LTE MIMO Performance Measurements
on Trains
MOHAMMAD ALASALI
Master of Science Thesis
Stockholm, Sweden 2012
LTE MIMO Performance Measurements
on Trains
MOHAMMAD ALASALI
Master of Science Thesis performed at
the Radio Communication Systems Group, KTH.
September 2012
Examiner: Professor Slimane Ben Slimane
Supervisor: Professor Claes Beckman
TRITA-ICT-EX-2012:206
c
Mohammad Alasali, September 2012
Tryck: Universitetsservice AB
Abstract
Providing passengers internet on board trains with continuous connectivity at high speeds and over large rural distances is a challenging issue. A promising solution to the problem is to use an on board WI-FI network combined with a cellular radio system connected to the LTE (long term evolution) 900/800MHz networks deployed outside the train. However, in order to reach the data rates that today’s business traveller is expecting, “Multiple Input Multiple Output” (MIMO) may be needed. In this project the plausibility of achieving LTE
MIMO functionality to moving trains is studied both theoretically and through measurements on a Swedish high speed train in a live LTE 900 network. The measurements were conducted along the main track between Stockholm and Gothenburg. Eight antennas were deployed at 4m height on the roof of a X2000 train that was moving at speeds between 100 and 200km/h. The average distance between the base stations and the track was 2.35km and the average base-station antenna height 45m. The measurement data was recorded using an LTE scanner provided by Rohde&Schwarz (TSMW). The simulations were conducted in Matlab and the above parameters were simulated for typical rural Line of Sight (LOS) radio channels. The simulation results show that MIMO may perform well on trains and reach higher rank in the above described radio channel, given that the inter-element distances are correctly chosen dependent on the base station-to-train distance. In the measurement a 2X2 MIMO system was tested and confirmed the model by showing that the condition number (CN), was less than 15 for 65% of the track covered by LTE. Hence, we find that it is indeed plausible to achieve MIMO functionality in LOS on the roof of a fast moving train. Furthermore, the report concludes that LTE-Advance with 8x8 MIMO functionality and carrier aggregation feature, may be a suitable solution in order to provide the internet capacity needed for all the passengers on board high speed trains.
Acknowledgements
By presenting this thesis, I will be finishing my wireless systems master program which I attended at the Royal Institute for Technology (KTH), Sweden. I would like to express my thanks and deep appreciation to my supervisor Prof. Claes Beckman for the great guidance and support he provided through project. I would like also to thank my examiner Prof. Slimane Ben Slimane for the comments and advices he provided for the conducted work. This thesis study has been part of a seed project sponsored by Wireless@KTH, conducted in collaboration with Icomera AB, SJ, Tele2 and Rohde&Schwarz.
I would like to thank MR.Mats Karlsson from Icomera for his support along the project, following up the data analysis and measurements and the ideas with very useful comments he provided. I would like to thank Henrik Andersson from Tele2 for his help and consideration while executing this project. Moreover, special thanks for Pontus Segerberg from Rohde & Schwarz for the support he provided regarding the equipment used in the project.
Last, but not for sure least, I would like to thank my family for their support and encouragement through my study at KTH. Special thanks go to my father Prof. Khalil Alasali for his continual support and enthusiasm regarding my study at KTH.
Contents
1 Introduction 1
1.1 Passenger Internet on board Trains . . . 1
1.2 Mobile Broadband . . . 3 1.2.1 LTE-A . . . 4 1.2.2 MIMO . . . 5 1.3 Related work . . . 6 1.4 Problem Formulation . . . 7 2 Channel Theory 9 2.1 Wireless Channel . . . 9 2.1.1 Pathloss . . . 10
2.1.2 Small Scale Fading . . . 11
2.2 MIMO . . . 11
2.2.1 SISO (Single Input Single Output) . . . 12
2.2.2 Receiver Diversity . . . 12
2.2.3 Transmitter Diversity . . . 13
2.2.4 Spatial Multiplexing . . . 13
2.3 MIMO in LTE-A . . . 14
2.4 MIMO Performance Indicator . . . 16
2.5 Mathematical Model . . . 16
3 Measurements Setup 21 3.1 Rohde & Schwarz Universal Radio Network Analyser equipment (TSMW) . . . 21
3.2 RF Antenna . . . 23
3.3 Cables . . . 25
3.4 Link-Budget and Sensitivity . . . 26
4 Simulation and Measurements Results 29 4.1 Simulation Results . . . 29
4.2 Measurements Results and Analysis . . . 31
4.3 Measurements and Simulation Comparison . . . 39
5 Discussion and Conclusion 45 5.1 Discussion . . . 45
5.1.1 Impact of Future LTE Release on Internet on board Trains 46 5.2 Conclusion . . . 47
5.2.1 Future work . . . 47 vii
References 49
A CoMP 53
A.1 CoMP . . . 53 A.2 CoMP Simulation . . . 54
List of Tables
2.1 Transmission Modes . . . 14 4.1 Simulation Parameters used for the simulator. . . 29 4.2 Relation between receiver distance elements and TX-RX distance 31 4.3 Spectral efficiency statical summary for the comparison between
measurements and simulation. . . 41 4.4 Comparison between simulation and measurements for base
station-train distance for 2X2 MIMO working points. . . 43
List of Figures
1.1 Icomera solution for providing internet on board trains . . . 2
1.2 LTE-A architecture and features [1] . . . 4
1.3 Top: example of continuous carrier aggregation. Bottom: example of non-contiguous carrier aggregation. . . 5
2.1 Illustration of multipath in Wireless channel. . . 9
2.2 Different MIMO scenarios. . . 12
2.3 Example of different receiver combining techniques performance [2]. 13 2.4 Multi-antenna modes and their relation to the antenna configura-tion and CSI modes. . . 14
2.5 Illustration of CN concept (R&S) [3]. . . 16
2.6 Antenna orientation demonstration . . . 18
3.1 Train Diagram. . . 21
3.2 The TSMW equipment [3]. . . 22
3.3 Train setup digram. . . 22
3.4 Train used for the measurements. . . 23
3.5 Installed equipment inside the train cabinet. . . 23
3.6 Antenna gain vs. frequency [4]. . . 24
3.7 Antenna radiation pattern at 900MHz [5]. . . 24
3.8 MIMO antennas installed on the train roof. . . 25
3.9 Antenna used in the measurements (HUBER+SUHNER). . . 25
3.10 Cable attenuation vs. frequency [6]. . . 26
3.11 Cables used in the test (LZSH) [6]. . . 26
3.12 System digram. . . 26
3.13 Link Budget Calculations. . . 27
4.1 Left: K=zero(NLOS),Right: K =∞(LOS) . . . 30
4.2 Comparison for 4 × 4 MIMO between fixed drand conditioned dr as proposed in the model (dr is the receiver inter-elements antenna distances). . . 30
4.3 Coverage between Stockholm-Gothunborg (Tele2 map). . . 32
4.4 Different scenarios coverage statistics. . . 33
4.5 Covered areas between Stockholm-Gothunborg (measured data). 34 4.6 Different Scenarios RSRP CDF. . . 34
4.7 Covered MIMO points statistics . . . 35
4.8 MIMO working. . . 35
4.9 MIMO working but unstable. . . 35
4.10 MIMO not working. . . 36 xi
4.11 MIMO performance in relation with the train speed. . . 36
4.12 Measured average RS-SINR vs. RSRP. . . 37
4.13 Measured average distance vs RSRP. . . 38
4.14 Average throughput along the track. . . 38
4.15 All measurements data statistics. . . 39
4.16 Spectral efficiency vs RS-SINR for both Measurements and simulation (2X2 MIMO) (a:1.5m scenario, b:4.5m scenario, c:7.5m scenario and d:10.5m scenario)(Blue: simulated, Green: Measured). 40 4.17 CN vs distance showing the 2X2 MIMO working points for both simulation and measurements (a:1.5m scenario, b:4.5m scenario, c:7.5m scenario and d:10.5m scenario)(Red: simulated, Black: Measured). . . 42
5.1 Future estimation for the internet on board trains by using LTE-A and its features. . . 47
List of Abbreviations
BER: Bit Error Rate BS: Base-station
CA: Carrier Aggregation
CDF: Cumulative Distribution Function
CN: Condition Number
CoMP: Coordinated Multiple Point Transmission and Reception CoMP-CS: CoMP-Coordinated Scheduling and or/ Beam forming CoMP-JP: CoMP-Joint Processing
CQI: Channel Quality Indicator CSI: Channel State Information
DL: Downlink
F: Noise Figure
GSM: Global System for Mobile Communications HSPA: High Speed Packet Access
HSPA+: Evolved High-Speed Packet Access ITU: International Telecommunication Union LOS: Line of Sight
LTE: Long Term Evolution
LTE-A: Long Term Evolution-Advanced MBB: Mobile Broad Band
List of Abbreviations xv NLOS: Non-line of Sight
OFDMA: Orthogonal Frequency-Division Multiple Access PDSCH: Physical Downlink Shared Channel
PMI: Precoding Matrix Indicator R&S: Rohde & Schwarz
RF: Radio Frequency RI: Rank Indication RN: Relay Nodes RS: Resource Block
RSINR: Reference Signal to Interference and Noise Ratio RSRP: Reference Signal Received Power
RX: Receiver
SFBC: Space Frequency Block Code
SINR: Signal to Noise and Interference Ratio SISO: Single Input Single output
SNR: Signal to Noise Ratio TX: Transmitter
UE: User Equipment
UL: Uplink
UMTS: Universal Mobile Telecommunications System VoIP: Voice Over Internet Protocol
WIMAX: Worldwide Interoperability for Microwave Access XMPP: Extensible Messaging and Presence Protocol
Chapter 1
Introduction
1.1
Passenger Internet on board Trains
Providing internet on board trains with a continuous connectivity at high speeds along the track is a challenging issue. The evolution in train technology is also currently rapid. For instance, China declared in 2011 the launch of the fastest and longest high speed rail in the world [7]. It connects Beijing to Shanghai, a distance of 1318 km. This train travels with a maximum speed of 380 km/h. The capacity of the rail road is estimated to about 220,000 passengers per day [7]. All these passengers will require a stable and high quality internet connection. Currently, internet is provided for passengers on trains using a number of different techniques as follows:
1. Personal dongle: the most common way to establish a mobile broadband connection on train is to use dongle on a personal device, the dongle connects directly to the 2G or 3G networks outside the train. The main problems using this method are:
• Firstly, using the dongle, the user will be limited by the provider coverage, and along the train track it is difficult to find a single operator that covers 100% of the whole track.
• Secondly, the train is often equipped with windows that have a metallic film coating, as the signal penetrates this glass a 20dB loss of the signal quality can be expected.
• Thirdly, as the train is moving with high speed, it is expected to have a distorted reception due to high Doppler shift and high inter-symbol-interference (ISI).
Using the above mentioned method, it is a difficult mission to have internet connection at all time.
2. RF (radio frequency) repeaters: A common solution in Europe, to provide a stronger 3G signal inside the train, is through the use of RF repeaters. Such repeaters, also known as Amplify and Forward, simply receives the signal from outside the train car, and amplifies it to the inside and forward
it to the user equipment(UE) or dongle. RF repeaters is a good way to improve the signal quality at the receiver, and to provide a denser infrastructure in the network. However, there are also numerous problems with this approach. The main problem is, that since each user is still communicating with one operator only, the coverage remains a problem as in the first point. Each user is still communicating with the repeater and reserve resources independently. Moreover, all the passengers are moving together and handover for the repeater resources will occur at the same time. This with the increasing number of passengers on board train, the users will start suffer from limitation on capacity and delay. Besides, the network will start suffer from high uplink signalling as a large number of passenger need to handover at the same time, this will be reflected on the overall network performance.
3. Third party solution: Nowadays it is common that a third party companies provide internet provided on board trains. Icomera (Sweden) is one of those companies that provides internet on board trains in 13 countries in Europe. An illustration of the Icomera solution is shown in Figure 1.1 [8]:
Figure 1.1: Icomera solution for providing internet on board trains
As indicated from the figure, the passengers are connected to the internet via a Wi-Fi connection. Each train car is equipped with a router and all routers communicate with centralized unit via a power line Ethernet installed in the train. The unit connected to multi-band antennas, installed at the roof of the train that can communicate with different technologies. Inside the unit different operators technologies communicate via an installed modems with the antennas installed on the roof of the train.
The antennas at the top communicate with different base stations and operators at the same time (according to the coverage), they receive resources from available provider to support the passengers demands on board.
This method is an efficient way of supporting internet on board. The passengers will not be aware of the coverage outside, and all the traffic from passengers is aggregated in one big pipe through the centralized unit, which splits out in the air to different available carriers. Also, the signal
1.2. Mobile Broadband 3
quality is expected to be high, as the antennas are installed on the roof of the train with approximately 4m hight and in LOS (line of Sight) with the base stations in the suburban areas. Thus, higher SINR (Signal to Noise and Interference ratio) readings are received, which will reflect on higher capacity calculations.
1.2
Mobile Broadband
The wireless market is continuously developing to reach higher speeds, capacity and lower latency to cope with new users demand (browsing, VoIP (Voice Over Internet Protocol), Streaming and XMPP (Extensible Messaging and Presence Protocol)). Different technologies are competing in the telecommunication market such as HSPA (High Speed Packet Access), UMTS (Universal Mobile Telecommunications System), HSPA+ (Evolved High-Speed Packet Access) , WiMax (Worldwide Interoperability for Microwave Access) and LTE (Long Term Evolution) [9]. Currently, internet is used to increase productivity and alter the way we live and work. It is also used to improve education, health service and connecting different part of the world over the internet.
At the same time, the number of users is increasing dramatically. Considering the Swedish market; the number of subscriptions for mobile broad band (MBB), as add-on service and stand-alone service, increased from 2.3 to 4.2 million between June 2010 and - June 2011, this represents an increase of about 80% [10]. It is becoming increasingly common for mobile terminals to be used for both mobile calls and MBB. On June 2011 the total number of Broadband (mobile and fixed) subscriptions was close to 7.2 million [10]. So the need for a reliable and easy access connection become an essential issue in the wireless market.
In the mean time, the 3GPP organization is working to release the new standards for LTE-Advance (LTE-A). Many features are being developed for this standard in order to meet the different market requirements. For example, higher rank of Multiple-Antenna system known as MIMO (Multiple input Multiple output) in both transmitter and receiver, which is an emerging cost-efficient technology that with the aid of other features could help LTE-A to reach up to 1Gbps in the downlink [11]. The LTE-A standard hosts the following features [12]:
• Advanced MIMO option up to 8×8 in Downlink and 4×4 in the Uplink. • Carrier aggregation (CA).
• Coordinated multiple point transmission and reception (CoMP). • Relay nodes (RN).
Integrating all the above mentioned features, LTE-A is a promising technology for better and more reliable connections in the near future. Figure 1.2 illustrate LTE-A with its different antenna configuration and features.
Figure 1.2: LTE-A architecture and features [1]
An introduction and related work about LTE-A, MIMO and carrier aggregation will be discussed in the next subsections, CoMP will be discussed and simulated in Appendix A.
1.2.1
LTE-A
LTE-Advanced is the evolution of LTE. It is supposed to support up to 1Gbps with low mobility in the downlink and 500Mbps in the uplink according to ITU (International Telecommunication Union) [13]. LTE-A should supports low latency<50ms in active mode instead of <100ms latency in LTE [13]. Besides, LTE-A supports spectrum flexibility, multi-antenna solution, coordinated multipoint transmission/reception, and the use of repeaters/relying [13]. Spectrum efficiency for LTE-A should support up to 30 bps/Hz with 8×8 antenna configuration in downlink, and 15 bps/Hz with 4×4 antenna configuration in the uplink [14].
LTE-A is considered to be a system immune to intra-cell interference, based on using OFDMA (Orthogonal Frequency Division Multiple Access). The main idea of this access method is to split high data stream into a number of lower data streams, then transmit them by set of orthogonal sub-carriers or resource blocks with 180KHz for each resource block [14]. Considering the channel matrix and the Eigenvalues of the channel, the system can offer parallel sub-channel, with different power gains depending on the Eigenvalues. Each resource block can experience different channel fading, and results in different SNR and capacity calculation at the receiver for each resource block [15].
The very high data rate require more bandwidth, as LTE already reach the Shanon limits with high SINR, higher throughput can be achieved with so-called carrier aggregation in LTE-A, where multiple LTE-A “component carriers” are aggregated on the physical layer, to provide up to 100 MHz instead of maximum 20 MHz in LTE [16]. Figure 1.3 shows an example of carries aggregation
1.2. Mobile Broadband 5
Figure 1.3: Top: example of continuous carrier aggregation. Bottom: example of non-contiguous carrier aggregation.
Bearing in mind that LTE-A supports non-contiguous spectrum component, this means that terminals and BS (Base-station) can support different frequency band simultaneously [13]. Regarding UE complexity, cost, capability, and power consumption, it is easier to implement continuous CA without making many changes to the physical layer structure of LTE systems [17]. In addition, compared to non-continuous CA, it is easier to implement resource allocation and management algorithms for continuous CA [17]. According to the existing spectrum allocation policies and the fact that the spectrum resources in the low frequency band(<4 GHz) is scarce, it is difficult to allocate continuous 100 MHz bandwidth for a mobile network [17]. Therefore, the non-continuous CA technique provides a practical approach to enable mobile network operators to fully utilize their current spectrum resources, including the unused scattered frequency bands and those already allocated for some legacy systems, such as GSM (Global System for Mobile Communications) and 3G systems [17]. Carries aggregation is considered as one of the most important features in LTE-A, since it will improve the system capacity up to five times the current throughput scenario [12].
1.2.2
MIMO
MIMO is a key feature in LTE-A to improve the peak data rate over the link. The capacity in any system increases linearly at low SINR (according to Shannon limits), and logarithmically at higher SNR. In MIMO systems, a given total transmitted power can be divided among multiple spatial paths, driving the capacity closer to linear regime, thus increasing the aggregate spectral efficiency even with low SNR [18].
Moreover, LTE-A supports up to 8 antennas in the downlink instead of 4 as defined in LTE. This will be reflected on the signal quality at the receiver when MIMO is used in diversity mode, and improves the peak data rates when different data streams are mapped to different antennas [19]. Similarly, in the uplink where LTE-A supports up to 4 × 4 transmission and reception.
According to the MIMO principle, the information-theoretic capacity can increase linearly with the number of antennas [19][16]. Simulating realistic MIMO radio
channel is a challenging issue, different number of parameters participate in MIMO performance, and any slight change in any of the parameters affects the system performance.
1.3
Related work
In [20], the report studied MIMO performance in LOS. The simulation used a Rician fading channel model, and the MIMO performance was evaluated with respect to ergodic capacity, outage capacity, and effective degree of freedom. The paper presented a study of the limitation and condition on the distances between the transmitter and the receiver antenna array, the alignment angle of the antenna, frequency, distance between TX-RX and their effect on Eigenvalues on the channel matrix. The results show that under certain condition MIMO can reach higher rank transmission, and enhance the spectral efficiency in LOS. In [19], the paper derives a generic model for MIMO wireless channel. It discuss the effects of interdependency of directions-of-arrival and directions-of-departure, large delay and angle dispersion by propagation via far clusters and rank reduction of the transfer function matrix. The paper propose a geometry based model that includes the propagation effects that are critical for MIMO performance as single scattering around the BS and UE, scattering environments and diffraction by roof edges. The results hows the capacity CDF (Cumulative Distribution Function) difference between micro cells, macro cells and independent Rayleigh fading at all antenna elements with 4X4 MIMO and fixed SNR.
In [15], MIMO improvements were compared to SISO (single input single output) in Downlink. This work take in consideration OFDMA access, modulation and coding schemes for the simulation. Besides, different power allocation (water filling and uniform distribution) for different resource blocks were simulated. The results shows that using 4×4 MIMO will gain 235 % better than SISO with water filling power allocation, and 159 % with uniform power allocation. Practical MIMO measurements were conducted to evaluate the MIMO per-formance in [21], the results focused on comparing UE throughput in different scenarios as follow:
• Applying high and low power in the transmitter
• Measure the performance in case of correlated and uncorrelated signals (this was accomplished by varying the distance between the transmitted antennas).
• Different MIMO schemes (1×2, 1×4, 2×4, 4×2 and 4×4) .
The results indicated that increasing the number of antennas will always improves the system performance. The correlated setup perform better for lower throughput until a certain point, the uncorrelated setup shows better performance, where different antennas transmit different data to the receiver. The same applies for power, as with lower throughput no need for higher received
1.4. Problem Formulation 7
power, as low SNR needed to support lower throughput.
The technical difficulties of implementing CA, such as: handover control, control signalling design and guard band settings was discussed in [17]. The results compares different modulation schemes of CA with and without the Doppler frequency shifts. Also, the performance of BER (bit error rate) in CA with and without aliasing effect. This work provides possible solution to implement CA in real mobile system and the problems the operator may face [17].
In [13], The paper discuss the different features and requirements for LTE-A to reach higher capacities. The report compares the mean user spectral efficiency according to the served traffic for LTE-A using CoMP with a conventional system where LTE-A features are not implemented, for both uplink (UL) and downlink (DL). Also, the report compares the performance difference with the different
number of sites coordinated by one eNodeB.
1.4
Problem Formulation
Multiple antennas can be used to serve several aspects in a radio system. It may be used for diversity, which will increase the signal-to-noise ratio at the receiver, beamforming which will improve the overall SNIR and MIMO which will increase the throughput over the link.
The main focus of this project will be to investigate the performance of MIMO when the antenna system is mounted on the roof of a 4m high train and the base stations are located in LOS in a typical rural or sub-urban environment at various distances. An important parameter when using this type of multi-antenna configuration is the distance between the antenna elements, since it will affect the correlation between the different paths in the MIMO radio channel [12]. Two other important issues are the distance between the base station and the train speed. The latter will cause a Doppler shift, generating a phase shift on different channel paths, that will be directly beneficial for the performance of the MIMO system since it will help to decorrelate the received different signal.
This thesis will, hence, address the following research questions:
How is the performance of a train deployed MIMO antenna system affected by: 1. The distance between the base station and the train.
2. The distances between the receive antenna elements. 3. The train speed.
The investigation will be conducted through both theoretical analysis and measurements using a real 2X2 MIMO LTE 900 system on board a Swedish X2000 train. The results will be used to estimate the future of internet on board trains using LTE-A and its related features.
Chapter 2
Channel Theory
2.1
Wireless Channel
In this section, a theoretical background is presented in order to understand the coordination between BS and the UE. In wireless systems, the transmission of data from a transmitter to a receiver is performed via propagation of electromagnetic waves over an unguided environment, referred to as the wireless channel [9]. The propagated waves experience different attenuation in the wireless channel caused by different factors, such as reflection, scattering, and diffraction by walls, buildings, trees and any other object that block the direct path between TX-RX as shown in Figure 2.1:
Figure 2.1: Illustration of multipath in Wireless channel.
This will result in the arrival of multiple replicas of the transmitted signal at the receiver, each with different characteristics (delay, gain, phase, etc.). Hence, the wireless channel is sometimes referred to as a multipath channel [9]. The variation the signal wave experience can be characterized as two time-scales:
1. Large-scale fading or Pathloss: Slow channel variations as the distance between the transmitter and the receiver changes significantly over a time-scale in the order of tens of seconds [22]. The variations in large-time-scale fading are mainly due to pathloss and shadowing. Pathloss describes the decay in the received signal power due to the distance between the transmitter and the receiver [9]. Shadowing is the attenuation in the received signal power as a result of absorption, reflection, scattering, and diffraction of the transmitted signal through large obstacles between the transmitter and the receiver.
2. Small scale fading: Fast channel variation due to small changes in the relative spatial position of the transmitter and the receiver over a time-scale on the order of a few milliseconds [22]. Small-scale fading channel variations result from the constructive and destructive addition of multipath signal components [22].
Large-Scale fading affect the average power of the received signal, and it is usually is known by UE (User Equipment), and can be measured and fed back to the BS. In contrast, small scale fading affects the details of the received signal, and it is not an easy mission to measure this variation, neither the BS nor the UE can measure small scale fading.
As this project focuses on the LOS case in the moving trains, it is important to study the pathloss model
2.1.1
Pathloss
Pathloss in wireless channel refers to the received signal power attenuation, as a result of distance between the transmitter and the receiver [9]. Pathloss in linear scale is defined as the ratio of the transmitted power to the received power:
P athloss = PT PR
(2.1) The Path gain (P.G) in wireless networks defined as the inverse of the pathloss, i.e P.G=P athloss1 . The pathloss is given according to Friis equation:
P.L = λ.G
(4π.R)α (2.2)
Where G is the antenna product gains, R is the TX-RX distance, and α is the pathloss exponent, (α)=2 in case of free space. This model is usually used in case there is a direct path between TX and RX. .
Usually, the received power and pathloss is presented in logarithmic scale as shown in the equations below:
P athloss = 32.44 + 10α log(R) + 20 log(f ) (2.3) PR= PT − pathloss + Gt+ Gr (2.4)
Where R (TX-RX distance) is in km and f (frequency) in MHz.
2.2. MIMO 11
according to the environment. In the simulation results section, it was chosen to be 2.6 in LOS case as we expect some diffraction and reflection from the surrounding environment, and 3.6 in NLOS (Non-line of sight) case. The NLOS was considered in this project, to compare the LOS results with NLOS, and evaluate the MIMO performance in LOS.
2.1.2
Small Scale Fading
The small-scale fading variations in the time domain are caused by the relative mobility of the transmitter and the receiver and are characterized through a parameter called coherence time [9]. This parameter represents the time interval over which the small-scale channel variation is negligible as the relative position of the transmitter and the receiver changes [9].
The coherence time Tc of the channel is connected to the Doppler spread
fD as Tc ∝f1D, where fDis equal to the maximum Doppler shift and is given by
fD=
v
λ (2.5)
Where v denotes the relative speed of the transmitter and the receiver and λ is the carrier wavelength. The shorter the coherence time, the faster the channel changes with time. Doppler effect will be considered in the simulation in case of high speed train. This parameter need to be considered in simulation as the Doppler shift will affect different paths at the receiver side, since different angles of arrival and speeds of different paths will reflect on different phase shifts on the propagated channel matrix. This will decrease the correlation between different received waves, which will be reflected on MIMO performance.
2.2
MIMO
Several different diversity modes are used to make radiocommunications more robust, even with varying channels. These include time diversity (different time slots and channel coding), frequency diversity (different channels, spread spectrum, and OFDM), and also spatial diversity [23]. Spatial diversity requires the use of multiple antennas at the transmitter or the receiver end [24]. Multiple antenna systems are typically known as (MIMO). Multiple antenna technology can also be used to increase the data rate (spatial multiplexing) instead of improving robustness [24].
In practice, both methods are used separately or in combination, depending on the channel condition, and to benefit from multiple antenna systems, different channel conditions need to be received on different receiver antennas [24]. Different channel paths means uncorrelated channel matrix (H), which is usually caused by a multipath fading (scattering, reflection and diffraction), that occurs due to the non-coherent combination of signals arriving at the receiver antennas [23]. Typically, this phenomenon is described as the constructive/destructive interference between signals arriving at the same antenna via different paths, and hence, with different delays and phases, resulting in random fluctuations of the signal level at the receiver [23]. Figure 2.2 illustrate the different MIMO techniques which will be discussed.
Figure 2.2: Different MIMO scenarios.
2.2.1
SISO (Single Input Single Output)
This represents conventional systems use one transmitter and one receiver antenna. Using this method the radio channel capacity depends on the SINR and B (bandwidth) according to Shannon theory [23]. Using SISO neither transmission robustness nor data rate increase can be gained.
2.2.2
Receiver Diversity
Receiver diversity uses more antennas on the receiver side than on the transmitter side. In Figure 2.2, RX-diversity represents SIMO case. Because special coding methods are not needed, this scenario is very easy to implement [24]. Any signal transmitted from the single transmit-antenna will arrive at all receiver antennas through different sub-channels [23]. Assuming that each sub-channel, and hence, each channel element is completely decorrelated, multiple independent copies of the same signal arrive at the receiver, it is possible to exploit the concept of spatial diversity, in this case receiver diversity [23]. Using RX-diversity, the SNR at the receiver will be higher than in SISO according to different combing technique, which results on a more robust system. Three combining techniques (Switched diversity, Equal gain combining and maximal ratio combining) can be used at the receiver to enhance the SNR. Switched diversity always uses the stronger signal, while maximum ratio combining uses the sum signal from the two branches, each branch weighted by their respective instantaneous SNR [23]. Equal gain combining is a simplified version of MRC where all branches experience equal gain at the receiver nevertheless of their instantaneous SNR [2]. Figure 2.3 shows an example of the different diversity techniques in relation with bit error rate (BER).
2.2. MIMO 13
Figure 2.3: Example of different receiver combining techniques performance [2].
2.2.3
Transmitter Diversity
When there are more TX than RX antennas, this is called TX diversity. In Figure 2.2, TX-diversity represents MISO case. Assuming that perfect channel knowledge is available at the transmitter, it is possible to achieve transmit-diversity using maximal ratio combining (MRC) [23]. Applying MRC at the transmitter, also known as beamforming, requires the use of a filter [23]. The filtered symbol, when transmitted through all antennas travels to the single receiver antenna. Sending the same data on different transmitter antennas has the advantage that the multiple antennas and redundancy coding is moved from the mobile UE to the base station [24]. To generate a redundant signal, space-time codes are used. Alamouti developed the first codes for two antennas [24]. Space-time codes additionally improve the performance and make spatial diversity usable. Besides, The signal copy is transmitted not only from a different antenna but also at a different time [24]. This delayed transmission is called delayed diversity. Space-time codes combine spatial and temporal signal copies. Using Transmit diversity will increase the system robustness and decrease the BER of the system [23].
2.2.4
Spatial Multiplexing
Spatial multiplexing is not intended to make the transmission more robust; rather it increases the data rate [24], by allowing users to transmit different symbols from different transmit antennas [23]. An important issue to increase the data rates, the different paths of the channel matrix need to be uncorrelated, so that different data streams can be received at different receiver antennas, which means the Singular value Decomposition of the channel matriz (H) gives non-zero Eigenvalues, and the receiver can differentiate between different paths.
2.3
MIMO in LTE-A
In LTE-A data are mapped to layers after been encoded and modulated, the number of transmitted layers is called transmission rank, which permits up to 8 layers in downlink, and 4 layers in the uplink [25]. These layers are precoded and mapped to antenna ports to use one of the seven modes shown in Table 2.1:
Transmission mode Description
1 Single antenna port
2 Transmit diversity
3 Open Loop spatial multiplexing 4 Closed loop spatial multiplexing
5 Multi-user MIMO
6 closed loop single layer precoding 7 Single antenna Port Specific RS
Table 2.1: Transmission Modes
The most important three aspects to multi-antenna transmission are: MIMO modes, the channel state information (CSI) provided as input to the MIMO modes, and the base station antenna configuration [26] as shown in Figure 2.4
Figure 2.4: Multi-antenna modes and their relation to the antenna configuration and CSI modes.
Modes 1 and 7 are identical from the UE perspective, except that in mode 7 the UE specifies which RB (resource block) to transmit [21]. In both modes a single layer is sent.
In mode 2, a single layer is encoded (rank-1) with a space frequency block code (SFBC) based on almouti code, and transmitted from multi antennas [21].
2.3. MIMO in LTE-A 15
This mode shows the best performance with high spatial correlation between antennas. Besides, this mode doesn’t need any precoder information, only channel quality indicator (CQI) to choose proper modulation and coding selection [26]. Mode 3 is similar to mode 2 when the rank is one. However for higher ranks, a predefined codebook of precoder is cycled in the frequency band, and the rank indication (RI) is sent from the UE in CSI. Mode 3 serves higher data rates, this is suitable in case of high speed scenario (high mobility) [26].
Mode 4 is similar to mode 3 but the UE feedback the channel condition. Also, the base station contributes on how many layers should be mapped and selects a precoder matrix from a predefined codebook with the help of the mobile’s feedback on PMI (precoding matrix indicator) sent in CSI. This mode require stable channel characteristics and is applicable in a low mobility scenarios [26]. Mode 5 enables single-layer transmission to several users simultaneously, to share the same frequency allocation [27].
Mode 6 is similar to mode 4 except the restriction to single layer [21].
The frequency and time granularity of the CSI, together with the delay in the reporting are reflected on the performance of MIMO and which mode should be selected [18].
In the closed-loop spatial multiplexing mode, the base station (also known as eNodeB) applies the spatial domain precoding on the transmitted signal [18]. Taking into account the (PMI) reported by the UE so that the transmitted signal matches with the spatial channel experienced by the UE [18]. To support the closed-loop spatial multiplexing in the downlink, the UE needs to feedback the RI, the PMI and the CQI in the uplink [18].
The RI indicates the number of spatial layers that can be supported by the current channel experienced at the UE. The eNodeB may choose the transmission rank M, taking into account the RI reported by the UE as well as other factors such as traffic pattern and available transmission power [26]. The CQI feedback indicates a combination of modulation schemes and channel coding rate that the eNodeB should use to ensure that the block error probability experienced at the UE will not exceed certain percentage, for instance 10% [26]. Depending on the PMI feedback the eNodeB selects a suitable precoding code book depending on number of TX-antennas and the number of layers it can support [27].
For MIMO, an important issue is the antennas configuration; they need to be arranged to be decorrelated either by separating them with several wavelength, or by cross-polarizing. Both configuration can serve in non-line-of-sight. In case of line-of-sight cross-polarized have better performance to operate MIMO, this is due to strong direct path and high correlated paths [26].
2.4
MIMO Performance Indicator
The Condition Number (CN) is a key parameter that will be used for both the simulation model and the measurements to evaluate the MIMO performance. It is calculated as the ratio of the largest Eigenvalue to the smallest one in case of 2X2 MIMO. A lower CN gives higher MIMO performance and closer different paths gain (uncorrelated paths).
The ideal channel conditions is to have CN=0 dB, so that different paths will have the same gain. In practical, CN≤ 10 dB is considered a well conditioned channel to support MIMO, where the strongest path have gain varies between 1-3.16 times the weakest one. If 10<CN≤15, then the channel is considered unstable MIMO as the first path has gain in a range of 3.16-5.6 times the second path, while if CN>15 is considered an ill condition to support MIMO. Figure 2.5 illustrates the idea of CN and how it affects the spectral efficiency.
Figure 2.5: Illustration of CN concept (R&S) [3].
It can be noticed from the above figure that CN=0 and 15dB SINR give the same performance with 20dB SINR and CN=15 dB. Thus, CN is a key parameter to monitor, which gives an indication about MIMO performance [3].
2.5
Mathematical Model
The complex baseband MIMO received signal can be presented as follows:
r = Hs + n (2.6)
where r, is the M×1 revived complex-value signal vector. s, is the N×1 transmitted complex-value signal vector. H, is the N×M complex-valued channel matrix.
n, is the M×1 white Gaussian complex-value signal vector zero mean and variance σn2, denoted CN(0, σn2).
In our simulation model, the different transmitted branches are provided with equal power (PN uniform power allocation, PN=PNT) [20]. The channel capacity
2.5. Mathematical Model 17
calculation according to Shanon limits, can be expressed as given in equation 2.7 : C = log2[det IM + PT N σ2 n HHH ] bit/s/Hz (2.7)
where HH is the Hermitian transpose.
In this project we assume that all receiver antennas experience the same average received power. This average received power, PR is a function of PT, path loss
model, the distance, and shadowing. The average received signal-to-noise ratio (SNR) at one receive antenna then becomes γ=PR
σ2
n [20]. H is considered to be
normalized channel matrix, which implies that each element has a unit average power. By applying singular value decomposition and using average SNR, the capacity can be given as:
C = M X 1 log21 + γ Nλi bit/s/Hz (2.8)
Here λi is the i-th Eigenvalue of W, which is defined as W=HHH for M≤N
and W=HHH for M>N.
Equation 2.8 shows that the system can be viewed as consisting of min(N,M) parallel single input-single output (SISO) channels, where each channel has a gain of λi.
In general, to model the channel matrix as sum of two components, a LOS component and NLOS component, then H can be expressed as :
H = r K 1 + KHLOS+ r 1 1 + KHN LOS (2.9)
Where K is the ratio between the power of the two components gives the Ricean factor K [20].
The NLOS part in H is a result of reflection, scattering and diffraction from the transmission environment. The channel matrix of this part was modelled by circularly symmetric complex Gaussian elements with zero mean and variance equal to one CN(0,1). The absolute values of these matrix elements will be Rayleigh distributed and normalized.
Placing the antenna in smart way, MIMO in LOS can reach higher modes, which corresponds to nonzero Eigenvalues λi and higher MIMO capacity. Figure
2.6 shows the transmitter inter-antenna distance dtwhich is considered to be
constant in our simulation, receiver inter-antenna distance dr designed for a
certain system depending on the average distance between the TX-RX (R). The relation between these parameter will be illustrated later in the text. (N-1).dt
and (M-1).dr presents total transmitter and receiver array length respectively
Figure 2.6: Antenna orientation demonstration
To reach higher rank using HLO S, a relation between dt and dr is given in [20],
the technique is based on ray tracing. The path length between the TX-n and the RX-m is approximated as found in [20]:
rm,n≈ R+mdrsin θrcos φr−ndtsin θt+
(mdrsin θrsin φr)2+ (mdrcos θr− ndtcos θt)2
2R
(2.10) Transmission in one of the n antenna will be received by M different antennas. As the train is moving, there will be different angles of arrival, and the distance R will be changing with time. Considering that each path will experience different path distance according to rm,nwhich will be reflected on different phase shift
of the channel matrix. The received vector from antenna n on the M receive antennas can be formulated as :
hn= [exp( j2π λ r1,n), ...., exp( j2π λ rM,n)] T (2.11)
where (.)T denotes the vector transpose and λ is the wavelength, λ will be
effected by the Doppler shift as discussed in Section 2.1.2, this will increase the probability of having different phases of the received signal on each antenna element. Then HLOS can be given as :
HLOS = [h1,h2,...,hN].
Deriving the optimal antennas separation for LOS channel component so that orthogonality between different columns in HLOS is achieved [20]. The condition
that should be satisfied is the inner product between two different transmit antenna vectors should equal to zero:
hhk, hli = M X m=1 exp(j2π λ(rm,k− rm,l)) (2.12) = M X m=1
exp(j2πdtdrcos θrcos θt
λR (l − k)m) (2.13) (2.14) =sin(π dtdr λR cos θrcos θt(l − k)M ) sin(πdtdr λR cos θrcos θt(l − k)) (2.15)
2.5. Mathematical Model 19
dtdr=
λR
max(N, M ) cos θrcos θt
(2.16)
As explained in [20], going from Eq.(2.13) to Eq.(2.15) they used the expression of a finite geometric series, and to Eq.(2.16) the solution with smallest distance were chosen.
From Eq.(2.16) many parameter can be modified to reach higher MIMO rank in LOS. In our simulation all the parameters are set to be constant, and the variable parameter is dr and the angle of arrival, which will be applicable to be
accomplished in real life, as there is enough space on the train roof to separate antennas for a longer distances.
Chapter 3
Measurements Setup
In this chapter, we describe the train system under test. It will describe the cable connections, RF antennas, the equipment used for the measurements and link-budget calculations for different tested scenarios, Figure 3.1 shows an illustration of the train setup.
Figure 3.1: Train Diagram.
3.1
Rohde & Schwarz Universal Radio Network
Analyser equipment (TSMW)
The TSMW is the equipment used to collect data on board the train. It is commonly used for optimizing conventional wireless communications networks. It supports input frequency from 30 MHz to 6 GHz [3]. It has two independent RF and signal processing inputs, each with 20 MHz bandwidth. The equipment supports both TDD and FDD LTE systems. Moreover, the TSMW is able to scan simultaneously GSM, WCDMA, LTE, CDMA2000, TETRA and networks [3]. For collecting RF data, the user specify the centre frequency of the LTE signal, then the TSMW can find all further information that is required, e.g. the bandwidth used, the physical cell ID, the eNodeB cell ID, RSRP (reference signal received power) and RS-SINR values. The measurement results can be output at a maximum rate of 200 measurements per second.
For MIMO, the two antennas simultaneously measure the LTE signal, making it possible to determine the degree of correlation of the MIMO channel. All
measurements are based on the channel matrix H with the complex amplitude and phase values. The TSMW generate matrix output for each measured resource block.
Figure 3.2: The TSMW equipment [3].
Firstly, to give an idea about the system dealt with while testing, Figure 3.3 illustrate the train schematic digram as implemented on the tested train.
Figure 3.3: Train setup digram.
It can be noted from the figure that the antennas and cables are already installed and connected in the train by Icomera to provide internet on board trains using
3.2. RF Antenna 23
the Icomera product. To start the measurements and collecting data, the only access provided was to the train cabinet in the which the TSMW equipment was installed. As can be seen in Figure 3.3, the TSMW needs two antenna connection to collect data. From the cabinet, only the cables tag number can be seen, which makes it possible to measure four test scenarios :
• Connecting both cables with tag 1 to TSMW: this will give a test measurement with 1.5m separation on the receiver antennas.
• Connecting both cables with tag 2 to TSMW: this will give a test measurement with 4.5m separation on the receiver antennas.
• Connecting both cables with tag 3 to TSMW: this will give a test measurement with 7.5m separation on the receiver antennas.
• Connecting both cables with tag 4 to TSMW: this will give a test measurement with 10.5m separation on the receiver antennas.
After connecting the antennas to the TSMW, data were recorded on a PC using a drive test software provided from R&S called ROMAS. Figure 3.4-3.5 shows the train used for the test and the setup inside the train cabinet.
Figure 3.4: Train used for the measure-ments.
Figure 3.5: Installed equipment inside the train cabinet.
3.2
RF Antenna
The receiver antennas are mounted 4m above the ground level. The base-stations are about at 45m height on the average along the track. The antennas were provided by HUBER+SUHNER and named “Sencity Rail Antenna”. The antenna is a broadband and supports GSM 900, GSM 1800, GSM 1900, UMTS, LTE, WiFi and WiMAX 2.4, 3.5, 5.3 and 5.8 GHz Bands [5]. The antenna has an
N-type (female) connector, and its dimensions are 153 X 100 X 256 mm (Height, Width and Depth) [5]. Figure 3.6 shows the antenna gain versus frequency.
Figure 3.6: Antenna gain vs. frequency [4].
The measurements were conducted on LTE 900MHz band. Figure 3.7 shows the antenna radiation pattern on this band [28]. Referring to the data sheet, the antenna has 4dBi gain and 50Ω nominal impedance on the 900MHz band.
Figure 3.7: Antenna radiation pattern at 900MHz [5].
Figure 3.8 shows how the antennas were installed on the train roof, while Figure 3.9 shows the antenna used in the test.
3.3. Cables 25
Figure 3.8: MIMO antennas installed on the train roof.
Figure 3.9: Antenna used in the measurements (HUBER+SUHNER).
3.3
Cables
The cables used to connect the antenna from the roof of the train to the cabinet inside the train were provided from LEONI Special Cables GmbH named “FlexLine, super flexible cables”. The inner conductor is a coppered-clad aluminium wire with 3.6mm diameter, and is insulated by foamed Polyethylene (PE) with skin of 9.1mm diameter [6]. Where the outer conductor is designed as copper-tape (thickness 0.25mm) longitudinal welded spirally corrugated with 12.3mm± 0.15mm diameter [6]. The jacket of the wire is thermoplastic copolymer (FRNC) BK with a thickness of 0.6mm, the jacket has a 13.5mm diameter [6]. The characteristic impedance of the wire is 50±1Ω. The attenuation of the cable for different frequencies is shown in Figure 3.10
Figure 3.10: Cable attenuation vs. frequency [6].
From the data sheet and the Figure 3.10, on 900MHz band the cables will have attenuation (loss) by 9.9 dB/100m [6], This will be used in te link-budget calculations in the next section. Figure 3.11 shows small part of the cable used in the test.
Figure 3.11: Cables used in the test (LZSH) [6].
3.4
Link-Budget and Sensitivity
From Figure 3.3, it can be noticed that different cables tag number have different lengths. This will affect the overall gain, since longer cables will have more attenuation. To calculate the noise added by the system, the over all noise figure(F) needs to be calculated for the system shown in Figure 3.12. The noise temperature for the antennas, cables and RF connectors was assumed to be 290K, which means F=1.
Figure 3.12: System digram.
The TSMW equipment received sensitivity is -127dBm for LTE as indicated in R&S data sheet with F= 7dB= 2.238. To calculate the SNR on each scenario, the overall F will equal
F = 1 + FT SM W − 1
3.4. Link-Budget and Sensitivity 27
the noise level=FKToB (Watt), where
• K is Boltzmann’s constant (1.38x10−23 J/K).
• To=290K.
• B is the bandwidth, in the conducted test the bandwidth used is 5MHz, but the TSMW measures the sensitivity on the Sync-carriers which is equal to 62 carriers, each carrier has 15KHz BW, which gives B=930KHz. Figure 3.13 shows the link-budget calculations on 900MHz where all losses were considered. Taking into consideration that the antenna gain is 4dBi and the cable attenuation is 9.9dB/100m, the RF connectors losses were provided by Icomera. As can be seen, the calculated overall F doesn’t have a significant change for the different scenarios, it is about 2.9dB for all scenarios, this shows how good are the cables used for the test.
Chapter 4
Simulation and
Measurements Results
4.1
Simulation Results
The MIMO system was simulated in Matlab. The mathematical model discussed in Chapter 2 was used to mimic the LOS component. The spectral efficiency was studied for both NLOS (K=0) and LOS (K=∞) to evaluate the MIMO performance in LOS compared to NLOS case, where K is the Rician factor discussed in chapter 2. The simulation parameters are as shown in Table 4.1 :
Simulation parameter Value
MIMO mode 1×1 , 2×2 , 3×3, 4×4 , 8×8 Frequency 900 MHz θt 0 θr π/3 φr π/6 dt 10λ PT 45dBm TX-gain , RX-gain 18dB , 4dB α 2.6 LOS , 3.5 NLOS Noise level -120 dBm Distance(R) 500m - 2500m Rician factor (K) 0 , ∞ Train Velocity 150-200 km/h BS , receiver hight 45m, 4m
Table 4.1: Simulation Parameters used for the simulator.
Simulating the above parameters and taking into the account that the average train velocity was 175km/h causing Doppler shift of the received signal. This will causes a phase shift in calculating the H-matrix components. The system simulate different cell radii, which reflects on different required receiver antenna-elements separation. The received power and SNR were calculated at the receiver
according to the path loss propagation model. The spectral efficiency versus the SNR was plotted as shown in Figure 4.1 for both cases LOS(K=∞) and NLOS (K=0)
Figure 4.1: Left: K=zero(NLOS),Right: K =∞(LOS)
Applying the above mentioned model, MIMO in LOS can reach higher rank and higher capacity as shown in Figure 4.1 where the LOS scenario performs with the same fashion as in the NLOS scenario. The NLOS perform better with higher SNR due to higher Eigenvalues gains, and less correlated paths.
To evaluate the difference between this model and the current deployed situation, where the receiver inter-elements antenna distances are fixed to a small distances. In most of the UE’s, dr (the distance between receiver elements) is fixed to λ/2.
Figure 4.2 shows a comparison in case of 4 × 4, the figure shows that at 20dB SNR the bandwidth efficiency increased from 10 to 30 bit/s/Hz.
Figure 4.2: Comparison for 4 × 4 MIMO between fixed drand conditioned dr
4.2. Measurements Results and Analysis 31
The receiver antenna separation distances were found from the above mentioned simulations. For different cell radii the simulation calculates the needed separation between receiver antenna elements so that the CN is between 0.1-2 dB i.e the different paths have close gains, the results are shown in Table 4.2. As can be noticed for higher distances we need higher separations between antennas, as the paths become more correlated and more difficult to resolve, so higher receiver antenna elements distances will affect the phase shifts on different paths.
Receiver elements distance (m) Distance between TX-RX (m) CN(dB)
8.2= 24.84λ 300 1.2 12.5 = 37.8λ 500 0.5 17.7 = 53.63λ 700 0.9 22.3 = 67.57λ 900 0.1 27.5 = 83.33λ 1100 1.3 33 = 100λ 1300 1.7 37.2 = 112.72λ 1500 2 42.6 = 129.09λ 1700 1.1 57.5 = 174.24λ 2300 1.5
Table 4.2: Relation between receiver distance elements and TX-RX distance
The above table shows the perfect inter-elements receiver antenna distances in a relation with TX-RX distance, to reach the best MIMO performance which results an uncorrelated radio paths with close Eigenvalue gain.
The above results show that the theoretical model simulation, conditions on the inter-elements distances at the receiver, capacity in case of LOS MIMO. The next section will go through the collected measurements. Afterwards, the results of the real life network will be compared with the proposed model.
4.2
Measurements Results and Analysis
Measurements are collected from the moving train using the TSMW. The idea is to connect the device to the antennas installed at the train, and collect different data including: RSRP, SINR, BS distances, RI, H matrix and CN. The RF (radio frequency) data collected are measured on the reference pilots sub-carriers within the channel bandwidth.
The channel matrix elements (h00, h01, h10and h11) are collected using TSMW
and exported for analysis. As explained in Chapter 2, CN is a key parameter to measure the performance of MIMO, the train setup discussed in Chapter 3 will be tested in real life network.
Firstly, before digging into data analysis, Tele2 provided us with the coverage map along the train track that will be tested, mainly to get an idea about the covered area while testing as shown in Figure 4.3.
Figure 4.3: Coverage between Stockholm-Gothunborg (Tele2 map).
The pink colour shows the coverage of LTE 933.6 MHz carrier (band 8, E-UTRA Absolute Radio Frequency Channel Number (EARFCN)= 3537). Tele2 currently uses 5MHz channel bandwidth on this band.
Secondly, data were collected using TSMW device, four different scenarios were simulated on the receiver inter-antenna distances (1.5m, 4.5m, 7.5m and 10.5m) to cover different aspects. Eight trips were conducted to collect the data between Stockholm-Gothunborg, each trip with different antenna separation. Then the data were exported to excel for analysis. The data was recorded each second along the track on each scenario. At each point the following where recorded:
• GPS of the base stations and the train. • The base station ID.
• RSRP (dBm).
• RS-SINR (reference signal-signal to interference and noise ration) dB. • H-matrix elements, where the H-matrix elements are collected on each
resource block (mW) with SINR on each resource block (dB). Within 5 MHz BW there is 25 resource blocks.
• CN on each resource block within each second. • RI of each resource block within each second. • Speed of the train km/h
To start with the analysis the distance between the train and the base station was calculated, the Spherical law of cosines were used as illustrated in the equation below:
Distance = arccos(sin(lat1)∗sin(lat2)+cos(lat1)∗cos(lat2)∗cos(lon2−lon1))∗R (4.1)
4.2. Measurements Results and Analysis 33
Where lat1, lat2, lon1, and lon2 are in radians. And R is earth radius = 6371 km.
The train is in LOS most of the time with the base station, so the TSMW will be able to capture a very far base station as the signal doesn’t suffer from much attenuation, but in reality the train will be moving with very high speed, and the far base-station will not be able to provide the train with connectivity, as far BS will have a high packet loss with high mobility. The following criteria was selected for data analysis so that a measured point is considered covered and enter the data analysis:
1. RSRP≥-110 dBm.
2. RS−SINR≥ -1.5 dB.
3. Distance≤6 km.
For the different scenarios the coverage changes with a small percentage, the percentage of the covered area along the track is illustrated in Figure 4.4 for the different scenarios:
Figure 4.4: Different scenarios coverage statistics.
The statistics from the figure above shows that there is LTE coverage around 47% along the track on average. To measure the accuracy of the measurements, the percentage of the covered track repeated for 10.5m and 7.5m with different trips, and the coverage changes by 1-2%.
The covered points were mapped using MapInfo to show the covered areas as shown in Figure 4.5:
Figure 4.5: Covered areas between Stockholm-Gothunborg (measured data).
From Figure 4.5 it can be seen that measured covered area agrees with the map provided by Tele2. This shows that the criteria chosen for considering a point covered agrees with Tele2 coverage prediction. Also, the points were mapped on Google earth, and most of the points could be considered in LOS with rural environment, except from the points near the cities (Stockholm, Gothunborg). To ensure that enough data were collected, the CDF of the RSRP for the different scenarios were compared with a Gaussian distribution that fit the data as shown in Figure 4.6:
Figure 4.6: Different Scenarios RSRP CDF.
It can be concluded from Figure 4.6 that most of the data fits with a Gaussian distribution with N(-80,12), the 7.5m doesn’t follow the Gaussian perfectly as the others, the data have a mean of -81.5 dBm and a standard deviation of 13.1. The small differentiation between different scenarios is referred to the fact that
4.2. Measurements Results and Analysis 35
the train may stop at different stations for a different period of times, this will be recorded and will be reflected on the data statistics. Furthermore, Tele2 were rolling new BS within the test period, this will be reflected on different scenarios statistics.
The purpose of the project is to study the MIMO performance in LOS scenario, so a further data analysis were performed to evaluate the MIMO performance on the covered points. As previously mentioned on each point the CN where calculated on each resource block using Singular Value Decomposition, also the TSMW records the RI for each resource block, which means if CN≤10 and RI is 2 then the receiver can support MIMO without any problem, meanwhile if CN≤ 10 and RI is 1 or 0 then the receiver can’t support MIMO.
The average on all resource blocks CNs within the same second (record) were calculated, to evaluate the CN within the whole channel at each point. The MIMO statistics for each scenarios is shown in Figure 4.7:
Figure 4.7: Covered MIMO points statistics
Recall that MIMO working but not stable is when the 10<CN≤15, to illustrate the concept for different MIMO naming used Figure 4.8, 4.9 and 4.10 shows different measurements snapshots collected within the test.
Figure 4.10: MIMO not working.
From the Figures shown, the y-axis represents a scale in dB, and the x-axis is the resource block number, and the red line is the CN on the whole BW. From Figure 4.8 all the resource blocks had CN≤10, which gives good MIMO performance, where in Figure 4.9 the CN at the beginning of the channel is high(>10) but after resource block number 10 it become<10, so around half of the channel resource blocks will support MIMO. Figure 4.10 doesn’t support MIMO as all the resource blocks have CN>15.
According to the theoretical model provided, the conditions for MIMO to work in LOS, depends on both the relation between TX-RX distance with inter-antenna receiver distance and the speed of the train. The MIMO working and MIMO unstable points where analysed to relate it with the train speed. Theoretically high speed will cause higher Doppler shift, which will reflects on decorrelated paths. Figure 4.11 shows the statistics on train speed for different scenarios:
Figure 4.11: MIMO performance in relation with the train speed.
It can be seen that most of the good MIMO performance points occurred while the train is moving with high speed, but in the same time most of the measured data are recorded while the train is moving with speed>100km/h. In this case, MIMO can work with high speeds, but we can’t declare that the MIMO performance is fully related with the speed and Doppler shift as in the simulated model.
4.2. Measurements Results and Analysis 37
To illustrate the above figures, let us take the 4.5m scenario, from Figure 4.4 it can be seen that this scenario had 46% coverage, within this covered points, there is 35% points support MIMO and 28% points with MIMO works but unstable as shown in Figure 4.7. Then from Figure 4.11, it can be concluded that 74% of MIMO working points occurs when the train is moving with a speed>100km/h, and 75% of MIMO working but unstable points occurs when the train is moving with speed>100km/h. The same way, all data for different scenarios can be read.
The RSRP of the collected data were taken as a reference for further data analysis, which means that the RSRP readings were split up into different interval with 5dBm interval i.e from -50 to -55 dBm was taken as an interval to average up other data, and from -55 to -60dBm was considered another interval and so on till -105 to -110 dBm.
Firstly, the RS-SINR were averaged within each RSRP interval, Figure 4.12 shows the relation between RS-SINR vs RSRP for all scenarios.
Figure 4.12: Measured average RS-SINR vs. RSRP.
It can be noticed from the Figure, all scenarios follow the same trend, where the RS-SINR decreases as the RSRP decreases, and it showed that there is no randomness on the RS-SINR collected data.
Secondly, the distance calculated were averaged within each RSRP interval, Figure 4.13 shows the relation between distance vs RSRP for all scenarios.
Figure 4.13: Measured average distance vs RSRP.
On the same fashion the distance increases as the RSRP decreases. The distance relation vs RSRP showed that the tested model agrees wit the Pathloss model used in the simulation. As noticed on higher and low RSRP interval there is not much data as for the middle intervals as can be concluded from Figure 4.6, which reflects on some trend changes.
Since the H-matrix and SINR are available for each resource block along with RI, the TSMW estimates the throughput by taking MIMO/decision per resource block, where in the UE it is taken per TTI (transmission time interval). The throughput measured by the TSMW will give a sense of the throughput can be achieved by the UE. Figure 4.14 shows the average throughput along the track for each scenario, a plot of spectral efficiency (capacity) vs SINR will be provided in the next section, when the measurements are compared to the simulated model.
Figure 4.14: Average throughput along the track.
The statistics (mean and standard deviation) of all above scenarios where calculated as shown in Figure 4.15
4.3. Measurements and Simulation Comparison 39
Figure 4.15: All measurements data statistics.
4.3
Measurements and Simulation Comparison
The above discussion shows the measured data analysis and results. These results need to be compared with the theoretical model. The theoretical model have a large SNR range as seen from Figure 4.1, and it shows up to 8X8 MIMO which make it difficult to compare it with the measurement data.
For comparison, the practical RS-SINR data were split up into intervals of 1dB and average on each interval starting from 1 up to 15 dB, i.e the last RS-SINR average interval is 14-15 dB, and the spectral efficiency were averaged on each interval. At the same time, the same theoretical model used before were used for comparison, but the simulation were manipulated to show only 2X2 scenario and the same SINR range were used as in the measurements range. The comparison results are shown in Figures 4.16:
Figure 4.16: Spectral efficiency vs RS-SINR for both Measurements and simulation (2X2 MIMO) (a:1.5m scenario, b:4.5m scenario, c:7.5m scenario and d:10.5m scenario)(Blue: simulated, Green: Measured).
4.3. Measurements and Simulation Comparison 41
The figures above show that both practical and simulation follows the same trend and according to the spectral efficiency within each RS-SINR range. The difference between simulation and practical on each RS-SINR value were calculated, the results are shown in Table 4.3
Description 1.5m 4.5m 7.5m 10.5
Scenario Scenario Scenario Scenario Max Difference between
measurement and simulation 2.8 2.2 2.32 1.06
bps/Hz
Min Difference between
measurement and simulation 0.52 0.08 0.6 0.01 bps/Hz
Average Difference on
whole RS-SINR range 1.58 1.06 1.18 0.76
bps/Hz
Table 4.3: Spectral efficiency statical summary for the comparison between measurements and simulation.
The difference may be explained as follows: in the theoretical model zero interference from other base stations were simulated, but in reality interference can’t be excluded. Also in the theoretical model the environments were considered ideally in LOS without any reflections and multipath (K=∞), but in the train situation ground and tree reflections is expected, which affects the signal quality and different paths correlation. Moreover, in the simulation part the receiver considered to receive all resource blocks at the same CN (same channel environment), where in measurements each resource block responds to the channel environment independently, which means different CN on each resource block, that will affect the capacity calculations.
Finally, as the speed were related to MIMO performance and the simulation were compared to the practical measurements, an important issue to investigate is the relation between RX-TX and RX-antenna separation in both theoretical model and the practical measurements.
The data provided in table 4.1 is the optimum separation to have CN ≤2dB. Obviously, in the simulation CN will be≤10dB with different TX-RX distances and fixed RX antenna-elements distances. The CN were averaged every 15m TX-RX distance for collected data, and the CN vs distance were plotted for both the measurements and simulation, only MIMO working range is shown in Figure 4.17:
Figure 4.17: CN vs distance showing the 2X2 MIMO working points for both simulation and measurements (a:1.5m scenario, b:4.5m scenario, c:7.5m scenario and d:10.5m scenario)(Red: simulated, Black: Measured).
![Figure 3.7: Antenna radiation pattern at 900MHz [5].](https://thumb-eu.123doks.com/thumbv2/5dokorg/5507234.143464/42.892.247.731.706.948/figure-antenna-radiation-pattern-at-mhz.webp)
