NHK Laboratories Note No. 469

A Loop Interference Canceller for the Relay Stations
in an Single Frequency Network for Digital Terrestrial Broadcasting

Hiroyuki HAMAZUMI, Koichiro IMAMURA, Naohiko IAI, Kazuhiko SHIBUYA and Makoto SASAKI
(Digital Broadcasting Networks Research Division)
    In each relay station in an SFN (single frequency network) for DTB (digital terrestrial broadcasting), the loop interference is caused by the coupling from the transmitting antenna to receiving antenna. The interference must be reduced to an allowable level in order to avoid problems with distortion and oscillation. We propose a frequency-domain adaptive cancellation algorithm which is operated by using Scattered Pilot information in the OFDM signal. In this paper, the principle of the proposed algorithm is outlined and the usefulness of the algorithm is verified by the results of computer simulations.
1.1 Set-up and Observation Method
    ln the world's leading industrial countries, recent progress in digital transmission has been dramatic, and technologies for digital terrestrial broadcasting are also growing accordingly [1, 2]. In Japan, OFDM (orthogonal frequency-division multiplexing) scheme has been selected as the transmission scheme of DTB, because of the possibility of SFN construction and its robustness against multi-path interference [3].
    Signal distributions to SFN stations can be achieved by using microwave links, optical fiber lines and broadcast-wave relays. Because of a lack of microwave frequencies and the excessive cost of fiber lines, the former two means would not become major schemes of signal distribution in Japan.
    The development of a broadcast-wave relay SFN is necessary for cost-effective introduction of DTB. In this case, the loop interference caused by the coupling from transmitting antenna to receiving antenna is a particular problem. The interference must be reduced to allowable level to avoid signal distortion and oscillation. So far the authors have been conducting a variety of measurements to clarify the nature of this effect [4, 5, 6], and have also been investigating countermeasures.
    In this report, we outline a way of using frequency-domain signal processing for adaptive loop interference canceller in an OFDM transmission scheme. The principle in which loop interference is reduced are first discussed in detail. An adaptive frequency- domain algorithm that uses scattered pilot (SP) signals in the broadcast OFDM signal is then described. Finally, the performance and utility of proposed canceller is described with the results of the computer simulation.

2. Loop Interference Reduction
    There are three basic types of countermeasures against loop interference in SFN relay stations:
(1) using topographical features like mountains and buildings lo reduce the interference;
(2) using the antenna's beam pattern to reduce the interference; and
(3) circuit-based techniques for canceling the interference.
The case of (1) is the traditional approach. The receiving and transmitting antennas must be quite distant from each other to keep coupling coefficient small enough, which means that the SFN relay station occupy a large area. In examining the three approaches, we feel that countermeasures of type (1) are not feasible, at least as the sole type of countermeasure, because of the wide variety of mountainous terrain in Japan, and the resulting variety of conditions for building. Countermeasures of type (2) alone are not able to reduce loop interference sufficiently. We therefore felt that the most effective approach would be to use countermeasures of type (3), circuit technology, in conjunction with countermeasures of types (1) and (2).
    There are two basic approaches to the cancellation of loop interference by using circuit technology: (a) a regeneration scheme at the relay station, in which the original signal is decided and regenerated; and
(b) a direct relay scheme in which neither decision of the signal nor signal regeneration.

Signal regeneration at relay points has been applied to radio-based pager systems [8]. Direct relay schemes have yet to be sufficiently investigated. When approach (a) above is applied to the relaying of the OFDM signals that are used for digital terrestrial broadcasting, a long delay is incurred because of the processing involved in regeneration.
    Here, a long delay means a delay of more than guard interval period in the OFDM signals. Consequently, in a broadcast-wave relay system in which the service areas of the host station and substations (SFN relay stations) overlap, the differences in arrival times will exceed the guard interval. In such a case, inter-symbol interference (ISI) occurs in the received OFDM signals. Regeneration at relay stations is thus impossible to apply. In this report, therefore, we focus on the direct relay system, approach (b) above.

3. Loop Interference Canceller For OFDM
3.1 Basic principle of proposed system
    The model for the loop interference canceller in a direct relay scheme that we have used in our study is shown in Fig.1. Here, because our focus is on canceling loop interference in OFDM signals, we assume that signals are translated into frequency domain signals by Fourier transform at estimation point P, and signals are expressed as the frequency domain signal at each point.
    To begin with, let the transmitted signal at the host station be X(), and system noise at the host station be Na(). The received signal R() then becomes the sum of these signals at the relay station.

Next, let the frequency characteristic of the coupling path be C(), the gain of the amplifier at the relay station be G(), noise added to the signal as received at the relay station be Nb(), the frequency characteristic of the cancellation path be W(), and the frequency spectrum of the signal at estimation point P be S(). The frequency spectrum of the each signal on this figure I() and S() can then be expressed by the following equations.

Figure 1. Model of broadcast-wave relay SFN used in our basic study

Now, substituting Eq. (2) in Eq. (3) and rearranging terms, we get Eq. (4) below.

Next, substituting Eq. (1) in Eq. (4), we get the following.

Accordingly, the transfer function of the entire system F() will be as follows.

The condition for loop interference to be canceled out is thus that G()C() = W() holds in Eq. (6). Here, if we transform Eq. (6) in terms of Er() = G()C() - W(), we get the following equation.

If control can be performed in such a way as to minimize the value of Er(), loop interference can be cancelled. In particular, if Na() and Nb() are sufficiently smaller than X(), we get the following expression.

However, to maintain tracking in the face of fluctuations in loop interference as the effects of Na() and Nb() are minimized, an adaptive algorithm that uses a special pilot signal must be introduced.

3.2 Introduction of a pilot signal
    The use of SPs (scattered pilots) as pilot carriers for coherent detection has been included in the current standard specifications for digital terrestrial broadcasting in Europe and Japan. The arrangement of SPs is as shown in Fig.2. Let k ( total number of carriers) be the carrier number, i the symbol number (integer), and p a non- negative integer. The position of an SP can then be defined by Eq. (9), and the condition that k = kp are SP carriers. Continual pilot (CP) carriers are also inserted in the right-hand edge (k = K-1).

3.3 Adaptive algorithm
    Figure 3 is a system diagram of the proposed loop inteference canceller. The signal at estimation point P, S(t), is converted into a frequency domain signal by using FFT (fast Fourier transform), and the resulting signal is indicated as S(n, k). Here, n is the iteration number for coefficient updating at the FIR (finite impulse response) filter. The tramsmitted signal X(n, k) is known because of the SP and CP carriers, and we get the following transfer function Fp(n, k).

Figure 3. System diagram of SFN relay station using a loop interference canceller
    After this calculation has been performed, interpolation processing is carried out on the data carrier. This interpolated signal is indicated as F(n, k) and consists of the following operation, making use of Eq.(8).

Here, er(n, k) is the impulse response of the residual loop interference, amd IFT indicates the inverse Fourier transform.
    To update the coefficients of the FIR filter, er(n, k) are multiplied by a rectangular window function r(k), and an coefficient is introduced to suppress noise. The coefficients are then sequentially updated by applying the following equation.

Here, is not greater than 1, amd M is the total number of taps on the FIR filter. Sequentially updating the coefficients allows tracking to be performed, even when the loop interference is fluctuating.

4. Computer Simulations
4.1 Overview of simulation
    The proposed system was computer simulated to evaluate its utility. Figure 4 is a system diagram of these simulations. Each signals in the simulation were treated as equivalent base-band signals and all calculations were at double precision. Ideal conditions were assumed for frequency synchronization and clock timing. In the Figure 4, AWGN-a addition represents system noise at the host station, and AWGN-b and AWGN-c addition represent head amplifier noises at the relay station and OFDM receiver respectively.

Figure 4. System diagram of computer simulations

(1) Canceling opejnalion
    Simulation parameters and conditions associated with the canceling of loop interference are given in Table 1. The loop interference canceller in this paper uses SPs in its estimation of the coupling transfer function. The FIR-filter coefficients are updated once every two OFDM symbols. The positions of data carrier require some type of interpolation processing. In these simulations, interpolation in the carrier direction was performed by linearly approximating every 12 carriers.

(2) QAM-OFDM modulation and demodulation
    The 64QAM-OFDM modulation and demodulation sections used to evaluate bit error rate for the SFN relay scheme. The modulation parameter was decided by the ISDB-T standard system description[3]. The error rate was evaluated without error correction. Parameters of these 64QAM-OFDM modulation zmd demodulation sections are shown in Table 2. In the two-dimensional interpolation filter [9, 10], a symbol filter that holds the most recent value is used. An ideal rectangular filter is used as a carrier filter, and decreases the bandwidth by 1/4 by using 2048-point FFT and IFFT.

Table 1. Parameters used for calculations in cancellation

Table 2. Parameters of OFDM modulation/demodulation sections

4.2 Learning Curves
    Learning curves were first computer simulated to verify the cancellation of loop interference by the proposed system. Here, we took averages in the carrier direction, of each symbol's signal-to-noise ratio at the receiver using ideal synchronous detection to evaluate the learning curves. The received signal at the receiver is given as u(i, k) and the decision value fTor the received signal u(i, k) as d(i, k), where k is the carrier number. The signal-to-noise ratio at the time of a symbol i is indicated as and defined by Eq.(14). Here, the error vector err(i, k) is given by the equation (15), and '*' indicates complex conjugation.

    Simulation results for the loop interference canceller in terms of learning curves are shown in Fig. 5. The vertical axis in the figure represents the average signal-to-noise ratio , and the horizontal axis indicates iteration number of the coefficient update at the FIR filter (n) that are performed once every two symbols. The host station system noise was taken to be C/Na = 40dB, and the head amplifier noise at the OFDM receiver was taken to be C/Nb = 40dB. Loop interference amplitude ratio pc = 2 dB, delay time = 10.09 s, and phase = 45 degrees.

Figure 5. Learning curves of loop interference canceller (simulation)

Figure 6. Averaged BER characteristics at receiver while broadcast-wave relaying is subjected to a loop interference (simulation)

4.3 CNR versus BER characteristics
    Computer simulations were performed to obtain average bit error rates under the Gaussian channel, to evaluate the utility of the proposed canceller in a broadcast-wave relay SFN scheme. The results are shown in Fig. 6, and were obtained under the single loop interference environment. Here, When C/Na = 40 dB for system noise at the host station and C/Nb = 40 dB for noise at the receiver section of the relay station, noise levels C/Nc at the receiver vary from 10 dB to 30 dB, the range shown along the horizontal axis. The vertical axis shows the average bit error rate (BER). Loop interference parameters were as follows: amplitude ratio pc = 2 dB, delay time = 10.09 s, and phase = 45 degrees. The coefficient was set to 1.0. Furthermore, on the basis of simulation results for convergence, 100 iterations of coefficient updating were performed, and the BER after convergence was calculated. It can be seen that the BER is significantly improved by the introduction of the loop interference canceller.

5. Conclusions
    In this paper, we have described a loop interference canceller for the broadcast- wave relays of an SFN. The canceller uses an adaptive FIR filter. The transfer function for the filter is determined from scattered pilot carriers that are contained in the QAM-OFDM signals. In order to reduce the loop interference between the transmitting zmd receiving antennas, this canceller employs a frequency-domain adaptive canceling scheme. The proposed scheme can cancel loop interference effectively, at least in the computer simulations, as long as the DU (desired-undesired) ratio of the loop interference is +1dB or over.
    In future research on this canceling scheme, we plan to study its characteristics in terms of tracking fluctuations in loop interference and the effects of nonlinear distortion, quantization errors, clock timing errors, frequency errors, and the like. An experimental canceller will be built soon after that, and field trials to demonstrate the utility of the scheme will follow immediately.


[1] ATSC Document A/53, "ATSC Digital Television Standard", Sep. l995
[2] ETS 300 744, "Digital Broadcasting Systems for Television, Sotuld and Data Services: Framing Structure, Channel Coding and Modulation for Digital Terrestrial Television", 1997
[3] M.Uehara, M.Takada, and T.Kuroda: "Transmission Scheme for the Terrestrial ISDB System", IEEE Transactions on Consumer Electronics, Vol.45, No.1, Feb.1999
[4] K.Imamura, N.Iai, K.Shibuya, and M.Sasaki: "A Measuring Method of Coupling Effects between Transmitting and Receiving Antenna in Relay Station - A Basic Study for Relay SFN-"(in Japanese), ITE Technical Report, Vol.22, No.7, pp.43-48, Jan.1998
[5] K.Imamura, N.IAI, K.Shibuya, and M.Sasaki: "A Basic Study for SFN -Measurement of Coupling between Transmitting and Receiving Antenna - "(in Japanese), ITE Technical Report, Vol.22, No.34, pp.13-18, Jun.1998
[6] K.Imamura, K.Shibuya, and M.Sasaki: "A Basic Study for Relay SFN Analyses of Fluctuations in Coupling" (in Japanese), ITE Technical Report, Vol.22, No.59, pp.37-42, Oct.1998
[7] Simon Haykin: "Adaptive Filter Theory", Third Edition, Prentice-Hall, 1996
[8] H.Suzuki, K.Itoh, Y.Ebine, and M.Sato: "A booster configuration with adaptive reduction of transmitter-receiver antenna coupling for pager systems", Proc. IEEE VTS 50th Vehic. Technol. Conf., pp.1516-1520, Sep.1999.
[9] Vittoria Mignone and Alberto Morello: "CD3-OFDM: A Novel Demodulation Scheme for Fixed and Mobile Receivers", IEEE Trans. on Communications, Vol.44, No.9, Sep.1996
[10] Y.Harada, T.Kimura, K.Hayashi, A.Kisoda, S.Kageyama, S.Sakashita, and H.Mori: "An Implementation of OFDM Receiver for Digital Terrestrial Television Broadcasting and Its Technologies", IEE, International Broadcasting Conference, No.447, Sep.1997

Hamazumi Mr. Hiroyuki Hamazumi
Hiroyuki Hamazumi graduated from the Fukui technical college, Fukui, Japan, in 1982, and joined NHK in April of the same year. After working with the field staff, he has been with NHK Science and Technical Research Laboratories since 1987. He has been engaged in the research on the digital transmission such as CDM, OFDM, adaptive equalization and diversity reception. He received a Dr.Eng. degree from Tokyo Institute of Technology, Tokyo, Japan, in 1999.
Imamura Mr. Kohichiro Imamura
Mr. Imamura received a B.E. and a M.E degrees in electrical engineering from Kagoshima University, Kagoshima, Japan, in 1992 and 1994. He joined NHK in 1994. Since 1997 he has been with NHK Science and Technical Research Laboratories, where he has been engaged in research on professional gap-fillers for digital terrestrial broadcasting in single frequency networks (SFN).
Iai Mr. Naohiko Iai
Naohiko Iai received a B.E. degree and a M.E. degree in Department of Applied Physics from Osaka University, Suita, Japan. He joined NHK in 1993. Since then he has been with NHK Science and Technical Research Laboratories, where he has been engaged in the research on millimeter-wave circuits, digital transmission techniques, digital terrestrial broadcasting systems and their applications to intelligent transport systems.
Shibuya Mr. Kazuhiko Shibuya
Kazuhiko Shibuya received a B.E. degree and a M.E. degree in Department of Electric and Electronic Engineering from Toyohashi University of Technology, Toyohashi, Japan, in 1980 and 1982. He joined NHK in 1982. Since 1995 he has been with NHK Science and Technical Research Laboratories, where he has been engaged in the research on digital transmission techniques and digital terrestrial broadcasting system.
Sasaki Mr. Makoto Sasaki
Makoto Sasaki received a B.E. degree in electronics engineering in 1972 from Tohoku University, Sendai, Japan. Since 1977 he has been with NHK Science and Technical Research Laboratories and engaged in the research on microwave tube for direct broadcasting satellite. His current research is terrestrial digital broadcasting systems.

Copyright 2000 NHK (Japan Broadcasting Corporation) All rights reserved. Unauthorized copy of the pages is prohibited.