 10 15 20 25 30 5 10 15 20 25 30 Number of users Number of users Figure 1(a): and 2 2 2 2 Figure 2(a): and 2 2 2 2 g = 0.3 σ p +σ q + σ d = 0d.5( B) g = 0.634 σ p +σ q + σ d = 0d.5( B)

0 0 10 10 5km/h 5km/h 40km/h 40km/h -1 100km/h -1 10 10 100km/h 300km/h 300km/h 350km/h 350km/h -2 -2 10 10 e e t r rat

-3 ra

r -3 10 10 rro erro t e i t i B B -4 10 -4 10

-5 10 -5 10

-6 10 -6 5 10 15 20 25 30 10 5 10 15 20 25 30 Number of users Number of users Figure 1(b): g = 0.3 and 2 2 2 2 σ p + σ q + σ d = 1d.5( B) Figure 2(b): and 2 2 2 2 g = 0.634 σ p + σ q + σ d = 1d.5( B)

determine the lower bound. Multi-user interference y is lnpo in equation 2.26. Cramer-Rao bound is modelled Gaussian process that increases one- provides lower bound of lnpo variance. Equation sided power spectral density No to Nt. 2.27 is obtained from this lower bound and equation 2.26. p 0 2.25 Nt = N0 + K(1+ g) 2 −1 W ⎧ T ⎫ 2 ⎪ 2 m ⎡ ∂ y / 2 ⎤ ⎪ 2.27 σ m1 ≥ ⎨ ⎢ e s(t)⎥ dt⎬ N ∫0 ∂y ⎩⎪ t ⎣ ⎦ ⎭⎪ po is K-1 interference signal power. W is the receiver bandwidth. g is the interference constant that represents other cell interference. Received signal is using equation 2.24 and 2.25 . p0 s(t) 200(loge) 2 ⎡ N K(1+ g)⎤ 2 0 2.28 σ m ≥ ⎢ + ⎥ T p W ⎛ y ⎞ 2.26 m ⎣ 0 ⎦ p0 s(t) = exp⎜ ⎟s(t) ⎝ 2 ⎠

112 CDMA2000 1X CAPACITY DECREASE BY POWER CONTROL ERROR IN HIGH SPEED TRAIN ENVIRONMENT

in KTX condition. Although coverage was not If we define T1 = Tp −Tm and use equation 2.23, we decreased, capacity was reduced severely in high obtain equation 2.29 from 2.21 mobile speed of 300 km/h. We don’t have to

−1 consider the mobile velocity in cell-planning ⎛ ⎞ ⎡ N ⎤ 2 2 α 2d 0 K(1+ g) 2 2 2 because capacity reduction is negligible in highway σ low ∆ 200(log e) ⎜ − − T1 ⎟ ⎢ + ⎥ + σ p + σ q + σ d ⎜ f c ⎟ p W ⎝ d ⎠ ⎣ 0 ⎦ and railroad. However, capacity is severely reduced 2.29 in KTX for its high velocity. We must consider the number of passenger carried by KTX when opposite train is met. Cell-planning without considering 3 SIMULATION RESULTS capacity can make the burst error in high traffic intensity. It causes not only quality degrade but also call drop. We must consider capacity as well as We simulated the BER of received signal at base coverage for cellular network planning. station when user number was 5 ~ 30 persons. Simulation circumstance was assumed to be pedestrian(5 km/h), urban vehicle(40 km/h), highway and railroad(100 km/h), KTX(300 km/h), REFERENCES and upgraded KTX(350 km/h). BER was calculated from equation 2.20. mr and A.M. Viterbi, A.J. Viterbi, and E. Zehavi, “Other-cell σ r of equation 2.20 was calculated from equation interference in cellular power-controlled CDMA,” 2.17 and 2.18. mk and σ k was calculated from IEEE Trans. on Communications, vol. 42, No. 2/3/4, equation 2.12 and 2.13. We used the lower bound of pp. 1501-1504, Feb./Mar./Apr. 1994. equation 2.28 when calculating equation 2.12 and B. Hashem and E. Sousa, “Increasing the DS-CDMA 2.13. Figure 1 and 2 shows user number and BER system reverse link capacity by equalizing the when mobile speed is 5 km/h, 40 km/h, 100 km/h, performance of different velocity users,” Pro. IEEE 300 km/h, and 350 km/h. We assumed Rb = 4.8 kbps, ICC’98, Atlanta, Georgia, USA. W = 1.2288 MHz, d = 4 km, α = 0.1, T1 = 100 us, N.C. Beaulieu, A.A. Abu-Dayya., and P.J. McLane, 2 2 No/po = 5 us, σ o = 3.9 (dB) . Figure 1 and 2 shows “Estimation the distribution of a sum of independent increase of mobile speed degrades the BER of lognormal random variables,” IEEE Trans. on received signal. This means the reverse link capacity Communications, vol. 43, No. 12, pp. 2869-2873, Dec. decreases to maintain the quality of service. Figure 1 1995. and 2 shows the result when interference constant g M.B. Pursley, “Performance evaluation for phase-coded is 0.3 and 0.634, respectively. Larger interference spread spectrum multiple-access communication – constant increases the receiver sensitivity with user part I : system analysis,” IEEE Trans. on number. (a) and (b) of each Figure shows the results Communications, vol. COM-25, No. 8, pp. 795-799, in the case of 2 2 2 2 and σ p + σ q + σ d = 0d.5( B) Aug.1977. 2 2 2 2 . High speed enlarges the σ p + σ q + σ d = 1d.5( B) J.M. Holtzman, “A simple, accurate method to calculate Doppler shift in equation 2.22. Doppler shift spread-spectrum multiple-access error probabilities,” increases the lower bound of error variance in IEEE Trans. on Communications, vol. 40, No. 3, pp. equation 2.29. This increases the BER and degrades 461-464, Mar. 1992. the quality of service. C.C. Lee and R. Steele, “Closed-loop power control in In urban vehicular (40 km/h) condition, BER CDMA systems,” IEE Proc.-Commun., vol. 143, pp. increase by Doppler shift was negligible. BER 231-239, Aug. 1996. degradation was not severe even though highway R.N. McDonough and A.D. Whalen, Detection of Signals and railroad (100 km/h) condition. We could plan in Noise, 2nd ed. San Diego : Academic Press, 1995. the cellular network assuming constant capacity with D. Torrieri, “Uplink capacity of a CDMA network,” mobile speed before KTX service. However, BER internal communication. was dramatically increased in KTX circumstance. User number in KTX was limited to 17 ~ 26 persons to maintain BER lower than 1 %.

4 CONCLUSION

We measured the coverage of CDMA2000 1X network experimentally and simulated the capacity

113