IEEE TRANSACTIONS ON BROADCASTING, VOL. 54, NO. 2, JUNE 2008 249 Modulation and Pre-Equalization Method to Minimize Time Delay in Equalization Digital On-Channel Repeater Heung Mook Kim, Sung Ik Park, Jae Hyun Seo, Homin Eum, Yong-Tae Lee, Soo In Lee, and Hyuckjae Lee Abstract—This paper presents novel modulation and pre-equal- ization methods to minimize a signal processing time delay in the Equalization Digital On-Channel Repeater (EDOCR) for the ATSC terrestrial digital TV system. The proposed modulation method uses Equi-Ripple (ER) filter for Vestigial Side Bands (VSB) pulse shaping instead of conventional Square Root Raised Cosine (SRRC) filter. And the proposed pre-equalization method calculates pre-equalizer filter coefficients by comparing a base- band signal as a reference signal and a demodulated repeater output signal, and then creates new VSB pulse shaping filter coef- ficients by the convolution of the ER filter and the pre-equalizer filter coefficients. The new VSB pulse shaping filter minimizes the time delay of EDOCR by adjusting the number of its pre-taps and also compensates the linear distortions due to the use of the ER filter and mask filter. Index Terms—ATSC, modulation, on-channel repeater, pre-equalization, SFN. I. INTRODUCTION TERRESTRIAL television broadcasters in general operate transmitters and translators according to the geographical locations of their coverage areas. In both analog and digital tele- vision broadcasting, Multiple Frequency Networks (MFNs) that assign different channels to each transmitter and translator have been used to cover service areas. However, the use of MFNs is very inefficient in the aspect of using frequencies since it is unable to share channels among a number of transmitters and translators unless the distance between two coverage areas is far enough. Therefore, Single Frequency Networks (SFNs) that operate multiple transmitters and repeaters on the same frequency is de- sirable for the efficient use of frequencies. Especially, in the re- cent transition period from analog to digital broadcasting, the need of SFNs is unavoidable due to the lack of frequencies for additional transmitters and repeaters. SFNs provide not only high Signal to Noise Ratios (SNR), but trigger the mobile DTV Manuscript received February 7, 2007; revised February 22, 2008. H. M. Kim, S. I. Park, J. H. Seo, H. Eum, and Y.-T. Lee are with the Ter- restrial Broadcasting Technology Research Team, ETRI, Yuseong-gu, Daejeon, 305-700, Korea (e-mail: hmkim@etri.re.kr; psi76@etri.re.kr; jhseo@etri.re.kr; hmeum@etri.re.kr; ytlee@etri.re.kr). S. I. Lee is with the Broadcasting System Research Department, ETRI, Yuseong-gu, Daejeon, 305-700, Korea (e-mail: silee@etri.re.kr). H. Lee is with the Radio & Communications Laboratory, ICU, Yuseong-gu, Daejeon, 305-714, Korea (e-mail: hjlee@icu.ac.kr). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TBC.2008.921371 Fig. 1. Block diagram of EDOCR. services [1], [2]. Recently, SFNs are considered for use in ter- restrial Advanced Television System Committee (ATSC) Dig- ital Television (DTV) services because of the performance im- provement of DTV receiver which are able to compensate for the long-time delay and high level ghost [3]. In the ATSC 8-VSB system, SFNs can be implemented with DTxT (Distributed Transmitters) that uses the same fre- quency among a number of transmitters, and/or with Digital On-Channel Repeaters (DOCRs) that uses the same frequency between transmitters and repeaters [4], [5]. The disadvantages of DTxT are that some devices maintaining the frequency synchronization between SFN transmitters must be added to existing transmitters and that the distance between the trans- mitters can be restricted by the limited equalization range of receivers. DOCRs do not need to change existing transmitters, but they produce limited output power and low quality of signal. As complementary to existing DOCRs, the Equalization DOCR (EDOCR) has been proposed [6], [7]. This paper presents the operational requirements of the EDOCR modulator and pre-equalizer, and proposes its config- uration to meet the requirements. The proposed modulation and pre-equalization method is analyzed by computer simulations, and it is also confirmed by laboratory tests. 0018-9316/$25.00 © 2008 IEEE 250 IEEE TRANSACTIONS ON BROADCASTING, VOL. 54, NO. 2, JUNE 2008 Fig. 2. Block diagram of VSB modulator. II. CHARACTERISTICS OF EDOCR DOCRs are used to fill in coverage gaps and to extend coverage areas which transmitter can not cover. Conventional DOCRs such as the RF processing DOCR and the IF processing DOCR offer a short processing time, but they provide limited transmitting power, low quality of output signal, and inadequate adjacent channel rejection. The EDOCR has been proposed to overcome such disadvantages of conventional DOCRs and its configuration is shown in Fig. 1. The EDOCR system takes the following advantages: • Since the EDOCR does not use Forward Error Correction (FEC) decoding and encoding, it does not have the ambi- guity problem in which the DOCR output symbol stream differs from its input symbol stream. • The EDOCR has good selectivity of the received signal due to utilizing a matched filter in demodulation. That is, it is capable of rejecting adjacent channels. • The EDOCR uses a blind Decision Feedback Equalizer (DFE), which includes the trellis decoder as a decision de- vice with TBD (Trellis Back Depth) of 1 [8]. The DFE is able to remove noise and multipath signals caused by the signal paths between the main transmitter and the EDOCR, so that the quality of output signal is better than that of the input signal. Also, since the equalizer rejects feedback signal due to low isolation between transmitting and re- ceiving antennas, the transmitting power of EDOCR can be increased more than 10 times higher than that of the conventional DOCRs. • Because of the re-modulation and pre-equalization, the EDOCR can transmit good quality of signal. The EDOCR involves a lot of digital signal processing, which possibly causes a long time delay between transmitted and received signals compared with conventional DOCRs. Due to the non-inclusion of FEC decoding and encoding, however, its signal processing time can be limited within 6 [5], [6]. The time delays of each module are 1 in the demodulator, 1 in the equalizer, 3 in the modulator, and 1 in the RF systems and cables. III. EDOCR MODULATOR AND PRE-EQUALIZER A. EDOCR Modulator The block diagram of a VSB modulator used in the ATSC terrestrial DTV transmitter or repeater is shown in Fig. 2, and its operation has following procedure: • Step 1: The data consisting of the equalizer output, the field sync and the segment sync is up-sampled after pilot insertion. • Step 2: The up-sampled data is filtered by a VSB I/Q pulse shaping filter. • Step 3: The VSB filtered I/Q components with the center frequency of 2.69 MHz is up-converted to the center fre- quency of , and combined to form the IF signal. A SRRC filter is generally used for VSB pulse shaping filter in the ATSC system and the VSB I/Q filters based on the SRRC filters are where is a time index, is a SRRC filter coefficient ac- cording to the time index, is 2.69 MHz, and is a symbol time (about 93 ns). The VSB modulated signal must meet the FCC emissions mask shown in Fig. 3 and maintain the output SNR greater than 27 dB [9]. Assuming that the up-sampling rate for VSB filtering is 4, Fig. 4 shows the output SNR and the spectrum shoulder amplitude according to the number of SRRC filter taps. The shoulder amplitude is the power difference between the ampli- tude of the spectral regrowth spectrum at the channel’s edge and the total average DTV power. To meet the emissions mask requirement, the shoulder ampli- tude must be greater than 47 dB. Suppose that the number of taps of the matched filter is 121 to measure the SNR while observing 200,000 symbols through an ideal channel. According to the Fig. 4, the VSB pulse shaping filter based on the SRRC filter should theoretically have more than 420 taps to meet the output SNR and emissions mask requirements simultaneously. How- ever, when the symbols are over-sampled at 4 times the ATSC KIM et al.: MODULATION AND PRE-EQUALIZATION METHOD TO MINIMIZE TIME DELAY IN EQUALIZATION DIGITAL ON-CHANNEL REPEATER 251 Fig. 3. FCC emissions mask. Fig. 4. SNR and shoulder amplitude according to the number of the SRRC filter taps. system symbol rate, it causes a time delay of about 4.9 which results in a relatively long delay in the EDOCR system. Since the time delay is critical, a new pulse shaping filter is required for the EDOCR modulator. Since the number of filter taps to satisfy the requirements is determined by the shoulder ampli- tude rather than the SNR according to the Fig. 4, the new pulse shaping filter must be designed to have large shoulder amplitude while maintaining the number of taps as small as possible. An ER filter that has good capability of out-of-band suppres- sion while allows relatively lots of in-band ripples can be used as a pulse shaping filter in the EDOCR system for a short time delay. The VSB I/Q filters based on the ER filter are Fig. 5. SNR and shoulder amplitude according to the number of the ER filter taps. where is a time index, is an ER filter coefficient according to the time index, is 2.69MHz, and is a symbol time. Fig. 5 shows the output SNR and the shoulder amplitude ac- cording to the number of ER filter taps, and ER filter coeffi- cients are calculated by Parks-McClellan algorithm [10], [11]. The ER filter with greater than about 140 taps can meet the output SNR and emissions mask requirements simultaneously according to the Fig. 5. When the symbols are over-sampled at 4 times the symbol rate, it causes a time delay of about 1.6 which is adequate as a pulse shaping filter in the EDOCR system. The ER filter has good capability of out-of-band suppression, but it causes lots of in-band ripples which are not ideal characteristic of Nyquist pulse shaping filter. Therefore, the output SNR of 252 IEEE TRANSACTIONS ON BROADCASTING, VOL. 54, NO. 2, JUNE 2008 Fig. 6. The modulator and conventional pre-equalizer. the ER filter is lower than that of the SRRC filter when the same number of taps is used. B. Pre-Equalization Method To meet the FCC emissions mask requirement, an EDOCR uses a mask filter which is capable of out-of-band suppression after a high power amplifier. The mask filter with good out-of- band suppression capability causes a lot of in-band group delay which degrades the output signal quality. Also, there is a pos- sibility of additional SNR degradation due to the use of an ER filter as a pulse shaping filter. To compensate these SNR degra- dations, a pre-equalizer is used. Fig. 6 shows the configura- tion of the modulator and the conventional pre-equalizer. In the pre-equalizer, its filter coefficients are calculated by com- paring the baseband signal to be transmitted and the demodu- lated channel filter output signal of the EDOCR. A pre-equalizer filter in general is a linear filter and its coeffi- cients can be calculated using Least Mean Square (LMS) algo- rithm. To update the coefficients, the following variables must be defined. : baseband signal to be transmitted at time , : demodulated signal after channel filtering at time , : pre-equalizer output signal at time , : -th filter tap coefficient of pre-equalizer at time . Thus, the pre-equalizer output is where is the number of the pre-equalizer filter taps. The number of taps is determined by the degree of linear distortion such as group delay. To obtain the update formula for filter tap coefficients, the error signal is defined The filter tap coefficients are updated as where is a step size which determines convergence speed and steady state Mean Square Error (MSE). For a large step size value the convergence speed is fast, but the steady state MSE is large. Otherwise, for a small step size value the steady state MSE is small, but the convergence speed is slow. To update the tap coefficients, the EDOCR uses known symbols as a training sequence, instead of the decision symbols of the VSB Demod- ulator output in Fig. 6. Therefore, it is recommended to use a small step size for a small steady state MSE although the convergence speed is slow [12]. The modulator including the pre-equalizer can compensate the linear distortions and reduce the ripples caused by the use of a mask filter and an ER filter, so that the output SNR of EDOCR can be improved. C. Combination of Pre-Equalizer Filter and Pulse Shaping Filter The symbol level pre-equalizer filter shown in Fig. 6 is one of the factors causing a time delay in the EDOCR. To minimize the time delay, the method of combining the pre-equalizer filter and the pulse shaping filter, and adjusting the number of the com- bined filter’s pre-taps is proposed in this section. Precisely, the time delay can be minimized by truncating the number of pre- taps after convolution of the pre-equalizer filter and the pulse shaping filter. The configuration of the EDOCR modulator in- cluding the proposed pre-equalizer is shown in Fig. 7, and the process of combining two filters and adjusting the number of pre-taps of combined filter is shown in Fig. 8. Assume that there are a pre-equalizer filter in which the total number of taps is and the main tap is positioned at , and a VSB I/Q filter in which the total number of taps is and the main tap is positioned at . After convolution of the pre-equalizer filter and the VSB I/Q filter, a combined VSB I/Q filter functioning pre-equalization in which the total number of taps is and the main tap is positioned at is created. And some of the left most filter coefficients of the combined VSB I/Q filter are truncated to reduce the processing time delay of EDOCR. So the truncated VSB I/Q filters have the total number of taps of and its main tap is positioned at where KIM et al.: MODULATION AND PRE-EQUALIZATION METHOD TO MINIMIZE TIME DELAY IN EQUALIZATION DIGITAL ON-CHANNEL REPEATER 253 Fig. 7. The configuration of modulator and proposed pre-equalizer. Fig. 8. Process of combining pre-equalizer filters and pulse shaping filter and adjusting the number of pre-taps. and . The post-taps can also be truncated to be accommodated in limited hardware re- sources. By such adjustment of the pre-taps, the truncated VSB pulse shaping filter can minimize the time delay in the EDOCR. Due to the pre-equalization, it can also compensate the linear distortions and reduce the in-band ripples so that the output SNR of the EDOCR can be significantly improved. IV. SIMULATION AND LABORATORY TEST RESULTS A. Simulation Results The computer simulations have been performed based on the configuration of the EDOCR modulator shown in Fig. 7. The up-sampling rate for VSB filtering was assumed as 4, and the ER filter with 191 taps was used as a pulse shaping filter. The linear distortions which can be caused by a high power amplifier were not considered, and the mask filter was modeled as the 8th order Chebyshev filter. Fig. 9 shows the magnitude and group delay characteristic of the designed mask filter. To calculate the pre-equalizer filter coefficients, the LMS algorithm was used. The total number of the pre-equalizer filter taps was set to 101 and its main tap was positioned at 51 to maintain the output SNR greater than 35 dB in symbol rate data. The time delay of the pre-equalizer filter itself is 4.74. that is relatively long. The number of taps of the matched filter was set to 121 to measure the SNR and an ideal channel with no multi-path and no additive noise was assumed while observing 200,000 symbols. Fig. 10 shows the simulation results of the pre-equalization when the pre-equalizer filter and the pulse shaping filter were used separately as shown in Fig. 6. Fig. 10(a) shows the EDOCR output constellation before pre-equalizing, in which the output SNR is about 14.1 dB, and Fig. 10(b) shows that after pre-equalizing, in which the output SNR is about 35.3 dB. The output SNR after pre-equalizing is greater than that of the ER filter with 191 taps (32.88 dB) in the Fig. 5 since the pre-equalizer can reduce in-band ripples due to the use of the ER filter. Fig. 11 shows the SNR and the shoulder amplitude in case of adjusting the number of the pre-taps after convolution of the pre-equalizer filter and the pulse shaping filter. In order to maintain the shoulder amplitude 254 IEEE TRANSACTIONS ON BROADCASTING, VOL. 54, NO. 2, JUNE 2008 Fig. 9. Magnitude and group delay characteristic of 8th order Chebyshev filter. (a) Magnitude characteristic. (b) Group delay characteristic. greater than 47 dB and the SNR greater than 27 dB, the number of the combined filter’s pre-taps should be greater than 95 and then the time delay becomes about 2.21. Therefore, the newly created filter by adjusting the number of the pre-taps after convolution of the pre-equalizer filter and the pulse shaping filter minimizes the time delay while maintaining the required SNR and shoulder amplitude. B. Laboratory Test Results To verify the performance of the proposed modulator and pre-equalizer in the EDOCR, a hardware was implemented and the EDOCR output was measured by RFA300A, the VSB test and measurement equipment. The implemented EDOCR system used the ER filter with 191 taps and the pre-equalizer filter in which its main tap was positioned at 51 in symbol rate was cal- culated by LMS algorithm. To reduce the time delay as possible without violation of the EDOCR requirements, the number of the pre-taps was adjusted as 95 which is the same number of the pre-taps of the ER filter after convolution of the ER filter and the pre-equalizer filter. Thus, the time delay in the modu- lator including the pre-equalization is 2.21 that are the same as in the ER filter only. The EDOCR output signal was verified by Fig. 10. Constellation of EDOCR output signal before and after pre-equal- ization. (a) Constellation of EDOCR output signal without pre-equalization (SNR = 14:1 dB). (b) Constellation of EDOCR output signal with pre-equal- ization (SNR = 35:2 dB). Fig. 11. SNR and shoulder amplitude according to the number of pre-taps after convolution of pre-equalizer filter and pulse shaping filter. RFA300A, and its spectrum, frequency response, group delay, and constellation before and after pre-equalization are shown in Fig. 12. Fig. 12 proves that the EDOCR output meets the KIM et al.: MODULATION AND PRE-EQUALIZATION METHOD TO MINIMIZE TIME DELAY IN EQUALIZATION DIGITAL ON-CHANNEL REPEATER 255 Fig. 12. Spectrum, frequency response, group delay, and constellation before and after pre-equalization. (a) Spectrum of EDOCR output signal (Left: before pre-equalization, Right: after pre-equalization). (b) Frequency response and group delay of EDOCR output signal (Left: before pre-equalization, Right: after pre- equalization). (c) Constellation of EDOCR output signal (Left: before pre-equalization, Right: after pre-equalization). spectrum mask and SNR requirements. Usually the linear dis- tortions caused by mask filter and other RF components in re- peater system are not so severe that they can be compensated by a linear filter with relatively small number of taps compared to that of the VSB pulse shaping filter. According to the labora- tory test results, it can be predicted that if the effective number of pre-taps of the combined filter is greater than that of the orig- inal pulse shaping filter, the proposed system would not perform as well as when the two filters are not combined but it still meets the FCC requirements. 256 IEEE TRANSACTIONS ON BROADCASTING, VOL. 54, NO. 2, JUNE 2008 V. CONCLUSIONS This paper presents the modulation and pre-equalization methods to minimize the time delay of the EDOCR. The proposed modulation method uses an ER filter as a VSB pulse shaping filter instead of a conventional SRRC filter. And the proposed pre-equalization method calculates the pre-equalizer filter coefficients by comparing a baseband signal as a reference signal and a repeater output signal, and then creates new VSB pulse shaping filter coefficients by the convolution of the ER filter and the calculated pre-equalizer filter coefficients. Ac- cording to the computer simulation and laboratory test results, the proposed methods have met the FCC requirements without causing significant system delay. REFERENCES [1] A. Mattsson, “Single frequency networks in DTV,” IEEE Trans. Broad- casting, vol. 51, no. 4, pp. 413–422, Dec. 2005. [2] Y. T. Lee, S. I. Park, S. W. Kim, C. T. Ahn, and J. S. Seo, “ATSC terres- trial digital television broadcasting using single frequency networks,” ETRI Journal, vol. 26, no. 2, pp. 92–100, April 2004. [3] O. Bendov, “Areas of cochannel interference and multi-path created by 8-VSB modulated distributed transmitters in flat terrain,” IEEE Trans. Broadcasting, vol. 52, no. 1, March 2006. [4] ATSC, “Standard A/110: Synchronization Standard for Distributed Transmission,” Advanced Television Systems Committee, Wash- ington, D.C., July 14, 2004. [5] ATSC, Recommended Practice A/111: Design of Synchronized Mul- tiple Transmitter Networks Advanced Television Systems Committee, Washington, D.C., Sep. 3, 2004. [6] S. W. Kim, Y.-T. Lee, S. I. Park, H. M. Eum, J. H. Seo, and H. M. Kim, “Equalization digital on-channel repeater in single frequency net- works,” IEEE Trans. on Broadcasting, vol. 52, no. 2, June 2006. [7] Y.-T. Lee, S. I. Park, H. M. Eum, J. H. Seo, H. M. Kim, S. W. Kim, and J. S. Seo, “A design of equalization digital on-channel repeater for single frequency network ATSC system,” IEEE Trans. on Broad- casting, accepted for publication. [8] H.-N. Kim, S. I. Park, and S. W. Kim, “Performance analysis of error propagation effects in the DFE for ATSC DTV receivers,” IEEE Trans. on Broadcasting, vol. 49, Sept. 2003. [9] ATSC, “Standard A/64-Rev.A: Transmission Measurement and Com- pliance for Digital Television,” Washington, D.C., May 30, 2000. [10] T. W. Parks and J. H. McClellan, “Chebyshev approximation for nonre- cursive digital filters with linear phase,” IEEE Trans. on Circuit Theory, vol. CT-19, pp. 189–194, 1972. [11] T. W. Parks and J. H. McClellan, “A program for the design of linear phase finite impulse response filters,” IEEE Trans. on Audio Electroa- coustics, vol. AU-20, pp. 195–199, 1972. [12] G. A. Clark, S. K. Mitra, and S. R. Parker, “Block implementation of adaptive digital filters,” IEEE Trans. on Circuits and Systems, vol. 28, no. 6, June 1981. Heung Mook Kim received the B.S. and M.S. degrees in electronics and electrical engineering from POSTECH, Pohang, Korea, in 1993 and 1995 respectively. From February 1995 to January 2002, he was with POSCO Technology Research Labo- ratory in the field of Measurement and Monitoring as research engineer. Since February 2004, he has been with the Broadcasting System Research Group, Electronics and Telecommunication Research Insti- tute (ETRI), where he is a senior member of research staff. Also, he is currently at Information and Com- munications University (ICU) pursuing Ph.D. degree. His research interests are in the areas of digital and RF signal processing and RF transmission for digital communications and digital television. Sung Ik Park received the BSEE from Hanyang University, Seoul, Korea, in 2000 and MSEE from POSTECH, Pohang, Korea, in 2002. Since 2002, he has been with the Broadcasting System Re- search Group, Electronics and Telecommunication Research Institute (ETRI), where he is a member of research staff. His research interests are in the areas of error correction codes and digital commu- nications, in particular, signal processing for digital television. Jae Hyun Seo received the BSEE and MSEE from Kyungpook National University, Daegu, Korea, in 1999 and 2001 respectively. Since January 2001, he has been with the Broadcasting System Research Group, Electronics and Telecommunication Re- search Institute (ETRI), Daejeon, Korea, developing advanced transmission and reception technology for terrestrial digital television. His research interests include digital signal processing, spatiotemporal signal processing, in particular, signal processing for digital television and digital communications. Homin Eum received the BSEE and MSEE from Korea University, Seoul, Korea, in 1998 and 2000 respectively. Since May 2000, he has been with Electronics and Telecommunication Research Insti- tute (ETRI), where he is a senior member of research staff. His main research interests are in the areas of digital communication systems, digital signal processing and DTV transmission systems. Yong-Tae Lee received the BSEE and MSEE from Hankuk Aviation University in 1993 and 1995 re- spectively and Ph.D. degree from Yonsei University, Seoul, Korea in 2007. Since 1995, he has been with the Radio Signal Processing Department and Broad- casting System Research Department, Electronics and Telecommunication Research Institute (ETRI), where he is a senior member of research staff. His research interests are in the area of digital signal processing and RF signal processing, in particular, signal processing for digital television, digital communication and analog narrow band communication. Soo In Lee received the M.S. and Ph. D degrees, all in electronics engineering from Kyungpook Na- tional University, Daegu, Korea, in 1989 and 1996. In 1990, he joined Electronics and Telecommunication Research Institute (ETRI), Daejeon, Korea, where he has been working on broadcasting system technolo- gies. Currently he serves as the Director for Broad- casting System Research Group. His research inter- ests include terrestrial DTV and DMB systems, dig- ital CATV systems, and 3DTV systems. Hyuckjae Lee was born in Inchon, Korea. He received B.S. degree in electronic engineering from Seoul National University, Korea, in 1970, and the Ph.D. degree in electrical engineering from Oregon State University, Corvallis, in 1982, where he spe- cialized in electromagnetic fields and microwave engineering. Since 1983, he has been with the Radio Technology Department, Electronics and Telecom- munications Research Institute (ETRI), and has been working in the fields of radio technology, IMT2000, broadcasting technology, and satellite communica- tions system. He is currently a professor of Information and Communications University, Daejeon, Korea.