Application of MPEG-2 Systems to Terrestrial ISDB (ISDB-T)

Application of MPEG-2 Systems to Terrestrial ISDB (ISDB-T) MICHIHIRO UEHARA Invited Paper NHK has developed the band segmented transmission orthogonal frequency division multiplexing (BST-OFDM) scheme for the transmission system of Integrated Services Digital Broadcasting- Terrestrial (ISDB-T). This scheme provides the great advantages of hierarchical transmission and partial reception. To provide commonality with other systems, the transport signal of ISDB-T adopts the MPEG-2 transport stream (TS). However, TS has been designed for neither hierarchical transmission nor partial reception. Thus, to fulfill the requirements of ISDB-T, the TS has been adapted to provide effective hierarchical transmissions and partial reception. This paper describes the TS generation methods used by the remultiplexer for minimizing receiver processing load. Briefly, they are: 1) a method enabling hierarchical transmission and partial reception of a single TS; 2) a method relating a TS packet to a segment of the OFDM signal; 3) a method for interfacing the remultiplexer with a modulator at a single constant clock; 4) a method for reconstructing a serial TS at receivers from hierarchical transmission signals allotted to layers in parallel by an OFDM multicarrier; and 5) a method for correctly recovering the program clock reference (PCR) at a partial reception receiver, even if the TS rate of the receiver is different from that of the transmission side. Keywords band segmented transmission OFDM (BST-OFDM), digital terrestrial broadcasting, hierarchical transmission, ISDB-T, MPEG-2 Transport Stream, OFDM, partial reception. I. INTRODUCTION NHK has been researching band segmented transmission orthogonal frequency division multiplexing (BST-OFDM) as a transmission system for both television and audio broadcasts in digital terrestrial broadcasting [1]. Utilizing the multicarrier modulation characteristics of OFDM, the system enables hierarchical transmission, which combines fixed-reception and mobile-reception modes in one group of segments, and partial reception whereby a receiver picks out Manuscript received March 3, 2005; revised August 24, 2005. The author is with the Planning Division, Engineering Administration Department, NHK (Japan Broadcasting Corporation), Tokyo 150-8001, Japan (e-mail: uehara.m-fu@nhk.or.jp). Digital Object Identifier 10.1109/JPROC.2006.859695 only a part of the segments. Here, a segment is the basic unit of BST-OFDM transmission in the frequency domain. MPEG-2 Systems [2] should be used as the multiplexing scheme for the transport layer to maintain interoperability with other digital broadcasting systems, such as ISDB-S, ISDB-C, and ISDB-T. However, the MPEG-2 transport stream (TS) takes neither hierarchical transmission nor partial reception into account. In the following, we examine of the use of a single TS for hierarchical transmission and partial reception in BST- OFDM for minimizing receiver processing load, instead of using multiple TSs for the corresponding hierarchical layers. First, we describe the frame structure of the OFDM signal and TS in ISDB-T. Next, we describe a method to interface the remultiplexer and a modulator by using a single constant clock, and a method to regenerate the same TS sent from the transmitting side at the receiver by using the frame structure. Finally, we describe a method for setting TS time stamps to accomplish partial reception. II. APPLICATION OF SINGLE TRANSPORT STREAM OF MPEG-2 SYSTEMS TO HIERARCHICAL TRANSMISSION The transport stream provided by MPEG-2 Systems (ISO/IEC 13818-1) is used as a multiplexing technique in many digital broadcasting systems around the world. To make up a program by synchronizing video, audio, and other monomedia components, MPEG-2 Systems uses time stamps that indicate the time for decoding and presenting each component. These time stamps are based on a program clock reference (PCR) that is transmitted in the header of a TS packet (TSP). MPEG-2 Systems does not, however, provide a way for synchronizing multiple TSs, which means that all program components need to be multiplexed in a single TS. In other words, the MPEG-2 TS as it is cannot be applied to hierarchical transmission. It is therefore desirable that a single TS be able to accommodate hierarchical transmission instead of using multiple TSs for corresponding hierarchical layers with BST-OFDM. 0018-9219/$20.00 2006 IEEE PROCEEDINGS OF THE IEEE, VOL. 94, NO. 1, JANUARY 2006 261

Table 1 OFDM Segment Parameters Table 2 Provisional Specifications of Digital Terrestrial Broadcasting Tables 1 and 2 list the specifications of the BST-OFDM system used in this development, and Fig. 1 shows the assignment of OFDM segments in the frequency domain. ISDB-T television broadcasting consists of 13 OFDM segments having a total bandwidth of 5.6 MHz. Although this 13-segment format enables a maximum of 13 hierarchical layers in theory, the number of layers is limited to three at most in practical applications. In addition to the television transmissions with the 13-segment format, ISDB-T can provide digital audio broadcasting in either of two formats; a 430-kHz bandwidth one-segment format consisting of a single OFDM segment; and a 1.3-MHz bandwidth three-segment format consisting of three OFDM segments. The one-segment format corresponds to segment no. 0 in Fig. 1, and the three-segment format corresponds to segment nos. 0, 1, and 2. In the three-segment format, segment no. 0 is always allocated for partial reception and the number of hierarchical layers is always two. A. Frame Structure of OFDM Signal and TS Fig. 2 shows the system diagram on the transmit side for hierarchical transmission of a single TS. In the transmission process, TSs inputted to the remultiplexer are combined into a single TS. Then, the single TS is separated into specific hierarchical layers TSP by TSP, and the TSPs are modulated through a channel coding process on each hierarchical layer. 262 PROCEEDINGS OF THE IEEE, VOL. 94, NO. 1, JANUARY 2006

Fig. 1. Arrangement of OFDM segments. Fig. 2. Transmit-side system diagram. Finally, the TSPs of each hierarchical layer are transmitted in parallel by using a multicarrier OFDM signal. In order to decode the signal with a very short time delay, the system has to periodically transmit synchronizing points indicating where the tops of the TSP data align in all hierarchical layers. Since the modulation scheme of each layer can be arbitrarily selected, the transmission capacity of each layer will vary depending on the transmission parameters. To cope with this issue, stuffing dummy bits have been introduced at the end of the aligned packet data array in all layers. However, the use of the dummy bits decreases transmission capacity; hence it would be desirable to make the transmission capacity compatible with original TS itself. For this purpose, the amount of data transmitted by using one OFDM segment in a certain period needs to correspond to an integral multiple of the number of TSP data in each segment. Letting denote the number of symbols, the total bits of one OFDM segment data can be written as follows: Table 3 Number of Symbols in OFDM Segment (1) where the number of effective carriers in one OFDM segment, the number of bits per carrier-modulated symbol and error-correcting code rate. When the value of for a certain number of symbols can be made times TSP ( b, where is a positive integer) for all OFDM-segment transmission parameters, sufficient compatibility between the OFDM segment and the number of TSPs can be achieved. Considering this requirement, the following equation is derived from (1): Table 3 lists for all combinations of transmission parameters for OFDM-segment Mode 1 in Table 1. This result (2) UEHARA: APPLICATION OF MPEG-2 SYSTEMS TO TERRESTRIAL ISDB (ISDB-T) 263

Table 4 Number of TSPs in 1-OFDM-Segment/1-OFDM-Frame Table 5 Results of Calculating Fm/Fs reveals that sufficient compatibility can be achieved with respect to the number of TSPs by giving the OFDM signal a frame structure where one frame consists of an integral multiple of 204 symbols. On the receiver side, a shorter frame length is desirable considering the time required for establishing synchronization. For this reason, the number of symbols composing a frame is chosen to be the minimum multiple of 204. This OFDM-signal frame is called an OFDM frame. Mode 2 and Mode 3, respectively, have twice and four times the number of carriers of Mode 1, which means that the number of symbols within the OFDM frame can be made half and quarter that of Mode 1. However, this number of symbols within an OFDM frame is set at 204 regardless of the mode, for the sake of simplifying the receiver s processing. Table 4 lists the number of TSPs in one OFDM segment per OFDM frame. B. Common Interface Clock for Various TS Bit Rates The TS can take on a variety of bit rates according to the transmission parameters of each hierarchical layer. This means that a variety of clock signals would have to be generated at the receiver to accommodate these various bit rates, which places a significant workload on the receiver. Adding an appropriate number of extra TSPs (null TSPs) to the TS (valid TSPs) so as to interface with a fixed clock is a solution to this problem. Letting denote the OFDM frame length and the interface clock of the TS, the total number of bits to be input from the interface point within the duration of one OFDM frame can be written as follows: Here, an interface can be achieved with a common clock by selecting to be at least the maximum number of bits that can be transmitted by the OFDM signal. The number of dummy bits to be inserted is equal to the difference between and the number of bits transmitted in an OFDM frame. (3) On the other hand, the process using common 204-B units needs a common interface clock, because RS(204,188) is employed as the outer code and the TSPs of the interfaced TS are arranged in equal intervals for the purpose of synchronizing and regenerating TS packets on the receive side. Considering the 204-B unit process, (3) can be rewritten in the following conditional form: where (4) and (5) (6) the maximum number of TSPs that can be transmitted by the OFDM signal, an integer equal to or greater than, guard-interval ratio and effective symbol length. If is related to the fast Fourier transform (FFT) sampling clock by a simple integer ratio, only one base signal clock will be needed in the receiver, thereby making the equipment simpler. Given that the number of FFT samples is, can be written as follows: (7) Equations (4), (6), and (7) give the following relationship: Substituting (5) and (8), we get (8) (the number of symbols) in (9) 264 PROCEEDINGS OF THE IEEE, VOL. 94, NO. 1, JANUARY 2006

Table 6 Number of TSPs Configuring a Multiplex Frame Fig. 3. Configuration of receiver model. To enable (9) to hold for all values of, we set to its minimum value and get (10) Table 5 lists the results of calculating (10) for all three modes in the 13-segment format, one-segment format, and three-segment format. These results show that a TS interface can be achieved by using a common interface clock that is four times the FFT sampling clock for the 13-segment and three-segment formats and two times that for the one-segment format. The TS here takes on a periodic structure corresponding to an OFDM frame. A TS interfaced by using the clocks given in the above in one OFDM frame period is called a multiplex frame. Table 6 lists the number of TSPs included in a multiplex frame. C. Agreement Between Transmit TS and Receive TS As shown in Fig. 2, the TS arranged with a multiplex frame structure, which is called transmit TS hereafter, is separated into appropriate hierarchical layers TSP by TSP, and each TSP is modulated with the assigned scheme. The modulated symbols are stored in the buffer of each hierarchical layer, mapped to a OFDM frame and transmitted in parallel by using a multicarrier OFDM signal; therefore, the information on the sequential order of the transmit TS gets lost in the serial-to-parallel process. At the TSP separator, all null TSPs that were inserted at remultiplexer so that transmit TS would be kept constant clock are dropped to maintain modulation efficiency in each layer. To recover the transmit TS at the receiver, the demodulated TSPs data of each hierarchical layer must be synthesized in the correct order with null TSPs inserted at the same positions as in the transmit TS. A typical method for doing the above is to add information that indicates the packet order so that the packets can be correctly regenerated at the receiver. The following describes one possible method of attaching a sequence number to all UEHARA: APPLICATION OF MPEG-2 SYSTEMS TO TERRESTRIAL ISDB (ISDB-T) 265

Fig. 4. Configuration of input signal to the hierarchical divider. TSPs in every multiplex frame. From Table 6, one sees that the total number of TSPs included in one multiplex frame is at most 5120, which means that this method would require a minimum of 13 bits for a sequence number to each TSP. To avoid adding extra data such as the sequence number, an algorithm to prescribe the TSP order within a multiplex frame so that the transmit TS can be recovered at the receiver is introduced. Since an actual receiver treats the data in serial order after the FFT process, the two relations in the following enable the order of TSPs in a multiplex frame, i.e., the multiplex frame pattern, to be uniquely determined from the OFDM signal configuration. 1) the relation between carrier symbols of the OFDM signal and the serial signal output from the FFT; 2) the relation between the serial signal and the TSP order in a multiplex frame. These relations are specified by the simple operation of the receiver model shown in Fig. 3. If the transmit TS is constructed according to this specification at the transmitter, the receiver can recover it correctly. Moreover, an actual receiver designed to operate according to this receiver model would have the advantage that it does not require any computation to know the multiplex frame pattern. 1) Prescribing the Relation Between OFDM Carrier and FFT-Output Serial Signal: After carrier demodulation and deinterleaving, the serial signal, i.e., the signal input to the hierarchical divider, is made in ascending order of the segment number, and also in ascending order of the carrier frequency of the information symbol within a segment (obtained by excluding the carriers of the control symbol). Fig. 4 shows the configuration of this serial signal. Here, dummy data included in one OFDM symbol correspond to the sum of sampling (equivalent to pilot signals), FFT sampling (sampling in excess of the net signal band), and guard-interval sampling. The serial signal for digital audio broadcasting is made in a similar manner; here, the segment data included in each OFDM symbol consist of segment 0 in the one-segment format and segments 0-2 in the three-segment format. 2) Prescribing the Relation Between the Serial Signal and TSP Order in a Multiplex Frame: The serial signal, divided into multiple hierarchical layers, is then subjected to depuncturing before being stored in the hierarchical buffer. At this time, the number of bits that are depunctured and stored in the hierarchical buffer is determined by the convolutional-code rate and the modulation scheme of each hierarchical layer. Here, the processing delay time is assumed to be the same in each hierarchical layer and can be treated as zero. Switch S1 changes the hierarchical buffer when the stored data reach the amount of one TS packet, and the data are instantaneously transferred to the TS buffer. The TS reproduction part checks the existence of TS-buffer data or not. If data exist, it switches S2 over to the TS-buffer position and reads out one TS packet data. If it does not, the TS reproduction part switches S2 over to the null TSP position and inserts a null TSP. This process forms a signal having consecutive TSPs. Readout of the TS is carried out at the TS bit rate, that is, at four times for digital television broadcasting and the three-segment digital audio broadcasting and at two times for one-segment digital audio broadcasting. Switch S3 is used to alternately move between two TS reproduction units for inputting a hierarchical-combiner-output-signal at the beginning of an OFDM frame. Switch S4 is used to move between TS reproduction-unit signal outputs. Depending on the pattern of the input serial signal, i.e., the hierarchical configuration of the OFDM signal, TSP data might be left in the TS buffer at the end of an OFDM frame. All data should be output within the OFDM frame duration, which is the same length as a multiplex frame. Accordingly, in order to delay outputting TS from the TS reproduction unit, the movement of S4 is delayed relative to S3. To examine the delay time, an algorithm was run to search for the number of residual TSP data in the TS buffers at the end of an OFDM frame for all configurations of the OFDM signal. Table 7 shows the results of this examination. The maximum number of residual TSPs is two, which means that S4 has to be delayed by at least two TSPs. III. CORRECTLY RECOVERING PCR DURING PARTIAL RECEPTION The partial reception mechanism picks out only one OFDM segment at position no. 0 from the 13-segment format signal or the three-segment format signal. Using the signal structure, low-power handheld receivers that receive only one OFDM segment can be implemented. In terms of the OFDM signal, partial reception means that the receiver demodulates only the carriers of OFDM segment 266 PROCEEDINGS OF THE IEEE, VOL. 94, NO. 1, JANUARY 2006

Table 7 Maximum Number of Residual TSPs Fig. 5. (a) PCR packet transmission in partial reception. (b) PCR packet transmission under transmission restrictions (Mode 1). no. 0 in the multicarrier signal. However, in terms of the TS, it extracts only TSPs assigned for the partial-reception hierarchical layer to reconstruct a low-bit-rate TS from the 13-segment or three-segment high-bit-rate TS. In general, when the UEHARA: APPLICATION OF MPEG-2 SYSTEMS TO TERRESTRIAL ISDB (ISDB-T) 267

TS bit rate changes, the PCRs must be replaced in accordance with the new TS bit rate. PCRs that were correctly stamped for a high-bit-rate TS do not have correct values for a reconstructed low-bit-rate TS. In Fig. 5(a), for example, TSP0 and TSPn, which carry PCRs, are extracted from the high-bit-rate TS and reconstruct a low-bit-rate TS. Here, 0 and n denote locations in the multiplex frame. Given that the beginning of the frame represents a reference time point, the PCRs of TSP0 and TSPn are received at time points different from their original time points, and therefore, the PCR interval between TSP0 and TSPn differs between the high-bit-rate and low-bit-rate TS. To avoid this PCR jitter, the location of the TSP that carries PCR, i.e., the PCR packet, is restricted in the multiplex frame when it is remultiplexed. In Mode 1, for example, only one PCR packet should be multiplexed per service for the duration of a multiplex frame, and the multiplexing position must remain constant for all multiplex frames [see Fig. 5(b)]. In this way, although some difference in offset may occur in the low bit rate TS, the PCR interval will always be equivalent to the multiplex frame period as shown in Fig. 5(b). Consequently, no PCR jitter occurs and no special processing such as PCR correction is necessary at the receiver side. IV. CONCLUSION In this paper, we presented the issues associated with hierarchical transmission by BST-OFDM using MPEG-2 TS and described how we deal with them by using methods such as building the OFDM signal and multiplex signal for ISDB-T and minimizing receiver workload. The following summarizes the issues that were dealt with in this paper. Compatibility between OFDM segments and transport stream packets can be ensured by introducing a frame configuration of the OFDM signal and multiplex signal. An interface can be established with a fixed transmission clock by inserting null packets in a multiplex frame. Packets can be correctly separated and synthesized with respect to their appropriate hierarchical layers by prescribing a packet arrangement in a multiplex frame based on the receiver operation. In partial reception, setting restrictions in the PCR transmission on the transmit side enables simplified reception without special processing such as PCR correction at the receiver. REFERENCES [1] T. Kuroda and M. Sasaki, Terrestrial ISDB system using band segmented transmission scheme, in Proc. Int. Television Symp. ITVS 20 1997, pp. 641 654. [2] Generic coding of moving pictures and associated audio: Systems, ISO/IEC 13818-1, Nov. 1994. Michihiro Uehara received the B.E. degree in physical electronics and the M.E. degree in electric and electronics engineering from the Tokyo Institute of Technology, Tokyo, Japan, in 1987 and 1989, respectively. In 1989, he joined NHK, Tokyo, and from 1994 to 2004 was with NHK Science and Technical Research Laboratories, where he worked on channel coding and signal multiplexing for satellite digital broadcasting systems and terrestrial digital broadcasting systems. He is now Senior Engineer of the Planning Division, Engineering Administration Department, NHK. 268 PROCEEDINGS OF THE IEEE, VOL. 94, NO. 1, JANUARY 2006