Performance Analysis of LTE Downlink System with High Velocity Users

Transcription

1 Journal of Computational Information Systems 10: 9 (2014) Available at Performance Analysis of LTE Downlink System with High Velocity Users Xiaoyue WANG, Di HE Department of Electronic Engineering, Shanghai Jiao Tong University, Shanghai , China Abstract This paper focuses on the performance analysis of LTE downlink system with high user-velocity. A novel simulator of Rician fading has been proposed for better simulation characteristics. Then Three channel models for different scenarios are presented to simulate the high speed LTE condition. One of the channel models is provided by 3GPP LTE and it is ideal. The other two models are more realistic, which consider the channel fadings and multipaths. The investigation compares the performance of Doppler-shift, bit error rate (BER), channel time delay of each model in different modulation modes. The results show that the performance of LTE with high velocity users is attractive and the two realistic models we provided have potential to use under high user-velocity condition. Keywords: LTE; High Velocity Users; Channel Model 1 Introduction With further advances in high speed conditions (such as High speed train communications condition, low level air-ground communications condition), the support of high data rate and voice communications for high velocity users has become urgent demands. Long Term Evolution (LTE), which can provide data rate up to 100Mbps and bandwidth up to 20MHz [1], has been considered as one of the technologies to support high velocity users. Quite a few works have been reported on the performance of LTE downlink system [2-3]. But the performance analysis with high velocity users (users speed is greater than 300km/h) was barely mentioned. Our work mostly focuses on analyzing LTE downlink system with high user-velocity. We firstly propose an improved Jakes simulator for propagation channel simulation. Then three channel models for different scenarios are provided and results of Doppler-shift, bit error rate (BER) and channel time delay are given. In this paper, the high speed train (HST) condition is considered. This paper is organized as follows: Section II describes the channel model. Followed by channel modeling in Section III and performance evaluation in Section IV, simulation result in Section V, the conclusions are drawn in Section VI. Corresponding author. address: wxyphoebe@sjtu.edu.cn (Xiaoyue WANG) / Copyright 2014 Binary Information Press DOI: /jcis10126 May 1, 2014

2 3646 X. Wang et al. /Journal of Computational Information Systems 10: 9 (2014) Channel Model Consider a single-user, single-antenna, turbo coded OFDM system for LTE downlink system. The Doppler shift in HST is given by 3GPP LTE [4] as f s (t) = f d cos θ(t). (1) Where f s (t) is the Doppler shift and f d is the maximum Doppler frequency, which can be calculated by f d = vf c c, v represents velocity of train, c is the speed of light, f c is the carrier frequency. θ(t) is the angle between the direction of the running train and the base station. The cosine of angle θ(t) is given by D s/2 vt D 2 min +(D s/2 vt) 2 cos θ(t) = 1.5D s+vt D 2 min +( 1.5D s vt) 2 cos θ(tmod(2d s /v)) when 0 t D s /v, when D s /v < t 2D s /v, when t > 2D s /v. Where D s /2 is the initial distance of the train from enodeb, and D m in is the minimum distance from enodeb to railway track. 3GPP-LTE provides HST scenario parameters for calculating Doppler shift [4], Fig. 1 gives the Doppler shift performance with high velocity users based on HST condition. (2) Fig. 1: Doppler shift of HST condition As can be seen from the figure, Doppler spread is not obvious when compared to Doppler shift. So only Doppler shift has been considered when we explain the effects of Doppler phenomenon. Since the channel in high speed conditions is time-varying and multipath, it can be represented by the channel impulse response (CIR) equation as follows [5] L 1 h(t, τ) = c l µ l (t)δ(τ τ l ). (3) l=0 Where c l is the path loss coefficients, µ l (t) represents time variation due to Doppler shift of l-th path and statistically independent, τ l is the delay of path l, and L is the number of paths between the transmitter and receiver.

3 X. Wang et al. /Journal of Computational Information Systems 10: 9 (2014) Because of the placement of the base stations, there is always a strong line-of-sight (LOS) path in the condition. The direct path has to be considered separately as a constant process while the other as Rayleigh process [6]. The HST channel model then can be modeled as Rician fading channel. The lowpass fading process of HST channel can be rewritten as h(t, τ) = h LOS (t, τ) + h Ray (t, τ) K = K + 1 c 0e j2πf st cos α 0 +φ 0 1 L 1 + c l e j2πf st cos α l +φ l δ(τ τ l ). K + 1 l=1 (4) where K is the Ricean factor, c 0 is the path loss coefficient of LOS path, c l (l = 1, 2,, L 1) are the scattered components which follow the Rayleigh distribution. φ l is the initial phase of each path, which is uniformly distributed over ( π, π]. α 0 represents the arrival angle of LOS path, α l (l = 1, 2,, L 1) are the arrival angles of other paths. 3 Channel Modeling A novel Rician channel simulator is presented in this section for better simulation performance. The Rayleigh channel simulator can be expressed as Y (t) = Y C (t) + jy S (t) N 1 2 Y C (t) = cos(ω d t cos α n + φ n ) N n=0 N 1 2 Y S (t) = sin(ω d t cos α n + φ n ) N n=0 (5) N sinusoids are used to generate an independent fading wave with α n can be calculated as α n = 2πn + θ n N π(n = 0, 1, 2,, N 1) (6) φ n and θ n are statistically independent and uniformly distributed over ( π, π] for all n. In order to propose L 1 independent Rayleigh channels, channel simulator can be improved as α n,k = 2πn + θ n,k N + 2πk LN π(n = 0, 1, 2,, N 1)(k = 1, 1, 2,, L 1) (7) where 2πk LN as h(t, τ) = is proposed to distinguish each channel. The novel channel simulator can be express K K + 1 c 0 exp[j2πf s t cos θ 0 + φ 0 ] + L 1 c l N(K + 1) l=1 N 1 n=0 exp[jω d t cos α n,l + φ n,l ]δ(τ τ l ) We present the comparison of the correlation statistics and PDF of envelop between our proposed model and Jakes model in Fig. 2 and Fig. 3. (8)

4 3648 X. Wang et al. /Journal of Computational Information Systems 10: 9 (2014) Cross correlation in one fading proposed method Jakes method Cross correlation in different fadings proposed method Jakes method cross correlation cross correlation Time[s] Time[s] Fig. 2: Cross-Correlation statistics in different model PDF of the simulators and reference fading envelop theoretic proposed method Jakes method f(r) Fig. 3: PDFs of different model r As can be seen from the figures, our model has better properties than Jakes model. The proposed model overcomes undesirable cross-correlation properties in Jakes model. And the PDF of envelop in our model is more close to Rayleigh fading. So the proposed model is more effective than Jakes model, and is used for simulation afterwards. We proposed three different channel models here based on distinct scenarios. The first HST channel given by [4] is a non-fading propagation channel with one LOS path. Here we call it as HST model in the following analysis. The second channel model we proposed is called improved- HST model. It s characterized by the number of paths, maximum time delay, fading of each path as shown in Table 1, referring the RA model in COST 207. It describes the rural area, which has little reflections or obstructions when the trains go by. The third model we presented is RMa-LOS model based on [7], the number of paths, maximum time delay, fading of each path are shown in Table 2. RMa-LOS model is considered as a typical high speed railway multipath channel.

5 X. Wang et al. /Journal of Computational Information Systems 10: 9 (2014) Table 1: Parameters of improved-hst channel No. of paths Delay [ns] Power [db] Table 2: Parameters of RMa-LOS channel No. of paths Delay [ns] Power [db] Table 3: Simulation main Parameters Parameter Value LTE frame structure FDD frame Carrier frequency 2.6GHz Bandwidth 20MHz Subcarrier Spacing 15kHz FFT size 2048 Sample Frequency 30.72MHz Coding Method Turbo Cyclic Prefix Normal Antenna Type SISO User speed 300km/h 400km/h Modulation Mode QPSK/16QAM/64QAM 4 LTE Performance Evaluation In this section, we would like to present the link level simulation results for the propagation models in Section III. The main parameters of the simulation are listed in Table 3. We provide performance of BER and transmission time delay of each channel model. System performance of different modulation modes are presented as well. The time delay of LTE downlink system can be calculated as T DL delay = T net enb + T enb + T enb UE. (9) Where the T net enb is the delay from the network side to enodeb, T enb represents the processing delay of enodeb, and T enb UE is the packet delay between the UE and the enodeb. For comparison, the user speed is set as 300km/h and 400km/h, respectively. 5 Simulation Results The simulation results are based on the parameters in previous section.

6 3650 X. Wang et al. /Journal of Computational Information Systems 10: 9 (2014) Performance of different channel models We provide performance of three channel models mentioned above by using 16QAM. We first analyze the BER performance of 3 different channel models in Fig LTE downlink BER of different channel model 10-1 BER 10-2 improved-hst 300km/h 10-3 improved-hst 400km/h RMa-LOS 300km/h RMa-LOS 400km/h HST 300km/h HST 400km/h SNR [db] Fig. 4: BER performance of different channel models We can configure from Fig. 4 that higher speed causes lager BER. But the BER performance changes slightly when the user speed changes, that is because the channel estimation, which can estimate the Doppler shift well. Comparing the performance of 3 channel models, the HST channel has the best BER performance. The reason is that the HST channel does not have channel interference. The improved-hst channel has better BER performance than RMa-LOS channel on account of less multipath number. The value of BER differs little within 3 channel models under low SNR environment (0-5dB) LTE downlink delay of different channel model[5db] 10 0 LTE downlink delay of different channel model[10db] delay probability 10-2 delay probability 10-2 improved-hst 300km/h 10-3 improved-hst 400km/h RMa-LOS 300km/h RMa-LOS 400km/h HST 300km/h HST 400km/h time [ms] improved-hst 300km/h 10-3 improved-hst 400km/h RMa-LOS 300km/h RMa-LOS 400km/h HST 300km/h HST 400km/h time [ms] Fig. 5: Time delay performance of different channel models Then, we take into account the time delay performance. Set the SNR values as 5dB and 10dB, respectively. According to the real LTE system of high speed train condition, the T net enb is 20ms and T enb takes 4ms in typical scenario. T enb UE can be calculated by packet transmission

7 X. Wang et al. /Journal of Computational Information Systems 10: 9 (2014) delay between the enodeb and UE and retransmission times. Packets transmission delay is fixed as 8ms. According to [9], we consider the maximum number of retransmission is 4. Then the packet delay performances of 3 channel models with 5 db and 10 db are shown in Fig. 5. We can realize from the above results that T DL delay changes mainly because of the packet retransmission times between the enodeb and UE. Comparing different SNR conditions in Fig. 5, it can be found that the SNR affect retransmission times a lot. For HST channel model, the proportion of packets that transmitting once increases about 13% when the SNR changes from 5 db to 10 db, and the percentage of packets that transmitting four times decreases to 0 when the SNR goes from 5 db to 10 db. Considering the improved-hst model and RMa-LOS model, the enhancement of packets that transmitting once is about 14% and 16%, respectively, the reduction of packets that transmitting four times is about 66% and 77%. 5.2 Performance of different modulation mode Due to the requirements of high data rate, high-order modulation modes are more likely to be used. We present the BER performance of QPSK, 16QAM, 64QAM in LTE downlink system, using channel models of improved-hst and RMa-LOS in Fig LTE downlink BER of different modulation mode 10-1 BER 10-2 improved-hst QPSK 10-3 improved-hst 16QAM improved-hst 64QAM RMa-LOS QPSK RMa-LOS 16QAM 10-4 RMa-LOS 64QAM SNR [db] Fig. 6: BER performance under different modulation methods From Fig. 6, it can be found that high-order modulation modes are more likely to have higher BER than low-order ones. Since the improvement of frequency bandwidth efficiency is at the expense of generating BER, modulation mode should be chosen carefully and properly. Combining with our provided channel models, 16QAM could find the balance between the BER and the efficiency. 6 Conclusion In this paper, the performance of LTE downlink system in high speed train condition has been presented. We firstly propose a novel Rician fading simulator and three channel in order to accommodate real propagations. The proposed simulator overcomes undesirable properties in

8 3652 X. Wang et al. /Journal of Computational Information Systems 10: 9 (2014) Jakes model. HST model is a one-path ideal model given by 3GPP. The improved-hst model is suitable for the rural area, which has little reflections and refractions. The RMa-LOS model is considered as a typical high speed railway channel with more reflections and refractions. Then we evaluate the performance of LTE downlink system, including BER, time delay and modulation scheme. For BER, with the same modulation method, the performances of three channel models are relatively close under low SNR environment, and differ from each other under high SNR environment. When considering different modulation methods, combined with our provided channel models, 16QAM could find the balance between the BER and the efficiency. For time delay performance, value of SNR affects it a lot. The results show that the performance of LTE with high velocity users is attractive. The two realistic models can perform quite close to the ideal model with proper modulation modes and good channel estimation methods. They have potential to use under high velocity users condition in different environments. Acknowledgement This research work is supported in part by the Important National Science and Technology Specific Project of China under Grant No. 2013ZX , by the National High Technology Research and Development Program of China under Grant No. 2013AA013602, and by the ZTE Corporation and University Joint Research Project under Grant No. CON References [1] Erik Dahlman, 3G evolution: HSPA and LTE for mobile broadband. Elsevier Ltd [2] A. Farajidana, et al., 3GPP LTE downlink system performance, IEEE Global Telecom. Conf., pp. 1-6, Nov [3] Virtej. E., Kuusela. M., Tuomaala. E, System Performance of Single-User MIMO in LTE Downlink, Personal, Indoor and Mobile Radio Communications, 2008, Page(s): 1-5. [4] 3GPP TS V ( ), 3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Evolved Universal Terrestrial Radio Access (E-UTRA); User E- quipment (UE) radio transmission and reception (Release 11). [5] J. D. Parsons and A. S. Bajwa, Wideband characterisation of fading mobile radio channels, IEE Proc., Vol. 129, No. 2, pp , April [6] P. A. Bello, Aeronautical channel characterization, IEEE Trans. Commun., vol. COM-21, pp. 548C563, May [7] ITU-R Document 5D/188-E, Proposed channel model parameter update for IMT-Advanced evaluation, 2nd meeting of Working Party 5D, Dubai, Jun. 24- Jul. 2, [8] 3GPP TS , 3rd Generation Partnership Project; Technical Specification Group Services and System Aspects; Policy and charging control architecture (Release 10). [9] 3GPP TS , 3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Evolved Universal Terrestrial Radio Access (E-UTRA); Radio Link Control (RLC) protocol specification (Release 9).