INTERNATIONAL CONFERENCE ON COMMUNICATION, COMPUTER AND POWER (ICCCP’09)
MUSCAT, FEBRUARY 15-18, 2009
Time-Domain Techniques for Impulsive Noise Reduction in OFDM-Based Power Line Communications: A Comparative Study Khalifa S. Al-Mawali1 , Amin Z. Sadik1 , MIEEE, and Zahir M. Hussain1 , SMIEEE
noise by minimum of 10-15 dB and may sometimes reach 50 dB. Due to the harsh channel properties, communication over power lines requires a transmission scheme that can effectively cope with hostile channel conditions. Orthogonal Frequency Division Multiplexing (OFDM) is a major candidate for PLC systems, owing to its robustness to multipath, selective fading and different kinds of interference. OFDM is a mature multicarrier transmission technique and have been adopted in several wideband digital communication systems. Examples of its applications include Digital Audio Broadcasting (DAB), Terrestrial digital TV (DVB-T), wireless LANs and Wi-Max. OFDM performs better than single carrier in the presence of impulsive noise [1]. This is due to the fact that OFDM spreads the effect of impulsive noise over multiple symbols due to Discrete Fourier Transform (DFT) operation. Moreover, the addition of Cyclic Prefix (CP) in OFDM symbols can mitigate the effect of multipath. Despite the superiority of OFDM in the presence of impulsive noise, it is necessary to employ mitigation techniques in order to further reduce its effect in data transmission. A simple technique of impulsive noise mitigation is to precede OFDM receiver with a memoryless nonlinearity. This method has been addressed for conventional OFDM receivers in wireless applications [7]. However, power line channels exhibit different channel characteristics to other communication channels. In this paper, we adopt a practically proven PLC channel model [3],[4] to compare the performance of three memoryless nonlinearity methods (Clipping, Blanking and Clipping/Blanking) for mitigation of impulsive noise in OFDM systems over PLC channel. Three different scenarios of impulsive noise, namely Heavily disturbed, medium disturbed and weakly disturbed are considered. The characteristic parameters of the three scenarios are obtained from practical measurement published in [3].
Abstract— Impulsive noise is one of the major challenges in power line communications and can degrade the performance of OFDM-based PLC systems significantly. In this paper, we adopt a practically proven channel model for broadband power line communication and utilize it to study and compare the BER performance of three nonlinear time-domain impulsive noise mitigation techniques: clipping, blanking and clipping/blanking. Three noise environments, namely ”heavily disturbed”, ”medium disturbed” and ”weakly disturbed”, that are based on practical measurements are used to investigate the performance of impulsive noise reduction schemes by means of computer simulations.
-Keywords: Power line Communications, Impulsive noise, OFDM, Blanking, Clipping. I. I NTRODUCTION In recent years, interest in using electric power lines for broadband multimedia communications has grown significantly. Power Line Communication (PLC) is becoming more and more attractive because it exploits the existing and widespread power distribution infrastructure to provide home networking as well as broadband services. However, due to hostile channel characteristics, communication over power lines faces serious challenges. The most influential channel properties degrading the performance of high-speed communications are noise, attenuation and multipath propagation [1]. Noise in power lines can not be described by Additive white Gaussian Noise (AWGN) as the case in many other communication channels [2]. It can be classified into five types [3]. These are coloured background noise, narrow-band noise, periodic impulsive noise asynchronous to the mains frequency, periodic impulsive noise synchronous to the mains frequency and asynchronous impulsive noise. The first three types of noise usually remain stationary over long periods of time (seconds, minutes or hours) and can be summarized as background noise. The last two types are time-variant and can be summarized as impulsive noise. Impulsive noise is mainly caused by power supplies (synchronous to mains frequency) or by switching transients in the network (Asynchronous to the mains power). It has a random occurrence and its duration varies from few microseconds to milliseconds. Practical experiments in power lines [3] show that the power spectral density (PSD) of impulsive noise exceeds the PSD of background
II. S YSTEM M ODEL A. OFDM System Orthogonal frequency Division Multiplexing (OFDM) is considered as the transmission scheme for Broadband over power lines (BPL) by most researchers. OFDM is an effective multi-carrier transmission technique that performs well in PLC channel conditions. In OFDM systems a high-speed serial data stream is split into a number of parallel slow data streams that are carried in multiple orthogonal subcarriers
1 School of Electrical and Computer Engineering, RMIT University, Melbourne, Victoria 3000, Australia. E-mails:
[email protected],
[email protected],
[email protected].
c SQU-2009 ISSN: 1813-419X °
1
INTERNATIONAL CONFERENCE ON COMMUNICATION, COMPUTER AND POWER (ICCCP’09)
by means of Inverse Discrete Fourier Transform (IDFT). the long symbol period of OFDM minimizes the effect of intersymbol interference (ISI) caused by multipath in the signal. The discrete time OFDM signal can be expressed as: N −1 k 1 X Sk ej2π N n s(n) = √ N k=0
impulsive noise. These five types of noise can be summarized into two major types: background noise and impulsive noise. To analyse their effect in OFDM-based PLC systems, the background noise (wk ) is modelled as AWGN with mean zero 2 and variance σw , and the impulsive noise (ik ) is given by:
(1)
ik = bk gk
B. Channel Model
λk , k = 0, 1, 2, ... (4) k! The amplitude of impulsive noise, on the other hand, follows Gaussian distribution with mean zero and variance σi2 . Impulse width is also assumed to be Gaussian. Fig. 1 shows the time domain representation of a sample impulse with an overall duration of 50 µs surrounded by background noise. In Fig. 2 a number of impulses with average impulse rate 5 s−1 and average inter-arrival time of 200 ms are illustrated. P (k) = P (X = k) = e−λ
Power line networks differ significantly in topology, structure, and physical properties from conventional communication channels such as twisted pair, coaxial, or fiber-optic cables [4]. Because they were not specifically designed for data transmission, power lines provide a harsh environment for higher frequency communication signals [2]. The most influencing properties of this hostile medium in the performance of high speed communications are signal distortion due to frequencydependant cable loses, multi-path propagation and noise [3]. These factors and others have to be taken into account in order for a successful communication over the powerline channel. As a result, the need for realistic and practical models for the power line channel transfer characteristics is inevitable. Zimmermann and Dostert [4] have proposed a practical channel model that is suitable for describing the transmission behavior of power line channels. The model is based on practical measurements of actual power line networks and is given by the channel transfer function: k
−j2πf (di /vp) 1 f )di ci . |e−(a0 +a {z } . |e {z } |{z} i=1 delay weighting attenuation portion portion factor
Example of a single impulse in background noise 1.5
1
tw = 50 µs 0.5
Amplitude
Np X
(2)
0
−0.5
−1
Where Np is the number of multipaths, ci and di are the weighting factor and length of the ith path respectively. Frequency-dependant attenuation is modelled by the parameters a0 , a1 and k. In the model, the first exponential presents attenuation in the PLC channel, whereas the second exponential, with the propagation speed vp, describes the echo scenario. The attenuation parameters for 15-path model were obtained using physical measurements [4]. In this paper we adopt this widely accepted model to simulate and study the performance of nonlinearity techniques in mitigating the impulsive noise encountered in practical PLC systems.
−1.5
100
100.2
100.4 Time (ms)
100.6
100.8
101
Fig. 1. Example of an impulse mixed with background (AWG) noise.
To study the performance of impulsive noise reduction techniques, three impulsive noise scenarios based on PLC environment are considered. These are ”heavily disturbed”, ”medium disturbed” and ”weakly disturbed”. They are characterized by the average impulse rate which is the number of impulse occurrences in a second and the disturbance ratio which indicates the actual disturbed time and is given by: PNimp tw,i (5) DR = i=1 Ttot
C. Impulsive Noise In contrast to most other communication channels, noise in power line channels can not be described as Additive White Gaussian Noise (AWGN). In PLC environments, data transmission may encounter different types of noise. These are coloured background noise, narrow-band noise, periodic impulsive noise asynchronous to the mains frequency, impulsive noise synchronous to the mains frequency and asynchronous
c SQU-2009 ISSN: 1813-419X °
(3)
Where bk is the Poisson Process which is the arrival of impulsive noise, and gk is white Gaussian process with mean zero and variance σi2 . This means that the arrival of impulsive noise follows a Poisson distribution with a rate λ units per second, so that the probability of event of k arrivals in a seconds is:
Where N is the number of sub-carriers, Sk is a sequence of QAM symbols. In order to eliminate inter-channel interference (ICI) and inter-symbol interference (ISI), OFDM uses a cyclic prefix (CP) that is appended at the start of OFDM symbols.
H(f ) =
MUSCAT, FEBRUARY 15-18, 2009
Where Nimp is the total number of impulses occurring in time Ttot and tw,i is the width of the ith impulse. Characteristics for all three scenarios are obtained from practical measurements in [3] and listed in table I. 2
INTERNATIONAL CONFERENCE ON COMMUNICATION, COMPUTER AND POWER (ICCCP’09)
1) Clipping ½ rk , |rk | ≤ Tc yk = , jarg(rk ) Tc e , |rk | > Tc
Example of Impulsive noise added together with AWGN 1.5
−1
λ=5s 1
Amplitude
k = 0, 1, ..., N − 1 (7)
where Tc is the Clipping Threshold. 2) Blanking ½ rk , |rk | ≤ Tb , k = 0, 1, ..., N − 1 yk = 0, |rk | > Tb
0.5
0
−1
0
0.2
0.4
0.6 Time (sec)
0.8
1
Example of an impulsive noise added together with background (AWG) noise. Fig. 2.
M EASURED C HARACTERISTIC
Scenario 1) Heavily disturbed 2) Medium disturbed 3) Weakly disturbed
Clipping and Blanking thresholds have to be carefully selected to minimize the BER in the receiver. When Tc or Tb is very small, most samples of the OFDM signal are clipped or replaced with zeroes. This introduces more errors in the signal and hence increases the BER. In contrast, for very large values of Tc or Tb , the nonlinear preprocessor will have no effect on the received signal, and therefore the system performance may be reduced significantly by impulsive noise.
TABLE I PARAMETERS OF T HE T HREE S CENARIOS
average DR (%) 0.327 0.00632 0.00135
(8)
where Tb is the blanking threshold. 3) Clipping/Blanking |rk | ≤ Tc rk , yk = T1 ejarg(rk ) , Tc < |rk | ≤ Tb , k = 0, 1, ..., N − 1 0, |rk | >≤ Tb (9) Where the clipping threshold (Tc ) is smaller than the blanking threshold (Tb ).
−0.5
−1.5
MUSCAT, FEBRUARY 15-18, 2009
average λ (s−1 ) 51.1 1.04 0.122
Let sk be the transmitted OFDM signal as in (1). The received signal after down-conversion, analog-to-digital conversion assuming perfect synchronization can be represented as:
BER against threshold value
0
10
Clipping Blanking Clipping/Blanking −1
rk = sk ∗ h + wk + ik
k = 0, 1, 2, ..., N − 1
10
(6)
where h is the PLC channel impulse response which is the Inverse Fourier Transform of the transfer function defined in (2). The symbol ∗ denotes convolution.
−2
BER
10
SNR=30 dB
−3
10
III. I MPULSIVE N OISE M ITIGATION
Optimal Clipping Threshold=0.18
−4
10
In OFDM systems, the long symbol period gives the signal more robustness to impulsive noise, since the impulse noise energy is spread over the subcarriers. Nevertheless, if mitigation techniques are not employed, impulsive noise can still have a significant effect in the performance of OFDM systems, especially in a medium like power lines. Different techniques for impulsive noise mitigation have been discussed in the literature [3][5][6][11]. Due to their simplicity, nonlinearity techniques are often used in practical applications for impulsive noise mitigation[7]. They are employed at the frontend of OFDM receiver before demodulating using Fast Fourier Transform (FFT). Fig. 4 depicts a block diagram of OFDM with a nonlinear block for impulsive noise reduction preceding demodulation using FFT. These techniques change only the amplitude of the signal according to a specific threshold without changing its phase according to the following:
c SQU-2009 ISSN: 1813-419X °
Optimal Blanking Threshold=0.32
−5
10
0
0.2
0.4
0.6 0.8 1 Clipping/Blanking Threshold
1.2
1.4
1.6
BER performance at the output of OFDM (QPSK) receiver in heavily disturbed PLC channel with three different types of nonlinearities for impulsive noise reduction as a function of threshold value. Tb = 1.6Tc for clipping/blanking. Fig. 3.
Fig. 3 illustrates how the BER changes with different threshold values. It can be observed from the figure that there exists an optimal threshold for clipping, blanking and the combined clipping/blanking nonlinearity. The results are for the ”heavily disturbed” scenario with SNR equals 30 dB. SNR in all figures is defined according to the following calculations: 3
INTERNATIONAL CONFERENCE ON COMMUNICATION, COMPUTER AND POWER (ICCCP’09)
Fig. 4.
MUSCAT, FEBRUARY 15-18, 2009
Block diagram of OFDM-based PLC system with impulsive noise reduction.
Let pn be the total noise power resulting from AWGN and impulsive noise, and px is the average PSK or QAM symbol power. the signal-to-noise ratio (SNR) can be defined as: px px SNR = = (10) pn pw + pi
Heavily disturbed Environment
0
10
−1
10
−2
10
where pw and pi are the mean AWGN power and mean impulsive noise power. The disturbance ratio DR given in equation (5) can be used to calculate the average impulsive noise duration in unit time. Since impulsive noise arrival is modelled by Poisson process with λ impulses per second, the following expression for the average disturbance ratio can be obtained: DR = λTnoise (11)
BER
−3
10
−4
10
−5
10
−6
10
Where Tnoise is the average impulse duration. The following expression for (SNR) is used in this paper: px (12) SNR = 2 σw + DR.σi2
5
10
15
20
25 SNR
30
35
40
45
Comparison of time-domain nonlinearity techniques for impulsive noise reduction (heavily disturbed PLC channel).
Medium disturbed Environment
0
10
(13)
λ = 1.04 s−1 DR = 0.00632% µ=3
−1
10
In our simulations we use µ = 3.
−2
10
BER
IV. E XPERIMENTAL R ESULTS AND D ISCUSSION Matlab was utilized to study the performance of the three types of nonlinear preprocessors used for impulsive noise mitigation in OFDM systems. A random signal is mapped into QPSK symbols and modulated using OFDM with 128 subcarriers and passed through a PLC multipath channel with 15 paths given in (2). The arrival of impulsive noise is assumed to follow a Poisson distribution as given by (3) while the background noise is assumed to be AWGN. The total noise resulting from impulsive and background noises is added to the OFDM signal. The characteristics of impulsive noise for each of the three scenarios are specified according to Table I. Nonlinear noise reduction techniques are then employed in OFDM receiver and their performance is compared. Figures 5-7 illustrate a comparison between the three simple nonlinearity techniques used for impulsive noise suppression in OFDM-Based PLC system. Results in these figures are obtained for ”heavily disturbed”, ”medium disturbed” and c SQU-2009 ISSN: 1813-419X °
PLC channel with Impulsive + AWGN Blanking Clipping Clipping/Blanking PLC channel with AWGN only
Fig. 5.
where σi2 is the mean power of a single impulse over its duration Tnoise and is modelled by Gaussian process as indicated in (3). A ratio µ between impulsive noise power and background noise power (AWGN) is defined as: pi DR.σi2 µ= = 2 pw σw
λ = 51.1 s DR = 0.327% µ=3
−1
−3
10
−4
10
−5
10
−6
PLC channel with Impulsive + AWGN Blanking Clipping Clipping/Blanking PLC channel with AWGN only
10
5
10
15
20
25
30
35
40
45
50
SNR
Comparison of time-domain nonlinearity techniques for impulsive noise reduction (medium disturbed PLC channel). Fig. 6.
”weakly disturbed” impulsive noise environments, respectively. It can be noticed from Fig. 5 that for such PLC channel conditions, although blanking nonlinearity performs 4
INTERNATIONAL CONFERENCE ON COMMUNICATION, COMPUTER AND POWER (ICCCP’09)
disturbed” impulsive noise environment. For ”weakly disturbed” scenario, clipping and blanking offer comparable performance with a slight advantage for blanking over clipping for higher SN R values.
Weakly disturbed Environment
0
10
λ = 0.122 s−1 DR = 0.00135% µ=3
−1
10
R EFERENCES
−2
BER
10
[1] Y. H. Ma, P. L. So and E.Gunawan, “Performance analysis of OFDM systems for broadband power line communications under impulsive noise and multipath,” IEEE Trans. Power Delivery, vol. 20, No. 2, pp. 674-681, April 2005. [2] M. S Yousuf, and M. El-Shafei, “Power line communications: an overview - part 1,” Innovations ’07, pp. 218-222. [3] M. Zimmermann and K. Dostert, “Analysis and modeling of impulsive noise in broad-band powerline communications,” IEEE Trans. Electromagn. Compat., vol. 44, No. 1, pp. 249-258, Feb. 2002. [4] M. Zimmermann and K. Dostert, “A multipath model for the pwerline channel,” IEEE Trans. Communications, vol. 50, No. 4, pp. 553-559, April. 2002. [5] F Abdelkefi, P. Duhamel, and F. Alberge, “Impulsive noise cancellation in multicarrier transmission,” IEEE Trans. Communications, vol. 53, No. 1, pp. 94-106, Jan. 2005. [6] S. Zhidkov, “Inpulsive noise suppression in OFDM based communication systems,” IEEE Trans. Consumer Electronics, vol. 49, No. 4, pp. 944-948, Nov. 2003. [7] S. Zhidkov, “Analysis and comparison of several simple impulsive noise mitigation schemes for OFDM receivers,” IEEE Trans. Communications, vol. 56, No. 1, pp. 5-9, Jan. 2008. [8] M. Ghosh,“Analysis of the effect of impulsive noise on multicarrier and single carrier QAM systems,” IEEE Trans. Communications, vol. 44, No. 2, pp. 145-147, Feb. 1996. [9] S. Zhidkov, “Performance analysis and optimization of OFDm receivers with blanking nonlinearity in inpulsive noise environment,” IEEE Trans. Vehicular Technology, vol. 55, No. 1, pp. 234-238, Jan. 2006. [10] M. Gotz, M. Rapp, and K. Dostert, “Power line channel characteristics and their effect on communication system design,” IEEE Communications Magazine, vol. 42, issue 4, pp. 78-86, April. 2004. [11] K. Al Mawali, A. Z. Sadik, and Z. M. Hussain, “Joint timedomain/frequency-domain impulsive noise reduction in OFDM-based power line communications,” Proceedings of the Australasian Telecommunication Networks and Applications Conference (ATNAC 2008), Adelaide, Australia, 6-10 Dec. 2008.
−3
10
−4
10
−5
10
−6
10
5
MUSCAT, FEBRUARY 15-18, 2009
PLC channel with Impulsive + AWGN Blanking Clipping Clipping/Blanking PLC channel with AWGN only 10
15
20
25
30
35
40
45
50
SNR
Comparison of tim-domain nonlinearity techniques for impulsive noise reduction (weakly disturbed PLC channel). Fig. 7.
better than clipping for small SNR, the latter technique outperforms the former by about 2 decibels for high SNR values (SN R > 35 dB) for fixed threshold values. For the same impulsive noise characteristics, simulations show that the best nonlinearity solution to impulsive noise in this case is the combined clipping/blanking. This technique can sometimes improve the performance over Clipping and blanking by more than 5dB. Note that for this environment, all three techniques provide negligible improvement for very high SNR values (SN R > 40 dB). In the ”medium disturbed” environment, depicted in Fig. 6, blanking nonlinearity performs the best of the three techniques using the same threshold values attained from Fig. 3. In this impulsive noise environment, all three techniques reduce the effect of impulsive noise significantly. Blanking nonlinearity can lessen the BER of OFDM based PLC receiver from 7 × 10−3 up to 10−5 when SNR is equal to 45 dB in a medium disturbed environment. Blanking and Clipping techniques provide comparable results in the case of ”weakly disturbed” power lines for fixed threshold values as can be observed from Fig. 7. Clipping/blanking achieved the worst performance under this impulsive noise environment. V. C ONCLUSION In this paper, the performance of nonlinear techniques for impulsive noise mitigation in OFDM under practical PLC channel characteristics was considered. Three different nonlinearities, namely clipping, blanking and clipping/blanking, were compared by means of computer simulations. The obtained simulation results show that for the ”heavily disturbed” scenario, the best performance is obtained by the clipping/blanking technique. In this hostile channel condition all three nonlinearities may have a negligible improvement in the system when SN R is high (SN R > 35 dB). Blanking technique achieves the best BER performance under ”medium c SQU-2009 ISSN: 1813-419X °
5
