 | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
4. Technical IssuesThere are number of issues in transporting video over wireless network. Actual data rate is much lower than the data rate at the air interface due to the overhead to deal with high error rates caused by various interference. Signal quality varies widely depending on where a receiver is located in a cell. Synchronization between audio and video is an issue when a receiver receives media from multiple different sources. Video compression methods are commonly used in transporting video over wireless network. The artifacts caused by communications errors are very visible in images processed by an inter-frame motion compensated technique. In this section, we will discuss issues in transporting video over wireless network. We will also introduce technologies for expanding bandwidth in HSDPA and EV-DO and methods of dealing with issues in transporting video over wireless networks. 4.1 Actual Throughput versus DistanceData rates advertised by providers are often misleading since wireless service providers typically specify theoretical peak data rates. Since many users share the bandwidth in a cell in the cellular network, the actual data throughput varies tremendously depending on where a subscriber is located in a cell of a cellular network. As is shown in Figure 4.1, as the coverage is extended, the downlink capacity decreases proportionally. The theoretical data rate of HSDPA drops from 14,400 Kbps at the tower to around 200Kpbs at the boundary of a cell. A cell range varies depends on the terrain. A typical cell range is about 300 meters to 750 meters in diameter in an urban area. In a rural area where cells are less congested, a typical cell range is three to five kilometers and it may extend up to several kilometers depending on a terrain.
4.2. Uplink Data RatesOne-way video services such as VOD and Live TV do not require a high bandwidth in the uplink since the data from a user to the head-end is minimal. Two-way real-time video service such as video telephony requires the same bandwidth for both directions. Users noticed that sending photos from a mobile device to other mobile users took long time in 3G networks since the uplink speed is much slower than the downlink speed and the time to transmit a photo is determined by the uplink speed. Currently no video telephony service is provided via the shared data channel like HSUPA or EV-DO because the shared channels do not offer a guaranteed QoS yet. Video telephony services in Korea Japan use a dedicated 64 Kbps CSD service, which provides the same up and downlink data rates. It is required to implement QoS schemes to provide real-time multimedia services such as video telephony and VoIP in HSDPA and EV-DO networks. The uplink speed of HSDPA is 384 Kbps even for Category 10 devices with a peak downlink rate of 14.4 Mbps. 3GPP Rel-6 introduced HSUPA (High Speed Up Packet Access), which provides higher uplink speed. Rel-6 defines six category devices with uplink speed ranging from 0.73 Mbps (Category 1) to 5.76 Mbps (Category 6). Unlike HSDPA using QPAK and 16 QAM, HSUPA's modulation technology is QPAK for all categories to minimize the power consumption by using lower modulation technology. Many carriers throughout the world has equipment ready for HSUPA. Some carriers will offer HSUPA services starting from 2007 but most will wait until mobile devices are ready and the market interest for video telephony is higher. The uplink speed of EV-DO Rev. 0 is 1.8 Mbps compared to downlink speed of 3.1 Mbps with a single channel. EV-DO allows aggregation of up to 15 channels, which can offer higher uplink bandwidths. 4.3. InterferencesOne of the major issues with wireless networks is interference. Sources of interference may be noises produced by automobiles, other radio devices, inter-user interference, multi-path interference, and inter-cell interference. To overcome interference problems, various technologies have been introduced at the physical layer (radio layer) and MAC layer. In this section, we will discuss problems and technologies used to overcome the interference problems 4.3.1. Forward Error Correction Adaptive Modulation and Coding (AMC) and Hybrid Automatic Repeat Request (HARQ)HSDPA and EV-DO uses Forward Error Correction (FEC). The two main categories of FEC are block coding and convolutional coding. Block codes work on fixed-size packets of bits of predetermined size and convolutional codes work on bit streams of arbitrary length. The most recent development in error correction is turbo coding, which were presented to the coding community in early 1990 for deep space and satellite channels. Turbo coding is a scheme that combines two or more convolutional codes and an interleaver to produce a block code. HSDPA uses block codes and EV-DO has option of using either turbo coding or convolutional coding. HSDPA uses Adaptive Modulation and Coding (AMC) and Hybrid Automatic Repeat Request (HARQ) for correcting errors in the air interface. Traditionally, WCDMA has used fast power control for link adaptation, but HSDPA holds the transmission power constant and uses AMC as an alternative link adaptation method to power control in order to improve the spectral efficiency. In AMC, the transport format including the modulation scheme and turbo code rate can be selected based on the downlink channel quality. Mobile devices provide the channel quality feedback to the transmission station by Channel Quality Indicator (CQI). The code rate can change between 1/4 and 3/4. If the code rate is k/n, for every k bits of useful information, the coder generates totally n bits of data, of which n-k are redundant. A lower code rate is used to handle higher error rates. For instance, a coding rate of ¼ means that error correction takes 75 percent of the bandwidth and the user data uses only 25 percent of the peak rate. A coding rate 4/4 means that there is no error correction. In theory, the protocol allows an un-coded link of 4/4 but that is not useful for real networks. The lab tests achieved the theoretical maximum of 14.4 Mbps using 16 QAM modulation with code rate 4/4. In a system with AMC, users close to the base transmission station (BTS or called Node-B) are typically assigned higher order modulation with higher code rates (e.g. 16 QAM with a 3/4 code rate), and the modulation-order and/or code rate generally decreases as the distance to the Node-B increases. Table 4.1 illustrates examples of throughput in HSDPA depending on code rates and the modulation method. Table 4.1: Adaptive Modulation and Coding Rates for HSDPA
The "Hybrid" in HARQ presents the method of combining FEC (Forward Error Correction) and ARQ (Automatic Repeat reQuest) by encoding the data block plus error-detection information (such as CRC) with an error-correction code (such as Turbo code). HSDPA adds HARQ function at BTS (Base Transceiver Station, commonly denoted as Node B in UMTS) to increase the transmission rate and reduce the time delay. In case of ARQ, the mobile device checks the CRC. If the CRC is the same as that received in the message ACK is sent back to the sender. In case if CRC does not match then NACK is sent back to request a retransmission. HARQ functionality is implemented at the MAC-hs layer at the BTS. Therefore, the retransmission delay is significantly shorter than using the higher layer. This method can use soft combining, which combines the information of the original transmission with the retransmitted information. This way the network does not re-request re-transmission of all the data. CDMA2000 provides the option of using either turbo coding or convolutional coding on the forward and reverse SCHs. Both coding schemes are optional for the base station and the mobile station, and the capability of each is communicated through signaling messages prior to the set up of the call. In addition to peak rate increase and improved rate granularity, the major improvement to the traffic channel coding in CDMA2000 is the support of turbo coding at rate 1/2, 1/3, or 1/4. The turbo code is based on 1/8 state parallel structure and can only be used for supplemental channels and frames with more than 360 bits. 4.3.2. MIMO (Multiple Input, Multiple Output)When a radio wave travels from a transmitter to a receiver, it takes many different paths and arrives at different time intervals. When the wave traveled in multiple paths, it arrives at the receiver out of phase with each other and creates fading, cut-out, and intermittent reception, which cause communication errors and lower the data rate. To eliminate the problem caused by multi-path wave propagation and even take advantage of this effect, MIMO technology was introduced. MIMO uses two or more antennas along with the transmission of multiple signals at the transmitter and the receiver. MIMO technology can be spatial multiplexing for enhancing the peak data transmission rate or beam-forming for improving received signal gain and reducing interference to other users. 3GPP is considering two MIMO proposals, Double Transmit Adaptive Arrays (DTxAA) and Per Antenna Rate Control (PARC), for HSDPA Release 7. In DTxAA, a partial beam-forming approach is used. Four transmit antennas are employed at the base station, the transmit antennas are divided into two sub-groups, and each sub-group transmits independent data steams with a TxAA operation of a pair of transmit antennas. PARC is a non beam-forming MIMO. Separately encoded data streams are transmitted from each antenna. 4.3.3. OFDM (Orthogonal Frequency Division Multiplexing)Most radio noises have narrow spectrum. Utilizing multiple small channels can confine the impact of noises within small number of sub-channels and maximize the capability of channels, which are not affected by the noises. OFDM also known as multi-carrier or Discrete Multi-tone Modulation (DMT) is based upon the principle of frequency-division multiplexing (FDM). It is the modulation technique used for DSL. The basic idea of multi-tone is to split the available bandwidth into a number of sub-channels. Each sub-channel carries a low data rate and the receiver combines data coming from each sub-channel. Using multiple sub-channels and aggregating data rates has advantages over sending data over a wide frequency band by confining errors within sub-channels . The sub-carrier frequencies in OFDM are chosen such that the cross-talk between sub-channels is eliminated. MIMO and OFDM are mostly used together. 3GPP Rel. 7 will use MIMO with WCDMA and LTE will employ MIMO with OFDMA. EV-DO Rev. C is expected to use CDMA/OFDM or a combination of OFDMA; MIMO/SDMA. OFDM and MIMO-OFDM are already key components in the wireless LAN (802.11a, 802.11g and 802.11n) and WiMAX. 4.4. OverheadThe published date rates are typically data rates over the air interface even without any error correction bits. The majority of HSDPA devices launched to date are Category12 devices, which support a maximum of 5 HS-DSCH codes and QPSK modulation. The maximum speed of Category12 devices over the air interface is 1.8Mbps. After removing radio-protocol overhead, the peak rate for MAC layer is 1.6Mbps. Since the HSDPA systems target a radio block error rate up to 10%, the net radio bandwidth available to MAC layer is 1.5Mbps. A HSDPA Category 6 device based on 16 QAM has a maximum speed of 3.6Mbps over the air interface and the MAC layer bandwidth is 3.1Mbps.
The peak data rate for applications decreases further from the MAC layer in IP network due to the overhead caused by packet headers. Figure 4.2 illustrates a multimedia packet such as a VoIP packet. The packet header is 40 bytes, which are composed of 20 bytes of IP v4 header, 8 bytes of UDP header and 12 bytes of RTP header. The IP v6 is gaining wide acceptance in cellular networks since it has to support a large number of mobile devices. The packet overhead with IP v6 header is 60 bytes, which is larger than a typical voice payload in the range of 20 to 60 bytes. To minimize the overhead, a header compression has been included in HSDPA Rel-5. A header compression can achieve 90% to 97.5% compression gain. A 60 bytes packet header composed of IPv6/UDP/RTP may be compressed to 3 bytes. An application with 40 bytes payload can utilized 2.88 Mbps (93% of 3.1 Mbps MAC layer bandwidth) in a HSDPA network with Category 6 device (Peak rate 3.6 Mbps). In a HSDPA network with Category 6 device, an effective bandwidth for applications is 80% of the peak data rate over the air interface for voice packets. Obviously utilization increases for larger payloads such as video packets. IETF formed "Robust Header Compression (ROHC)" Working Group to develop new header compression protocols, which were to take into account typical needs presented by various wireless link technologies, and perform well for cellular links built using technologies such as WCDMA, EDGE, and CDMA-2000. The WG has specified a number of RFCs for header compression. 4.5. QoS (Quality of Service)Circuit switching networks like the PSTN network allocate a bandwidth for each call regardless whether it is used or not. Shared networks based on statistical multiplexing like IP networks and wireless networks need to provide preferential treatment for real-time traffics like audio and video from the time insensitive data traffics. QoS is a commonly used term for describing various methods to provide preferential treatment for real-time traffics and higher priority traffics. UMTS defined four different QoS classes * Conversational class, * Streaming class, * Interactive class, and * Background class. The main distinguishing factor between these QoS classes is how delay sensitive the traffic is. Conversational class is meant for delay sensitive traffics while Background class is the most delay insensitive traffic class. Conversational and Streaming classes are intended to be used to carry real-time traffic flows. The main divider between them is how delay sensitive the traffic is. Video telephony and VoIP services are the most delay sensitive applications, which should be carried in Conversational class. VOD services should be carried in Streaming class. Interactive class and Background class are meant to be used for web browsing, email, file transfer, etc. Due to looser delay requirements unlike Conversational and Streaming classes, both classes take advantages of retransmission and higher code rates to achieve lower error rates. The main difference between Interactive and Background class is that Interactive class is mainly used by interactive applications, e.g. text chat or interactive Web browsing, while Background class is meant for background traffic, e.g. background download of emails or files. Traffic in the Interactive class has higher priority in scheduling than Background class traffic, so background applications use the bandwidths only when interactive applications do not need them. Table 4.2 summarizes the defined UMTS bearer attributes and their relevancy for each bearer traffic class. Note that traffic class is an attribute itself. Table 4.2: Radio Access Bearer attributes defined for each bearer traffic class
EV-DO Rev A utilizes a flow-based QoS to manage different types of traffics. Flow-based QoS prioritizes applications and flows within applications based on the sensitivity to delay. A flow is defined by several parameters including its sensitivity to delay, acceptable packet error rate, and data rate changes. For instance, the voice application would use a delay sensitive flow and would get priority over email or web access. The video telephony application would use a delay sensitive flow for voice and a rate sensitive flow to maintain acceptable quality video. The video telephony is an example of prioritizing flows within applications for better control, where the audio stream in a video telephony can be given a higher priority than the accompanying video streams. EV-DO Rev. A also permits the use of higher power in mobile devices for higher priority packets in order to reduce the number of re-transmission necessary to successfully send these packets on the uplink. 4.6. Broadcast and Multicast ServicesAt present mobile TV services are provided by unicast where each user establishes a separate connection with a server. As a consequence, a content server has to establish and manage a separate connection with each user and each user uses a bandwidth to receive the same contents shared with others. To achieve more efficient use of air interface and network resources when sending the same information to multiple users in wireless networks, 3GPP and 3GPP2 began addressing broadcast and multicast services. Broadcast service is considered un-addressed messaging such as weather or traffic alerts to mobile devices in a cell or region. Multicast service is defined by the ability to send information to a specific set of users based on their subscription and is considered addressed messaging. 3GPP called broadcast and multicast services, Multimedia Broadcast and Multicast Service (MBMS) and 3GPP2 named them, Broadcast and Multicast Service (BCMCS). Open Mobile Alliance (OMA) BCAST is working on the specification of broadcast and multicast related service layer functions such as content protection, service and program guides, and transmission scheduling. In 3GPP, Technical Specification TS 22.146 defines MBMS User Services including service related information, requirement in terms of data rates, quality of service requirements, typical volumes of data and some guidance on applications services and bit rates. MBMS User Services are classified into three types as follows: * Streamingservices: A service that provides a stream of continuous media * File downloadservices: A service that delivers binary data (file data) over an MBMS bearer * Carousel services:A service that combines aspects of both the Streaming and File download services Table 4.3 shows examples of use cases of MBMS in 3GPP. The specification says that MBMS content providers shall be able to invoke DRM to prevent unauthorized copying and forwarding of content. The capabilities (e.g. memory size) required to receive a particular transmission shall be notified in advance by the network or service center. Table 4.3: Examples of use cases on MBMS in 3GPP.
The first version of BCMCS in EV-DO is Gold Multicast and is supported by EV-DO Rev 0. Gold Multicast supports a physical data rate of 614 Kbps. The next version of BCMCS, Platinum Multicast, is supported by EV-DO Rev A and triples the physical data rate up to 1.8Mbps. The broadcast-multicast channel can carry downlink signaling messages but does not have uplink messages. Downlink messages are sent for transmission to signaling application and lower layer protocols negotiated when the session starts. 4.7. Sound/Video Synchronization>In media processing, sounds signals and video signals are processed separately by different coding and decoding methods as shown in Figure 7 by sequentially by one processor or in parallel by multiple processors. Video processing is more computationally intensive than sound processing. As a result, sound has to be delayed until the video is available even when processed in parallel with video. Without the synchronization, what viewers hear will not match with what they see on the screen. For instance, the timing of lips moving is off from the words being heard. Audio/video synchronization is often known as "Lip Sync" A 20 ms sound advance and a 40 ms delay with respect to picture are detectable in speech and a sound advance of more than 40 ms or a sound delay of more than 160 ms is annoying. Various methods have been introduced to handle lip sync problems. A simple method may add a fixed amount of delay to sound but it does not work well in some cases. It is because the compression delay in video coding varies widely depending on amount of changes between frames in encoding methods based on motion compensated frame difference such as MPEG/H.264 and H.263. The amount of delay between sound and video also varies depending on the capability of a processor. The problem may exacerbate when the processor in a transmitting device is different from the one in the receiving device or processes by one processor sequentially.
One of the methods used to synchronize sounds and video is defined in RTP (A transport protocol for real time applications) specified by IETF RFC 3550. The RTP header contains 32 bits of timestamp, which reflects the sampling instance of the RTP packet. The timestamp is used to place the audio and video packets in the correct timing order to provide lip sync. RTP timestamps work well when compression and decompression steps are not involved in transporting signals. In video encoding and decoding engines, RTP timestamps merely indicate the sampling instance of RTP packets after the compression. The encoding and decoding engines may have to add appropriate delays to minimize the differences caused in the video compression/decompression and the audio compression/decompression as shown in Figure 4.3. Digital watermarking may be used for lip sync. Digital watermarking technology uses an arrangement of digital bits as patterns to encode identification information. The ID information is imperceptible to viewers when it is embedded within the video. A reference code called signature is derived from the audio envelope. Then the signature is embedded as a watermark into the video to indicate a reference point. The embedded reference point is not visible to viewers and can be extracted and used for correcting lip sync errors. It was demonstrated that watermarking withstood compression and de-compression in MPEG and MJPEG in filed trials. 4.8. Power ManagementWith the convergence of communication, entertainment, and computing on mobile devices, power demands are ever increasing and yet the batteries cannot keep up with the demands. Manufacturers of mobile devices are challenged to reduce power consumption while enhancing the performance of the device. Various techniques are used to minimize the use of power and maximize performance including low power processors, variable-speed clocks, automatic standby or shutoff, low voltage circuits, software controlled power management, power-management ICs, efficient display technology, and time sliced reception and so on. In this section, we will describe two notable technologies - the time-sliced reception in DVB-H and the RGBW LCD display. 4.8.1. Time Slicing in DVB-HThe broadcast television standard, DVB-H, uses a power-saving algorithm based on the time-multiplexed transmission of different services. Time slicing consists of sending data in bursts at high bit rates and indicating to the receiver when to expect the time to the beginning of the next burst within the burst. Time slicing enables a receiver to stay active only a fraction of the time. Figure 4.4 shows a cut-off of a data stream containing time-sliced services. One quarter of the total capacity of the DVB-T channels is assigned to DVB-H services shared among users. A mobile device receiving DVB-H Service 1 receives and stores the burst of data in a buffer and reads out the buffer to play the contents. Practically the duration of one burst is in the range of 1/10 of each cycle, which results in over 90% battery power saving effect. |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
