Introduction to Multimedia Communications 231 of video frames that are likely to be corrupted by packet losses or the number of speech segments that will be delivered past playback time. The reverse mapping is also of great interest for QoS negotiation purposes. A user needs to be able to map his/her performance requirements into equivalent metrics at the layer immediately below the application layer in order to determine what QoS guarantees to request from that layer. The process repeats across corresponding layers offering QoS guarantees. When packets are transmitted across heterogeneous networks, QoS guarantees may be achieved separately for each subnetwork. Therefore, concatenation must be performed to translate between point-to-point and end-to-end guarantees. Point-to-point delay within each network is additive, to form end-to-end delay distribution. To calculate the effective bandwidth available on the end-to-end connection, the minimum bandwidth available in each domain that forms the connection is taken. Different subnetworks making up the end-to-end path might offer distinctly-defined QoS parameters. In addition, several issues will impact the process of mapping and concatenation, such as [8]: . Segmentation, fragmentation and reassembly: packets often get segmented into smaller, lower-layer data units such as data link layer units. If QoS guarantees are given for these lower-layer data units, how do they map into performance guarantees at the higher layer? In addition, packets may get fragmented and reassembled multiple times as they pass through possibly heterogeneous subnetworks with different maximum transfer units (MTUs). . Absolute and relative guarantees: some protocol layers or subnetworks may offer absolute (or quantitative) service differentiation; others may offer relative (or qualitative) guarantees. For example, in a priority-based policy, QoS is relative. Some users are assured of receiving preferential treatment over others. However, there is not necessarily any minimum quality level to be guaranteed. Other QoS mechanisms offer absolute QoS parameters. For example, certain ATM service classes offer strict assurances of maximum loss rates and delay that are independent of other traffic currently traversing the network. Characterizing the QoS delivered to the user may need to combine the effect of both quantitative and qualitative mechanisms. . Flow aggregation: in some QoS architectures, streams with similar performance require- ments are aggregated into a single flow for scalability purposes; QoS guarantees then apply to the aggregate flow. QoS translation is denoted by the process of translating QoS requirements such as delay, packet loss, and throughput into specific resources such as buffers, bandwidth, and so on that need to be allocated to meet those requirements [9]. 7.2.2 Constraints on Wireless Multimedia Communications Reliable transmission of multimedia, especially video, over wireless links is becoming an increasingly important application requirement in mobile communications. However, support- ing robust multimedia communications over wireless networks is a significant problem, primarily because of three factors: . Scarcity of bandwidth. . Time-varying error characteristics of the transmission channel. . Power limitations of wireless devices.
232 Visual Media Coding and Transmission While the above constraints are present in many communications systems, the challenges they impose are particularly acute for video communications. Voice applications are delay sensitive by definition; however, voice traffic requires low-channel bandwidth in the 5 24 kbps range. On the other hand, data applications require high bit rates but tolerate high transmission latency, leaving the processor freedom to retransmit the packet if necessary, in order to achieve adequate quality. In contrast to the above, low-bit rate video coding algorithms operate at rates ranging from tens to hundreds of kbps. Moreover, video applications are delay sensitive and it may not be easy to use retransmission as much as in data applications. Unlike wire-line networks, where an increase in the number of users can be met by adding more fiber, cable, or other similar wired media, the system capacity in wireless networks cannot be increased arbitrarily. Thus, due to the large quantity of data involved in video commu- nications, compression is almost always used in the transmission and management of digital video over wireless networks. Video compression technologies typically apply predictive coding and variable-length coding to achieve a higher degree of compactness. However, this has the undesirable side effect of increasing the susceptibility of the media streams to errors. Predictive coding inevitably results in frame-to-frame error propagation. Variable-length decoding, on the other hand, tends to lose code word synchronization in the presence of bit errors. This causes a large number of following symbols to be incorrectly decoded, regardless of their correct reception. The problems arising from scarcity of bandwidth and the use of highly error-susceptible video coding techniques are aggravated by the fact that radio channels are highly unpredict- able. This unreliability results from distance propagation phenomena such as multipath fading, shadowing, path loss, noise, and interference from other users, all of which have a multiplica- tive effect on the transmitted signal, causing it to deteriorate. Multipath propagation, caused by the superposition of radio waves reflected from surround- ing objects, gives rise to frequency-selective fading, resulting in rapid fluctuations of the phase and amplitude of the signal. Shadowing, caused by the presence of large physical objects which preclude a direct line of sight between the radio transmitter and the receiver, is a medium-scale effect. Path loss causes the received power to vary gradually due to signal attenuation, determined by the geometry of the path profile in its entirety. The combined effect of these phenomena is that the receiver has to deal with a bitstream that is corrupted by both random bit errors and burst errors. Unless the video encoder and decoder take proper action to deal with a bitstream that is corrupted by random bit and burst errors, the video communication system may totally break down. Video coders are currently designed independently of channel coders. They do not take into account the joint source channel coding adaptation and optimization possible for resource allocation in a time-varying channel environment. When independently-encoded video is transmitted over a time-varying channel such as a typical wireless link, poor video perfor- mances result. As shown in [10], joint source and channel coding can be implemented to optimize the performances of video communication over error-prone environments. At a higher layer, error-protection schemes that use sophisticated convolutional or block codes may be employed to alleviate the problem associated with channel errors, but they aggravate bandwidth problems since more bits have to be added to the video bitstream.
Introduction to Multimedia Communications 233 Similarly, automatic repeat request-type retransmission procedures improve error recovery against burst errors, but aggravate latency problems. Equalization and interleaving techniques that reduce channel interference and the effect of burst errors can also be used to alleviate some of these problems [11]. However, performance improvements come at the expense of reduced system efficiency, increased system cost, and increased processing delay. Evaluations of current cellular communications standards are biased towards integrated packet voice and data communications, even though digital video communications were considered desirable. The characteristics of digital video were not accommodated in the development of these standards. For example, the selection procedure for the choice of radio technologies of UMTS specifies only speech and data traffic models to be used in the testing procedures, but no video traffic model is given [12]. This provides inadequate systems for transmission of compressed video over wireless channels in terms of resource-allocation algorithms, traffic-scheduling mechanisms, and channel-access protocols. In second-generation cellular standards, the radio frame length is set to 20 ms, which is equivalent to the voice codec packet-generation period. The connection is also guaranteed one time slot per radio frame for the duration of a talk spurt in time division multiplex access (TDMA)-based communication systems. Therefore, the resource allocation and quality management can be optimized for voice services. However, resource allocation based on 20 ms frame length or time slots has no meaning for video communication, as video data arrival intervals depend on the video frame rate setting of the encoder. In addition, the data within a video frame is larger than a radio frame, and the video frame size varies from frame to frame due to the variable-bit rate nature of video codecs. Thus, multiple-slot resource allocation is necessary, and it should be designed carefully to guarantee the quality requirement while maximizing the system utilization. Current cellular standards use separate protocol-layer architectures to reduce implementa- tion complexity. Each protocol layer is individually optimized for the defined functions and the protocol-layer interaction is conducted through specified interfaces. The data within each protocol layer is transparent to the others. Therefore, layered protocol architecture may not provide optimal processing for the application data and may add some redundant information at the lower layers. This reduces the amount of available radio resources, making cross-layer optimization important in multimedia communications. Currently-available compressed video, such as MPEG-4-coded video streams, can be separated into multiple streams, with different QoS requirements according to their importance to the perceptual quality of the decoded video. Most of the popular channel access protocols are unable to guarantee these multiple QoS requirements, resulting in poor system performances and sometimes intolerable video quality. These multiple quality requirements should be incorporated into the channel access protocol design. Video is characterized by high data rates and low delay tolerance, and it has different system requirements to voice and data services. Thus, video cannot be treated similarly to voice and cannot be given a high priority in the network resource allocation. On the other hand, it is not appropriate for video to be treated as data, as this would lead to unacceptable delays. Therefore, video has to be treated as a separate entity with its own set of requirements in network resource-allocation schemes. While it is true that it is impossible to eliminate all the problems stated above, the work described here has made some attempt to achieve improved- quality video transmission over third-generation wireless communication systems.
234 Visual Media Coding and Transmission 7.2.3 Multimedia Compression Technologies 7.2.3.1 Video Compression Technologies Among many video codec standards, only MPEG-4 and H.263 are recommended by 3GPP [13] for use in video communications over third-generation mobile communication systems. Both standards share a common basic approach, which uses hybrid motion compensation-based DCT in compression. A number of different techniques introduced in the above standards are used to enable robust transmission of compressed video data over noisy communication channels. Most of the error-resilience features introduced in H.263 are described in annexes, but they are not always implemented. Due to its enhanced error- resilience capabilities, included as core features, MPEG-4-encoded sequences are used in the experiments carried out here. Thus, no further discussion of H.263 is given. Interested readers can refer to [14,15]. MPEG-4 was originally aimed at low-bit rate video communication. Later, attention was focused on producing a multimedia coding standard, providing a variety of new features. MPEG-4 follows an object-based compression methodology, where a scene is considered to consist of a number of audio-visual objects (AVOs). These AVOs may represent a complete video frame, an object present in a video scene, a computer-generated graphic, text, an image, an audio stream, or speech. AVOs can be encoded and transmitted as separate entities or in a combined form according to the requirement specified by the applications. For example, in an interactive application session, the user may require object modifications, thus separate encoding and transmission of AVOs are necessary. A system layer, which is incorporated into the standard, describes the way these AVOs are synchronized, multiplexed, and presented at the receiver, providing appropriate QoS for various applications. The delivery layer is further divided into a delivery multimedia integra- tion framework (DMIF) layer and a TransMux layer. The synchronization layer specifies the synchronized delivery of streams from the source to the destination, while exploiting the different QoSs offered by the network. Media synchronization is guaranteed by the insertion of a time stamp into elementary streams. The syntax of the synchronization layer can be reconfigured according to the application requirement, allowing its operation over a wide range of services and systems. The DMIF layer manages the efficient delivery of multimedia over the selected network. In particular, stream grouping based on quality requirements, classification, and multiplexing are performed at the DMIF layer using the MPEG-4-defined FlexMux tool. Low multiplexing overhead is achieved in these processes. Interfaces to a set of transport protocols are specified within the TransMux layer [16]. The choice of appropriate transport protocol is left to the end users or the service providers. The RTP/UDP/IP interface is investigated for the transmission of video over the packet-switched connection in the work described here. In addition to the features capable of supporting delivery of multimedia efficiently and flexibly over a variety of transport networks, many techniques have been incorporated into MPEG-4 in order to make the coded video stream more resilient to channel degradation when operating in error-prone environments [17]. These techniques can be categorized as encoder error-resilience tools and decoder error-concealment tools, considering their implementation. There are four main error-resilience tools introduced in the MPEG-4 encoder: video packet resynchronization, data partitioning, reversible variable length codes (RVLC), and header extension codes [18].
Introduction to Multimedia Communications 235 7.2.3.2 Speech Compression Technologies The adaptive multi-rate wideband (AMR-WB) codec was selected as a harmonized wideband codec for GSM, 3G WCDMA, and ITU-T in 2001Rapporteur’s meeting, and was approved by the ITU in January 2002 and named ITU-T G.722.2. The acceptance of a single harmonized speech code allows easy implementation of wideband voice applications and services across a wide range of communication systems and platforms without the use of transcoding between wireless and wired infrastructure. The AMR-WB codec is based on the algebraic code-excited linear prediction (ACELP) technology. ACELP has been successfully used in a wide range of speech-compression standards, such as 3GPP AMR, ETSI EFR, NA-TDMA IS-641, NA-CDMA-IS-127, ITU-T G.729, and ITU-T G.723.1 codecs. However, these were developed for narrow-band signals. AMR-WB includes much functionality to make the speech signal robust to wideband channel errors. A detailed overview of the AMR-WB codec can be found in [19]. The AMR-WB codec has been developed for use in several applications, including GSM full-rate channel, GERAN, UTRAN, and voice-over-IP applications. A wide range of applications is envisioned for the AMR-WB codec, which includes ISDN wideband telephony, audiovisual teleconferencing, voice-over-IP, IP video conferencing, voice mail, voice chat, broadcast, and voice streaming. The AMR-WB codec includes a set of fixed-rate speech and channel codec modes, a voice activity detector, discontinuous transmission functionality in GSM, UMTS and source-controlled rate functionality in 3G, in-band signaling for codec mode transmission, and link adaptation to control the mode selection. The AMR-WB codec adapts the bit-rate allocation between speech and channel coding, optimizing speech quality to prevailing radio-channel conditions; AMR-WB is also very robust against transmission errors due to the multi-rate operation and adaptation. 7.2.4 Multimedia Transmission Issues in Wireless Networks In addition to the encoder error-resilience and decoder error-concealment, there are other techniques designed to improve the received quality of transmitted multimedia over band- width-limited error-prone environments. This section examines some existing efficient transmission methodologies proposed by other researches. The review is split into five areas: . Joint source channel coding. . Advanced forward error correction. . Feedback-based techniques. . Multiplexing. . Link adaptation. 7.2.4.1 Joint Source–Channel Coding In order to reduce the effect of channel noise on the transmitted bitstream, some form of forward error correction (FEC) is applied to information transmission over error-prone channels. However, the appropriate level of FEC coding generally depends upon many parameters, such as the type of media, the available bandwidth, and the channel bit error rate. Especially for video applications, the quality distortion seen in a decoded video sequence is determined by the distortion due to channel errors, as well as the quantization distortion
236 Visual Media Coding and Transmission at the source encoder. Therefore, a combined source channel coding approach is necessary in optimal allocation of rates between source and channel protection, subject to a fixed constraint on overall transmission bandwidth. Much research has been conducted in joint source channel coding for image and video applications. Studies reveal that lower source rate and extra channel protection are favorable for image and video transmission over low-quality channels [20]. The algorithms which determine the optimal source channel coding ratio for a particular source-coding algorithm, channel-coding scheme, and set of channel characteristics have been proposed in [10,21,22]. The communi- cation channel states can often be modeled as a hidden Markov process. Based on this statistical framework, the design of an optimal joint source channel coding is discussed in [23]. The method proposed in [23] uses state estimation and a minimum mean-square-error estimation procedure in the determination of optimal source channel code ratio. The theoretical perfor- mance bounds in distortion-rate characteristics based optimal source channel bit allocation have been derived in [10]. 7.2.4.2 Advanced Forward Error Correction Advanced forward error correction methods aim to reduce the effect of channel error on video quality by minimizing the amount of important information loss during the transmission. The output bitstream from the video encoder is modified or restructured in such a way as to achieve maximum perceptual quality at the receiver. The techniques used can be categorized into three main areas: . Unequal error protection. . Unequal power allocation. . Multiple descriptive transmission. Unequal error protection (UEP) basically uses prioritized transmission. The encoder output bitstream is separated into several substreams, divided according to their importance in perceiving video quality. Then high-priority streams are transmitted using highly-protected channels, while low channel protection is used for low-priority streams. A common way of splitting encoded data into separate streams is the MPEG-4 data partition method [24]. This method is simple to implement, yet provides a significant improvement in received video quality. Another obvious way of separating video information into multiple streams is scalable or layered video coding. A layered video encoder generally generates a number of bitstreams. The most critical information is contained in the base layer. Other layers contain information that is needed for enhancing the quality of the base layer. Another technique used in bitstream splitting is information separation in the transform domain [25]. Unequal power-allocation (UPA) techniques for enhancing video quality have recently been proposed, especially for video communications over CDMA-based cellular networks. The operation of unequal power-allocation techniques is similar to that of the unequal error- protection techniques described above. However, network compatibility is essential for practical implementation. In addition to the constraints imposed by the limited bandwidth or channel throughput, the transmit power also provides a limiting factor in the UPA algorithm design. Several UPA techniques for wireless video applications have been proposed in [26 28]. All of
Introduction to Multimedia Communications 237 these techniques are optimized to achieve a target video quality while minimizing the transmission power. Multiple descriptive coding (MDC) is intended to achieve quality improvements by exploring the diversity of transmission links. Similar to layered coding, a multiple descriptive code generates multiple bitstreams. However, unlike in layered coding, these multiple streams have equal priority. The streams are transmitted over separate transmission channels with equal channel characteristics in terms of average channel quality. The transmitted streams can be decoded individually at the receiver. When they are combined, improved quality is achieved because of the transmission diversity gain [29]. For example, assume two streams are transmitted over radio channels with similar channel parameters. Even though both channels have similar average channel characteristics, the instantaneous channel qualities experienced are very different, due to the time, space, and frequency parameters. Therefore, appropriate stream combinations can result in better received quality. MDC is especially favorable in ad hoc networks. 7.2.4.3 Feedback-based Techniques In all of the above error-resilience techniques, the encoder and the decoder operate indepen- dently of each other. If a feedback channel can be set up, the decoder can inform the encoder about the instantaneous behavior of the transmission channel and any corrupted packets. Thus the encoder can adjust accordingly to suppress or minimize the effect of channel errors on the video stream. Feedback-based techniques can be described in three main areas: . Retransmission techniques. . Optimal encoder mode selection. . Error tracking. The automatic repeat request (ARQ) method is a powerful technique, commonly used in conventional data transmission to retransmit the error packets when bit errors are not allowed. However, retransmission control as used in data transmission creates unacceptable delay and is not suitable for real-time applications. Modifications of ARQ techniques so that they are suitable for real-time applications have been proposed in [30,31]. A common feature of these modifications is the use of hybrid ARQ, which is a combination of ARQ and other error-resilience techniques. The combination of FEC and ARQ for mobile video commu- nications has been studied in [30]. These schemes request that the encoder retransmit any incorrectly-received packets. Even with one bit error in a packet, the encoder is told to retransmit the entire packet. Therefore, algorithm efficiency is significantly low, especially under burst errors. Matoba [31,32] has proposed selective-repeat ARQ/FEC (SR-ARQ), which is a modifica- tion of the FEC/ARQ method. Using the properties of layered coding, and considering different QoS requirements of different data sections within a video stream, further modifications to SR- ARQ, called QoS-aware selective repeat ARQ (QSR-ARQ), are proposed by Wang in [33 35]. Results show that QSR-ARQ improves the data-link protocol performance, network through- put, and bandwidth efficiency. Other approaches based on combined adaptive QoS control, optimal mode selection, and delay-constrained hybrid ARQ can be found in [36].
238 Visual Media Coding and Transmission Another way of achieving performance improvements with the use of feedback channel information is the reference picture-selection method. If a reference frame is corrupted due to channel errors, the corruption propagates over subsequent video frames until the reference frame is refreshed by an intra-coded frame. The error propagation can be mitigated by reference to a correctly-received video frame for the encoding of future video frames. This concept is employed in the optimal reference picture-selection methods proposed in [37]. The algorithm takes into account the channel condition and the error-concealment method used by the decoder to optimize video coding mode selection in the compressed bitstream. However, the algorithm should be carefully designed to avoid performance degradation caused by parameter mismatch between the parameters used by the encoder, the parameters associated with the actual channel condition, and the decoder error-concealment methods. Error-tracking methods also utilize a feedback channel to adapt the system to varying channel conditions. In this method, an error is localized to an image region. Decoder error concealment is employed to make transmission errors less visible, and unacceptable residual errors are compensated for by coding the distorted regions in intra-mode. Negative acknowl- edgments (NACK) are sent back to the transmitter for any image part that could not be decoded successfully. The encoder evaluates the NACKs and makes the intra-mode decision on a macro- block basis. The reconstruction of spatio-temporal error propagation at the encoder permits compensation for errors that have propagated [30,38,39]. 7.2.4.4 Data Stream Multiplexing Since video traffic is delay sensitive, a resource-reservation scheme seems to be the right choice for guaranteeing quality requirements for real-time video communications. However, due to the unpredictably bursty nature of variable bit rate (VBR) video, the resource reservation task is complicated. If resources are reserved according to peak rates, the network is under-utilized most of the time. VBR not only causes low transmission efficiency, but also causes difficulties for QoS provisioning in multi-user communication systems. Many research activities concentrate on minimizing the effect of VBR on resource allocation. These are based on different multiplexing schemes and can be categorized into three main classes [40]. The bursty traffic can be smoothed to become near-constant bit rate traffic by using a buffer for each data stream. This method is called temporal statistical multiplexing and is adopted in some real-time MPEG encoders. Although this is very easy to implement, it can cause delay variation, leading to inconsistent video quality. A second method is called pre-fetching. It allows users to get data before the playback start time. This can be used for non-real-time streaming applications, utilizing the high link capacity at non-peak times. Spatial statistical multiplexing, the third method, allows sharing of bandwidth between several streams to achieve an aggregated constant bit rate like throughput. In this way, the streams running at peak rate borrow the bandwidth from streams running at low rates. Delay variation is not observed with this method, but the arrangement of traffic is critical and difficult. A scene-based multiplexing method proposed in [40] belongs to the spatial statistical multiplexing methods described above. Several video bitstreams are multiplexed into a single bitstream, resulting in better bandwidth utilization without introducing delay varia- tion. Scene information is used to decide the bandwidth requirement for the aggregated video streams. The basic assumption in this system is that a significant video traffic bandwidth change could result from a visual scene change. Video trace synchronization is used to avoid
Introduction to Multimedia Communications 239 several I-frames being transmitted simultaneously. Simulation results show that scene-based multiplexing achieves improved bandwidth utilization when transmitting MPEG-1 and MPEG-2 video streams. Another form of spatial statistical multiplexing is described in [41], namely inter-frame statistical multiplexing. This method uses region segmentation or the MPEG-4 video object concept to segment the video frame into regions or video objects of differing importance. The peak bit rate for the most important region is reserved at connection-establishment time. It is likely that most of the time the compressor will produce bits far below this peak value. The bandwidth left over after the important region has been transmitted is used to transmit the remaining regions, and to retransmit them if required. Priority is given to the lowest- frequency sub-band of the main regions, followed by the lowest-frequency sub-band of the remaining regions, followed by the subsequent higher-frequency sub-bands. Any bits left over after the reserved bandwidth is used up are transmitted using any available unreserved bandwidth. Experimental results illustrate the efficient bandwidth allocation. As the most important regions always reach the receiver, transmitted image frames can be displayed in any case. Object prioritization based on the statistical multiplexing method is described in [42]. Using the object-oriented features of MPEG-4, the video frame is divided into separate objects. These objects are transmitted independently, assigning higher priorities to the most important objects, and are multiplexed at the receiver to get the combined frame. The use of object-based prioritization schemes has been shown to give a noticeable perceptual quality improvement in received video signals over bandwidth-limited and noisy channels. 7.2.4.5 Link Adaptation Link adaptation is employed to mitigate the effect of time-varying radio channel conditions on the received media quality. This involves the modification of source and network parameters in response to the instantaneous channel and interference conditions. The concept of source- and network-parameter adaptation for system throughput maximization has been considered for data communication in [43 47]. Ideally, the design of link-adaptation algorithms for multimedia services should optimize the delivery of acceptable video quality, facilitating good-quality video services with the minimum possible demand on network resources. Application of link-adaptation techniques for video communication has been introduced by Hanzo et al. in their series of paper publica- tions [48 50]. In these schemes, adaptive modulation is considered. The appropriate modulation mode for the transmission is selected from a set of different modulation schemes according to the instantaneous quality of the transmission channel. Adaptive schemes have been shown to provide better received video quality than the non- adaptive schemes for video transmission over narrowband and wideband channels. 7.2.5 Resource Management Strategy in Wireless Multimedia Communications In multimedia communications, there are several source and system parameters to be taken into account in the provision and maintenance of adequate service quality to multiple users in the
240 Visual Media Coding and Transmission system. These are: . multimedia quality . source bit rate . error resilience . channel quality . channel protection . transmission power . processing power . mobility . system capacity . system load . coverage . total system transmit power . cell interference/cell planning . complexity. Radio resource management algorithms in wireless systems are designed to utilize the air- interface resources in such a way as to guarantee the required service quality while maintaining high capacity over the specified coverage area. They can adjust link-level parameters such as the spreading factor (SF) in UMTS, transmit power, number of time slots, and channel coding scheme for each radio bearer according to varying traffic, system interference, and channel conditions. Conventionally, these parameters are adapted based on the estimated received channel quality at the receiver. All these parameters are interrelated and provide contradictory constraints on system perfor- mances. In conventional (system-level) resource-management schemes, only system capacity, system load, coverage, transmission power, and interference are considered. Average received channel quality is used to indicate the QoS (multimedia quality) received by the user. However, for multimedia communications, the average received channel condition does not always provide an accurate performance figure for perceptual quality. Moreover, a combination of source bit rate, error resilience, channel protection, and transmission and processing power should be taken into account in optimizing the received multimedia quality in systems with limited resources. Figure 7.4 presents a graphical interpretation of the tradeoff problem between source and system parameters in allocating resources for multimedia communications. Figure 7.5(a) illustrates the relationship between distortions, source bit rate, and channel quality for video transmission over a fixed-bandwidth channel. Video distortion can be considered as a combination of quantization distortion and channel-induced distortion. Quantization distortion decreases with increases in source rate. The distortion due to the channel errors increases with increase of bit rate. This is because the amount of channel protection depends on the channel coding rate, which is decided by the available channel bandwidth and the source rate. A higher source rate requires a reduction in channel protection when operating over a fixed-bandwidth channel. The resulting total distortion is shown in Figure 7.6. The optimal source rate which gives minimum distortion varies according to the channel quality. The setting of optimal source and channel-coding rates for a fixed bandwidth channel is provided by joint source channel coding.
Introduction to Multimedia Communications 241 User mobility Interference System capacity Total Transmission Coverage power System load Channel quality Transmission power (a) Source rate Channel protection Error Processing resilience power Channel quality Multimedia quality Complexity (b) Figure 7.4 Resource tradeoffs in wireless multimedia communication: (a) conventional resource tradeoffs diagram; (b) multimedia quality management Mobile terminal power consumption can also be divided into two parts: processing power and transmission power. The transmission power is directly proportional to the source rate if the transmission bit energy is kept constant during transmission. Processing power is often ignored in resource-allocation schemes. However, processing power is significant in current and future multimedia source coding as much signal processing is involved in compression and resilience algorithms. Mobile computation and communication platforms must rely on
242 Visual Media Coding and Transmission Distortion Trans. & proc. Power Source rate Source rate Channel condition (a) Bit energy (b) (BER) Quality Error resilience BER Channel coding Channel SNR (d) Channel SNR Channel SNR (c) Modulation System load (e) & Interference Figure 7.5 Graphical interpretation of resource allocation in wireless multimedia communications: (a) rate distortion characteristics; (b) rate power characteristics; (c) characteristics of channel coding and modulation; (d) effects of error resilience; (e) channel quality vs. system load and interference battery power, and future mobile terminals should offer longer battery life in an attractive smaller terminal. Therefore, the processing power should be considered as a parameter in resource allocation, and the combination of processing and transmission power should be minimized for a given transmission [51]. Figure 7.7 shows the relationship between processing power, transmission power, and source rate. A three-dimensional illustration is given in Figure 7.5(b). Channel coding and modulation provide different degrees of channel protection and source throughput in a given channel environment. Variation of bit error rate characteristics with channel coding and modulation is shown in Figure 7.5(c). Figure 7.5(d) depicts the effects of
Introduction to Multimedia Communications 243 Distortion Total distortion Quantization Channel 1 distortion Channel 2 Channel distortion Source rate Figure 7.6 Rate distortion characteristics for video communication error resilience on multimedia quality, while influences of system load and interference on channel quality are shown in Figure 7.5(e). Resource allocation in multimedia communication should consider all these factors in achieving optimal received quality. The complexity of such a system is considerably high, thus a sub-optimal approach is often followed. Moreover, due to the dynamic nature of communi- cation systems, the system parameters fluctuate with time. Therefore, adaptive resource allocation must be implemented in order to maintain the required service quality at an acceptable level. These resource management issues may be directly and indirectly exploited for the transmission of video over third-generation wireless communication systems. Power Total Noise Processing power Transmission Source rate power Figure 7.7 Rate power characteristics
244 Visual Media Coding and Transmission 7.3 Conclusions Mobile and multimedia communication technologies have experienced rapid growth and commercial success during the last decade. Driven by the powerful vision of being able to communicate from anywhere, at any time, and with any type of data, the integration of multimedia and mobile technologies is currently underway. Third/fourth-generation commu- nication systems will support a wide range of communication services for mobile users from any geographical location, in a variety of formats such as voice, data, images, and video. Among those, video communication is particularly demanding, due to the stringent require- ments on QoS and the enormous amounts of data involved, for example in 3D video. This chapter has given an overview of the technologies and concepts involved in provision- ing multimedia communications over wireless access networks. The first half dealt with recent advances in supporting QoS in wireless networks and multimedia technologies. A description of the constraints involved in wireless multimedia communications was given, describing the scarcity of bandwidth, the error characteristics of wireless channels, and the lack of network support. The role of media compression technologies, in particular error-robust low-bit rate video codec design, was also discussed. A variety of techniques have been proposed for increasing the quality of the transmitted media stream over error-prone wireless links. These techniques are categorized into joint source channel coding, advanced forward error correc- tion, feedback-based techniques, stream multiplexing, and link adaptation. A discussion of each of these techniques and their use in multimedia communication was presented in Subsection 7.2.4. The second half of the chapter discussed the resource-management issues in wireless multimedia communications. The tradeoffs between source rate, channel protection, error resilience, transmission power, processing power, complexity, and perceptual quality were discussed in relation to the allocation of network resources for transmission of multimedia over given channel conditions. The effects of system parameters such as system capacity, coverage, interference, mobility, and system load on multimedia performances in a multi-user scenario should also be taken into account in system resource allocation. References [1] M. Naghshineh and A.S. Acampora, “QoS provisioning in micro cellular networks supporting multiple classes of traffic,” Wireless Networks, Vol. 2, No. 3, pp. 195 203, 1996. [2] P. Ahluwalia and U. Varshney, “A link and network layer approach to support mobile commerce transactions,” IEEE 58th Vehicular Technology Conference, Vol. 5, pp. 3410 3414, 2003. [3] A. Nagshineh and M. Acampora, “QoS provisioning in micro cellular networks supporting multiple classes of traffic,” ACM/Baltzer Journal on Wireless Networks, Vol. 2, pp. 195 203, 1996. [4] D.A. Levine, I.F. Akyildiz, and M. Naghshineh, “A resource estimation and call admission algorithm for wireless multimedia networks using the shadow cluster concept,” IEEE/ACM Transactions on Networking, Vol. 5, pp. 1 12, 1997. [5] S. Khan, S. Khan, S.A. Mahmud, and H. Al Raweshidy, “Supplementary interworking architecture for hybrid data networks (UMTS WiMAX),” Proc. Int. Multi Conf. on Computing in the Global Info. Technol. (ICCGI’06), pp. 57 61, 2006. [6] H. Holma and A. Toskala, WCDMA for UMTS Radio Access for Third Generation Mobile Communications, John Wiley & Sons, Ltd., 2000. [7] O. Abdul Hameed, S. Nasir, H. Karim, T. Masterton, and A.M. Kondoz, “Enhancing wireless video transmissions in virtual collaboration environments,” 16th IST Mobile and Wireless Communications Summit, 2007.
Introduction to Multimedia Communications 245 [8] L.A. DaSilva, “QoS mapping along the protocol stack: discussion and preliminary results,” Proc. ICC 2000, Vol. 2, pp. 713 717, Jun. 2000. [9] K. Kim and K. Nahrstedt, “QoS translation and admission control for MPEG video,” in A. Campbell and K. Narhrstedt (eds), Building QoS into Distributed Systems, Champan and Hall, Sep. 1997. [10] M. Bystrom and J.W. Modestino, “Combined source channel coding schemes for video transmission over an additive white Gaussian noise channel,” IEEE Journal on Selected Areas in Communications, Vol. 18, No. 6, pp. 880 890, Jun. 2000. [11] F. Freiha, K. Chandra, V. Mehta, and C. Thompson, “Performance of VBR video with equalization on wireless fading channels,” Future Wireless Communication System Globecom’99, 1999. [12] Universal Mobile Telecommunications System (UMTS), “Selection procedures for the choice of radio trans mission technologies of the UMTS (UMTS 30.03 version 3.2.0),” TR 101 112, v. 3.2.0., Apr. 1998. [13] 3GPP TS 26.235, “Packet switched conversational multimedia applications: default codecs,” v. 4.0.0., Apr. 2001. [14] ITU T, “H.263þþ , video coding for low bit rate communications,” ITU T Draft Recommendation, November 2000. [15] ITU T, “H.263þ , Video coding for low bit rate communications,” ITU T Draft Recommendation, H.263 Version 2, Sep. 1997. [16] R. Koenen,“Overview of the MPEG 4 standard,” Doc. N3747, La Baule, ISO/IEC JTC1/SC29/WG 11, MPEG, Oct. 2000. [17] R. Schafer, “MPEG 4: a multimedia compression standard for interactive applications and services,” IEE J. Electronics & Communications Eng, Dec. 1998. [18] R. Talluri, “Error resilient video coding in the ISO MPEG 4 standard,” IEEE Communications Magazine, pp. 112 119, Jun. 1998. [19] “AMR wideband speech codec: general description,” v. 6.0.0. [20] J.W. Modestino, D.G. Daut, and A.L. Vickers, “Combined source channel coding of images using the block cosine transform,” IEEE Transactions on Communications, Vol. 29, No. 9, pp. 1261 1273, Sep. 1981. [21] M. Bystrom and J.W. Modestino, “Combined source channel coding schemes for transmission of video over a slow fading rician channel,” IEEE Int. Conf. Image Proc., ICIP, Chicago, IL, Vol. 2, pp. 147 151, Oct. 1998. [22] R. Kurceren and J.W. Modestino, “A joint source channel coding approach to scalable delivery of digital video over ATM networks,” IEEE Int. Conf. Image Proc., ICIP, Vancouver, Canada, Vol. 1, pp. 371 374, Sep. 2000. [23] R.E. Dyck and D.J. Miller, “Transport of wireless video using separate, concatenated, and joint source channel coding,” Proc. IEEE, Vol. 87, No. 10, pp. 1734 1750, Oct. 1999. [24] M. Ghanbari and V. Seferidis, “Cell loss concealment in ATM video codecs,” IEEE Trans. Circuits Syst. Video Technolo., Vol. 3, pp 238 247. [25] H. Gharavi and S.M. Alamouti, “Multipriority video transmission for third generation wireless communication systems,” Proc. IEEE, Vol. 87, No. 10, pp. 1751 1763, Oct. 1999. [26] Y. Eisenberg, C.E. Luna, T.N. Pappas, R. Berry, and A.K. Katsaggelos, “Joint source coding and transmission power management for energy efficient wireless video communications,” IEEE Trans. Circuits and Systems for Video Technology, Vol. 12, No. 6, pp. 411 424, Jun. 2002. [27] I. M. Kim and H. M. Kim, “An optimum power management scheme for wireless video service in CDMA systems,” IEEE Trans. Wireless Comm., Vol. 2, No. 1, Jan. 2003. [28] S. Zhao, Z. Xiong, and X. Wang, “Joint error control and power allocation for video transmission over CDMA networks with multiuser detection,” IEEE Trans. Circuits and Systems for Video Technology, Vol. 12, No. 6, pp. 425 437, Jun. 2002. [29] Y.C. Lee, Y. Altunbasak, and R.M. Mersereau, “A coordinated multiple description scalar quantizer and error concealment algorithm for error resilient video streaming over lossy channels,” IEEE International Conference on Communications, ICC 2002, Vol. 1, Apr. May 2002. [30] H. Liu and M. El Zarki, “Performance of H.263 video transmission over wireless channels using Hybrid ARQ,” IEEE J. Select. Areas Commun., Vol. 15, No. 9, Dec. 1997. [31] N. Matoba, Y. Kondo, and T. Tanaka, “Performance of real time video transmission using error control over Rayleigh and Rician fading channels,” 9th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, Vol. 3, pp. 1061 1065, 1998. [32] N. Matoba, Y. Kondo, and T. Tanaka, “Low delay, error resilient error control for mobile video communication,” GLOBECOM ’97, IEEE, Vol. 2, pp. 1032 1036, 1997. [33] S. Wang, H. Zheng, and J.A. Copeland, “An error control design for multimedia wireless networks,” VTC 2000 Spring Tokyo 2, IEEE 51st, Vol. 2, pp. 795 799, 2000.
246 Visual Media Coding and Transmission [34] S. Wang, H. Zheng, and J.A. Copeland, “A QoS enhanced hybrid SR ARQ for mobile video communications,” IEEE International Conference on Communications, ICC 2000, Vol. 1, pp. 526 530, 2000. [35] H. Zheng, S. Wang, and J.A. Copeland, “QoS aware mobile video communications,” Proc. Military Commu nications Conference Proceedings, MILCOM 1999, Vol. 2, pp. 1231 1235, 1999. [36] D. Wu, Y.T. Hou, Y. Q. Zhang, W. Zhu, and H.J. Chao, “Adaptive QOS control for MPEG 4 video communication over wireless channels,” ISCAS 2000, Geneva, Vol. 1, pp. 48 51, 2000. [37] G. Cote, S. Shirani, and F. Kossentini, “Optimal mode selection and synchronization for robust video communications over error prone networks,” IEEE Journal on Selected Areas in Communications, Vol. 18, No. 6, pp. 952 964, Jun. 2000. [38] B.S. Choi, S.I. Chae, and J.B. Ra, “A feedback channel based error compensation method for mobile video communications using H.253 standard,” Proc. International Conference on Image Processing, Vol. 2, pp. 555 559, 1999. [39] E. Steinbach, N. Farber, and B. Griod, “Standard compatible extension of H.263 for robust video transmission in mobile environments,” IEEE Trans. Circuits and Systems, Video Techno, Vol. 4, No. 6, Dec. 1997. [40] S. Wang, H. Zheng, and J.A. Copeland, “Video multiplexing with QoS constraints,” IEEE SPIE Conf. on Internet Routing and Quality of Service, Oct. 1998. [41] P. Bahl, “Supporting digital video in a managed wireless network,” IEEE Communications Magazine, Vol. 36, No. 6, pp. 94 102, Jun. 1998. [42] A. Cellatoglu, S. Fabri, S. Worrall, and A.M. Kondoz, “Use of prioritised object oriented video coding for the provision of multiparty video communications in error prone environments,” IEEE Vehicular Technology Conference, Amsterdam, The Netherlands, Vol. 1 pp. 386 390, Oct. 1999. [43] S. Nanda, K. Balachandran, and S. Kumar, “Adaptation techniques in wireless packet data services,” IEEE Comms. Mag., pp. 54 64, Jan. 2000. [44] K. Balachandran, S.R. Kadaba, and S. Nanda, “Channel quality estimation and rate adaptation for cellular mobile radio,” IEEE J. Select. Areas Comms., Vol. 17, pp. 1244 1256, Jul. 1999. [45] N.B. Mehta and A.J. Goldsmith, “Performance analysis of link adaptation in wireless data networks,” Proc. IEEE GLOBECOM ‘00, pp. 1422 1426, 2000. [46] P. Gutierrez, J. Wigard, P. Andersen, H. Damgaard, and P. Mogensen, “Performance of link adaptation in GPRS networks,” IEEE Proc. VTC 2000, Vol. 2, pp. 492 499, Fall 2000. [47] L. Zhao, J.W. Mark, and Y.C. Yoon, “A combined link adaptation and incremental redundancy protocol for enhanced data transmission,” GLOBECOM ’01, Vol. 2, pp. 1277 1281, 2001. [48] L. Hanzo, C.H. Wong, and P. Cherriman, “Channel adaptive wideband wireless video telephony: design tradeoffs and performance comparisons,” IEEE Signal Processing Magazine, pp. 10 30, Jul. 2000. [49] L. Hanzo, P. Cherriman, and E. L. Kuan, “Interactive cellular and cordless video telephony: state of the art system design principles and expected performance,” Proc. IEEE, Vol. 88, No. 9, pp. 1388 1413, Sep. 2000. [50] P. Cherriman, C.H. Wong, and L. Hanzo, “Turbo and BCH coded wide band burst by burst adaptive H.263 assisted wireless video telephony,” IEEE Trans. Circuits, Sys. Video Tech., Vol. 10, pp. 1355 1363, Dec. 2000. [51] T. Lan and H.T. Ahmed, “A resource management strategy in wireless multimedia communications: total power saving in mobile terminals with a guaranteed QoS,” IEEE Transactions on Multimedia, Vol. 5, No. 2, pp. 267 281, Jun. 2003.
8 Wireless Channel Models 8.1 Introduction In order to investigate the performance of novel audiovisual services and applications over future heterogeneous systems, an appropriate system simulator model should be designed and implemented. The main objective of this chapter is to present some possible such simulator designs for future communication systems. GPRS/EGPRS, WCDMA UMTS, and WiMAX IEEE 802.16e wireless communication technologies have been selected as the major com- munication systems for performance investigation of future audiovisual applications. 8.2 GPRS/EGPRS Channel Simulator The simulator model described in this section characterizes GPRS and EGPRS radio access network functionality. The main issues tackled are data flow across the radio access protocol stack and the effects of the physical link layer, which include forward error correction, modulation, transmission over fading channels, equalization and reception, and detection of correctable and uncorrectable errors. Power control mechanisms are not implemented in this model. 8.2.1 GSM/EDGE Radio Access Network (GERAN) The GERAN system architecture is shown in Figure 8.1. GERAN is connected to the core network through the Iu and A interfaces. Services based on second-generation (2G) systems are supported by the 2G SGSN (Serving GPRS Support Node) and the 2G MSC (Mobile Switching Centre). The 3G SGSN and 3G MSC take responsibility for the support of third-generation mobile services. Each SGSN is in charge of several base station subsystems (BSSs). A BSS contains a base station controller (BSC) and one or more base transceiver stations (BTSs). The BCS monitors and controls the BTSs in its BSS. Mobile terminals (MTs) are connected to a BTS via the Um air interface [3]. GERAN uses the same air interface specified in GSM. Visual Media Coding and Transmission Ahmet Kondoz © 2009 John Wiley & Sons, Ltd. ISBN: 978-0-470-74057-6
248 Visual Media Coding and Transmission Figure 8.1 GERAN system architecture Multiple access is based on the combination of TDMA and FDMA techniques [4]. Frequency bands have been reserved around 900 MHz for GSM 900 and around 1800 MHz for GSM 1800 for use in Europe. Allocated frequency bands are: 890 915 MHz (uplink) and 935 960 MHz (downlink), and 1710 1785 MHz (uplink) and 1805 1880 MHz (downlink). These bands are divided by frequency into a number of carriers, each spaced 200 kHz apart. One or more carrier frequencies are assigned to each base station. The selection of carrier frequencies for each base station and the carrier reuse distance are decided by the cell planning procedure. Optionally, frequency-hopping capabilities can be used. The advantage is that the quality on all communication links is averaged through the interferer’s diversity. Each of the carrier frequencies is divided into eight time slots. The communication channel is described by its time-slot number and a carrier frequency. One time slot is 577 ms long. Within a time slot, data is transmitted in bursts. GERAN introduces a set of control and traffic channels to be used in packet-switched connections. The packet common control channel (PCCCH) comprises the common control and signaling channels used for the transfer of packet data. PCCCH contains the packet random access channel (PRACH), packet paging channel (PPCH), packet access grant channel (PAGCH), and packet broadcast control channel (PBCCH). PRACH is used by the MT to initiate an uplink data transmission. PPCH is used to page the MT prior to downlink data transfer. PAGCH is used to send resource assignment information to the MT, while PBCCH is used to broadcast packet data-specific system information. The packet data traffic channel (PDTCH) and packet-associated control channel (PACCH) form the traffic channels in GERAN. PDTCH is used to convey the traffic data to a specific MT. PDTCH is temporally dedicated to a user. Users are allowed to use multiple simultaneous PDTCHs to transfer data in multi-slot operation. Resource assignment/reassignment, adjustment in allocated capacity for PDTCH, and other necessary signaling information for each user are conveyed via PACCH [5]. The user plane protocol layers of GERAN include the transport/network layer, subnetwork- dependent convergence protocol (SNDCP) layer, logical link control (LLC) layer, radio link control (RLC) layer, and medium access control (MAC) layer. Transmission/reception data
Wireless Channel Models 249 Transport & RTP/UDP/IP PDU Network layer H Transport layer PDU SNDC LLC H SNDC PDU RLC/MAC RLC/MAC Block … RLC/MAC Block Physical Link Burst Burst Burst Burst Layer Figure 8.2 E GPRS data flow diagram flow over the GERAN protocol stack is shown in Figure 8.2. The transfer of LLC PDUs on the packet data physical channel is supported by the use of a temporary block flow (TBF). TBF is responsible for allocating radio resources on PDTCHs. As the name suggests, resource allocation based on TBF is temporary and is maintained only for the duration of the LLC PDU transfer. LLC PDUs are segmented into RLC/MAC blocks, and each RLC/MAC block is augmented with an 8 bit MAC header and a 16 bit block check sequence at the RLC/MAC layer. The radio frame (or RLC/MAC block) is 20 ms long and is transmitted over the wireless link using four bursts. EGPRS introduces new modulation-coding schemes (MCS) to achieve enhanced throughput capacity. These modulation-coding schemes are referred to as MCS-1 9. MCS-1 4 are based on the Gaussian minimum shift keying (GMSK) modulation scheme, as used in GSM speech services and GPRS PDTCHs. Schemes MCS-5 9 employ a higher-rate, eight-state phase shift keying (8-PSK) modulation scheme. Convolutional codes with different code rates are used for Application Relay IP IP SNDCP GTP-U GTP-U SNDCP LLC Relay LLC UDP UDP RLC RLC BSSGP BSSGP IP IP MAC L2 GSM RF MAC Network Network L2 Service Service L1 L1 MS GSM RF L1bis L1bis GGSN Um Gb Gn Gi BSS SGSN Figure 8.3 GERAN user plane protocol stack [6]. Reproduced, with permission, from 3GPP Specifi cation no. 23.060 Ó2001 3GPP. Ó1998 3GPP. Reproduced by permission of Ó European Telecommu nications Standards Institute 2008. Further use, modification, redistribution is strictly prohibited. ETSI standards are available from http://pda.etsi.org/pda/
250 Visual Media Coding and Transmission Table 8.1 Modulation coding parameters for E GPRS [7]. Reproduced, with permission, from 3GPP TS 03.64 2002, “GSM/EDGE; Overall description of the GPRS radio interference; Stage 2, Release 1999”, v 8.10.0. Ó1998 3GPP. Reproduced by permission of Ó European Telecommunications Standards Institute 2008. Further use, modification, redistribution is strictly prohibited. ETSI standards are available from http://pda.etsi.org/pda/ Scheme Convolutional Modulation RLC blocks per Raw data Data rate code rate for radio block within one kbps data (20 ms) radio block MCS 1 0.53 GMSK 1 176 8.8 MCS 2 0.66 GMSK 1 224 11.2 MCS 3 0.80 GMSK 1 296 14.8 MCS 4 1.0 GMSK 1 352 17.6 MCS 5 0.37 8 PSK 1 448 22.4 MCS 6 0.49 8 PSK 1 592 29.2 MCS 7 0.76 8 PSK 2 2 Â 448 44.8 MCS 8 0.92 8 PSK 2 2 Â 544 54.4 MCS 9 1.0 8 PSK 2 2 Â 592 59.2 the channel protection. A summary of the coding parameters for the EGPRS coding schemes and the resulting information data rates for single-slot downlink transmission are given in Table 8.1. Allowing a maximum of eight-time-slot operation, this channel structure supports a maximum data rate of 473.6 kbps for EGPRS downlink. 8.2.2 GPRS Physical Link Layer Model Description The model developed simulates the physical-layer characteristics of the channel between a GPRS mobile terminal and a base station. An outline of the simulator is shown in Figure 8.4. It can be seen that the transmitted signal is subjected to a multipath fast-fading environment and a single co-channel interferer. In addition, an additive white Gaussian noise source is present at the receiver. This allows for the bit error and block error characteristics to be determined for a range of carrier-to-interference (C/I) ratios and signal-to-noise (Eb/No) ratios using the different GPRS and EGPRS channel coding schemes. Note that a static C/I profile is implemented in this model and no shadowing or slow-fading effects are described in the simulator model. However, these may be easily implemented by concatenating the data sets describing the channel bit error characteristics of different, static, carrier-to-interference levels. The simulator model was built using the COSSAP [8] stream simulation environment. The modulator and receiver block employed for the GMSK schemes was the standard COSSAP block found in the GSM reference design kit. A decision-feedback equalizer [9] was used for the 8-PSK modulation-coding schemes. All other elements of the simulation model were either created or adapted from other lower-level COSSAP blocks. 8.2.2.1 Channel Coding GPRS employs four channel protection schemes, offering flexibility in the degree of protection and traffic capacity available to the user. Varying the channel coding scheme allows for an optimization of the throughput across the radio interface as the channel quality varies. The channel coding schemes are defined in [10] and are outlined here.
Wireless Channel Models 251 Source Data Carrier AWGN Source Data Forward Error Profile Source Fast Fading Error correction Encoding and block error Interleaving Fast Fading Burst Formatting detection Modulation De-Interleaving Interfering Signal Burst Formatting Equalizer Modulation Channel Estimator Rx Filter Interference Profile Figure 8.4 E/GPRS physical link layer simulator model Packet Data Block Type 1 (CS-1) The coding scheme used for CS-1 is the same as used for the GSM SACCH (slow associated control channel). One hundred and eighty-four source data bits are coded using a shortened binary cyclic (FIRE) code to give a block of 224 bits, to which four tail bits are added. These 228 bits are then encoded using the 1/2 rate convolutional encoder used to protect class 1 bits in the TCH/FS speech channel, thereby producing a block of 456 coded bits. The code polynomials are: G0 ¼ 1 þ D3 þ D4 ð8:1Þ G1 ¼ 1 þ D þ D3 þ D4 Packet Data Block Type 2 (CS-2) The data is forwarded to the encoder in fixed-sized blocks of 271 bits. The first three bits, representing the uplink status flag (USF), are precoded into six bits, and a further sixteen parity bits and four tail bits are added to the information block. The resulting 294 bits are encoded with the same 1/2 rate code as used in CS-1, thereby giving a block of 588 bits that is reduced to 456 bits by means of puncturing. Packet Data Block Type 3 (CS-3) In this coding scheme the information bits are forwarded to the encoder in blocks of 315 bits. As in scheme CS-2, the three USF bits are precoded into six bits, and sixteen parity and four tail
252 Visual Media Coding and Transmission bits are added, to form a block of 338 bits. Convolutional encoding using the CS-1 encoder produces a block of 676 bits that is reduced to 456 bits by puncturing. Packet Data Block Type 4 (CS-4) This is a transparent scheme in which no forward error correction is carried out on the information bits. The three USF bits are block coded into twelve bits for protection, while sixteen parity bits are added to the end of the 431 bit information block for error-detection purposes. Once again, the result is a 456 bit radio block. 8.2.2.2 Interleaving The GPRS packet data channels are interleaved using a block rectangular scheme identical to that used for the SACCH [10]. In this scheme, the depth of interleaving is four blocks. This means that each data block is spread over four consecutive TDMA frames, which is equivalent to 18.46 ms. 8.2.2.3 Channel Decoding The channel decoder uses a soft-decision Viterbi algorithm to decode the convolutional codes. In scheme CS-1, the FIRE code is used for both error correction and detection, whereas in codes CS-2 4, the cyclic codes are used for error-detection purposes only. 8.2.3 EGPRS Physical Link Layer Model Description The structure of the EGPRS physical link layer model is similar to that discussed for the GPRS simulator. 8.2.3.1 EGPRS Channel Coding Schemes EGPRS supports nine joint modulation-coding schemes, referred to as schemes MCS-1 9 (see Table 8.2). MCS-1 4 are based on the same GMSK modulation scheme used in GSM speech services and GPRS PDTCHs, whereas schemes MCS-5 9 employ a higher-rate 8-PSK modulation scheme [11]. One very noticeable difference between the schemes used in EGPRS and those employed by the GPRS PDTCHs is that the radio block headers are encoded separately from the data payload. Another difference is that in schemes MCS-7 9, two RLC/ MAC blocks are inserted into a single radio block. In GPRS a one-to-one block mapping is always maintained. The coding schemes and block formatting for uplink and downlink scenarios vary slightly. Schemes and results are shown for the downlink only. There is negligible difference in error performance for the uplink and downlink of the same scheme. The schemes are divided into three families. Each of these families has a different unit of payload (37 and 34 for A, 28 for B, and 22 for C) into which the data payload is divided. Each radio block may contain either two or four of these payload units, and in the case where four units are present, the data is split into two RLC/MAC blocks. The details of the MCS-1 and MCS-5 schemes, which form the basis of all the other punctured codes, and which may therefore be regarded as being representative of the other schemes, are shown below.
Wireless Channel Models 253 Table 8.2 EGPRS channel coding schemes [12]. Reproduced, with permission, from ETSI/SMG, GSM 03.64 1998, “Overall description of the GPRS radio interface stage 2”, V. 5.2.0., 1998. Ó2002 3GPP. Ó1998 3GPP. Reproduced by permission of Ó European Telecommunications Standards Institute 2008. Further use, modification, redistribution is strictly prohibited. ETSI standards are available from http:// pda.etsi.org/pda/ Scheme Code Header Modulation RLC blocks Raw data Family BCS Tail HCS Data rate code rate per radio within one payload rate block radio block kbps (20 ms) MCS 9 1.0 0.36 8 PSK 2 2 Â 592 A 2 Â 12 2 Â 6 8 59.2 MCS 8 0.92 0.36 8 PSK 2 2 Â 544 A 12 6 54.4 MCS 7 0.76 0.36 8 PSK 2 2 Â 448 B 44.8 MCS 6 0.49 1/3 8 PSK 1 592 A 29.6 544 þ 48 27.2 MCS 5 0.37 1/3 8 PSK 1 448 B 22.4 MCS 4 1.0 0.53 GMSK 1 352 C 17.6 MCS 3 0.80 0.53 1 296 A 14.8 272 þ 24 13.6 MCS 2 0.66 0.53 1 224 B 11.2 MCS 1 0.53 0.53 1 176 C 8.8 Note: the italic captions indicate the padding. MCS-1 Block size: 209 bits Information payload: 176 bits Both the header and the payload sections of the radio block are encoded using a 1/3 rate convolutional mother code, which has a constraint length of 7. The polynomials used to generate the codewords are: G4 ¼ 1 þ D2 þ D3 þ D5 þ D6 ð8:2Þ G7 ¼ 1 þ D þ D2 þ D3 þ D6 G5 ¼ 1 þ D þ D4 þ D6 Both the header and the payload sections are punctured so as to produce a 1/2 rate convolutional code. An 8 bit parity check is added to the header section to provide for error detection, while a 12 bit check is added to the data payload. See Figure 8.5. MCS-5 Block size: 478 bits Information payload: 448 bits (56 octets) An 8 bit cyclic redundancy check sequence is added to the 25 bit header for error detection. The resulting block is then encoded using a 1/3 rate convolutional code defined by the
254 Visual Media Coding and Transmission 3 bits 36 bits 196 bits Data = 22 octets = 176 bits BCS TB USF RLC/MAC HCS FBI E Hdr 12 bits 108 bits Rate 1/3 convolutional coding puncturing 588 bits puncturing SB = 12 12 bits 68 bits 372 bits 372 bits P1 P2 464 bits Figure 8.5 MCS 1 data flow polynomials: G4 ¼ 1 þ D2 þ D3 þ D5 þ D6 ð8:3Þ G7 ¼ 1 þ D þ D2 þ D3 þ D6 G5 ¼ 1 þ D þ D4 þ D6 This results in a 99 bit block to which one spare bit is added. The 3 bit USF is mapped to a 36 bit sequence. The burst-mapping and interleaving mechanisms ensure this 36 bit precoded USF is spread evenly over the four bursts containing the MCS-5 radio block. A 12 bit CRC and six tailing bits (all 0) are added to the data payload, as in MCS-1. Two puncturing schemes are used to produce an output code rate for the user data of 0.37. See Figure 8.6. 8.2.3.2 Propagation Model The channel model used in the simulator follows the description of the GSM mobile radio multipath propagation model described in GSM 05.05 [13]. In this model, it is assumed that the mobile radio environment is dispersive, with several reflectors and scatterers and different distances from the line-of-sight path between the mobile terminal and the base station. For this reason, the transmitted signal may reach the receiver via a number of distinct paths, each having different delays and amplitudes. This phenomenon is best described as a power-delay profile of the propagation environment. This is essentially a number of individual taps representing a single beam, and has a gain that varies with time according to fast-fading characteristics. GSM 05.05 defines three such multipath propagation models for use in simulation models, namely the typical cases for rural area (RA), hilly terrain (HT), and urban area (TU). All of these three models are implemented in the simulator, and the delay-spread characteristics for the urban area model are shown in Table 8.3.
Wireless Channel Models 255 3 bits 33 bits 468 bits Data = 56 octets = 448 bits BCS TB USF RLC/MAC HCS FBI E Hdr. SB = 8 36 bits 99 bits Rate 1/3 convolutional coding P2 36 bits +1 bit 1248 bits 100 bits 1404 bits puncturing P1 1248 bits 1392 bits Figure 8.6 MCS 5 data flow Within each path followed by the transmitted waves, there occurs narrowband fading. The first-order statistics of such fading are described by the Rayleigh distribution, whereas the second-order statistics are characterized by the classical Doppler spectrum. This describes the spread of frequencies that occurs when there is a relative difference in velocity between the mobile terminal and the base station, and is a function of the speed of the mobile terminal and of the carrier wavelength. In the simulator model, two main propagation models are used, one for ideal frequency hopping (IFH) and one for no frequency hopping (NFH). When no frequency hopping is used, it is assumed that the radio bursts are transmitted on a single Table 8.3 Typical case for urban area (Tux) [13]. Reproduced, with permission, from 3GPP “Technical specification 3rd Generation Partnership Project; Technical Specification Group GERAN; digital cellular telecommunications system (phase 2 þ ); radio transmission and reception, (release 1999)”, 3GPP TS 05.05 V8.8.0. January 2001. Ó2001 3GPP. Ó1998 3GPP. Reproduced by permission of Ó European Telecommunications Standards Institute 2008. Further use, modification, redistribution is strictly prohibited. ETSI standards are available from http://pda.etsi.org/pda/ Tap number Relative time Average relative power Doppler spectrum (ms) (dB) (1) (2) (1) (2) 1 0.0 0.0 3.0 3.0 Classical 2 0.2 0.2 0.0 0.0 Classical 3 0.5 0.6 2.0 2.0 Classical 4 1.6 1.6 6.0 6.0 Classical 5 2.3 2.4 8.0 8.0 Classical 6 5.0 5.0 10.0 10.0 Classical
256 Visual Media Coding and Transmission carrier with continuous second-order fast-fading characteristics. This means that if a burst is currently experiencing a fade in received power, then it is likely that subsequent bursts may experience similar fading. However, when slow frequency hopping is used, the mobile station will transmit or receive on a fixed frequency for one timeslot (%577 ms) and then must hop before the timeslot in the next TDMA frame. This provides interference diversity at the receiver, as all bursts experience fading characteristics with greatly reduced correlation with the characteristics of any previous bursts. The simulator model implements ideal frequency hopping by using four uncorrelated Rayleigh processes, which are applied alternately to each successive burst. As the interleaving depth of the PDTCH is only four bursts, the fading characteristics of all bursts within a single radio block are uncorrelated, emulating ideal frequency hopping. In addition to fading and multipath characteristics, the received signal is corrupted by co- channel interference and noise at the receiver. The interference is simulated by a single co- channel carrier that is added to the transmitted carrier. Eb/No characteristics are represented by an additive white Gaussian noise source added at the receiver. 8.2.3.3 Simulator Parameters Table 8.4 is a list of all the parameters that are user-definable, either by modifying the parameters of COSSAP hierarchical models, by changing the building blocks that constitute the model, or by using different schematics. 8.2.4 GPRS Physical Link Layer Simulator Reference performance figures for the GSM physical channel are given in GSM 05.05 [13]. These allow for the setting of reference transmitter and receiver performance figures for nominal error rates, sensitivity levels, interference levels, and erroneous frame indication performance levels. Techniques for specifying reference interference performance of GPRS are tackled in Annex L of GSM 05.50 [14], where methods on how to report GPRS performance in GSM 05.05 are described. As GPRS is designed to provide error-free transmission in bursty error environments, the most appropriate metric for evaluating channel error performance of the different channel coding schemes is the block loss error ratio (BLER). This gives the rate at which uncorrectable errors are detected in a decoded RLC/MAC radio block, which essentially translates into the rate of discarded blocks. In [14], two methods for reporting GPRS performance are proposed. One is to evaluate the BLER and C/I ratios for all coding schemes corresponding to the ranges of highest throughput. The second is to evaluate the C/I ratios for a fixed reference BLER value of 10% for coding schemes CS-1 4 so as to try to maximize the throughput performance. For all tests a 2 dB implementation margin is included when quoting C/I values (as shown in Table 8.5). As the physical link layer simulator model described does not implement any RLC/MAC functionality, the latter approach is selected for validating the GPRS simulator performance. This is the method adopted in GSM 05.05. 8.2.4.1 GPRS Model Validation at TU50 IFH 900 MHz As mentioned above, Annex L of [14] presents reference interference performance figures for GPRS systems operating at 900 MHz using the TU50 multipath fading channel model. Table 8.6 shows a comparison of the results obtained using the simulation model designed
Wireless Channel Models 257 Table 8.4 E/GPRS simulator parameters Settings Parameter Channel Coding Scheme Supported GPRS PDTCH CS 1,CS 2,CS 3,CS 4 HSCSD TCH/F9.6, TCH/F14.4, EGPRS PDTCH MCS 1, Interleaving MCS 2, MCS 3, MCS 4, MCS 5, MCS 6, MCS 7 Block rectangular over four frames for GPRS Training Sequence Codes As specified in GSM 05.01 for EGPRS Modulation Eight codes available Interference Characteristics GMSK, 8 PSK User definable static C/I ratio for single co channel Fading Characteristics interferer. May also be disabled. No frequency offset Rayleigh fading for each path (Rice for one component of Multipath Characteristics RA). Fading varies during one burst TU, RA, HT propagation environments supported, as in Transmission Capabilities GSM 05.05 User definable. Can simulate no frequency hopping and Mobile Terminal Velocity ideal frequency hopping (no correlation between suc Carrier Frequency cessive bursts) Antenna Characteristics User definable. Static > 250 kmph (for 900 MHz) User definable to 900 MHz or 1800 MHz Signal to Noise Characteristics 0 dB gain for both transmitter and receiver. No antenna Burst Recovery diversity AWGN source at receiver. User definable Eb/No ratio Equalizer Synchronization based on the cross correlation properties of the training sequence Channel Decoding 16 state soft output MLSE equalizer for GMSK 16 state decision feedback MLSE equalizer for 8 PSK Performance Measures Soft decision Viterbi convolutional decoder. Fire cor Simulation Length rection and detection for CS 1 and CRC detection for CS 2 4 and MCS 1 9 Bit error patterns and block error patterns User definable. Most experiments run for 15 000 blocks per timeslot for these experiments, which will be referred to as the CCSR (Centre for Communication System Research) model, and those quoted in [13]. From the results in Table 8.6 it can be seen that for coding schemes CS-2 and CS-3 the performance of the CCSR model is about 1.5 dB better than the reference value suggested in Table 8.5 Reference interference performance values; implementation margin of 2 dB included Coding scheme C/I at BLER ¼ 10% GSM 900 TU50 ideal FH GSM 900 TU3 no FH CS 1 9 dB 13 dB CS 2 13.8 dB 15 dB CS 3 16 dB 16 dB CS 4 23 dB 19.3 dB
258 Visual Media Coding and Transmission Table 8.6 Comparison of reference performance at TU50 IFH 900 MHz; Conditions: ideal frequency hopping, receiver noise floor Eb/No ¼ 28 dB (2 dB implementation margin assumed) Coding scheme C/I at BLER ¼ 10% CS 1 [14] [13] CCSR model CS 2 CS 3 9 dB 9 dB 8.5 dB CS 4 13.8 dB 13 dB 12.3 dB 16 dB 15 dB 14.6 dB 23 dB 23 dB 23.5 dB Annex L of GSM 05.50. The difference in performance for schemes CS-1 and CS-4 is less pronounced at around 0.5 dB. However, comparison of BLER interference performance traces obtained using the CCSR model with those presented in Annex P of the same document [13], which gives the results obtained by Ericsson, shows that the performances of the two simulated systems are virtually identical. This is shown in Figure 8.7. 8.2.4.2 GPRS Model Verification at TU1.5 NFH 1800 MHz [13] Annex P also presents proposals for GPRS reference performance results for the TU3 multipath model at 1800 MHz with no frequency hopping implemented. The results of the CCSR model at a BLER of 10% under these conditions are shown in Table 8.7. The results in Table 8.7 clearly show that under the propagation conditions described in TU1.5, the performance of the CCSR model at a resulting BLER of 10% is very close to the reference interference performance levels specified in [14] Annex L. In fact, the obtained C/I ratio differs by no more than 0.5 dB for CS-1 and CS-3. No suitable results for comparison were 1.0E+00 1.0E–01 BLER 1.0E–02 CS-1 CS-2 1.0E–03 CS-3 CS-4 EricssonCS-1 EricssonCS-2 1.0E–04 EricssonCS-3 EricssonCS-4 3 4567 8 9 10 11 12 13 14 15 16 17 18 19 C/I (dB) Figure 8.7 GPRS interference performance TU50 IFH 900 MHz
Wireless Channel Models 259 Table 8.7 Comparison of reference performance at TU1.5 NFH 1800 MHz; Conditions: no frequency hopping, receiver noise floor Eb/No ¼ 28 dB (2 dB implementation margin assumed) Coding scheme C/I at BLER ¼ 10% CS 1 [14] [13] CCSR model CS 2 CS 3 13 dB 13 dB 13.5 dB CS 4 15 dB 15 dB 15.2 dB 16 dB 16 dB 16.5 dB 19.3 dB 19 dB obtained from CS-4 as simulations were carried up to C/I ¼ 18 dB, at which value the resulting BLER was still in excess of 10%. The performance traces under these conditions can be seen in Figure 8.7, where the results obtained with the CCSR model closely match those obtained by Ericsson in [14] Annex P. 8.2.4.3 GPRS Model Verification at TU1.5 IFH 1800 MHz Results obtained using the CCSR simulator were compared with those specified in GSM 05.05 and are shown in Table 8.8 and Figures 8.8 8.10. The simulations were run for lengths equivalent to 2 Â 106 information bits, which is equivalent to roughly 10 800 RLC/MAC blocks for CS-1 and 4600 RLC/MAC blocks for CS-4 coding schemes. This length of information bits can be used to provide a continuous channel error pattern for a 64 kbps video stream for over 30 seconds before looping over to the beginning of the error sequence. This is because the bursty nature of the GSM fading channel, coupled with the differing visual susceptibility to errors of different parameters that constitute the video bitstream, requires that sufficiently long error patterns are used to obtain meaningful results. The performance of the CCSR model was validated for a variety of conditions. In particular, the model was seen to give accurate results with respect to variations in interference levels, transmission modes (frequency hopping enabled/disabled), and carrier frequency. Although the results for GPRS were presented as block error ratio values, the output from the simulators characterizes the physical link layer in terms of both bit and block error patterns. Although the results obtained with the designed model closely match the quoted reference performance Table 8.8 Comparison of reference performance at TU1.5 IFH 1800 MHz; Conditions: no frequency hopping, receiver noise floor Eb/No ¼ 28 dB (2 dB implementation margin assumed) Coding scheme C/I at BLER ¼ 10% CS 1 [13] CCSR Model CS 2 CS 3 9 dB 9.7 dB CS 4 13 dB 13 dB 15 dB 15 dB 23 dB
260 Visual Media Coding and Transmission 1.0E+00 BLER 1.0E–01 CS-1 CS-2 CS-3 CS-4 EricssonCS-1 EricssonCS-2 EricssonCS-3 EricssonCS-4 1.0E– 02 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 C/I (dB) Figure 8.8 GPRS interference performance TU1.5 NFH 1800 MHz levels, it must be appreciated that much of the system performance relies on implementation- dependent factors. This is particularly true for the receiver’s correlator and equalizer, and, to a lesser extent, the channel decoding mechanisms. These factors lead to variations in GPRS physical layer performance figures released by different manufacturers of up to 2 dB [14]. In 1.0E+00 1.0E–01 BLER 1.0E–02 CS-1 CS-2 CS-3 CS-4 1.0E–03 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 C/I (dB) Figure 8.9 GPRS interference performance TU1.5 NFH 1800 MHz
Wireless Channel Models 261 1.0E+00 1.0E–01 BLER 1.0E–02 1.0E–03 CS-1 CS-2 CS-3 EricssonCS-1 EricssonCS-2 EricssonCS-3 1.0E–04 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 C/I (dB) Figure 8.10 GPRS interference performance at TU1.5 IFH 1800 MHz addition, it must be noted that the assumption made in [13] that the TU1.5 1800 MHz and TU3 900 MHz propagation models are identical is retained here. 8.2.5 EGPRS Physical Link Layer Simulator As for the GPRS PDTCHs, the reference performance of EGPRS links is specified in terms of the carrier-to-interference ratio or energy per modulated bit required to obtain a 10% radio block error ratio. A block is considered erroneous in the simulation when any of the following occur: . Uncorrectable bit errors in the data field after decoding (including CRC bits). . Uncorrectable bit errors in the header field after decoding (including CRC bits). . Erroneous decoded stealing flag code word. [15] and [16] also specify that erroneous modulation detection, which is also referred to as blind detection error, should be simulated. This type of error was not included in the simulation models used, as different models were used for the GMSK and 8-PSK modulation schemes, with the receivers knowing a priori the type of modulation to expect. It is not expected that this assumption would cause excessive deviation from the required results. When examining the results obtained by different manufacturers and specified in [17], it can be seen that the specified values for the 8-PSK schemes vary widely. For example, in the co- channel interference case at TU1.5 NFH at 1800 MHz given in [17], there exists a 4.9 dB difference between the worst and best quoted values for MCS-5, and of 5.1 dB for MCS-6. In [18], fewer results from fewer manufacturers are given, but there still remains a spread of around 3 dB in the given values. As a result, average-based reference performance values are given in [17] and [18], which are seen to be quite similar to each other. The performance of the CCSR simulation model was therefore compared with these average figures. The variance in
262 Visual Media Coding and Transmission values of the GMSK coding schemes (MCS-1 4) is shown to be considerably lower than for the 8-PSK schemes. This is probably due to the greater maturity of GMSK-based technology for fading channels as compared to 8-PSK. 8.2.5.1 EGPRS Model Validation at TU 50 NFH 900 MHz The EGPRS models were simulated for co-channel interference, with a single interferer being used. When examining the results obtained in [16], it was seen that MCS-8 and MCS-9 could not reach 10% BLER even at C/I values in excess of 30 dB. For this reason, reference performance figures at 30% BLER were used in these cases. As will be shown in this chapter, channel error ratios that result in BLER values of around 10% are considerably too high to produce acceptable video quality. For this reason MCS-8 and MCS-9 were not used in the tests carried out. The results obtained are shown in Tables 8.9 8.12. The reference figures given in [17] and [19], which are both published by the EDGE drafting group, are seen to be extremely similar to each other. Under these propagation conditions, the CCSR model is seen to produce results to within 0.5 dB for all coding schemes, with the exception of MCS-2, where the performance of the CCSR model is inferior by 1.5 dB, and MCS-3, where a discrepancy of 2 dB is noted. At MCS-7, and MCS-5, the CCSR model’s performance is superior to the reference figures. Reference performance figures for MCS-1 4 for these propagation conditions were not available in the given references. However, the performance is fairly similar to the equivalent results obtained at TU50 NFH 900 MHz, indicating that the performance is close to the expected values. Indeed, comparing the values for CS-1 4 for GSM 900 MHz and DCS 1800 MHz at TU50 NFH in [13], it can be seen that the reference figures at 10% BLER do not vary by more than 1 dB. There does however exist a considerable discrepancy between the results for MCS-5 7 as given in [20] and [18], where differences of up to 6 dB can be seen. Although the CCSR model performs well at MCS-5, giving results of within 1 dB of the value quoted in [20], codes MCS-6 and MCS-7 perform considerably worse. In fact at MCS-7 a figure of 10% BLER was not achieved at all. 8.2.5.2 EGPRS Model Validation at TU50 NFH 1800 MHz 8.2.5.3 EGPRS Model Validation at TU1.5 NFH 1800 MHz The CCSR EGPRS model performs to within 1 dB of the reference figures at all C/I ratios, except for MCS-3 where the discrepancy is of 1.5 dB. Indeed, at MCS-5 the CCSR model Table 8.9 Performance comparison at TU50 NFH 900 MHz CCSR [17] [19] 9 dB MCS 1 8.5 dB 15.5 dB 12 dB MCS 2 10.5 dB 18.0 dB 17 dB MCS 3 15.0 dB 25.0 dB 20.5 dB MCS 4 20.0 dB 15.3 dB MCS 5 15.5 dB 18.5 dB MCS 6 18.0 dB 24.0 dB MCS 7 23.0 dB
Wireless Channel Models 263 Table 8.10 Performance comparison at TU50 NFH 1800 MHz [18] [20] CCSR MCS 1 [1] 15 dB 13.0 dB 8 dB MCS 2 18 dB 15.5 dB 12 dB MCS 3 27.5 dB 21.5 dB 17 dB MCS 4 21 dB MCS 5 16 dB MCS 6 24 dB MCS 7 exceeds the values given in [17] and [19] by about 4 dB. It is evident that under these conditions, the CCSR model performs very similarly to the reference models. See Table 8.11. 8.2.5.4 EGPRS Model Validation at TU1.5 IFH 1800 MHz When using ideal frequency hopping, the CCSR model on average performs around 1.5 dB worse than the reference figures. One reason for this is that, in order to simplify implemen- tation and reduce the complexity of the simulation models, even though the degree of correlation between consecutive bursts is greatly reduced when compared to the non- frequency hopping case, the de-correlation is not perfect. This slightly reduces the efficacy of the interleaving mechanisms. However, the difference between the obtained values and the reference figures is only in excess of 1.5 dB at MCS-3, and is less than 1 dB for MCS-4, MCS- 5 and MCS-7. The performance of the model under these conditions can consequently be considered adequate. Table 8.11 Performance comparison at TU1.5 NFH 1800 MHz [19] [17] CCSR MCS 1 [1] 19.5 dB 11.0 dB 11.3 dB MCS 2 21.5 dB 13.0 dB 13 dB MCS 3 26.5 dB 14.5 dB 16 dB MCS 4 17.0 dB 18 dB MCS 5 19.0 dB 15.2 dB MCS 6 21.0 dB 21 dB MCS 7 24.0 dB 24 dB Table 8.12 Performance comparison at TU1.5 IFH 1800 MHz [19] [17] CCSR MCS 1 14.5 dB 7.5 dB 9.0 dB MCS 2 17.0 dB 10.0 dB 11.5 dB MCS 3 23.5 dB 14.5 dB 16.5 dB MCS 4 19.5 dB 20.0 dB MCS 5 14.0 dB 15.3 dB MCS 6 17.0 dB 18.5 dB MCS 7 22.5 dB 22.0 dB
264 Visual Media Coding and Transmission The EGPRS physical link layer model was validated for four propagation conditions, all of which assumed the typical urban multipath model. Both the GMSK and 8-PSK modulator/ demodulator structures were seen to give the expected results, and the relative differences between the performances of the two receiver structures as compared to the reference values were rather small. There does however exist a considerable difference between the performance of the designed model and those of the better-performing models given in [17], particularly for the 8-PSK modulation-coding schemes. As already mentioned, there exists a spread of around 4 5 dB in the figures given for different receiver implementations. There may be various reasons for this, including possible differences in the ways the propagation models are simulated. However, the most likely reason lies within the different implementation strategies of the receivers. There exist several techniques for carrying out channel and noise estimation, and for implementing equalization at the receiver. Some methods can adapt dynamically to differing channel conditions, whereas others perform optimally under certain conditions and not so well under others. The differences in the 8-PSK values are greater than for GMSK, where the technology is fairly stable and consolidated. As a result, although the CCSR model matches the reference performance figures or comes very close to them under practically all conditions tested, they should be regarded to a certain extent as the worst-case performance figures. In fact, as already described, the CCSR model used a custom-made EDGE receiver mechanism, which although employing a very effective nonlinear direct-feedback equalizer, cannot be considered the most optimal solution. In particular, no automatic frequency-correction mechanisms were implemented in the receiver. The comparison with the reference performance figures is however only part of the story. The 10% BLER figure chosen for measuring performance compared to reference values was selected on the basis of being around the position where optimal throughput is achieved when operating with block retransmissions [14]. However, real-time services require information integrity without the use of retransmissions, and consequently the error performance requirements are much more stringent. Typically, error ratios in the order of 10 3 to 1.0E+00 1.0E–01 BLER 1.0E–02 1.0E–03 MCS-1 MCS-2 MCS-3 MCS-4 1.0E–04 5 10 15 20 25 30 0 C/I (dB) Figure 8.11 GMSK EGPRS interference performance TU50 NFH 1800 MHz
Wireless Channel Models 265 1.0E+00 BLER 1.0E–01 MCS-5 MCS-6 MCS-7 1.0E–02 0 5 10 15 20 25 30 35 C/I (dB) Figure 8.12 8 PSK EGPRS interference performance TU50 NFH 1800 MHz 10 4 are required for video communications. For this reason, the performance of the various modulation-coding schemes at lower BLER values and relatively high C/I values are more critical than they would otherwise be for typical data transfer applications. The equalizer used for the GMSK modulation schemes is a 16-state Viterbi equalizer. Examination of results above C/I values of around 15 20 dB show a considerable deviation from the quoted figures. Figures 8.11 8.18 show the BER sensitivity performance of the GMSK receiver for 1.0E+00 1.0E–01 BLER 1.0E–02 MCS-5 MCS-6 MCS-7 1.0E–03 0 5 10 15 20 25 30 35 C/I (dB) Figure 8.13 GMSK EGPRS interference performance TU1.5 IFH 1800 MHz
266 Visual Media Coding and Transmission 1.0E+00 BLER 1.0E– 01 MCS-1 1.0E– 02 MCS-2 MCS-3 MCS-4 1.0E– 03 5 10 15 20 25 30 0 C/I (dB) Figure 8.14 8 PSK EGPRS interference performance TU1.5 IFH 1800 MHz the TU50 multipath model at 900 MHz and 1800 MHz. As these figures display raw error rates with no forward error correction, the results are not affected by frequency hopping. For Eb/No values below 18 dB, the equalizer used in the CCSR model outperforms the reference figures, but then levels off to a higher asymptotic BER value. The probable reason for such a deviation is that the equalizer was optimized for operation of the speech channels and low-bit 1.0E+00 BLER 1.0E– 01 MCS-1 MCS-2 MCS-3 1.0E– 02 5 10 15 20 25 30 0 C/I (dB) Figure 8.15 GMSK EGPRS interference performance TU1.5 NFH 1800 MHz
Wireless Channel Models 267 1.0E+00 1.0E– 01 BLER 1.0E– 02 MCS-5 MCS-6 MCS-7 1.0E– 03 0 5 10 15 20 25 30 35 C/I (dB) Figure 8.16 8 PSK EGPRS interference performance TU1.5 NFH 1800 MHz rate data. These typically operate at C/I values below 12 dB. Indeed, differences in bit error ratios below 10 3 have a negligible effect on speech quality or data throughput using the TCH/9.6 or TCH/4.8 channels. This difference is consequently more conspicuous when using schemes MCS-3 and MCS-4, which for real-time services are the coding schemes that would operate at such interference levels when using GMSK. The major deviations are visible for MCS-3, where for example the BLER values at C/I ¼ 20 dB for TU50 NFH 1.0E+00 1.0E– 01 1800 MHz 900 MHz 1800 MHz [SMG2- 274/99] BER 1.0E– 02 1.0E– 03 1.0E– 04 5 10 15 20 25 30 35 40 0 Eb/No (dB) Figure 8.17 Raw bit error ratio GMSK TU50
BER268 Visual Media Coding and Transmission 1.0E+00 1800 MHz 1.0E– 01 900 MHz 1.0E– 02 1800 MHz [SMG- 2 274/99] 1.0E–03 5 10 15 20 25 30 35 40 0 Eb/No (dB) Figure 8.18 Raw bit error rate 8 PSK 1800 MHz are 0.05 and 0.015, respectively. A similar trend is visible when examining the sensitivity performance of the 8-PSK receiver in the TU50 multipath propagation conditions. At low Eb/No values, while inferior to the reference performance figures, the performance of the equalizer used in these experiments is quite close to the reference performance figures. However, at low noise values, this discrepancy increases considerably. There is also a large difference between the performance at 900 MHz and at 1800 MHz, where the difference between asymptotic bit rates is nearly in the order of one magnitude. This is caused by the greater susceptibility of 8-PSK to intersymbol interference, and the greater symbol spreading that occurs at 1800 MHz, particularly at high mobile speeds. The residual bit error patterns obtained for both GPRS and EGPRS are nevertheless suitable for use in the audiovisual transmission experiments, as they exhibit relative performance figures between coding schemes that are consistent with the relative power of the schemes. Moreover, the obtained results were shown to display a high degree of correlation with the performance results given by several manufacturers. 8.2.6 E/GPRS Radio Interface Data Flow Model The design of the EGPRS physical link layer model was restricted to examining the effects of varying channel conditions upon bits exiting the channel decoders. In order to carry out more extensive and detailed examinations of the effect of channel errors upon end-applications, such as video coding implementations, a GPRS data flow simulator was implemented. The model was implemented in Matlab, as this language provides a rapid development environment and comprehensive data analysis tools. The layers implemented included an MPEG-4 video codec with rate control functionality, RTP/UDP/IP transport layers, and GPRS SNDC, LLC, RLC/ MAC layer protocols. This layout is shown in Figure 8.19. It must be emphasized that only the data flow properties of the protocols have been implemented in this model. This means that none of the protocol signaling mechanisms have actually been included in the model, but only
Wireless Channel Models 269 Application Layer: MPEG-4 Codec Video Packet Transport Layer: Rate Control: Segmentation Some Video Packets Discarded Transport PDU Transport PDU Transport PDU Add Headers IP PDU IP PDU RTP-UDP-IP Headers Header Compression (optional) GPRS SNDC Layer SNDC Payload SNDC Header GPRS LLC Layer LLC Payload LLC Header GPRS RLC/MAC Layer Select Channel Coding Scheme CS-1 CS-2 CS-3 CS-4 RLC/MAC RLC/MAC RLC/MAC RLC/MAC RLC/MAC block block block block block Physical Link Layer Select Number of Timeslots 12 .......... 8 Figure 8.19 GPRS data flow model
270 Visual Media Coding and Transmission the resulting effect on header sizes, packet and stream segmentation procedures, and flow control effects. For example, when describing the RTP layer, sequence numbering is not actually implemented; only its effect on the resulting RTP-PDU header size is modeled in the simulator. The application layer consists of a traffic source emulating an MPEG-4 video codec that employs error-resilience functionality as described in [21] and rate control mechanisms that place an upper limit on the output throughput from the encoder, calculated on a frame-by-frame basis. The maximum allowable throughput is set according to the resources allocated across the radio interface. The output from the MPEG-4 codec is forwarded to the transport layers in units of discrete video frames, which will be referred to as video packets. These packets are then split up into transport layer PDUs according to the maximum IP packet size defined by the user. Each packet is then encapsulated into an independent RTP/UDP/IP [22] packet for forwarding down to the GPRS network. Header compression [23], [24], which is a user-definable feature of this simulator model, is implemented at the transport layer, although it is not actually used in the experiments. This means that the compression protocol that is employed in the end terminals must be supported in all the intermediate nodes in the core network, which potentially includes the Internet. This is hardly a realistic scenario, and a more appropriate implementation would be to include such functionality in the SNDC layer [25]. This would allow for headers to be compressed for transmission across the GPRS radio interface, only to be restored to their initial size at the SGSN for forwarding across the remainder of the network. However, as this model is solely concerned with the performance across the Um interface, the location of the compression algorithm has no effect upon any results obtained. At the SNDC layer, the 8 bit SNDC header [25] is added to the transport packet before forwarding on to the LLC layer. Here a 24 bit frame header and 24 bit frame check sequence are added to each LLC-PDU. A check is also carried out to ensure that no LLC frames exceed the maximum size of 1520 octets specified in [26]. The error-control and flow-control functions of the LLC layer are not implemented in this model. The LLC frames are then passed on to the RLC/MAC layer, where they are encapsulated into radio blocks according to the forward error correction scheme selected. In practice, the choice of coding scheme depends upon the carrier- to-interference ratio at the terminal and the resulting throughput that can be sustained at that C/I level using the different channel coding schemes. The major side-effect of varying the protection afforded to the user data is the modification in the size of the GPRS radio blocks in terms of the number of information bits per block. The model therefore segments the incoming PDU into data payloads for the output radio blocks according to the selected channel coding scheme, and forwards these blocks to a FIFO buffer. Once in the output buffer, the blocks wait for one of the timeslots allocated to their associated terminal to become available and are then transmitted over the given timeslots. The model allows any number of timeslots from one to eight to be allocated to the source terminal. This layered model design provides for error occurrences at the physical layer to be mapped on to the actual application-layer payload. Each channel error can therefore be mapped on to an individual video information bit, or the header or checksum section of any protocol in the GPRS stack. 8.2.7 Real-time GERAN Emulator The GPRS and EGPRS physical link layer simulation model is extremely computationally intensive, largely due to the modeling of the multipath propagation model, where a Rayleigh fading filter is used to represent each pathway. For example, a simulation representing data
Wireless Channel Models 271 encoded using the CS-1 scheme at TU1.5 IFH 1800 MHz at a carrier-to-interference ratio of 15 dB run on a 296 MHz Ultra SPARC processor, runs at an average rate of 138 information bits per second. Although the exact processing speed depends upon several factors, including the propagation conditions modeled, the C/I ratio present, and the modulation-coding scheme used, the obtained rates are way below those necessary to support a real-time simulation environment. In order to create a real-time testing environment for video communications applications, a real-time emulator was built using Visual C þþ for Microsoft Windows. The emulator implemented the data flow model and allowed for up to eight-slot allocation. The emulator program made use of a table look-up method to allow for real-time emulation. Data sets of bit error patterns at the physical link layer were created with the E/GPRS simulator for a wide range of interference and propagation conditions for each coding scheme. These were then used by the real-time emulator and fed into the GPRS radio interface data flow model described in Figure 8.19. Multi-threaded programming techniques were used to build the model, and a graphical user interface (Figure 8.20) was designed to allow for interactive manipulation of the coding scheme, interference level, carrier frequency, timeslot allocation, and frequency- hopping capability. The emulator was used in conjunction with a real-time MPEG-4 video encoder/decoder application using RTP and is shown in Figure 8.21. 8.2.8 Conclusion The design and validation of the EGPRS and GPRS physical link layer simulation model has been described. The GPRS models were seen to give a performance that closely matches the GSM reference performance figures. Although the performance of the EGPRS model was seen to match the figures given by different terminal manufacturers when operating at low terminal Figure 8.20 GPRS emulator graphical user interface
272 Visual Media Coding and Transmission MPEG-4 Server MPEG-4 Client RTP RTP UDP UDP SNDC SNDC LLC LLC RLC/MAC RLC/MAC Physical Link Layer Figure 8.21 GPRS radio access emulator structure velocities, a significant divergence from the reference figures was obtained using the TU50 propagation models at 1800 MHz. This may be attributable to the increased Doppler spread at high terminal velocities and carrier frequencies, and corresponding limitations in the equalizer and receiver architecture in dealing with the resulting intersymbol interference. However, at lower terminal velocity figures and carrier frequencies, the simulator model closely matched the reference performance figures, and may therefore be considered suitable for use in the media transmission experiments. 8.3 UMTS Channel Simulator This section presents the design and implementation procedure for a multimedia evaluation test- bed of the UMTS forward link. AWCDMA physical link layer simulator has been implemented using the Signal Processing WorkSystem (SPW) software simulation tools developed by Cadence Design System Inc. [27]. The model has been developed in a generic manner that includes all the forward link radio configurations, channel structures, channel coding/decoding, spreading/de-spreading, modulation parameters, transmission modeling, and their correspond- ing data rates according to the UMTS specifications. The performance of the simulator model is validated by comparison with figures presented in the relevant 3GPP documents [28], [29] for the specified measurement channels, propagation environments, and interference conditions in [30]. Using the developed simulator, a set of UMTS error pattern files is generated for different radio bearer configurations in different operating environments, and the results are presented. Furthermore, UMTS link-level performance is enhanced by implementing a closed-loop power- control algorithm. A UMTS radio interface protocol model, which represents the data flow across the UMTS protocol layers, is implemented in Visual C þ þ . It is integrated with the physical link layer model to emulate the actual radio interface experienced by users. This allows for interactive testing of the effects of different parameter settings of the UMTS Terrestrial Radio Access Network (UTRAN) upon the received multimedia quality. 8.3.1 UMTS Terrestrial Radio Access Network (UTRAN) The system components of UTRAN are shown in Figure 8.22. Functionally, the network elements are grouped into the radio network subsystem (RNS), the core network (CN), and the
Wireless Channel Models 273 Figure 8.22 Systems components in a UMTS user equipment (UE). UTRAN consists of a set of RNSs connected to the core network through the Iu interface. The interface between the UE and the RNS is named Uu. An RNS contains a radio network controller (RNC) and one or more node Bs. The RNS handles all radio-related functionality in its allocated region. A node B is connected to an RNC through the Iub interface, and communication between RNSs is conducted through the Iur interface. One or more cells are allocated to each node B. The protocol within the CN is adopted from the evolution of GPRS protocol design. However, both the UE and UTRAN feature completely new protocol designs, which are based on the new WCDMA radio technology. WCDMA air interfaces have two versions, defined for operation in frequency division duplexing (FDD) and time division duplexing (TDD) modes. Only the FDD operation is investigated in this chapter. The modulation chip rate for WCDMA is 3.8 mega chips per second (Mcps). The specified pulse-shaping roll-off factor is 0.22. This leads to a carrier bandwidth of approximately 5 MHz. The nominal channel spacing is 5 MHz. However, this can be adjusted approximately between 4.4 and 5 MHz, to optimize performance depending on interference between carriers in a particular operating environment. As described in Table 8.13, the FDD version is designed to operate in either of the following frequency bands [30]: . 1920 1980 MHz for uplink and 2110 2170 MHz for downlink. . 1850 1950 MHz for uplink and 1930 1990 MHz for downlink.
274 Visual Media Coding and Transmission Table 8.13 WCDMA air interface parameters for FDD mode operation Operating Frequency Band 2110 2170 MHz 1930 1990 MHz downlink Duplexing Mode 1920 1980 MHz Chip Rate 1850 1910 MHz uplink Pulse shaping Roll off Factor Frequency Division Duplex (FDD) Carrier Bandwidth 3.84 mega chips per second 0.22 5 MHz All radio channels are code-division multiplexed and are transmitted over the same (entire) frequency band. WCDMA supports highly variable user data rates with the use of variable spreading factors, thus facilitating the bandwidth-on-demand concept. Transmission data rates of up to 384 kbps are supported in wide-area coverage, and 2 Mbps in local-area coverage. The radio frame length is fixed at 10 ms. The number of information bits or symbols transmitted in a radio frame may vary, corresponding to the spreading factor used for the transmission, while the number of chips in a radio frame is fixed at 38 400 [30]. 8.3.1.1 Radio Interface Protocol The radio interface protocol architecture, which is visible in the UTRAN and the user equipment (UE), is shown in Figure 8.23. Layer 1 (L1) comprises the WCDMA physical Control plane User plane RRC Network Layer L3 H SDU Control Interface PDCP/ Infor. Field Logical channels BMC L2 RLC/ Transport channels MAC Physical L1 Radio Frame Radio Frame Physical channels Layer SDU – Service Data Unit Figure 8.23 Radio interface protocol architecture
Wireless Channel Models 275 layer. Layer 2 (L2), which is the data-link layer, is further split into medium access control (MAC), radio link control (RLC), packet data convergence protocol (PDCP), and broadcast multicast control (BMC). The PDCP exists mainly to adapt packet-switched connections to the radio environment by compressing headers with negotiable algorithms. Adaptation of broad- cast and multicast services to the radio interface is handled by BMC. For circuit-switched connections, user-plane radio bearers are directly connected to the RLC. Every radio bearer should be connected to one unique instance of the RLC. The radio resource control (RRC) is the principal component of the network layer layer 3 (L3). This comprises functions such as broadcasting of system information, radio resource handling, handover management, admission control, and provision of requested QoS for a given application. Unlike the traditional layered protocol architecture, where protocol layer interaction is only allowed between adjacent layers, RRC interfaces with all other protocols, providing fast local interlayer controls. These interfaces allow the RRC to configure char- acteristics of the lower-layer protocol entities, including parameters for the physical, transport, and logical channels [31]. Furthermore, the same control interfaces are used by the RRC layer to control the measurements performed by the lower layers, and by the lower layers to report measurement results and errors to the RRC. See Figure 8.24. UTRAN supports both circuit-switched and packet-switched connections. In order to transmit an application’s data between UE and the end system, QoS-enabled bearers have to be established between the UE and the media gateway (MGW). Figure 8.25 shows the user plane protocol stack used for data transmission over packet-switched connection in Release 4. Measurements RRC Measurement reports RRC Control Control Measurements RLC RLC Control MAC Radio resource MAC assignment: Code, TF set, power step, etc. RLC re-transmission control Control PHY PHY UE UTRAN Figure 8.24 Interactions between RRC and lower layers [32]. Reproduced, with permission, from 3GPP TS 25.301, “Radio interface protocol architecture”, Release 4, V4.4.0. (2002 09). Ó2002 3GPP. Ó1998 3GPP. Reproduced by permission of Ó European Telecommunications Standards Institute 2008. Further use, modification, redistribution is strictly prohibited. ETSI standards are available from http://pda.etsi.org/pda/
276 Visual Media Coding and Transmission Application Relay Relay E.g. IP, E.g. IP, PDCP GTP-U GTP-U GTP-U PPP PPP GTP-U PDCP RLC RLC UDP/IP UDP/IP UDP/IP UDP/IP MAC MAC L2 L2 L2 L2 L1 L1 L1 L1 L1 L1 MS Gn Gi Uu Iu-PS 3G-GGSN UTRAN 3G-SGSN Figure 8.25 User plane UMTS protocol stack for packet switched connection [6]. Reproduced, with permission, from “Technical specification, 3rd Generation Partnership Project; technical specification group services and systems aspects; general packet radio service (GPRS); service description stage 2 (release 4)”, 3GPP TS 23.060 V4.0.0, March 2001. Ó2001 3GPP. Ó1998 3GPP. Reproduced by permission of Ó European Telecommunications Standards Institute 2008. Further use, modification, redistribution is strictly prohibited. ETSI standards are available from http://pda.etsi.org/pda/ 8.3.1.2 Channel Structure Channels are used as a means of interfacing the L2 and L1 sub-layers. Between the RLC/MAC layer and the network layer, logical channels are used. Between the RLC/MAC and the PHY layers, the transport channels are used, and below the PHY layer is the physical channel (see Figure 8.23). Generally, logical channels can be divided into control and traffic channels. The paging control channel and the broadcast control channel are for the downlink only. The common control channel is a bi-directional channel shared by all UEs, while the common transport channel is a downlink-only shared channel. Dedicated control channels and dedicated transport channels are unique for each UE. Transport channels are used to transfer the data generated at a higher layer to the physical layer, where it gets transmitted over the air interface. The transport channels are described by a set of transport channel parameters, which are designed to characterize the data transfer over the radio interface. Each transport channel is accompanied by the transport format indicator (TFI), which describes the format of data to be expected from the higher layer at each time interval. The physical layer combines the TFI from multiple transport channels to form a transport format combination indicator (TFCI). This facilitates the combination of several transport channels into a composite transport channel at the physical layer, as shown in Figure 8.26, and their correct recovery at the receiver [31]. In UTRA, two types of transport channel exist, namely dedicated channel and common channel. As the names suggest, the main difference between them is that a common channel has its resources divided between all or a group of users in a cell, while a dedicated channel reserves resources for a single user.
Wireless Channel Models 277 CH3 CH1 Information Transport TFI Data Channels TFCI Multiplexing & Physical Physical layer processing Layer Physical Control Physical Data Physical Channel Channel Channel Figure 8.26 Transport channel mapping The transmission time interval (TTI) defines the arrival period of data from higher layers to the physical layer. TTI size has been defined to be 10, 20, 40, and 80 ms. Selection of TTI size depends on the traffic characteristics. The amount of data that arrives in each TTI can vary in size, and is indicated in the transport format indicator (TFI). In the case of transport channel multiplexing, TTIs for different transport channels are time-aligned, as shown in Figure 8.27. The physical channels are defined by a specific set of radio interface parameters, such as scrambling code, spreading code, carrier frequency, and transmission power step. The channels are used to convey the actual data through the wireless link. The most important control information in a cell is carried by the primary common control physical channel (PCCPCH) and secondary common control physical channel (SCCPCH). The difference between these two is that the PCCPCH is always broadcast over the whole cell in a well-defined format, while Bit rate in TTI 40 (ms) 30 20 10 0 10 20 30 40 50 60 70 80 Transmission time (ms) Figure 8.27 Transmission time intervals (TTIs) in transport channel multiplexing
278 Visual Media Coding and Transmission the SCCPCH can be more flexible in terms of transmission diversity and format. In the uplink, the physical random access channel (PRACH) and physical common packet channel (PCPCH) are data channels shared by many users. The slotted ALOHA approach is used to grant user access in PRACH [33]. A number of small preambles precede the actual data, serving as power control and collision detection. The physical downlink shared channel (PDSCH) is shared by many users in downlink transmission. One PDSCH is allocated to a single UE within a radio frame, but if multiple PDSCHs exist they can be allocated to different UEs arbitrarily: one to many or many to one. The dedicated physical data channel (DPDCH) and dedicated physical control channel (DPCCH) together realize the dedicated channel (DCH), which is dedicated to a single user [31]. 8.3.1.3 Modes of Connection Figure 8.28 shows the possible modes of realizing the connections of the radio bearers at each layer. PDCP, RLC and MAC modes must be combined with physical-layer parameters in a way that satisfies different QoS demands on the radio bearers. However, the exact parameter setting is a choice of the implementer of the UMTS system and of the network operator. The radio bearer can be viewed as either packet switched (PS) or circuit switched (CS). A PS connection passes the PDCP, where header compression can be applied or not. The RLC offers three modes of data transfer. The transparent mode transmits higher-layer payload data units (PDUs) without adding any protocol information and is recommended for real- time conversational applications. The unacknowledged mode will not guarantee the delivery to the peer entity, but offers other services such as detection of erroneous data. The acknowledged mode guarantees the delivery through the use of automatic repeat request (ARQ) [34]. Network CS PS layer PDCP H. compress No H. compress RLC Transparent Unacknowledged Acknowledged MAC Dedicated Shared Broadcast PHY Dedicated Shared Figure 8.28 Interlayer modes of operation
Wireless Channel Models 279 The MAC layer can be operated in dedicated, shared, or broadcast mode. Dedicated mode is responsible for handling dedicated channels allocated to a UE in connected mode, while shared mode takes the responsibility of handling shared channels. The broadcast channels are transmitted using broadcast mode. The physical layer follows the MAC in choosing a dedicated or shared physical channel [35]. In UTRA, spreading is based on the orthogonal variable spreading factor (OVSF) technique. Quadrature phase shift keying (QPSK) modulation is used for downlink transmission. Both convolutional and turbo coding are supported for channel protection. The maximum possible transmission rate in downlink is 5760 kbps. It is provided by three parallel codes with a spreading factor of 4. With 1/2 rate channel coding, this could accommodate up to 2.3 Mbps user data. However, the practical maximum user data rate is subject to the amount of interference present in the system and the quality requirement of the application. 8.3.2 UMTS Physical Link Layer Model Description The physical link layer parameters and the functionality of the downlink for the FDD mode of the UMTS radio access scheme are described in this subsection. The main issues addressed are transport/physical channel structures, channel coding, spreading, modulation, transmission modeling, and channel decoding. Only the dedicated channels are considered, as the end application is real-time multimedia transmission for dedicated users. The implementation closely follows the relevant 3GPP specifications. A closed-loop fast power control method is also implemented. The developed model simulates the UMTS air interface. Figure 8.29 is a block diagram of the simulated physical link layer. It can be seen that the transmitted signal is subjected to a multipath fast-fading environment, for which the power-delay profiles are specified in [36]. In addition, an AWGN source is presented after the multipath propagation model. Co- channel interferers are not explicitly presented in the model because the loss of orthogonality of co-channels due to multipath propagation can be quantified using a parameter called the “orthogonality factor” [31], which indicates the fraction of intra-cell interfering power that is perceived by the receiver as Gaussian noise. The multipath-induced intersymbol inter- ference is implicit in the developed chip-level simulator. By changing the variance of the AWGN source, the bit error and block error characteristics can be determined for a range of carrier-to-interference (C/I) ratios and Signal-to-Noise (S/N) ratios for different physical layer configurations. The simulator only considers a static C/I and S/N profile, and no slow fading effects are implemented. However, slow fading can easily be implemented by concatenating the data sets describing the channel bit error characteristics of different, static, C/I levels. Each radio access bearer (RAB) is normally accompanied by a signaling radio bearer (SRB) [37]. Therefore, in the simulator, two dedicated transport channels are multiplexed and mapped on to a physical channel. 8.3.2.1 Channel Coding UTRA employs four channel-coding schemes, offering flexibility in the degree of protection, coding complexity, and traffic capacity available to the user. The available channel-coding
280 Visual Media Coding and Transmission Signalling Data Signalling Data Source Data Source Data CRC Attachment CRC Removal Channel Decoding Channel Coding Rate Dematching Rate Matching De-interleaving Interleaving TrCh Demux Radio Frame Format TrCh Mux AWGN De-interleaving Source Interleaving Demodulation & Despreading Modulation & Spreading RAKE Combiner Tx Filter Multipath Mobile Rx Filter Channel Figure 8.29 UMTS physical link layer model methods and code rates for dedicated channels are 1/2 rate convolutional code, 1/3 rate convolutional code, 1/3 rate turbo code, and no coding. 1/2 rate and 1/3 rate convolution coding is intended to be used with low data rates, equivalent to the data rates provided by second-generation cellular networks [31]. For high data rates, 1/3 rate turbo coding is recommended, and it typically brings performance benefits when sufficiently large input block sizes are achieved. The channel-coding schemes are defined in [38], and are outlined here. Convolutional Coding Convolutional codes with constraint length 9 and coding rates 1/3 and 1/2 are defined. Channel code block size is varied according to the data bit rates. The specified maximum code block size for convolutional coding is 504. If the number of bits in a transmission time interval (TTI) exceeds the maximum code block size then code block segmentation is performed (Figure 8.30). In order to achieve similar size code blocks after segmentation, filler bits are added to the beginning of the first block.
Search
Read the Text Version
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
- 17
- 18
- 19
- 20
- 21
- 22
- 23
- 24
- 25
- 26
- 27
- 28
- 29
- 30
- 31
- 32
- 33
- 34
- 35
- 36
- 37
- 38
- 39
- 40
- 41
- 42
- 43
- 44
- 45
- 46
- 47
- 48
- 49
- 50
- 51
- 52
- 53
- 54
- 55
- 56
- 57
- 58
- 59
- 60
- 61
- 62
- 63
- 64
- 65
- 66
- 67
- 68
- 69
- 70
- 71
- 72
- 73
- 74
- 75
- 76
- 77
- 78
- 79
- 80
- 81
- 82
- 83
- 84
- 85
- 86
- 87
- 88
- 89
- 90
- 91
- 92
- 93
- 94
- 95
- 96
- 97
- 98
- 99
- 100
- 101
- 102
- 103
- 104
- 105
- 106
- 107
- 108
- 109
- 110
- 111
- 112
- 113
- 114
- 115
- 116
- 117
- 118
- 119
- 120
- 121
- 122
- 123
- 124
- 125
- 126
- 127
- 128
- 129
- 130
- 131
- 132
- 133
- 134
- 135
- 136
- 137
- 138
- 139
- 140
- 141
- 142
- 143
- 144
- 145
- 146
- 147
- 148
- 149
- 150
- 151
- 152
- 153
- 154
- 155
- 156
- 157
- 158
- 159
- 160
- 161
- 162
- 163
- 164
- 165
- 166
- 167
- 168
- 169
- 170
- 171
- 172
- 173
- 174
- 175
- 176
- 177
- 178
- 179
- 180
- 181
- 182
- 183
- 184
- 185
- 186
- 187
- 188
- 189
- 190
- 191
- 192
- 193
- 194
- 195
- 196
- 197
- 198
- 199
- 200
- 201
- 202
- 203
- 204
- 205
- 206
- 207
- 208
- 209
- 210
- 211
- 212
- 213
- 214
- 215
- 216
- 217
- 218
- 219
- 220
- 221
- 222
- 223
- 224
- 225
- 226
- 227
- 228
- 229
- 230
- 231
- 232
- 233
- 234
- 235
- 236
- 237
- 238
- 239
- 240
- 241
- 242
- 243
- 244
- 245
- 246
- 247
- 248
- 249
- 250
- 251
- 252
- 253
- 254
- 255
- 256
- 257
- 258
- 259
- 260
- 261
- 262
- 263
- 264
- 265
- 266
- 267
- 268
- 269
- 270
- 271
- 272
- 273
- 274
- 275
- 276
- 277
- 278
- 279
- 280
- 281
- 282
- 283
- 284
- 285
- 286
- 287
- 288
- 289
- 290
- 291
- 292
- 293
- 294
- 295
- 296
- 297
- 298
- 299
- 300
- 301
- 302
- 303
- 304
- 305
- 306
- 307
- 308
- 309
- 310
- 311
- 312
- 313
- 314
- 315
- 316
- 317
- 318
- 319
- 320
- 321
- 322
- 323
- 324
- 325
- 326
- 327
- 328
- 329
- 330
- 331
- 332
- 333
- 334
- 335
- 336
- 337
- 338
- 339
- 340
- 341
- 342
- 343
- 344
- 345
- 346
- 347
- 348
- 349
- 350
- 351
- 352
- 353
- 354
- 355
- 356
- 357
- 358
- 359
- 360
- 361
- 362
- 363
- 364
- 365
- 366
- 367
- 368
- 369
- 370
- 371
- 372
- 373
- 374
- 375
- 376
- 377
- 378
- 379
- 380
- 381
- 382
- 383
- 384
- 385
- 386
- 387
- 388
- 389
- 390
- 391
- 392
- 393
- 394
- 395
- 396
- 397
- 398
- 399
- 400
- 401
- 402
- 403
- 404
- 405
- 406
- 407
- 408
- 409
- 410
- 411
- 412
- 413
- 414
- 415
- 416
- 417
- 418
- 419
- 420
- 421
- 422
- 423
- 424
- 425
- 426
- 427
- 428
- 429
- 430
- 431
- 432
- 433
- 434
- 435
- 436
- 437
- 438
- 439
- 440
- 441
- 442
- 443
- 444
- 445
- 446
- 447
- 448
- 449
- 450
- 451
- 452
- 453
- 454
- 455
- 456
- 457
- 458
- 459
- 460
- 461
- 462
- 463
- 464
- 465
- 466
- 467
- 468
- 469
- 470
- 471
- 472
- 473
- 474
- 475
- 476
- 477
- 478
- 479
- 480
- 481
- 482
- 483
- 484
- 485
- 486
- 487
- 488
- 489
- 490
- 491
- 492
- 493
- 494
- 495
- 496
- 497
- 498
- 499
- 500
- 501
- 502
- 503
- 504
- 505
- 506
- 507
- 508
- 509
- 510
- 511
- 512
- 513
- 514
- 515
- 516
- 517
- 518
- 519
- 520
- 521
- 522
- 523
- 524
- 525
- 526
- 527
- 528
- 529
- 530
- 531
- 532
- 533
- 534
- 535
- 536
- 537
- 538
- 539
- 540
- 541
- 542
- 543
- 544
- 545
- 546
- 547
- 548
- 549
- 550
- 551
- 552
- 553
- 554
- 555
- 556
- 557
- 558
- 559
- 560
- 561
- 562
- 563
- 564
- 565
- 566
- 567
- 568
- 569
- 570
- 571
- 572
- 573
- 574
- 575
- 576
- 577
- 578
- 579
- 580
- 581
- 582
- 583
- 584
- 585
- 586
- 587
- 1 - 50
- 51 - 100
- 101 - 150
- 151 - 200
- 201 - 250
- 251 - 300
- 301 - 350
- 351 - 400
- 401 - 450
- 451 - 500
- 501 - 550
- 551 - 587
Pages: