IEEE TRANSACTIONS ON BROADCASTING, VOL. 46, NO. 2, JUNE 2000 101 Comparison of Terrestrial DTV Transmission Systems: The ATSC 8-VSB, the DVB-T COFDM, and the ISDB-T BST-OFDM Yiyan Wu, Ewa Pliszka, Bernard Caron, Pierre Bouchard, and Gerald Chouinard Abstract—this paper compares the performances of the ATSC 8-VSB, the DVB-T COFDM, and the ISDB-T BST-OFDM digital television terrestrial transmission systems under different impairments and operating conditions. First, a general system level description is presented. It is followed by comparisons based on laboratory test results and theoretical analyzes. The differences in the system threshold definitions are discussed. In addition, a brief performance and implementation analysis is also presented for the three transmission systems under different network infrastructures. Whenever possible, the impact on the broadcasters or consumers is discussed. Possible performance improvements are also identified. Index Terms—COFDM, data broadcasting, DTV, DTV terrestrial broadcasting, mobile reception, SFN, VSB. I. INTRODUCTION A FTER a decade of intense research and development, Digital Television Terrestrial Broadcasting (DTTB) has finally reached the implementation stage. DTTB services have been available in North America and Europe since November 1998. Many countries have announced their choice for a DTTB system and their implementation plan. Currently, there are three DTTB transmission standards: 1) The Trellis-Coded 8-Level Vestigial Side-Band (8-VSB) modulation system developed by the Advanced Television Systems Committee (ATSC) in the USA [1]; 2) The Coded Orthogonal Frequency Division Multiplexing (COFDM) modulation system adopted as the Digital Video Broadcasting—Terrestrial (DVB-T) standard in Europe [2]; and 3) The Band Segmented Transmission (BST)-OFDM system adopted in Japan for Terrestrial Integrated Service Digital Broadcasting (ISDB-T) [3]. Since there are more than one DTTB approaches, many countries and administrations are now engaged in the process of selecting a DTTB system. Each country has specific characteristics and needs. The selection of a DTTB system must be based upon how well each modulation system meets specific conditions such as the use of the spectrum resource, coverage requirements and transmission network structure, reception conditions, type of service required, policy, objectives for program exchange, cost to the consumers and broadcasters, etc. Manuscript received May 10, 2000; revised July 7, 2000. Y. Wu, B. Caron, P. Bouchard, and G. Chouinard are with the Communications Research Centre Canada, 3701 Carling Avenue, Ottawa, Ontario, Canada K2H 8S2 (e-mail: yiyan.wu@crc.ca). E. Pliszka is with the Telekomunikacja Polska S.A. Research and Development Centre, Obrzezna 7, 02-691 Warszawa, Poland (e-mail: epliszka@cbr.tpsa.pl). Publisher Item Identifier S 0018-9316(00)08224-X. This paper compares the performances of the ATSC 8-VSB, the DVB-T COFDM, and the ISDB-T BST-OFDM transmission systems under different impairments and operating conditions. First, a general system level description is presented. It is followed by comparisons based on laboratory test results and results from theoretical analyzes. The differences in the system threshold definitions are discussed. A calculated performance comparison of the three transmission systems for a 6 MHz channel is provided. The 7 and 8 MHz systems should yield the same performance, except for higher bit-rate capacity, since identical modulation and channel coding schemes are used. In addition, a brief performance and implementation analysis is also presented for the three transmission systems under different transmission network infrastructures. Whenever possible, the impact on the broadcasters or consumers is discussed. Possible performance improvements are also identified. It should be pointed out that the performance benchmarks quoted in this paper are representative of present technologies. Meanwhile, the tests have been conducted in different laboratories, under different test environments and using receivers from different manufacturers over more than one generation of products. This might result in some discrepancies due to actual equipment implementation. On the other hand, with the technical advances, all systems will improve in performance up to close to their theoretical limits. II. SYSTEM DESCRIPTIONS A. General System Descriptions The main characteristics of the three DTTB systems are listed in Table I. 1) The ATSC 8-VSB System: The ATSC Digital Television Standard was developed by the Advanced Television Systems Committee in the USA [1]. The ATSC system was designed to transmit high-quality video and audio (HDTV) and ancillary data over a single 6 MHz channel. The system was developed for terrestrial broadcasting and for cable distribution. It can reliably deliver 19.4 Mbit/s of data throughput in a 6 MHz terrestrial channel and 38.8 Mbit/s in a 6 MHz cable television channel. Two modes of operation are available: the 8-VSB “simulcast terrestrial mode” intended to be more immune to the NTSC interference, and the 16-VSB “high data rate mode” primarily developed for the cleaner—compared to terrestrial—cable channels. Although the system was developed and tested with 6 MHz channels, it can be scaled to any channel bandwidths (6, 7, or 8 MHz) with corresponding scaling in the data capacity. For terrestrial broadcasting, the system was designed to allow the allocation of an additional digital transmitter for 0018–9316/00$10.00 © 2000 IEEE 102 IEEE TRANSACTIONS ON BROADCASTING, VOL. 46, NO. 2, JUNE 2000 TABLE I MAIN CHARACTERISTICS OF THREE DTTB SYSTEMS each existing NTSC transmitter with comparable coverage, and minimum disturbance to the existing NTSC service in terms of both area and population coverage. This capability is met and even exceeded as the RF transmission characteristics of the system are carefully chosen to cope with an NTSC environment. Various picture qualities can be achieved with 18 video formats (SD or HD, progressive or interlaced, as well as different frame rates). There is a great potential for data-based services utilizing the opportunistic data transmission capability of the system. The system can accommodate fixed (and possibly portable) reception. The system is quite efficient and capable of operating under various conditions, i.e. clear channel availability or, as implemented in the US, constrained to fit 1600 additional channel allocations into an already crowded spectrum, and reception with roof-top or portable antennae. The system is designed to withstand many types of interference: existing analog NTSC TV services, white noise, impulse noise, phase noise, continuous wave and passive reflections (multipath). The system is also designed to offer spectrum efficiency and ease of frequency planning. The system uses a single carrier modulation scheme, eightlevel Vestigial-SideBand (8-VSB) modulation. It is designed for single transmitter (Multi-Frequency Network, MFN) implementation. However, limited on-channel repeater and gap-filler operation are possible. The main characteristics of the ATSC 8-VSB system are listed in Table I. 2) DVB-T COFDM System: The DVB-T system was developed by an European consortium of public and private sector organizations—the Digital Video Broadcasting Project [2]. The DVB-T specification is part of a family of specifications also covering satellite (DVB-S) and cable (DVB-C) operations. This family allows for digital video and digital audio distribution as well as transport of forthcoming multimedia services. For terrestrial broadcasting, the system was designed to operate within the existing UHF spectrum allocated to analogue PAL and SECAM television transmissions. Although the system was developed for 8 MHz channels, it can be scaled to any channel bandwidth (8, 7, or 6 MHz) with corresponding scaling in the data capacity. The net bit rate available in 8 MHz channel ranges between 4.98–31.67 Mbit/s, depending on the choice of channel coding parameters, modulation types, and guard interval duration. The system was essentially designed with built-in flexibility, in order to be able to adapt to all types of channel. It is capable of coping not only with Gaussian channels, but also with Ricean and Rayleigh channels. It can withstand high-level (up to 0 dB) long delay static and dynamic multipath distortion. The system is robust to interference from delayed signals, either echoes resulting from terrain or building reflections, or signals from distant transmitters in a single frequency network (SFN) arrangement. The system features a number of selectable parameters that accommodate a large range of carrier-to-noise ratios and channel behaviors. It allows fixed, portable, or mobile reception, with a consequential trade-off in the usable bit rate. This range of parameters allows the broadcasters to select a mode appropriate to the application foreseen. For instance, a moderately robust mode (with a correspondingly lower data rate) is needed to ensure reliable portable reception with a simple set-top antenna. A less robust mode with a higher WU et al.: COMPARISON OF TERRESTRIAL DTV TRANSMISSION SYSTEMS data rate could be used where the service planning uses frequency-interleaved channels [2], [12]. The less robust modes with the highest payloads can be used for fixed reception and if a clear channel is available for digital TV broadcasting. The system uses a large number of carriers per channel modulated in parallel via an FFT process, a method referred to as Orthogonal Frequency Division Multiplexing (OFDM). It has two operational modes: a “2k mode” which uses a 2k FFT; and an “8k mode” which requires an 8k FFT. The system makes provisions for selection between different levels of QAM modulation and different inner code rates and also allows two-level hierarchical channel coding and modulation. Moreover, a guard interval with selectable width separates the transmitted symbols, which allows the system to support different network configurations, such as large area SFN’s and single transmitter operation. The “2k mode” is suitable for single transmitter operation and for small SFN networks with limited distance between transmitters. The “8k mode” can be used both for single transmitter operation and for small and large SFN networks. The main characteristics of the DVB-T COFDM system are listed in Table I. 3) ISDB-T BST-OFDM: The ISDB-T system was developed by the Association of Radio Industries and Businesses (ARIB) in Japan [3]. ISDB (Integrated Services Digital Broadcasting) is a new type of broadcasting intended to provide audio, video and multimedia services. The system was developed for terrestrial (ISDB-T) and satellite (ISDB-S) broadcasting. It systematically integrates various kinds of digital contents, each of which may include multi-program video from Low Definition TV (LDTV) to HDTV, multi-program audio, graphics, text, etc. Since the concept of ISDB covers a variety of services, the system has to meet a wide range of requirements that may differ from one service to another. For example, a large transmission capacity is required for HDTV service, while a high service availability (or transmission reliability) is required for data services such as the delivery of a “key” for conditional access, downloading of software, and so on. To integrate different service requirements, the transmission system provides a range of modulation and error protection schemes which can be selected and combined flexibly in order to meet each requirement of these integrated services. For terrestrial broadcasting, the system has been designed to have enough flexibility to deliver digital television and sound programs and offer multimedia services in which various types of digital information such as video, audio, text and computer programs will be integrated. It also aims at providing stable reception through compact, light and inexpensive mobile receivers in addition to integrated receivers typically used in homes. The system uses a modulation method referred to as Band Segmented Transmission (BST) OFDM, which consists of a set of common basic frequency blocks called BST-Segments. Each segment has a bandwidth corresponding to 1/14th of the terrestrial television channel spacing (6, 7, or 8 MHz depending on the region). For example, in a 6 MHz channel, one segment 103 occupies 6/14 MHz = 428.6 kHz spectrum, seven segments ocMHz = 3 MHz. cupy In addition to the properties of OFDM reviewed in the previous section, BST-OFDM provides hierarchical transmission capabilities by using different carrier modulation schemes and coding rates of the inner code on different BST-segments. Each data segment can have its own error protection scheme (coding rates of inner code, depth of the time interleaving) and type of modulation (QPSK, DQPSK, 16-QAM or 64 QAM). Each segment can then meet different service requirements. A number of segments may be combined flexibly to provide a wideband service (e.g., HDTV). By transmitting OFDM segment groups with different transmission parameters, hierarchical transmission is achieved. Up to three service layers (three different segment groups) can be provided in one terrestrial channel. Partial reception of services contained in the transmission channel can be obtained using a narrow-band receiver that has a bandwidth as low as one OFDM segment. Thirteen OFDM spectrum segments are active within one terrestrial television channel. The useful bandwidth is BW , corresponding to 5.57 MHz for a BW MHz channel, 6.50 MHz for a 7 MHz channel, and 7.43 MHz for an 8 MHz channel. The system was developed and tested with 6 MHz channels but it can be scaled to any channel bandwidth with corresponding variations in the data capacity. The net bit rate for one 428.6 kHz segment in a 6 MHz channel ranges between 280.85–1787.28 kbit/s. The data throughput for a 5.57 MHz DTV channel ranges between 3.65–23.23 Mbit/s. The main characteristics of ISDB-T BST-OFDM system are listed in Table I. B. System Performance Summary Generally speaking, each system has its own unique advantages and disadvantages. The ATSC 8-VSB [1] system is more robust in an Additive White Gaussian Noise (AWGN) channel, has a higher spectrum efficiency, a lower peak-to-average power ratio, and is more robust to impulse noise. It also has comparable performance to DVB-T and ISDB-T systems on low-level ghost ensembles and against analog TV interference. Therefore, the ATSC 8-VSB system could be more advantageous for Multi-Frequency Network (MFN) implementation and for providing HDTV service within a 6 MHz channel to fixed receivers. The DVB-T COFDM [2] system has performance advantages with respect to high-level (up to 0 dB), long-delay static and dynamic multipath distortion. It could be advantageous for services requiring large-scale Single Frequency Network (SFN) (8k mode) and for mobile reception (2k mode). Hierarchical channel coding and modulation, which uses multi-resolution constellation on OFDM carriers (16QAM or 64QAM), is also available to provide two-tier services within one DTTB channel. The ISDB-T BST-OFDM [3] system, which uses the same modulation and channel coding scheme as the DVB-T system, has similar performance advantages as the DVB-T system. It was designed to operate under large-scale SFN and, particularly, in a mobile reception environment. The depth of the time interleaver can optionally be selected to improve the quality of 104 IEEE TRANSACTIONS ON BROADCASTING, VOL. 46, NO. 2, JUNE 2000 mobile reception and the immunity against impulse noise. The band segmentation transmission allows the use of up to three different modulation schemes and coding rates on different channel segments to meet various service requirements and interference conditions. However, it should be pointed out that large scale SFN, mobile reception, and HDTV service cannot be achieved concurrently without severe data rate or C/N penalties for all existing DTTB systems over any channel spacing, whether 6, 7, or 8 MHz. Specific system parameters should be selected for particular implementations. III. SYSTEM PERFORMANCE COMPARISON A. DTTB Signal Peak-to-Average Power Ratios The COFDM signal can be statistically modeled as a twodimensional Gaussian process [4]. Its Peak-to-Average power Ratio (PAR) is somewhat independent of the filtering. On the other hand, the 8-VSB PAR is largely set by the roll-off factor of the spectrum shaping filter, i.e., 11.5% for the ATSC 8-VSB signal. Studies show that the PAR of the DVB-T and ISDB-T signals, for 99.99% of the time, is about 2.2 dB higher than that for the 8-VSB system [4]–[6]. For 99.9% of the time, the PAR of COFDM signal is about 2 dB higher. In order to achieve a 35 or 36 dB spectrum attenuation at the channel edges [13], a 6 dB output back-off (OBO) is required for ATSC 8-VSB signals, whereas for COFDM, the OBO should be 7.5 to 8 dB. For the same level of adjacent channel spill-over, which is the major source of adjacent channel interference, the DVB-T and ISDB-T systems require a transmitter that can handle a higher peak power to accommodate the 2 dB additional output power back-off, or a better channel filter with additional side-lobe attenuation. However, the high PAR has no impact on system performance. It just increases the start-up investment cost and operational power consumption cost for the broadcasters. In some cases, however, the PAR may have an impact on the transmitting antenna and transmission line, depending on their respective peak power-handling capabilities. B. Thermal Noise 1) C/N Thresholds: Theoretically, from a modulation point of view, the OFDM and single carrier modulation schemes, such as VSB and QAM, should have the same C/N threshold over Additive White Gaussian Noise (AWGN). It is the channel coding, channel estimation, and equalization schemes, as well as other implementation margins (phase noise, quantization noise, intermodulation products), that result in different C/N thresholds. All three DTTB systems use concatenated forward error correction and interleaving. The DVB-T and the ISDB-T outer code ) code with 12 R–S is a Reed–Solomon (R–S) (204, 188, block interleaving. The R–S (204, 188) code, which is shortened from the R–S (255, 239) code, can correct 8-byte transmission errors and is consistent with the DVB-S (satellite) and DVB-C (cable) standards for commonality and easy interconnectivity. The ATSC system implements a more powerful R–S ) code, which can correct 10-byte errors, (207,187, and uses a much larger (52 R–S block) interleaver to mitigate impulse and co-channel NTSC interference. The different R–S code implementations may result in a small difference in C/N performance. Computer simulations show that the ATSC system has a small (0.3–0.5 dB) advantage over the DVB-T and ISDB-T systems [7]. The ATSC system implements a trellis-coded modulation (TCM) as the inner code, while the DVB-T/ISDB-T system used a punctured convolutional code (the same as the one used in the DVB-S standard for commonality). Again, this gives a slight coding advantage in favor of the ATSC system, which is estimated to be around 0.5–1.0 dB. Therefore, the implementation difference in forward error correction gives the ATSC system an estimated total C/N advantage of between 0.8–1.5 dB. This difference could, in the long term, be reduced with technical advances or system improvements, for example using iterative de-coding schemes in the DVB-T/ISDB-T systems. Although it is not mandatory, all of the ATSC receivers on the market implemented a Decision Feedback Equalizer (DFE). The DFE causes a very small noise enhancement, but it also results in a very sharp Bit Error Rate (BER) threshold, because of the error feedback. In a DVB-T/ISDB-T receiver which is properly implemented to allow fast multipath channel tracking and interference rejection, various degradations arise which are well understood and quantified [33]. These degradations add up to some 1.5–2 dB [8], [9]. Therefore the aggregate C/N performance difference, based on today’s technology, is estimated to be around 2–3 dB in favor of the ATSC system over an AWGN channel [6], [10], [11]. However, the AWGN channel C/N performance is only one benchmark for a transmission system. It is an important performance indicator, but it might not represent a real-world channel model. Meanwhile, the channel estimation, equalization and Automatic Gain Control (AGC) systems designed to perform well in an AWGN channel might be slow to respond to signal variations and/or moving echoes. In Europe and Japan, the Ricean channel model was used in the DTTB spectrum planning process [9], [12]. The computer simulation results show that the C/N threshold difference between Gaussian channel and Ricean channel (direct path to muldB) is about 0.5–1 dB, depending on tipath power ratio the modulation and channel coding used [2]. The actual C/N threshold values recommended for the planning process factored in a 2 dB noise degradation caused by the channel estimation and equalization over the receiver noise floor [9]. The frequency planning for the ATSC system has been done using different approaches. In the USA, the FCC used a Gaussian channel performance [6]. In Canada, a generous 1.3 dB C/N margin was set aside for multipath distortion (direct dB), which is similar path to multipath power ratio to the European approach [13]. Table II presents the C/N thresholds (AWGN channel) for the three DTTB systems based on computer simulations [1]–[3] and the most recent laboratory RF back-to-back test results available [6], [9], [11], [14]–[17]. It should be pointed out that the C/N threshold measurement is somewhat dependent on the measurement conditions. For example, changing the RF channel frequency and signal levels might result in minor (a few tenths of a dB) C/N threshold differences. WU et al.: COMPARISON OF TERRESTRIAL DTV TRANSMISSION SYSTEMS 105 TABLE II C/N THRESHOLDS BASED ON TEST RESULTS 2) Fair Comparison of the System C/N Performances: It should be pointed out that the threshold values presented in Table II are not a fair comparison, because the systems have different data rates, and their thresholds are also defined differently (TOV vs. QEF). For the DVB-T and ISDB-T systems, selection of different guard intervals, while maintaining the same channel coding, will result in different data throughputs, despite the same C/N threshold. , or carrier-to-noise ratio One alternative is to use the per unit of data capacity (bit) and unit of bandwidth (Hz), to evaluate the system performance, as it takes account of the is a digital transmission system data rate and bandwidth. system’s spectrum and power efficiency measurement, which is widely used in digital communications’ literature. The smaller value, the more efficient the transmission system. the It is defined as: dB (1) is the system data throughput and is the system where bandwidth. For the 6 MHz ATSC system, the data rate is Mbit/s [1]. The comparable DVB-T and ISDB-T systems coding and 1/32 guard interval) (6 MHz system, with Mbit/s [2] and 17.7 Mbit/s [3], redata rates are spectively. Thus, for the DVB-T and ISDB-T systems, using the same coding but a different guard interval length, the system will be different, due to C/N should be the same, while the different data throughput. The DVB-T and ISDB-T systems’ threshold was defined at before the R–S decoding a Bit Error Rate (BER) of [2], [3]. After R–S decoding, this corresponds to a BER of less , or Quasi Error Free (QEF) reception, which than is equivalent to one error hit every few hours. This threshold definition is typical of performance required for high-speed data transmission. The ATSC threshold was actually derived subjectively from the video picture “Threshold Of Visibility” (TOV), assuming that some video error concealment or resilient techniques are implemented in the receiver. The corresponding objective mea, or Segment Error surement was defined at BER Rate (SER) = , after the R–S decoding. This SER translates into an 8-VSB symbol error rate after the equalizer (before trellis decoding) of 0.2. It also indicates a byte error rate of after the trellis decoding [18]. It can be seen about that the ATSC threshold is more relaxed than that of the DVB-T and ISDB-T systems. A correction factor should be added for a fair comparison. For an AWGN channel, the ATSC system correction factor, between TOV and QEF, should be around 0.8 dB [19]. For the DVB-T and ISDB-T systems, the correction factor was reported as 1.3 dB [15]. It should be pointed out that measurement on different receivers may result in different values depending on their implementation. Based upon the above discussion, factoring in the data rate and the threshold definition differences, the calculated system thresholds for AWGN channels are presented and in Table III. Two convolutional coding rates, are selected for DVB-T and ISDB-T systems, the latter one providing comparable data rates as with the ATSC system. From the RF back-to-back test data, the ATSC system presently has a few dB advantage for an AWGN channel. Again, it should be mentioned that improvements are possible for all systems, and the AWGN channel might not be the best channel model for DTTB, especially for indoor reception. Since all three systems can be scaled for different channel spacings, i.e., 6, 7, and 8 MHz, without changing the channel values presented in Table III coding scheme, the system are generally valid for 6, 7, and 8 MHz systems. C. Multipath Distortion The COFDM modulation system used by DVB-T and ISDB-T has a strong immunity against multipath distortion. It can withstand echoes of up to 0 dB relative to the most powerful received signal. High level echoes are usually found in urban areas, where direct line of sight to the transmitter is blocked, and when indoor or set-top antennae are used. The implementation of a guard interval can totally eliminate the inter-symbol interference, except for echoes with excess delays larger than the extent of this guard interval. However, the in-band fading will still have an impact on the required C/N, especially when high order modulations are used on the COFDM carriers. A strong inner error correction code and a good channel estimation system are required for the DVB-T and ISDB-T systems to withstand 0 dB echoes and a higher C/N will be needed to deal with such strong echoes. With the convolutional coding, it needs about 6 dB more signal power to deal with the 0 dB echoes [5], [10]. However, some of this increased C/N requirement will be compensated by the signal power arriving from the echoes [20]. The balance of these requirements will depend on the code rate selected. Soft decision decoding using an eraser technique can significantly improve the performance [21]. For static echoes with levels less than 4 to 6 dB, the 8-VSB system, using a Decision Feedback Equalizer (DFE), yields a smaller noise enhancement [11], [18]. 106 IEEE TRANSACTIONS ON BROADCASTING, VOL. 46, NO. 2, JUNE 2000 TABLE III SYSTEM E =N THRESHOLDS For very close-in echoes, produced by nearby structures, wide-band selective fading will be experienced. These close-in echoes will affect more predominately narrow-band systems, such as the segments used in the ISDB-T system, which might cause reception failure on some segments or the entire system. Wide-band systems, such as the ATSC and the DVB-T will be more immune to these types of channel impairment. The DVB-T and ISDB-T system guard interval can be used to deal with both advanced or delayed multipath distortions. This is important for SFN operation. The ATSC system cannot handle long pre-echoes, as it was designed for an MFN environment where they usually do not happen in outdoor fixed reception conditions. SFN operation can provide significant savings in spectrum requirements as all the transmitters in one area can operate on the same frequency (see Section IV-F). It can also provide for significant savings in total transmission power because of the increased probability of receiving the signal from a number of transmitters, the so-called “network gain.” is necessary to consider interference from more than one analog television system and, in such cases, a fixed set of notch filter frequencies may be less appropriate. Considerable attention was paid to this aspect during the design and development of the DVB-T system. COFDM systems using 8k FFT should outperform systems using smaller sizes of FFT. D. Co-Channel Analog TV Interference F. Impulse Noise Co-channel analogue television interference (with its energy concentrated around the visual carrier and, to a lesser extent, the color sub-carrier and the aural carriers) will interfere with a limited number of COFDM carriers in specific portions of the DTTB band. A good channel estimation system combined with soft decision decoding using an eraser technique should result in good performance against the analog TV interference. The ATSC system uses a much different approach. A carefully designed comb-filter or notch filter is implemented to notch out the analog TV’s visual, aural and color sub-carriers to improve the system performance. Actually, the OFDM system can also implement a notch filter to improve the performance against co-channel analog TV interference. The ATSC approach is relatively simple to adopt in the case where there is only one interfering analog television system to be considered. In Europe and in some other parts of the world, it The impulse noise interference usually occurs in the VHF and low UHF bands, and is caused by industrial equipment and home appliances, such as microwave ovens, fluorescent lights, hair-dryers, and vacuum cleaners. High-voltage power transmission lines, which often generate arcing and corona, are also a source of impulse noise. Theoretically, OFDM modulation should be more robust to time-domain impulse interference, because the FFT process in the receiver can average out the short duration impulses. Therefore, an OFDM system with larger FFT size, e.g., 8k FFT, will perform better against impulse than the system with smaller FFT size, e.g., 2k FFT [41]. However, as mentioned previously, the channel coding and interleaver implementation also play an important role. The stronger R–S(207, 187, 10) code with the 52-segment interleaver makes the ATSC system more immune to impulse interference than the DVB-T and ISDB-T systems E. Co-Channel DTV Interference Good co-channel DTV C/I performance will result in less interference into the existing analog TV services. It will also mean better spectrum efficiency once the analog services are phased out. All three DTV signals behave like additive white Gaussian noise. Therefore, the co-channel DTV interference performance should be highly correlated with the C/N performance, which is largely dependent upon the channel coding and modulation used. There is about a 3–4 dB advantage for the ATSC system, see Tables III and IV, as it benefits from its better forward error correction system. WU et al.: COMPARISON OF TERRESTRIAL DTV TRANSMISSION SYSTEMS 107 TABLE IV DTV PROTECTION RATIOS FOR FREQUENCY PLANNING. which use R–S(204, 188, 8) code with a 12-segment interleaver [11]. With respect to the inner code, the shorter constraint length of 2 for the ATSC system (7 for DVB-T and ISDB-T) also results in shorter error bursts, which are easier to correct by the outer code. However, the ISDB-T system with the option of using a long time interleaver also proved to be robust to impulse noise [41]. The robustness of the carrier recovery and synchronization circuits against impulse noise can also limit the system performance. This is however primarily a receiver design issue and not a system issue. G. Continuous Wave (CW) Interference Since a COFDM system is a frequency-domain technique, which implements a large number of closely spaced carriers, a single CW or narrow band interference will destroy only a few of these carriers, but the lost data can be easily recovered by the error correction code. On the other hand, CW interference will cause eye closing for the 8-VSB modulation. The adaptive equalizer could reduce the impact of the CW interference, but, in general, DVB-T and ISDB-T systems should outperform the ATSC system on CW interference by a large amount ( 10 dB) [5], [11]. However, tone interference is just another performance benchmark. In the real world, a DTTB system should never experience a tone interference dominated environment, as a well-engineered spectrum allocation plan is designed to avoid such a problem. A badly designed receiver front-end might generate inter-modulation products falling in the signal band, which would result in “CW-like” interference. But this is a receiver design issue, not a system issue. H. Phase Noise Performance For a single carrier modulation system, such as 8-VSB [40], the phase noise generally causes constellation rotation and jitter that can mostly be tracked by a phase-locked loop. Theoretically, the OFDM modulation is more sensitive to the tuner phase noise. The phase noise impact can be modeled into two components [28], [29]: 1) a common rotation component that causes a phase rotation of all OFDM carriers; 2) a dispersive component, or inter-carrier interference component, that results in noise-like defocusing of the carriers’ constellation points. The first component can easily be tracked by using in-band pilots as references. However, the second component is difficult to compensate. It will slightly degrade the DVB-T and ISDB-T system noise threshold. A tuner with a more optimized phase noise performance will be needed for the DVB-T and ISDB-T systems [30]. Using a single conversion tuner or a double conversion tuner will also cause differences in performance. Single conversion tuners have less phase noise and better dynamic range, but are less tolerant to adjacent channel interference, especially on the image channel. A tuner that covers both VHF and UHF bands will be slightly worse than a single-band tuner. IV. DISCUSSION ON THE MEANING OF THE PERFORMANCE SYSTEM A. Indoor Reception Indoor reception of DTTB systems needs more investigation. There is no published large-scale field trial data to support a meaningful system comparison. In general, indoor signals suffer from strong multipath distortion, due to reflections from indoor walls, as well as from outdoor structures. The movement of 108 human bodies and even pets can significantly alter the distribution of indoor signals, causing time varying echoes and field strength variations. The indoor signal strength and its distribution are related to many factors, such as building structure (concrete, brick, wood), siding material (aluminum, plastic, wood), insulation material (with or without metal coating), and window material (tinted and metal coated glasses, multi-layer glass). Typical building attenuation for VHF/UHF signals is around 10 and 25 dB. Measurements on indoor set-top antennae showed that gain and directivity depend very much on frequency and location [12]. For “rabbit ear” antennae, the measured gain varied from about 10 to 4 dBi. For five-element logarithmic antennae, the gains are between 15 to 3 dBi [12]. Meanwhile, indoor environments often experience high levels of impulse noise interference from power lines and home appliances. B. Mobile Reception Distribution of multimedia services (digital TV, audio, data, etc.) to portable and vehicular receivers may become an important application for terrestrial broadcasters. DVB-T and ISDB-T systems can be used to provide mobile services, but a lowerorder modulation on the OFDM carriers and a lower rate of conor ) are recommended to volutional coding (e.g., provide reliable services [22], [23]. There is a penalty in data throughput for mobile reception in comparison to fixed recepor , or 16 QAM with tion. Usually, QPSK with are the preferred modes for reliable mobile reception with data rates in the order of 5 to 12 Mbit/s [22], [23]. With higher order modulation, the system will be sensitive to fading/shadowing and Doppler effects, which, in turn, would require more transmission power. In these conditions, it is not possible to achieve a 19 Mbit/s data capacity required for a single good quality HDTV program and associated multi-channel audio and data services in the envisaged channel bandwidths (6–8 MHz) in a mobile environment [24]. However, providing mobile multimedia services (multi-program SDTV, audio and data services) seems to be a possible option [22], [23], [25], [34], [42]. The DVB-T system was originally designed for fixed and portable reception, but some of its rugged modes, e.g., 2k FFT with QPSK or 16-QAM modulation and strong coding, can provide for mobile reception of DTV services [2], [23]. Theoretical analysis, computer simulations, laboratory tests and field trials show that the channel Doppler spread is not the fundamental limitation of the DVB-T system, although such Doppler spread needs to be compensated for by a somewhat higher C/N. It is the lack of proper time interleaving which restricts the performance of the system in a mobile environment [23]. Field tests of mobile reception with the DVB-T system have shown that increasing the field strength is the prime means to counter this lack of time interleaving and provide a satisfactory mobile service. In MFN situations, an additional 6–8 dB is needed in the case of a Ricean channel to provide for mobile reception of the [23]. 2k DVB-T mode with Since the ISDB-T system was designed from the outset for mobile reception, it can optionally use a large time interleaver— IEEE TRANSACTIONS ON BROADCASTING, VOL. 46, NO. 2, JUNE 2000 up to approximately 0.5 second—to improve the mobile reception service quality [3], [25], [42]. The robust DQPSK modulation can be selected for the OFDM carriers and a 2k or 4k FFT mode can be used for mobile service. In the DQPSK modulation mode, channel estimation is not necessary. This can reduce the receiver complexity and power consumption. It could be advantageous for small hand-held battery-operated receivers. In the case of mobile reception in an SFN environment, where the mobile receiving terminal is travelling at a different speed relative to each SFN transmitter, there will likely be strong Doppler spread effects on strong multipath signals which have to be dealt with by channel estimation and error correction systems. The “nonpunctured” convolutional inner code, , is recommended for mobile implementation [23]. One potential problem with mobile services is the spectrum availability. Since mobile reception requires different modulation and channel coding from the fixed services, it might be advantageous to provide mobile services in other dedicated channels rather than share the same channel with the fixed reception DTV/HDTV services, which usually opt for the maximum data throughput. Since many countries have difficulties allocating one fixed service DTV channel to every existing analog TV broadcaster, finding additional spectrum for mobile service might be difficult. One alternative to provide fixed and mobile services within one channel is, however, to use the hierarchical channel coding and modulation (DVB-T and ISDB-T systems), which will be discussed in Section IV-H. Another alternative is to time-share the transmission facility for fixed and mobile services. For example, the HDTV service may be provided during prime time, while the mobile service is offered during traffic rush hour. However, in the case of several broadcasters sharing a multiplex of SDTV programs on a terrestrial channel, this approach might be difficult to implement. It should be mentioned that digital audio and video services are more robust to transmission errors than data services. With error concealment and muting capabilities, a BER in the order can probably meet the DTV service requirements [35]. of On the other hand, a data service might need error-free transor less). A mobile reception environment mission (i.e., might have an irreducible transmission error floor, which can not achieve error free transmission. Mobile data services, may use either a return link (e.g., via cellular phone) to acknowledge the transmission error and initiate data re-transmission, or blind multiple data re-transmission, via so called data carousels, which will significantly improve the transmission performance over error prone channels. It should be pointed out that since mobile services are mostly intended to deliver audio, data, and low-resolution video services to car drivers or passengers on public transportation systems (buses and trains) [25], [42], they are in direct competition with Digital Audio Broadcasting (DAB) and the third generation Personal Communications System services (IMT-2000). They might also need special approval from the proper regulatory authorities. WU et al.: COMPARISON OF TERRESTRIAL DTV TRANSMISSION SYSTEMS C. Spectrum Efficiency OFDM, as a modulation scheme, is slightly more spectrum-efficient than single carrier modulation systems, since its spectrum can have a very fast initial roll-off even without an output spectrum-shaping filter. For a 6 MHz channel, the useful %) (3-dB) bandwidth is as high as 5.7 MHz (or %) for the DVB-T system [2], and 5.6 MHz (or for the ISDB-T system [3], in comparison with the 5.38 MHz %) useful bandwidth of the ATSC system [1]. (or OFDM modulation has, therefore, an advantage of up to 5% in spectrum efficiency. However, the guard interval that is implemented to provide enhanced freedom from multipath distortion and the in-band pilots inserted for fast channel estimation reduce the data capacity for the DVB-T and ISDB-T systems. For example, the DVB-T offers a selection of system guard intervals, i.e., 1/4, 1/8, 1/16 and 1/32 of the active symbol duration. These are equivalent to data capacity reductions of 20%, 11%, 6%, and 3%, respectively. The 1/12 in-band pilot insertion results in an 8% loss of data rate. Overall, the data throughput reductions are up to 28%, 19%, 14%, and 11% for the different guard intervals. Subtracting the previous 5% bandwidth efficiency advantage for the OFDM system, the total data capacity reductions for the DVB-T system, in comparison with the ATSC system, are 23%, 14%, 9%, and 6%, respectively. This means that, for a 6 MHz system, assuming equivalent channel coding and modulation schemes ), the DVB-T system will offer data rates (64 QAM, of 14.9, 16.6, 17.6, and 18.1 Mbit/s for the guard interval ratios identified above. The ISDB-T system will provide data rates of 14.6, 16.2, 17.2, and 17.7 Mbit/s. The corresponding ATSC system data rate is a fixed 19.4 Mbit/s. Actually, the best indication of spectrum efficiency and values listed in transmission power requirement is the Table III. The apparent loss of spectrum efficiency for DVB-T and ISDB-T systems must be considered in the light of the performance improvement in: a) severe multipath environment, b) fast moving multipath environment, c) Single Frequency Networks (SFN), d) mobile reception and e) nondirectional receiving antennae situations. The above analysis of spectrum efficiency is based on an MFN approach. In an SFN environment, it is possible to have a number of transmitters re-use the same frequency (channel) to cover a large geographical area, which results in overall saving of spectrum and transmission power for DVB-T and ISDB-T systems. D. HDTV Capability Research on digital video compression showed that, based on current technology, a data rate of at least 18 Mbit/s is required to provide a satisfactory HDTV picture for sports and fast action programming (1920 by 1080 format) [24]. Additional data capacity is required to accommodate multi-channel audio and ancillary data services. The ATSC system data rate is 19.4 Mbit/s. Based on the DVB-T and ISDB-T standards, with modulation and channel coding schemes equivalent to the ATSC 8-VSB ), the 6 MHz DVB-T and ISDB-T system (64 QAM, systems’ data throughput ranges from 14.9–18.1 Mbit/s and 109 14.6–17.7 Mbit/s, respectively, depending on the guard interval selected. To achieve a higher data rate, a weaker error correction coding would need to be selected. For example, by increasing the convolutional coding rate to , the range for the data rates becomes 16.8–20.4 Mbit/s for the DVB-T system, and 16.4–20.0 Mbit/s for the ISDB-T system. However, this approach would require about 1.5 dB of additional signal power [2], [16], [17]. The estimated system performance is listed in Table III. Increasing the coding rate will also compromise the performance against multipath distortion, especially for indoor reception and SFN environments. E. Interference into Existing Analog TV Services Since both VSB and COFDM signals behave more or less like white noise. They have the same impact to the analog TV systems. In many countries, the government policy requires analog television and DTTB to co-exist for an extended period of time, and no additional spectrum resources are available for DTTB implementation. DTTB services can only be implemented in channels that cause limited interference into existing analog television reception. It is expected that one of the limiting factors will be the DTTB interference into the existing analog television services during the analog television-to-DTV transition period. In an MFN environment, the current 4 dB C/N difference in planning parameters (see Table IV) requires that DVB-T or ISDB-T systems transmit 4 dB more power than the ATSC system to achieve the same noise limited coverage area. This might make the frequency planning more difficult and cause additional interference into analog television systems. Extra measures would need to be taken to increase the co-channel spacing, or reduce the DTV transmission power (or coverage). F. Single Frequency Network (SFN) and On-Channel Repeater Capability The 8k mode in the DVB-T and ISDB-T systems was included for large scale (nation-wide or region-wide) synchronous SFN operation, where a cluster of transmitters, all fed from the same source, is used to cover a designated service area [38]. It uses a small carrier spacing, which can support very long guard intervals. It can also sustain 0 dB multipath distortion, if a strong ). However, about 6 dB convolutional code is selected ( more signal power is required to deal with the 0 dB multipath distortion [5], [10], [23]. One alternative to reduce the excess transmission power is to use a directional receiving antenna, which would likely eliminate most 0 dB multipath distortion conditions. Such an antenna would also improve the reception of the ATSC 8-VSB signals in such circumstance. It should be noted that the probability of finding a location with two equal power signals in a coverage area is relatively small. The SFN approach can provide stronger field strength throughout the core coverage area and can significantly improve the service availability. The receivers have more than one transmitter from which they can receive the signal (diversity gain). They have better chances to have a strong signal path to a transmitter to achieve reliable service. 110 By optimizing the transmitter network (transmitter density, tower height and location, as well as the transmission power at each transmitter) SFN can yield better coverage with lower total transmit power and provide better spectrum efficiency. Tighter control of the rate of decay of the field strength outside the service area is also possible through the design of the SFN in order to maintain a satisfactory level of interference to and from close neighboring networks [26]. Special measures must be taken to minimize the frequency offset among the repeaters and flexibly address each transmitter with respect to its exact site, power, antenna height and the insertion of specific local signal delays. For DVB-T systems, this can be achieved by using the DVB MIP (Mega-frame Initialization Packet) system. All transmitters in an SFN network can, then, be synchronized in time and frequency. One problem that might impact a large-scale SFN implementation is co- channel and adjacent channel interference. In many countries, it might be difficult to allocate a DTV channel for large-scale SFN operation that will not generate substantial interference into existing analog TV services during the analog TV to DTV transition period. Finding the additional tower sites at desired locations and the associated expenses (such as property, equipment, legal, construction, operation, and environmental studies) might not be practical or economically viable. The DVB-T and ISDB-T systems also allow the use of on-channel repeaters to improve the coverage at the edge of the coverage area (coverage extenders) and fill holes within the coverage area (gap-fillers). In this case, these on-channel repeaters are designed to capture the signal off-air as emitted by the main transmitter, amplify it and re-transmit it. The maximum transmission power is limited by the amount of isolation that can be achieved between the directional receiving antenna and the transmit antenna. Such limitation could be alleviated by regenerating the DTTB signal at the repeater, i.e., the received off-air signal is demodulated, decoded, and re-modulated. In such a case, the transmission errors generated in the first hop could be corrected. The signal transmitted from such an on-channel repeater would normally result, at the receiver, in the creation of an advanced active echo (from the main transmitter) with an excessive time difference. In such a case, sufficient isolation from the main transmitter would need to be provided by the home receiving antenna. On the other hand, the ATSC system was not designed for synchronous SFN implementation. Nevertheless, implementation of on-channel repeaters and gap-fillers for coverage extension is possible, if enough isolation between the reception of the original signal and the re-transmitted signal can be achieved [27]. The limitations in designing the on-channel gap fillers to avoid signal feedback between their transmit and receive antennae are the same for all systems. Local signal re-generation would also be beneficial for all systems but results in a need for increased home antenna discrimination toward interfering transmitters resulting from the additional excess delay generation. The key difference between a DTV and an analog TV system is that the DTV can withstand at least 20 dB of co-channel interference, while the analog TV co-channel threshold of visibility is around 50 dB (30 to 35 dB for CCIR Grade 3). In other words, IEEE TRANSACTIONS ON BROADCASTING, VOL. 46, NO. 2, JUNE 2000 DTTB is 10 to 30 dB more robust than analog TV, which provides more flexibility for the repeater design and siting. In the case of an ATSC system repeater implementation [27], using a directional receiving antenna will increase the location availability as well as reduce the impact of fast-moving or longdelay multipath distortions. The operational parameters will depend on the population distribution, terrain environment, and intended coverage area. It should be pointed out that under any circumstances, an RF transmission system (ATSC, DVB-T or ISDB-T; SFN or MFN), can not realistically provide a service with a 100% location availability. G. Noise Figure Generally speaking, the noise figure of a receiver is an implementation issue. It is system-independent. A low noise figure receiver front end can be used in any DTTB system to reduce the minimum signal level required. The critical parameter for planning purposes is sensitivity, which accounts not only for the noise figure, but also the susceptibility of the system to effects such as self-interference and inter-modulation products. A single conversion tuner has a better dynamic range, and low phase noise, but its noise figure, which tends to be lower, is not consistent over the full VHF and UHF TV bands. Single conversion tuners provide less suppression on adjacent channel interference. This suppression is also not consistent over all the channels. On the other hand, a double conversion tuner has a higher noise figure, less dynamic range, and higher phase noise. It can achieve better adjacent channel suppression, especially on the image channel. Its noise figure and adjacent channel suppression are also much more consistent across the frequency bands. Tuner performance is very much linked to the cost (materials, components, frequency range, etc.). With today’s technology, for low-cost consumer-grade tuners, the single conversion tuner has a noise figure of about 7 dB. A double conversion tuner typically achieves 9 dB. However, tuner noise figure only impacts the system performance at the fringe of the coverage, where signal strength is very low and there is no co-channel interference present. This situation might only represent a very small percentage of the intended coverage areas, since most of the coverage is interference-limited. However, some countries do regulate the noise figure of the receivers. Different countries might opt to use different noise figures in their frequency planning process. For example, DVB recommended the use of a 7 dB noise figure in their UHF band frequency planning [9], [12]. In the USA, the FCC used a 10 dB noise figure as requested by the receiver manufacturers [31]. In Canada, a 5 dB noise figure was used taking into account future technology developments [13]. H. Hierarchical Modulation and Services Hierarchical modulation can, among other things, help overcome the problem of a total signal disruption in areas with low field strength by providing a lower rate, more reliable bit stream as part of the transmitted signal. The DVB-T system can use multi-resolution constellations on the OFDM carriers (16 QAM or 64 QAM) to provide two tiers of service within one DTTB WU et al.: COMPARISON OF TERRESTRIAL DTV TRANSMISSION SYSTEMS channel [2]. It can offer a robust channel with a low data rate and ), and a high error protection (e.g., 4.5 up to 6.3 Mbit/s, simultaneously a less robust channel with a higher data rate and ), deweaker error protection (e.g. 15 up to 20 Mbit/s, pending on the channel bandwidth (6, 7, or 8 MHz) [32]. These two channels, the high priority (HP) stream and the low priority (LP) stream can be used, either for independent services or for simulcasting of the same services. But they are not intended to provide scalable or hierarchical video coding, since, presently, all three DTTB standards only implement MPEG-2 Main Profile (see Table I) which does not support hierarchical or multi-resolution video source coding. For example, the HP stream can provide the basic DTV service and some audio programs, while the LP stream is used for additional DTV or data services or for one service of HDTV-like quality. The difference in the required C/N between the HP and LP for fixed reception is in the order of 10 dB [32], [34]. From the coverage point of view, the LP data stream can be used to cover the core service area, but a fixed roof-top directional antenna would be required in most locations. Generally, within the LP stream coverage, the HP stream should always be available. The HP data stream can be used to provide three classes of services/coverage: • Class I: The HP stream, due to its low C/N requirement, can be used to provide extended coverage using a fixed roof-top antenna. • Class II: The HP stream can be used to provide service to mobile terminals, e.g., with a car-mounted omni-directional antenna. • Class III: The HP stream can be used to provide service to a portable indoor set-top antenna with limited directivity. However, coverage depends on many factors, such as terrain, transmitting tower height and power, receiving antenna height and gain/directivity. For example, for terrain-limited coverage area (due to mountains, valleys, horizon or man-made structures), the core-coverage (LP data) and Class I (HP data) coverage areas might be quite close. They can achieve different service areas only if the coverage is limited by signal power or interference. For relatively flat terrain, a 10-dB difference in C/N requirement will result in a coverage difference of about 10 to 15 km in radius. For the mobile reception case, gain/directivity of the receiving antennae are usually much lower than in the case of the roof-top antennae (a difference of about 10 dB in the UHF band). The C/N requirement for mobile reception is usually at least 6 dB higher than for fixed reception (Rayleigh vs. Gaussian channels) [23]. The reduced receiving antenna height, 1.5 m car-mounted antennae vs. 10 m roof-top antennae, will also reduce the received signal strength due to the increased presence of blockage and shadowing. For indoor reception, the main problem is building penetration loss, which ranges between 10–25 dB. The coverage of mobile and indoor portable reception of the HP stream (Class II and III coverage) is likely to be less than the core coverage area, i.e., where the LP data stream can be received with fixed roof-top antennae. It should also be pointed out that, due to terrain blockage and shadowing, the real coverage areas for different classes of 111 service are usually not as simple as a series of concentric circles, but rather are akin to a Swiss cheese—with many “holes” within the coverage—and terrain-limited end-of-coverage contours with irregular shapes. In the case of the ISDB-T system, the band segmentation transmission concept allows the use of up to three different modulation schemes and coding rates on different channel segments to meet various service requirements and interference conditions. It can also independently decode the signals in the different band-segments. The penalty for providing hierarchical services, using hierarchical modulation and channel coding (DVB-T) or band segmented modulation and channel coding (ISDB-T), is that it either increases the C/N requirement (DVB-T case) or reduces the total data rate (ISDB-T case), in comparison with the nonhierarchical approach. More studies and tests are required to demonstrate the viability of providing tiered services [34] and to establish the required minimum C/N difference between two layers of services to support two distinctive or meaningful services. I. System Flexibility The DVB-T and ISDB-T standards offer broadcasters a large selection of operational modes. The reasons for providing this wide range of choices are the different applications foreseen and the different introduction scenarios expected in various countries. The parameters that can be chosen for a given application are: a) The size of FFT, which specifies the number of OFDM carriers (for DVB-T: 2K and 8K FFT; for ISDB-T: 2k, 4k and 8k FFT); b) Carrier modulation (for DVB-T: QPSK, 16 QAM and 64 QAM; for ISDB-T: DQPSK, QPSK, 16 QAM, and 64 QAM); c) Coding rate for the inner error correction code (1/2, 2/3, 3/4, 5/6, and 7/8); d) Guard interval width (1/4, 1/8, 1/16, and 1/32 of the duration of an OFDM symbol); e) Non-hierarchical or hierarchical modulation and channel coding. These selections result in a large number of nonhierarchical and hierarchical mode combinations. The standard receiver should be able to automatically detect which mode is being used. In addition, these two standards are fully specified for all three existing channel spacings: 6 MHz, 7 MHz, and 8 MHz. The choice of transmission mode will set the system data capacity and will affect the C/N performance, and, therefore, the coverage corresponding to different types of receiving installations—fixed roof-top antennae, indoor portable receivers, and mobile receivers at high speeds. Depending on the chosen combination, the DVB-T and ISDB-T systems will be able to accommodate SFN’s of various transmitter densities resulting in improved coverage and spectrum efficiency. One special feature for the ISDB-T system is that it can independently decode signals in part of the DTTB band, i.e., in band-segments [3]. This can be used, for example, to facilitate narrow-band audio/data broadcasting. 112 IEEE TRANSACTIONS ON BROADCASTING, VOL. 46, NO. 2, JUNE 2000 The ATSC 8-VSB system was designed to maximize the data capacity. The data rate for a 6 MHz channel is 19.4 Mbit/s. However, if there is demand from the broadcasters, 2-VSB and 4-VSB versions of the modulation are possible, which, at the expense of reduced data rate, would provide more robust reception [40]. J. Systems Scaled for Different Channel Bandwidths The DVB-T system was originally designed for 7 and 8 MHz channels. By changing the system clock rate, the signal bandwidth can be adjusted to fit 6–8 MHz channels. The corresponding hardware differences are the channel filter, IF unit, and the system clock. On the other hand, the ATSC and ISDB-T systems were designed for a 6 MHz channel. These systems can operate over 7 and 8 MHz bandwidths also by changing the system clock, as in the DVB-T case. However, the ATSC system implemented a comb-filter or a notch filter to reduce the impact of co-channel NTSC interference. This would need to be changed to deal with different analog TV systems. The use of a comb-filter or a notch filter is not mandatory and might not be needed, if co-channel analog TV interference is not a major concern. For instance, some countries might implement DTV on dedicated DTV channels where there is no analog co-channel interference. Generally speaking, a narrower channel results in a lower data rate for all three modulation systems, due to slower symbol rate. However, it also means a longer guard interval for DVB-T and ISDB-T systems and longer echo correction capability for the ATSC system. One minor weak point for the 6 MHz COFDM 8k systems is that their narrow carrier spacing of less than 1 kHz (75% compared to 8 MHz system) might cause the systems to be more sensitive to phase noise and Doppler spread in the case of mobile reception. The 5.7 MHz useful bandwidth of the 6 MHz DVB-T system might also need steep RF filtering [2] to reduce any adjacent channel interference into analog television services produced by nonlinearity in the high power amplification stages at the transmitters (ISDB-T 6 MHz system useful bandwidth is 5.6 MHz). V. DTV IMPLEMENTATION PARAMETERS Planning parameters for digital terrestrial TV are given in ITU Recommendation 1368 [39]. Countries choosing the same DTTB system could still use different implementation approaches, emission masks, and technical parameters in their spectrum allotment process, depending on their spectrum resources and policy, population distribution, required service quality, etc. For example, Canada adopted the ATSC DTTB system, but implemented different DTV technical parameters and emission masks [13] than the USA. Table IV lists the Canadian [13], the American [6], the European [9], [12] and the Japanese [36], [37] DTV technical parameters, more specifically the protection ratios, used in DTV planning. In the Canadian plan, a generous 1.3 dB C/N margin has been allocated for multipath distortion, which is similar to the EBU approach that uses a Ricean channel performance threshold as a planning parameter [9]. Since noise and co-channel DTV interference are additive, a total dB was allocated in Canada as the system dB, threshold [in Table IV, dB]. Also in Table IV, a Canadian co-channel NTSC to DTV interference threshold of 7.2 dB was used. This allows the system to withstand a co-channel NTSC interference of 7.2 dB, and, at the same time, either a C/N of 19.5 dB or a co-channel DTV interference of 19.5 dB. The adjacent channel DTV interference parameters are generally the same as the American ones, as shown in Table IV. It should be pointed out that the protection ratios for DTV interference into analog TV systems depend on many factors, such as the analog TV standards (NTSC, PAL and SECAM) and the system bandwidths (6–8 MHz), as well as the subjective evaluation methods (CCIR Grade 3, Threshold of Visibility, continuous or tropospheric interference). Other nontransmission or spectrum-related factors, such as audio coding (stereo, or surround sound), video format (SDTV or HDTV, progressive or interleaved), error concealment, and reception conditions (fixed, portable or mobile) should also be considered. VI. CONCLUSIONS There are three DTTB standards available. The final choice of a DTV modulation system is based on how well the system can meet the particular requirements or priorities of each country, as well as other nontechnical (but critical) factors, such as geographical, economical, and political relations with the surrounding countries and regions. Each country needs to clearly establish its needs, then investigate the available information on the performances of different systems to make the best choice. It is hoped that the information provided in this paper will be helpful in reaching that goal. ACKNOWLEDGMENT The authors would like to acknowledge the help and guidance from W. Luplow, W. Bretl and R. Citta of Zenith Electronics Corp.; E. Williams of PBS, T. Gurley of MSTV, Dr. C. Weck of IRT, E. Wilson of DigiTAG/EBU, P. MacAvock of DVB, E. Stare of Teracom, B. Sueur of CCETT, Dr. P. Pogrzeba of Deutsche Telecom, C. Nokes and A. Oliphant of BBC R&D; Dr. S. Moriyama, Dr. O. Yamada and Dr. T. Kuroda of NHK Science and Technical Research Laboratories. REFERENCES [1] ATSC Digital Television Standard, ATSC Standard A/53, September 16, 1995. [2] ETS 300 744, “Digital broadcasting systems for television, sound and data services; Framing structure, channel coding and modulation for digital terrestrial television,”, ETSI Draft EN 300 744 V1.2.1, 1999-1. [3] ARIB, “Terrestrial integrated services digital broadcasting (ISDB-T)—Specifications of channel coding, framing structure, and modulation,”, September 28, 1998. [4] A. Chini, Y. Wu, M. El-Tanany, and S. Mahmoud, “Hardware nonlinearities in digital TV broadcasting using OFDM modulation,” IEEE Trans. Broadcasting, vol. 44, no. 1, March 1998. [5] Y. Wu, M. Guillet, B. Ledoux, and B. Caron, “Results of laboratory and field tests of a COFDM modem for ATV transmission over 6 MHz channels,” SMPTE Journal, vol. 107, Feb. 1998. WU et al.: COMPARISON OF TERRESTRIAL DTV TRANSMISSION SYSTEMS [6] ATTC, Digital HDTV Grand Alliance System Record of Test Results. Alexandria, VA: Advanced Television Test Centre, October 1995. [7] CRC, R&D, “Codes de correction d’erreurs pour la transmission de la télévision Numérique (Phase 2),” Industry Canada, Contract 67qsg-56181, with the University of Ottawa, March 1996. [8] J. E. Salter, “Noise in a DVB-T system,”, BBC R&D Technical Note, R&D 0873(98), Feb. 1998. [9] ITU-R SG 11, Special Rapporteur—Region 1, “Protection ratios and reference receivers for DTTB frequency planning,”, ITU-R Doc. 11C/46-E, March 18, 1999. [10] A. Morello et al., “Performance assessment of a DVB-T television system,” in Proceedings of the International Television Symposium 1997, Montreux, Switzerland, June 1997. [11] N. Pickford, “Laboratory testing of DTTB modulation systems,” Australia Department of Communications and Arts, Laboratory Report 98/01, June 1998. [12] ETSI, TR 101 190v1.1.1, “Implementation guidelines for DVB terrestrial services: Transmission aspects,”, ETSI Technical Report TR 101 190v1.1.1, 1997-12. [13] Y. Wu et al., “Canadian digital terrestrial television system technical parameters,” IEEE Transactions on Broadcasting, vol. 45, no. 4, December 1999. [14] C. R. Nokes, “Results of tests with domestic receiver IC’s for DVB-T,” in Proceedings of the Int’l. Broadcasting Convention, Amsterdam, Sept. 1998, pp. 294–299. [15] “Results of initial performance tests with the LSI Logic L64780 DVB-T COFDM demodulator chip,” BBC R&D, http://www.bbc.co.uk/rd/projects/chipset/l64780.html. [16] ITU-R WP, 11A/69, “Transmission performance of ISDB-T,”, ITU-R WP 11A/69-E, May 17, 1999. [17] S. Kaneko, “Transmission performance of ISDB-T,” in Proceedings of 6th International Symposium on Broadcasting Technology, ISBT’99, Beijing, Aug. 22–24, 1999, pp. 60–74. [18] M. Ghosh, “Blind decision feedback equalization for terrestrial television receivers,” Proceedings of the IEEE, vol. 86, no. 10, pp. 2070–2081, Oct. 1998. [19] G. Sgrignoli, W. Bretl, and R. Citta, “VSB modulation used for terrestrial and cable broadcasts,” IEEE Transactions on Consumer Electronics, vol. 41, no. 3, pp. 367–382, Aug. 1995. [20] A. K. Salkintzis and P. T. Mathiopoulos, “On the combining of multipath signals in narrow band Rayleigh fading channels,” IEEE Trans. Broadcasting, vol. 45, no. 2, pp. 192–195, June 1999. [21] J. H. Stott, “Explaining some of the magic of COFDM,” in Proceedings of the International TV Symposium 1997, Montreux, Switzerland, June 1997. [22] P. Pogrzeba, R. Burow, G. Faria, and A. Oliphant, “Lab and field test of mobile applications of DVB-T,” in Proceedings of the International TV Symposium 1999, Montreux, Switzerland, June 1999. [23] R. Burow and K. Fazel et al., “DVB-T and DMB in mobile environments,” MINT Project Report by Task Force DVB-DMB, Feb. 1998. [24] D. Lauzon, A. Vincent, and L. Wang, “Performance evaluation of MPEG-2 video coding for HDTV,” IEEE Trans. Broadcasting, vol. 42, no. 2, June 1996. [25] T. J. Thong, “Digital TV field trials in Singapore,”, ABU Technical Review, no. 178, 1998. [26] A. Ligeti and J. Zander, “Minimal cost coverage planning for single frequency networks,” IEEE Trans. Broadcasting, vol. 45, no. 1, March 1999. [27] W. Husak et al., “On-channel repeater for digital television implementation and field testing,” in Proceedings of the 1999 Broadcast Engineering Conference, NAB’99, Las Vegas, April 17–22, 1999, pp. 397–403. [28] J. H. Stott, “The effect of phase noise in COFDM,” EBU Technical Review, Summer 1998. [29] Y. Wu and M. El-Tanany, “OFDM system performance under phase noise distortion and frequency selective channels,” in Proceedings of Int’l Workshop of HDTV 1997, Montreux, Switzerland, June 10–11, 1997. [30] C. Muschallik, “Influence of RF oscillators on an OFDM signal,” IEEE Trans. Consumer Electronics, vol. 41, no. 3, pp. 592–603, August 1995. [31] FCC, 11A/69, “Sixth further notice of proposed rule making,”, MM Docket No. 87-268, Aug. 14, 1996. [32] G. Faria, “DVB-T hierarchical modulation: An opportunity for new services,” in Proceedings of the Int’l. Broadcasting Convention, Amsterdam, Sept. 10–14, 1999, pp. 582–588. 113 [33] M. Speth, S. A. Fechtel, G. Fock, and H. Meyr, “Optimum receiver design for wireless broad-band systems using OFDM: Part I,” IEEE Trans. Communications, vol. 47, no. 11, pp. 1668–1677, Nov. 1999. [34] MOTIVATE, 11A/69, “Report on laboratory tests for mobile DVB-T,”, http://b5www.berkom.de/Motivate/PR9903.html, March 17, 1999. [35] C. S. Lee, T. Keller, and L. Hanzo, “OFDM-based turbo-coded hierarchical and nonhierarchical terrestrial mobile digital video broadcasting,” IEEE Trans. Broadcasting, vol. 46, no. 1, March 2000. [36] S. Nakahara, M. Okano, T. Kurakake, and M. Ishida, “Experimental results for protection ratios of analog television signal interfered with by digital terrestrial television signal,” ITE Technical Report, vol. 23, no. 7, pp. 29–34, Jan. 1999. [37] S. Nakahara, M. Okano, N. Iai, and S. Kimura, “Performance of DTTB signal interfered with by analog television signal,” ITE Technical Report, vol. 23, no. 34, pp. 7–12, May 1999. [38] ITU-R, 11A/69, “Terrestrial and satellite sound broadcasting to vehicular, portable and fixed receivers in the VHF/UHF bands,” ITU-R, Geneva, Switzerland, Special Publication, 1995. [39] ITU-R, 11A/69, “Planning criteria for digital terrestrial television services in the VHF/UHF bands,” ITU, Recommendation ITU-R BT. 1368-1. [40] C. Eilers and G. Sgrignoli, “Digital television transmission parameters—Analysis and discussion,” IEEE Transactions on Broadcasting, vol. 45, no. 4, December 1999. [41] SET/ABERT, 11A/69, “Digital television systems Brazilian tests final report,”, http://www.set.com.br/testing.pdf. [42] T. J. Thong, “Singapore digital television technical committee—Final report,”, May 1999. Dr. Yiyan Wu is a Senior Research Scientist with the Communications Research Centre, Ottawa, Canada. His research interests include digital video compression and transmission, signal and image processing, LMDS/MMDS, satellite and mobile communications. He is actively involved in the ATSC technical and standard activities and ITU-R digital television and data broadcasting studies. He is an Adjunct Professor of Carleton University, Ottawa, Canada, a member of the IEEE Broadcast Technology Society Administrative Committee and an Associate Editor of the IEEE TRANSACTIONS ON BROADCASTING. Ewa Pliszka received her M.Sc. from the Electronics Faculty, Technical University of Warsaw in 1973. She also completed post-graduate studies in 1977 at the Electronics Faculty, Technical University of Warsaw and, in 1993, at the National Institute of Telecommunications, Evry, France. In 1973, she joined Research and Development Centre, Telekomunikacja Polska S.A. as a Research Engineer. Her first activities covered the design and construction of measurement equipment for transmission parameters of audio and video analog signals. Later on she became interested in the transmission of digital audio and video signals. Her work now covers many aspects of the transmission of digital TV and sound signals. Her current work also involves—among other things—collaboration with the Radiocommunication Sector of International Telecommunication Union (ITU), participation in Study Group 11 (Television Broadcasting) and Study Group 10 (Sound Broadcasting), and also participation in the activities of the Polish Standardization Committee, especially in the area of digital TV. Bernard Caron received a B.Sc. in electrical engineering from Laval University, Quebec in 1978 and a M.Sc. from University of Ottawa in 1984. Since 1979, he has been with the Communications Research Centre Canada and has worked on Teletext, mobile data transmission, video channel characterization and simulation. He is now the Manager of the Television Systems and Transmission program in the Broadcast Technologies branch. He is currently involved in the introduction of digital television terrestrial broadcasting in Canada. Pierre Bouchard is a Research Engineer in the Television Systems and Transmission group at the Communications Research Centre Canada, in Ottawa. He is currently involved in research on DTV transmission, DTV coverage, and on broadband broadcasting technologies (MMDS and LMCS/LMDS). Gerald Chouinard received a B.Sc. in electrical engineering in 1975 from University of Sherbrooke. He worked five years for the Canadian Broadcasting Corporation (CBC) in the field of International Technical Relations. He joined the Communication Research Centre (CRC) in 1981, where he became involved in technical and spectrum-orbit utilization studies related to satellite DBS. In 1992, he became Director, Radio Broadcast Technologies Research. Since April 1998, he reports to the CRC President as a Senior Advisor on Broadcast Technologies. In his work, Dr. Chouinard has been closely involved in standards setting activities for both Digital Television and Radio Broadcasting in Canada, internationally through the ITU-R, as well as in North-America by his participation in the work of the ATSC and the NRSC.