Seminar. Ausgewählte Kapitel der Nachrichtentechnik, WS 2009/2010. LTE: Der Mobilfunk der Zukunft. Broadcast Operation. Christopher Schmidt

Transcription

1 Seminar Ausgewählte Kapitel der Nachrichtentechnik, WS 2009/2010 LTE: Der Mobilfunk der Zukunft Broadcast Operation Christopher Schmidt 27. Januar 2010 Abstract Long Term Evolution (LTE) provides an improved Multimedia Broadcast Multicast Service (MBMS) compared to the Universal Mobile Telecommunications System (UMTS), to distribute multimedia content to multiple users. To this end, LTE uses unicast communication as well as broadcast and multicast communication combined with a Single Frequency Network (SFN). The first part of this report describes the fundamentals needed for MBMS transmission in LTE. Afterwards, the second part deals with the Broadcast Operation itself, concluding with an outlook to following releases of the LTE standard. 1 Introduction The number of application areas for high-bandwidth multimedia services, which require high data rates, is rising. Thus, in 3GPP Long Term Evolution (LTE), the successor of Universal Mobile Telecommunications System (UMTS), the support of an enhanced version of Multimedia Broadcast Multicast Service (MBMS) has been an essential requirement. Therefore, an efficient broadcasting mode is needed, to distribute the multimedia content to multiple users; the MBMS transmissions can be realised as single-cell transmission or as multi-cell transmission. For multi-cell transmission, all participating cells transmit the same signal under the constraint of tight time-synchronization using the same spectral resources to enable the User Equipment (UE) to handle the received superimposed signal like one that has passed through a multipath propagation channel; this is the concept of a Single Frequency Network (SFN). For this purpose, LTE provides a system architecture, which is able to switch between unicast,

2 2 Christopher Schmidt multicast and broadcast communication on different carriers, namely dedicated and mixed carrier, to transmit the multimedia content always in the most efficient way to multiple users. Besides, LTE is based on an air interface, which is robust against the multipath character of the wireless communiation channel. Hence, in LTE efficient MBMS transmission is possible due to the properties of the Orthogonal Frequency Division Multiplex (OFDM) air interface [1]. 2 Air Interface 2.1 Signal Propagation LTE is a mobile terrestrial communication system, designed to increase capacity and data rates compared to UMTS Release 6. UMTS is a third-generation mobile communications technology and was the successor of the Global System for Mobile communications (GSM). Like every terrestrial communication system, LTE has to deal with the typical terrestrial propagation channel, which is dominated by path loss, multipath propagation and shadowing. Path loss is the reduction of power flux density resulting from the propagation in free space without any reflections or additional attenuation caused by obstacles. The transmitted power is distributed over the surface of a sphere, so it decreases proportionally with the square of the radius. In addition to the free space propagation, there is also a propagation loss due to obstacles. The received power can be described by the power law P r d n, with d being the distance between transmitter and receiver and n the path loss exponent [6]. In case of free space propagation, the path loss exponent is equal to 2, as a result of the quadratic decrease of power flux density. In a realistic environment there are e.g. buildings, mountains, and forests between transmitter and receiver, so the path loss exponent is usually larger than 2 (cf. Tab. 1). The Okumura-Hata model provides a mathematical description of the signal attenuation in different morphological environments which is caused by path loss and shadowing. This model is based on a large number of measurements in rural, urban, and suburban areas to get information about attenuation, dependence on distance, carrier frequency, and transmitter and receiver antenna height. Environment Path loss exponent Free Space 2 Urban area 2,7... 3,5 Urban area with shadowing Inside buildings with direct line-of-sight 1,6... 1,8 Inside buildings with shadowing Inside factory buildings with shadowing Table 1: Environment and associated path loss exponent. [7]

3 Broadcast Operation 3 The transmitted signal is reflected at buildings, mountains and other terrestrial obstacles, so additional propagation paths with different delays occur (cf. Fig. 1); this phenomenon is called multipath propagation. The result is a longer Channel Impulse Response (CIR), since these paths arrive at different delays at the receiver. The Power Delay Profile (PDP) describes the intensity of a received signal as a function of the time, which was transmitted over a multipath channel. There is constructive and destructive interference dependent on the phase relations of the signal components. The resulting minima and maxima alternate at approximately half of the wavelength of the carrier frequency. Thus, there are changes in the strength of the received signal within a few meters, so multipath fading is also called fast fading. The multipath fading can be modelled by a stochastic process. Figure 1: Multipath Propagation [10] Terrestrial obstacles also cause further attenuation of the transmitted signal; this phenomenon is called shadowing. The direct line-of-sight is interrupted, so only a part of the transmitted power can be received in the shadowed areas. The size of the shadowed areas and the attenuation of the signal within this area are dependent on the morphological structure of the environment. Shadowing is usually considered as constant in a range of approximately 10 to 100 meters, depending on the size of the obstacles, so the attenuation due to shadowing changes slowly compared to multipath effects; this effect is referred to as slow fading. Shadowing can also be modeled by a stochastic process. This process describes the statistical characteristic of the mean received power of the multipath fading. Usually, the channel is modelled by static coefficients for path loss and shadowing. The timevariance of the CIR is dependent on the speed of the user, and the CIRs result as realizations from their associated stochastic process [1] [6].

4 4 Christopher Schmidt 2.2 Orthogonal Frequency Division Multiplex (OFDM) LTE uses Orthogonal Frequency Division Multiplexing (OFDM) in the downlink, as shown in [8]. OFDM is a digital multi-carrier modulation method, which distributes the modulated data symbols to N orthogonal sub-carriers, with N being the length of the Discrete Fourier Transform (DFT). The per subcarrier symbol rate is very low compared to a single-carrier modulation, because the symbol duration in each narrowband sub-carrier is N times longer. The insertion of a Guard Interval (GI) between the discrete-time OFDM symbols enables a constructive combination of delayed paths at the receiver due to multipath propagation without Inter-Symbol Interference (ISI). Therefore, OFDM is particularly suitable for broadband mobile communication systems like LTE. The mobility of the UE can also cause problems: Inter-carrier Interference (ICI) occurs, when the orthogonality of the sub-carriers is lost, e.g. as a result of Doppler shift. Figure 2: Insertion of a CP in OFDM [1] In OFDM, the GI is realised as a Cyclic Prefix (CP) which repeats a defined number of discrete time samples from the end of an OFDM symbol at its beginning. The length of the CP also determines the longest acceptable path delay; there is no ISI at the receiver if the maximum delay path in the CIR arrives within the GI (cf. Fig. 2). The received signal for Single-Input Single-Output (SISO) transmission is given by J r i [k] = h j,i [k] s i [k] + n i [k] (1) j=1 in the discrete-time domain. Provided that the GI is not exceeded the received signal in the DFT domain is given by R i [m] = H i [m] S i [m] + N i [m], (2) where i represents the OFDM symbol index, S i [m] the transmitted signal, N i [m] the noise, m the subcarrier index 0 m N 1 for N being the DFT size, and H i [m] = J j=1 H j,i [m] as well as H j,i [m] = DF T N {h j,i [k]} the CIR.

5 Broadcast Operation Impact of Guard Interval Excess In the following, the effect of an insufficient length of the CP is discussed. N is defined as the DFT size, G is the number of samples of the CP, and L is the number of non-zero taps of the channel Power Delay Profile (PDP) ρ[k]. The ISI power can be calculated by and the ICI power results in P ISI = N+G 1 k=g 2 (k G)2 h[k] (3) N 2 N+G 1 2 N(k G) (k G)2 P ICI = 2 h[k], (4) N 2 k=g with being h[k] the CIR. The mathematical relation of P ISI, accumulating the overlapping energy for k G, is plotted in Fig. 3, as well as P S and P ICI. The power of the ISI and the ICI increases when the boundary of the CP is exceeded. Hence, the useful signal power P S is reduced. The useful signal power P S is given by P S = G 1 k=0 h[k] 2 + N+G 1 k=g The resulting Signal-to-Interference Ratio (SIR) is given by SIR = 2 (N k + G)2 h[k]. (5) N 2 P S P ISI + P ICI. (6) Fig. 4 illustrates the SIR and shows the influence of an exceeded CP on the SIR. The SIR decreases rapidly, because the signal power P S is reduced by the sum of P ISI and P ICI. The SIR plotted in Fig. 4 is based on a rectangularly shaped PDP; this means, that all received delayed paths arrive at the UE with the same power. In a realistic multipath environment, the path power would decrease with an increasing delay, so the SIR would decrease slowlier with increasing value of L, because the paths, which exceed the CP would have less power on average [1] [7]. 2.4 Multicast/Broadcast Single Frequency Network (MBSFN) In a Multi Frequency Network (MFN), the use of different carrier frequencies in different cells is necessary to prevent high levels of interference especially for UEs located at the cell edges. Accordingly, the frequency reuse is low, because no surrounding cell is allowed to reuse the frequency. A Single Frequency Network (SFN) uses one common carrier frequency for all enodebs for the simultaneous transmission in a multi-cell network (cf. Fig. 5). The enodebs have to be

6 6 Christopher Schmidt Figure 3: P S, P ISI, and P ICI as a result of an exceeded CP [1] tightly time-synchronised to guarantee that the same information is transmitted at the same time on the same set of OFDM subcarriers at every base-station. Fig. 6 illustrates an SFN, consisting of J = 3 Base Stations (BS)s. The composit CIR h[k] is given by the sum of J CIRs linking the UE with the BS, i.e. h[k] = J j=1 h j [k]. A receiver in this geographical SFN area will pick up multiple versions of the same signal from different enbs with different time delays due to the different distances of the UE from the enbs. The synchronisation enables the receiver to handle the multiple signal versions like multipath propagation, because they appear as echoed paths from a single transmitter in an SFN; so there is no additional complexity required for discrete-time channel equalization as long as the CP is not exceeded. In LTE, SFN transmission can be easily realized due to the OFDM air interface. The length of the CP of the OFDM modulation approximately restricts the maximum ISI/ICI free distance between the enbs and the UE. As long as the sum of the propagation delay and maximum delay of the CIR does not exceed the CP, there is no ISI and ICI. The PDP is dependent on the morphology of the propagation environment. Thus, the length of the CP has to be adapted to the environment of the SFN. In urban areas, the attenuation of the signal is higher than in rural areas, because, the path loss exponent is larger. This means that the effective range of a transmitted signal is shorter compared to rural areas and therefore the CP in urban areas can

7 Broadcast Operation 7 Figure 4: SIR as a result of an exceeded CP [1] be shorter which is exploited for unicast transmission. The main advantage of an SFN is the transformation of inter-cell interference at the cell edges into useful signal energy by combining the different components constructively. The size of the covered area can easily be increased by adding further enbs on the same OFDM subcarriers to the SFN instead of allocating further spectrum for other frequencies in an MFN. The main drawback of an SFN is the requirement for time-synchronisation of all enbs belonging to the SFN. This requires a synchronisation with an accuracy which should be considerably smaller than the length of the CP. Otherwise, the CP will be exceeded and there will be destructive interference due to delayed transmission, resulting in lower SIRs (cf. Fig. 4). LTE also provides unicast transmission, because the enbs used for the unicast cells do not have to be synchronized in FDD mode, which means that the synchronization complexity is reduced [1] [7] [2]. One of the main advantages of LTE, compared to UMTS Release 6, is the possibility of using a multi-cell single frequency network for data broadcast. An SFN increases the spectral efficiency, due to the enhanced signal to interference-plus-noise ratio (SINR) in comparison to unicast transmission. The combination of the different received signal paths from different enbs increases the received signal energy, while the inter-cell interference is reduced particularly for users which are located at the all border. Fig. 7 shows the coverage vs. the spectral efficiency; the size of the MBSFN area is used as parameter. When the SFN-area size increases, then the spectral efficiency improves at equal coverage: adding a first ring around the central cell

8 8 Christopher Schmidt Figure 5: Multi Frequency Network vs. Single Frequency Network [9] generates a significant gain in spectral efficiency as a result of the increased distance to the interferers, i.e., the surrounding enbs which transmit other data services. A further increase of the number of MBSFN cells surrounding the central cell results in a further but smaller benefit, caused by a decreased interference power and a smaller received useful signal energy, due to larger distances between the UE and the additional enbs. In LTE, three different lengths of the CP are standardized: the normal CP, the extended CP, and an extended CP in combination with a smaller sub-carrier spacing paired with a doubled OFDM symbol length (cf. Fig. 8). Tab. 2 shows the different CP durations and the corresponding propagation distances. CP mode GI duration Corresponding propagation T g distance d p = c T g Normal CP ( f = 15 khz) 4,69 µs 1406 m Extended CP ( f = 15 khz) 16,67 µs 5000 m Extended CP ( f = 7,5 khz) 33,33 µs m Table 2: CP mode and corresponding propagation distances and GI durations. [1] In urban areas, the reach of the signal is reduced compared to a rural area, due to a larger value of the path loss exponent as a result of the urban morphology; the use of a CP with a duration of 16,7 µs duration is satisfactory. In contrast, the signals in a rural areas have a larger range due to the smaller path loss exponent, so the usage of the extended CP with a duration of 33,3 µs may be necessary. The corresponding propagation distance d p of m is sufficient, because at distances greater than m the path loss exponent is considerably increasing due to the earth curvature; hence, these enbs have only negligible influence on the UE. In the 33,3 µs extended mode, the sub-carrier spacing ( f) is halved to reduce the overhead, resulting of the doubled OFDM symbol duration in comparison to the extended OFDM symbol duration. The

9 Broadcast Operation 9 Figure 6: System model of an SFN [1] 7,5 khz sub-carrier spacing mode results in a higher sensitivity to Doppler spread and shift, due to the smaller distance between the sub-carriers. Generally, in large MBSFNs the appearance of non-negligble channel paths with large delays is likely as a result of shadowing [1]. The spacing of the Reference Symbols (RSs) is modified for MBSFN transmission, compared to the non-mbsfn transmission. Fig. 9 shows the cell-specific RS for the case of the normal CP length and the RS for the MBSFN mode. The Root Mean Square (r.m.s.) delay spread σ τ is a parameter which characterizes the wide-band multipath channel. The delay spread can be derived from the PDP. The coherence bandwidth is a statistical measure of the range of frequencies over which the channel can be considered flat and is approximately B C 1 50σ τ for a frequency correlation above 0.9. With this approximation, coherence bandwidth and r.m.s. delay spread are inversely proportional. Thus, an increasing delay spread causes a decreasing coherence bandwidth. MBSFN channels exhibit a larger delay spread due to long propagation delays from possibly distant enbs. Therefore, the coherence bandwidth of the channel is smaller than e.g. for the unicast case, i.e. the channel is more frequency-selective. So, a closer spacing of the RSs in the frequency domain is necessary to improve the channel estimation at the UE which receive the MBMS service. The MBSFN RS patterns use every other sub-carrier for a pilot instead of every sixth as in the non-mbsfn mode [4].

10 10 Christopher Schmidt Figure 7: Performance of an MBSFN network illustrated by means of coverage vs. spectral efficiency for various MBSFN area sizes [1]. The enbs belonging to the SFN have to be tightly time-synchronised. This requires a synchronisation with accuracy considerably smaller than T G, so the adoption of satellite-based services like GPS is necessary for synchronisation [1] [5]. 3 Broadcast Operation 3.1 Broadcast Modes The transmission of data to multiple users can be realised by three different transmission modes, which vary e.g. in exploitation of feedback, scheduling of UEs, and spectral efficiency. Unicast provides a bidirectional point-to-point transmission, so the same data has to be transmitted to each user individually. Hence, the network burden is increasing with the number of transmitters, due to multiple transmissions of the same data. So, the overall spectral efficiency is very poor when a broadcast to many UEs is realised by unicasting, especially for high-bandwidth multimedia applications. Due to the bidirectional communication, there is a return channel for Channel Quality Information (CQI), which is needed for link adaptation. Broadcast is a downlink-only transmission mode which uses point-to-multipoint connections. The unidirectional communication comes with the absence of a return channel. Therefore, the enb does not know anything about the individual channel conditions; accordingly, there is no individual link adaptation for the broadcast mode. The data is only transmitted once for all

11 Broadcast Operation 11 Figure 8: Standardized LTE slots and associated CP lengths [1]. receivers in a geographical area, so the resource consumption is to a large extent independent of the number of users but relatively large, since it has to be ensured that a large percentage of all UEs can successfully decode the received data packets. Multicast is a special case of a broadcast. It is also a downlink-only point-to-multipoint transmission, but the content can only be received by a managed group of subscribers in the geographical area. Multicast is, on the one hand, the most efficient mode to transmit the same information to a limited number of users without using an individual connection to each user, but there is also no link adaptation due to the missing return channel on the other hand. In unicast communication, link adaptation due to the received CQI enables higher reliability or higher throughput, but the number of resources which have to be scheduled for each unicast user increases linearly with the number of recipients, so there is a break even point, where the MBSFN throughput exceeds unicast throughput [1]. Unfortunately, MBSFN transmission in LTE Release 8 does not allow for spatial multiplexing, in contrast to the unicast transmission mode. This means that unicasting can potentially realize much higher per user throughputs than MBSFN broadcast transmission in general. The Multimedia Broadcast Multicast Service (MBMS) in LTE is based on the UMTS Release 6 MBMS Network Architecture, which is shown in Fig. 10. MBMS, which was specified for the first time in 3GPP (3rd Generation Partnership Project) UMTS Release 6, provides the options unicast and broadcast/multicast to transmit data from a single point to multiple users by using the core network and the UMTS Terrestrial Radio Access Network (UTRAN).

12 12 Christopher Schmidt Figure 9: Reference symbols for non-mbsfn and MBSFN mode [1] Figure 10: UMTS Release 6 unicast and broadcast/multicast modes The network architecture can be divided into three parts: content provision, core network, and radio access network. The content providers are normally external to the core network, e.g. television broadcasters. The Broadcast-Multicast Service Center (BM-SC) is the interface between the content providers and the core network, which receives the content, schedules the services and manages the group membership. The Gateway GPRS Support Node (GGSN) and the Serving GPRS Support Node (SGSN) are the entry points to the core network respectively Radio Access Network (RAN). Finally, the RAN transmits the MBMS data in the most efficient way; dependent on the number of receivers within a cell, a decision between unicast and broadcast/multicast is made. UMTS Release 6 uses different scrambling codes in each cell for CDMA, also for multi-cell broadcast transmission which prevents SFN-like broadcasting. Release 6 defines a single-cell transmission with different scrambling codes for different cells. The successor, Release 7, provides the opportunity to realise multi-cell SFN transmission by using the same scrambling codes in all participating cells [3].

13 Broadcast Operation MBMS Network Architecture The architecture of the MBMS network in LTE is based on the UMTS Release 6 MBMS network architecture (cf. Fig. 10). Fig. 11 illustrates the design of the LTE network, including SFN transmission. Obviously, the basic structure of an LTE network is identical to that of the UMTS network. The main difference lies in the air interface designs, which is based on OFDM for LTE and CDMA for UMTS, respectively. Figure 11: LTE MBMS network Architecture [11] LTE enables the simultaneous deployment of unicast and multicast/broadcast communication. Therefore, different areas are defined within an MBMS service area, which can be seen in Fig. 12. The MBMS service area is a geographical area, where MBMS can be applied. Within this area, there is the MBSFN synchronization area, where all enbs can be time-synchronised and therefore can offer MBSFN services. Inside the MBSFN synchronization area, there are different MBSFN areas, in which MBSFN transmission is enabled (cf. Fig. 12, Groups 2, 3, 4 and 5). The overlapping of different MBSFN areas (cf. Fig. 12, Group 5) is a special case of the MBSFN and necessitates a more complex configuration for the allocation of separate resources and signalling to support the different MBMS services which each involved area provides. There can also be cells inside the MBMS service area which do not belong to the MBSFN synchronization area. These cells transmit the data by unicast communication, so there is no need to time-synchronise the relevant enbs (Fig. 12, Group 1). LTE provides two kinds of carriers: a dedicated and a mixed carrier. The dedicated carrier

14 14 Christopher Schmidt Figure 12: MBMS services areas in LTE [1] uses all subframes for MBSFN transmission, whereas the mixed carrier shares the subframes between MBSFN and unicast communication. In every 10 ms radio frame, the subframes 0, 4 and 5 are used for the unicast transmission and therefore they are not allowed to be used for MBSFN transmission. LTE is defined as a mobile standard, which is based on a cellular network. Accordingly, procedures are required, which provide a handover for the UE which is connected to an enb when it crosses a cell border. Several scenarios exist in LTE, involving intra-frequency or inter-frequency handover. Leaving an MBSFN area implies a change from a dedicated carrier to a mixed carrier; this can be complex and requires special procedures. In contrast, basic handovers which do not require the change of the carrier follow the traditional handover procedures. Further details on MBSFN-related handovers can be found in [1]. 3.3 Comparison of Mobile Broadcast Modes LTE provides MBMS for an efficient transmission of multimedia data to multiple users, but there are also other options apart from cellular networks. The content delivery by a cellular network has the advantage of reuse of already existing infrastructure and therefore no allocation of additional spectrum. This reduces the bandwidth available for other mobile services, but the cellular structure also offers a return channel for CQI or interactive services. This is an improvement compared to a pure broadcast system, where the enb has no information about the UEs. Broadcast content can be delivered by standalone broadcasting networks, e.g. DVB-H, which cover a wide geographical area with a relatively small number of high-power transmitters. The drawback is the missing CQI due to a missing return channel, but the broadcasting systems reach a relatively high data rate in a wide area. An obvious solution is to split the delivery of the content to a broadcast and a multicast transmission. For mobile television, the main channels with a high usage could be transmitted via broadcast and the niche channels, which are only demanded by a smaller number of users,

15 Broadcast Operation 15 are transmitted via multicast or unicast (cf. Fig. 13). Hence, the resource consumption would be optimised by combining unicast, multicast and broadcast transmission [1]. Figure 13: The "Long Tail of Content" [1] 4 Outlook and Conclusion 4.1 Outlook MBMS in LTE is not specified in Release 8, because other aspects of the LTE standard had higher priority for the commercial deployment; thus, not all innovations were specified in the first LTE release. However, the essential components for deployment of MBMS are already specified in Release 8 to ensure forward-compatibility. Furthermore, Multiple Input Multiple Output (MIMO) MBSFN transmission is unlikely to be feasible in following releases. 4.2 Conclusion The initialization of considerably higher data rates enables high-bandwidth multimedia applications via MBSFN. Due to MBSFN transmission, the r.m.s. delay spread increases, so the

16 16 Christopher Schmidt coherence bandwidth is reduced which requires a higher reference symbol density in frequency direction. Additionally, LTE provides different CP modes for the OFDM air interface to respond to the signal propagation effects. LTE combines unicasting with broadcast/multicast on different carriers and provides handover procedures for mobile users. Therefore, LTE provides a higher cell edge throughput for MBMS services in comparison to UMTS. References [1] S. Sesia, I. Toufik and M. Baker: LTE The UMTS Long Term Evolution, Wiley [2] M. Konrad, W. Gerstacker and W. Koch: Robust MBSFN Transmission Using the Golden Code, IEEE Workshop on Mobile Computing and Networking Technologies (WMCNT), St. Petersburg, [3] 3GPP TS V7.7.0, Introduction of the Multimedia Broadcast Multicast Service (MBMS) in the Radio Access Network, Technical Specification Group Radio Access Network, [4] 3GPP TS V8.4.0, Evolved Universal Terrestrial Radio Access (E UTRA); Physical Channels and Modulation, Technical Specification Group Radio Access Network, [5] T. Rappaport: Wireless Communications: Principles and Practice, 2nd Edition, Prentice Hall, [6] W. Koch: Lecture Notes: Fundamentals of Mobile Communication, Lecture in WS 2009/2010. [7] A. Heuberger: Lecture Notes: Digital Broadcasting Systems, Lecture in SS [8] S. Zarai: OFDM & Downlink Physical Layer Design, Seminar Ausgewählte Kapitel der Nachrichtentechnik: LTE Der Mobilfunk der Zukunft, [9] [10] [11] G. Fettweis: LTE Netzwerk und Systemaspekte für die die digitale Rundfunkübertragung,