Hardware-in-the-loop Simulation of LTE System in High-speed Environment

Transcription

1 203 8th International Conference on Communications and Networking in China (CHINACOM) Hardware-in-the-loop Simulation of LTE System in High-speed Environment Dan Fei, Lei Xiong, Haoruo Yan and Shihan Huang State Key Laboratory of Rail Traffic Control and Safety Beijing Jiaotong University Beijing 00044, P.R. China Abstract The hardware-in-the-loop (HIL) simulation is an advanced simulation method. It overcomes the weakness of bad reliability and lower efficiency in traditional pure computer simulation method, by the combination of simulation computers and field equipments. Meanwhile, the communication quality in high-speed environment has become a shared concern of related researchers, with the wireless communication system working in higher frequency and broader bandwidth. In this paper, we build our HIL simulation system with the software-defined radio platform and a radio channel emulator. And then we analyse the RF performance of LTE uplink received signal in the conditions of different mobile speed and SNR. Finally, we get the conclusion that mobile speed has an important effect on EVM, and affects ACLR, OBW, and SEM obviously, but has little influence on IQ quadrature skew, IQ offset, and IQ gain imbalance. While SNR affects all the parameters. I. INTRODUCTION In high-speed environment, the wireless channel has its particularity compared to ordinary environment, such as variable propagation scenes, complex large-scale fading, and of most important, server Doppler shift caused by terminal moving in high speed and handing over frequently. As a consequent, there are more stringent requirements on the wireless communication system. Therefore, it is necessary to seek a more advanced simulation method than the traditional methods (mainly field testing and pure software simulation) to meet the flexibility, efficiency and real-time requirements. The hardware-in-the-loop simulation technology[] described in this paper is such a method. It refers to the simulation methods with practical equipments to replace parts of the mathematical models in the simulation system. This approach can get the maximal approximation of the real-world system in speed and authenticity by a seamless connection between the simulation equipments. In this paper, we choose the software defined radio (SDR) platform[2] of National Instruments(NI) and the radio channel emulator of Elektrobit(EB) to build our simulation system as the main equipments. The advantage of the SDR platform lies in the fact that there is no need for the user to have in depth hardware knowledge and it may greatly reduces the time involved to build LTE test system in conjunction with the LabVIEW graphical programming language. Meanwhile, the channel emulator is able to emulate flexibly various static and high-speed scenes, as well as supports co-simulation with MATLAB and field data import. Based on this, we get the measurement results for both the demodulation and spectral Fig.. HIL Simulation System parameters, and then evaluate the system performance in highspeed environment. This paper is organized as follows. Section II describes the structure of the hardware system as well as the simulation scheme. Section III explores the theory and methods of radio channel simulation for standard channels and dynamic Doppler shift. Section IV implements the analysis for measurement results on the basis of performance parameters introduction. II. SYSTEM DESCRIPTION The simulation system mainly consists of a vector signal generator (VSG), a vector signal analyzer (VSA), a radio channel emulator and control terminals, as illustrated in Figure. All the subsystems can be connected via GPIB to easily implement automated test with LabVIEW program on the control terminal. The NI SDR platform, including the VSG and VSA, are based on the PXI/PXIe bus technology which is an extension of PCI for instrumentation. It is convenient to configure and program them through National Instruments s LabVIEW programming environment. LabVIEW is a data flow based graphical programming language which could provide the user with simple programming and rapid prototyping capabilities. On the platform, National Instruments have developed the LTE toolkits used to test and measurement the signals that conform to the 3GPP specifications. The radio channel emulator (EB Propsim C8) is an outstanding product of Elektrobit in Finland. It supports multiple channel, wide RF bandwidth, and high dynamic frequency range with very high accuracy. Especially, equipped with IEEE

2 (a) The high-speed railway scenario Center 2.6 GHz 2 khz/ Span 20 khz (b) Leaving the base station Center 2.6 GHz 2 khz/ Span 20 khz (c) Approaching the base station Fig. 2. The simulation of dynamic Doppler shift internal and interference generators, it can provide realistic channel emulation of all typical propagation effects affecting wireless receiver and system performance. In our simulation system, the VSG generates a LTE FDD uplink signal in the center frequency of 2.6GHz, then the signal is processed by the radio channel emulator which emulates the high-speed propagation environment and finally received by the VSA for measurement. The configuration of the LTE uplink signal generated by VSG is listed in the following table. TABLE I. SPECIFICATIONS OF THE SIGNAL Num. of Antennas Carrier Frequency 2.6GHz Power Level -25dBm System Bandwidth 0MHz IQ Rate 5.36MS/s Num. of Resource Block 50 Taking into consideration the specific characteristic of highspeed environment, we choose the following channel models with dynamic Doppler shift in our simulation, COST207 Rural Area 6 Taps, WINNER II D2a, and channel as comparison. The part about channel simulation will be detailed in the next section. III. CHANNEL SIMULATION As we known, the radio channel has an important influence on the real-world wireless communication systems due to multipath propagation, interference, noise and jamming. Therefore, the channel modeling is a key procedure in this simulation system. In regard to statistical modeling, radio channel is modeled by time variant impulse response h(t, τ). If the receiver is moving, impulse responses are different in every position of the receiver. Equation () shows the impulse response h(t, τ). L h(t, τ) = β i (t)e jφi(t) δ[τ τ i (t)] () i= where β i (t) and φ i (t) represent the amplitude and phase of the ith path arriving at delay τ i (t) and at time t. Received signal y(t) can be calculated by transmitted signal x(t) and system s impulse response h(t, τ). Equation (2) shows that received signal y(t) is convolution of the x(t) and h(t, τ). y(t) =x(t) h(t, τ) (2) In the following section, two channel simulation methods are introduced. Simulation of dynamic Doppler shift is based on the impulse response matrix processing, and the others are based on standard channel models. A. Dynamic Doppler Shift In high-speed environment, there are many reasons resulting in a serious decline in the reliability of the communication system, one of the most is the severe Doppler shift[3]. Doppler shift refers to the frequency offset between the transmitted signal and the received signal due to the relative movement between the transmitter and the receiver. The Doppler shift will cause the channel to encounter significant time-varying fading that will result in bad performance of the channel estimation and equalization processing. Equation (3) show the formula to calculate the Doppler shift (f d ). f d = f v cos θ (3) c Where, f is the carrier frequency, v is the velocity of the mobile station, c is the velocity of light, and θ is the angle between the wave propagation direction and the mobile station movement direction. In high-speed environment, the mobile station s Doppler frequency shift is a variable value. For instance, in high-speed railway scene, it depend on the speed of the train, and the relative position of base station and the train (see figure 2(a)). In this situation: d v t θ = arccos (4) (d v t)2 + h 2 so f d = d v t (d v t)2 + h v f 2 c Where, d is the radius of base station coverage, t is the time when the mobile station enters the coverage area of the base station time and h is the distance of base station to the railway. According to the modulation theorem, the shift in frequency domain F (j(ω ω 0 )) corresponds to the shift in time domain f(t)e jω0t where ω 0 =2πf d. Therefore, we can multiply the corresponding elements of the impulse response matrix by e j2πf dt, if we want to add Doppler shift of f d in the frequency domain. Especially, it is necessary to consider the positive and negative transformation of f d when emulating the handing-over scenario. For example, when Doppler frequency (5) 324

3 (a) (b) Rural Area 6 Taps (c) WINNER II D2a Fig. 3. EVM measured with different channel models shift converts from negativity to positivity, it should be just the time the train passing by the base station. Figure 2(b) and (c) show the emulation result of dynamic Doppler shift in the conditions as follows: input signal: 2.6GHz single carrier; channel model: COST207 Rural Area 6 Taps; mobile speed: 350km/h. As shown in the Figures, the center frequency shifts to lower frequency when the train leaves away from the base station, and shifts to higher frequency when the train approaches the base station on the contrary. B. Standard Channel Models In standard channel modeling, each channel model has one or several taps and each tap is specified by a relative (with respect to the first path) delay, a relative power, and a Doppler spectrum category. At present, there is not yet a standard model which can be specifically used in high-speed environment. But by revising the existing models with some measurement data, some of them can be applied to guide our theoretical analysis and engineering applications. Here, we introduce two of them, COST207 and WINNER II. ) COST207: COST-207[4], is a working group established by CEPT (Conference of European Post and Telecommunication) in 984. The group has standardized a common set of channel models for mobile radio that can be used by different communications designers to simulate their systems. There are four propagation models defined: rural area (RA), bad urban area (BU), typical urban area (TU), and hilly terrain (HT). The RA case comprises two distinct channel models, while the other cases each comprise four channel models, for a total of 4 channel models. 2) WINNER II: WINNER (Wireless World Initiative New Radio) II[5] is a wideband MIMO channel model based on the geometric statistical modeling method. The statistical parameters include delay spread, angular spread, angle of arrival, angle of divergence and so on. WINNER II channel model contains 5 kinds of scenarios, and supports the communication frequency of 2GHz to 6GHz with bandwidth up to 00MHz. Among the scenarios, D2a (the first portion of the rural moving network scene) has the parameters that are consistent with the high-speed environment of open field or viaduct scene. Fig. 4. IV. Specification of EVM calculation SIMULATION RESULTS AND DATA ANALYSIS In this section, we describe and analyse the measurement results for the error vector magnitude (EVM); the demodulation impairments including quadrature skew, IQ offset, IQ gain imbalance; and the spectral parameters including adjacent channel power (ACP), occupied bandwidth (OBW) and spectrum emission mask (SEM). The variable is SNR or mobile speed. Each measured value is obtained by averaging 00 frames of received data. It should be noted that it s difficult to synchronize in the simulation when the SNR is below 2dB, and that s why we determine the minimum SNR at 2dB in the measurement of demodulation parameters. A. EVM EVM[6], [7] is a measurement of demodulator performance. As shown in Figure 4, the symbol decision generated by the demodulator is given by z. However, the ideal symbol location (using the symbol map) is given by r. Therefore, the resulting error vector is the difference between the actual measured and ideal symbol vectors given by e = z r. Analytically, RMS EVM over a measurement window of N symbols is defined as N N EV M = e 2 00% (6) N r 2 N The measurements of EVM are shown in Figure 3, each of the three subfigures corresponds the EVM variation in a certain simulation environment with different channel models, 325

4 Quadrature Skew(deg).4.2 RA6 WINNER-D2a IQ Offset(dB) RA WINNER-D2a IQ Gain Imbalance(dB) RA6-0.4 WINNER-D2a (a) Quadrature Skew (b) IQ Offset (c) IQ Gain Imbalance Fig. 5. Demodulation impairments measured with different channel models, COST207 Rural Area 6 Taps or WINNER II D2a. Each subfigure describes the EVM at variable speed or SNR. From these figures, we could find that the EVM reduces gradually with SNR increasing whatever the channel model is. And it can be speculated in Figure 3(a) that the reduction is dramatic when the SNR is below 20 db, which means the SNR is the main factor affecting the EVM in this situation. Meanwhile, Figure 3 (b) and (c) show the impact of mobile speed on the EVM in corresponding situation. The EVM increases gradually with mobile speed increasing. For instance in Figure 3(b), when the SNR is 8dB, The EVM is 42.9% at the mobile speed of 80km/h, then it increase to 58.3% when the speed is up to 200km/h. When the speed rises as high as 350km/h, the EVM undergoes a dramatic change to 68.9%. With the mobile speed continuing rising to 500km/h, the EVM turns to 76% which has been an unacceptable value. For the case of Figure 3(a), the impact of mobile speed on the EVM is not obvious, because there is only one tap with pure Doppler shift in model. B. Demodulation Impairments In this section, we analyse the demodulation impairments caused by the propagation environment. All the parameters, including quadrature skew, IQ offset, and IQ gain imbalance are briefly introduced as follows. - In a quadrature modulated (QM) system, quadrature skew, also referred to as quadrature error, describes a complex signal impairment such that the I and Q components are not perfectly orthogonal. Quadrature skew can be either positive or negative, with the sign indicating the orientation of the error. IQ offset is a complex signal impairment that shifts the locus of ideal symbol coordinates off center in the IQ plane. A IQ offset can be added to the baseband I component, the Q component, or both. IQ gain imbalance refers to a difference in scaling between the I and Q components of IQ data. When expressed in db, IQ gain imbalance can be either positive or negative, with the sign indicating which component has been impaired. The above three parameters are measured in the condition that the mobile speed is fixed to 280km/h, since the impact of mobile speed on these parameters is not obvious in the view of our measurements. For example, when the SNR is 20dB, the quadrature skew is 0.74deg at the mobile speed of 0km/h and it just turns to 0.78deg as the mobile speed rise to 500km/h. Similarly, the IQ Offset is respectively -59.3dB and -59.5dB as well as IQ gain imbalance is -.35dB and -.24dB at the mobile speed of 0km/h and 500km/h. As illustrated in Figure 5(a) and (b), both quadrature skew and IQ offset reduce gradually with SNR increasing whatever the channel model is. And when the SNR rise high enough, the reduction speed becomes slower. Respectively, the IQ Gain Imbalance increases gradually with SNR increasing in all the channel models. C. Spectral Parameters In this section, we analyse how the spectral parameters are affected by the propagation environment. All the parameters, including ACLR, OBW, and SEM are briefly introduced as follows. The ACLR is the ratio of the filtered mean power centred on the assigned channel frequency to the filtered mean power centred on an adjacent channel frequency. ACLR requirements are specified for two scenarios for an adjacent UTRA (Universal Terrestrial Radio Access) and E-UTRA (Evolved-Universal Terrestrial Radio Access) channel. The OBW is defined as the bandwidth containing 99% of the total integrated mean power of the transmitted spectrum on the assigned channel. The occupied bandwidth for all transmission configurations shall be less thanthe channelbandwidthspecified. The SEM[8] measures the out of band emissions power shifting from the center frequency between 2.5MHz to 2.75MHz. It mainly considers the leak power and nonlinear distortion out of the channel made by modulation and radio frequency equipment. The measure of SEM makes the center frequency to be symmetry, dividing the band into 8 parts. The LTE uplink signal band in our simulation is 0MHz. Its SEM specification is listed in table II, the Δ f is the scope frequency offset. The sign Limit is the corresponding power limit for transmitter. TABLE II. THE SEM SPECIFICATION WITH BANDWITH OF 0MHZ Δ f (MHz) ±0- ±-2.5 ±2.5-5 ±5-6 ±6-0 ±0-5 Limit(dBm) Figure 6 shows the impact of SNR variation on the spectral parameters in different mobile speed. As we can see, the ACLR of UTRA is affected by both the SNR and mobile speed. The higher the SNR or the lower the mobile speed, the lower the 326

5 ACLR(dB) km/h km/h km/h Ocuppied Bandwidth(MHz) km/h km/h 0km/h Power(dBm) no noise dB -50 2dB Frequency(GHz) (a) ACLR (UTRA) (b) OBW (c) SEM Fig. 6. Spectral measurements with different channel models ACLR. For example, at the speed of 350km/h, the ACLR is -4.60dB when the SNR is 2dB, then it turn to 30.4dB when the SNR is 30dB. And at the same SNR, the ACLR has a difference of about 2dB when the speed changes from 0km/h to 500km/h. Respectively, the mobile speed and SNR have an impact on the OBW. For instance, at the speed of 350km/h, when the SNR changes from 8dB to 2dB, the OBW rapidly decreases from 0.8MHz to 9.6MHz. At lower SNR, the OBW changes with relatively slower speed. When the SNR is high enough (above 24dB),the OBW is fixed about 9MHz which is the ideal OBW. The impact of mobile speed is obvious as illustrated in Figure 6(b). Finally, let s pay our attention to the SEM at different SNR. Although with the measurement results in only three SNRs, it is easy to distinguish the impact of SNR from Figure 6(c). In the case of useful signal almost without change, the out of band emissions power undergoes a significant transformation from -83.5dB to -57.2dB at corresponding SNR. And the mobile speed has an impact on it like OBW in view of measurement results that is not shown in the figure. V. CONCLUSION AND FUTURE WORK In this paper, we build the hardware-in-the-loop simulation system for LTE with NI SDR platform and EB Propsim C8. Through the test for RF signal quality and the simulation in the conditions of different SNR and mobile speed, it is obtained that the mobile speed has an important effect on EVM, the higher the mobile speed, the higher the EVM. And the impact of mobile speed on ACLR, OBW and SEM is obvious. But it has little influence on IQ quadrature skew, IQ offset, and IQ gain imbalance. In addition, we also get the conclusion that the SNR has an impact on all the parameters measured in this paper, the higher the SNR, the better the performance, and when the SNR is high enough, the performance gain reduces. It guides us that it is need to consider the impact of both the mobile speed and SNR on the communication quality when we practically use the LTE system in high-speed environment. In addition, there is a need to emphasize the HIL simulation method used in this paper. The NI SDR platform frees us from low-level hardware implementation problems to allow more intensive study on the algorithm and system design. And the EB Propsim C8 has a powerful channel simulation ability with flexible user-defined functions. We combine the two platforms together organically to form the high-speed simulation system which could meet the requirements on speed and reliability. In future work, we plan to implement algorithm research to improve the communication quality in high-speed environment on the platform. And the research work in this paper has laid a good foundation for us. ACKNOWLEDGMENT This research was supported by the National Natural Science Foundation of China under Grant(No ), the Key grant Project of Chinese Ministry of Education(No.33006), the Program for Changjiang Scholars and Innovative Research Team in University under Grant (IRT0949), the Beijing Jiaotong University basis research Grant (200JBZ008), the State Key Laboratory of Rail Traffic Control and Safety (RC- S20ZZ002), Beijing Municipal Natural Science Foundation under Grant (42048), Interdiscipline cooperation projects of the New-Star of Science and Technology supported by Beijing Metropolis, under grant no. xxhz2020. REFERENCES [] M. Zhu, Z. Zhong, L. Xiong, J. Ding, and S. Lin, Hard-in-the-loop simulation of high-speed railway mobile communication, in Communications and Networking in China (CHINACOM), 20 6th International ICST Conference on, 20, pp [2] A. Gupta, A. Forenza, and R. Heath, Rapid mimo-ofdm software defined radio system prototyping, in Signal Processing Systems, SIPS IEEE Workshop on, 2004, pp [3] W. Xin, Z. Gang, C. Xia, and T. Zhen-hui, Doppler diversity for ofdm high-speed mobile communications, in Communications, ICC 06. IEEE International Conference on, vol. 0, 2006, pp [4] COST207, Digital land mobile radio communications. Office for Official Publications of the European Communities, 989. [5] P. K. et.al, WINNER II Channel Models. IST-WINNER, D..2v.2, Sept [6] F. Xiu-li, Z. Jian-hong, and C. Li, Evm testing and analysing in tdscdma, Journal of Chongqing University of Posts and Telecommunications, vol. 2, pp , Apr [7] C. Fatang and Z. Xiaoxian, Evm measurement and dsp realization for ue of td-lte systems, STUDY ON OPTICAL COMMUNICATIONS, vol. 6, pp , Dec. 20. [8] 3GPP TS 36.0, User Equipment (UE) radio transmission and reception Std., Rev. V8.6.0, Jul