You are on page 1of 6

G3-PLC Specification for Powerline Communication: Overview, System Simulation and Field Trial Results

Kaveh Razazian, Maher Umari, Amir Kamalizad, Victor Loginov, Michael Navid Signal Processing and Conversion Business Unit Maxim Integrated Products Irvine, CA, U.S.A.
AbstractAn overview of a new physical layer specification for powerline communication, named G3-PLC, is presented. The specification uses OFDM for data transmission and is intended to operate in the low-frequency band, from 35.9 kHz to 90.6 kHz. Simulation results are presented showing the expected performance of a typical G3-PLC-compliant modem when operating in a powerline channel, followed by field trial results showing performance of several G3-PLC modems operating on actual MV and LV powerline channels. In particular, results are presented for the first time showing performance of modems residing on opposite sides of a 20kV 220V transformer that interfaces an MV powerline to an LV powerline. Keywords- PLC; powerline; OFDM; Smart Grid

channels. He then developed a simplified powerline channel model useful for comparative analysis of modulation schemes. Since many of the models available for powerline channels are parametric models that use an average of the measured frequency responses or impulse responses of typical channels as the basis for estimating the rest of the parameters in the channel model [2], [4], [5], an accurate measured frequency response or impulse response of the channel can result in better estimates of the rest of the parameters used in the channel model. Experimental results obtained from field trials can be of great use for such models. Due to significant interest in utilizing the low frequency regions from 20 kHz to 500 kHz for data communication on powerlines, standards have been put in place by regulatory bodies to regulate the use of this low-frequency band. As an example, the European Standard CENELEC 50065-1 [6] has divided the 3 kHz to 148.5 kHz low-frequency powerline spectrum into four different frequency bands. These frequency bands are referred to as CENELEC A, B, C and D bands. On the other hand, in the United States, the FCC (Federal Communications Commission) has allocated the whole spectrum between 14 kHz and 480 kHz to one wideband channel often referred to as FCC band [7]. While over the past several years there has been several physical layer standards developed for high data rate communication on powerlines in the high-frequency region that extends to several MHz [8], [9], the authors are aware of three physical layer specifications that have been recently developed to address low-data rate transmission in the low-frequency band. The first specification is referred to as the "PRIME" draft specification [10] and is intended for modems operating in the 42 kHz to 88 kHz frequency range. The second specification is the HomePlug Command and Control [11] which operates at frequencies up to 400 kHz. Finally, the third specification is referred to as the "G3-PLC" Physical Layer Specification [12] and is intended for modems operating in the 35.9 kHz to 90.6 kHz frequency range with provisions to extend it to 480 kHz. This last specification is a new specification that was published in August 2009 and was developed by Maxim Integrated Products and the French Electric utility company ERDF. It is the objective of this paper to provide an overview of this specification along with simulation and experimental results showing the expected performance of modems compliant with this specification.

I.

INTRODUCTION

The use of powerline distribution grid for data communication has gained heightened interest over the past several years. It is well known that a powerline channel is far from being an ideal channel for data transmission. Powerline channel characteristics and parameters typically vary with frequency, location, time and the type of equipment connected to the channel. While the high-frequency regions above 1 MHz exhibit significant multipath frequency selective fading accompanied by severe attenuation on distances above several 100 ft, the lower frequency regions from 20 kHz to 500 kHz are especially susceptible to narrowband interference and impulsive noise. There has been a lot of research done on modeling and simulation of a powerline channel [1]-[5] over the past several years. In particular, Zimmermann et. al [1] presented a simplified analytical model describing complex transfer functions of typical powerline networks using a limited set of parameters. The model included numerous branches and impedance mismatching. In a different approach, Tlich et al. [2] presented a deterministic model describing the magnitude and phase of complex transfer functions of a powerline channel using a single parameter. The various powerline channels on which measurements were obtained were classified into different groups and a separate model was developed for each group by using the average magnitude and phase response of the channels in that group. Most recently, Galli [3] compared some statistical properties of powerline channels with those of wireless

978-1-4244-5010-7/10/$26.00 2010 IEEE

313

The rest of the paper is organized as follows: Section II gives an overview of the G3-PLC specification, followed by simulation results in section III. Experimental results obtained from field trials performed on powerline networks in France are presented in section IV, followed by a conclusion in section V. II. G3-PLC SPECIFICATION OVERVIEW

frame that follows it. It has information about the type and length of the frame as well as the tone map index used for that frame. The FCH data integrity is protected with a 5-bit CRC. Following FCH, a data frame consisting of a maximum of 252 data symbols is transmitted.
Preambe ACK Frame Preambe Data Frame
Figure 1. G3-PLC frame structure

FCH

The G3-PLC supports communication in the portion of the spectrum between 35.9 kHz to 90.6 kHz of the CELENEC-A band, with the option for extending the upper frequency to 180 kHz. As a result, the sampling frequency is selected to be 400 kHz in order to provide some margin above the Nyquist frequency for signal filtering in the transmitter (for spectrum shaping and removing the signal images) and at the receiver (for frequency-band selection). OFDM with DBPSK and DQPSK modulation schemes per carrier are selected to support up to 34.1 kbps data rate in normal mode of operation. Initial field trials showed that the best performance is achieved when an IFFT size of 256 is used. This results in frequency spacing between the OFDM carriers to be equal to 1.56 kHz (fs / N), where fs is the sampling frequency and N is the IFFT size. As a result, 36 subcarriers are used in each symbol. The PHY supports two types of frames: Data frame and ACK frame, both shown in Fig. 1. A typical frame starts with a preamble, which is used for synchronization, detection and AGC adaptation. The preamble is composed of 8 identical P symbols and 1 M symbols. Each of the P and M symbols are 256 samples that are generated from a PN sequence [12]. The M symbols are identical to P symbols except that all the carriers are phase shifted by . At the receiver, the phase distance between P symbol and M symbol waveforms is used for frame synchronization. The preamble is followed by symbols allocated for Frame Control Header (FCH). FCH has important control information required to demodulate the data

FCH

Data (multiple of 4 symbols)

A block diagram of the system is shown in Fig. 2. The system works in three different modes, namely, normal DBPSK mode, normal DQPSK mode, and ROBUST mode. In both normal modes, the forward error correction (FEC) is composed of a Reed Solomon encoder and a convolutional encoder. In ROBUST mode the FEC is composed of Reed Solomon encoder and a convolutional encoder that is followed by a Repetition Coder (RC) that introduces three bits of redundancy for each data bit. There is also an option for a super ROBUST mode which is used to transmit the FCH. It uses convolutional code combined with RC that introduces 5 bits of redundancy for each data bit. The convolutional encoder is a rate encoder with an octal generator polynomial G = [133 171] and a constraint length of 7. Six zeroes are added to the data to terminate the trellis to state 0. As for the RS encoder, standard RS(255,239) and RS(255,247) are used to further enhance the performance of the system in hostile powerline channels. A shortened code can be used to accommodate different sizes of data.

Mapping DBPSK DQPSK


FCH

IFFT

Add CP

Windowing

Data

Scrambler

Reed-Solomon Encoder

Convolutional Encoder

Bit Interleaver Robust Interleaver S-Robust Interleaver

AFE

FEC Encoder

Power Line

Figure 2. Block diagram of the G3-PLC System

314

Following the convolutional encoder, an interleaver block is used. The purpose of the interleaver is to provide protection against two different sources of error: A burst error that can corrupt a few consecutive OFDM symbols A frequency deep fade that can corrupt a few adjacent frequencies for a large number of OFDM symbols

can be seen that ROBUST mode has about 4 dB coding gain over DBPSK, which in turn has about 1.5 dB coding gain over DQPSK. The drawback is that the data rate in ROBUST mode is about 5.6 kbps vs. 20 kbps for DBPSK and 34 kbps for DQPSK. It should be noted that a soft Viterbi decoder was used to achieve the performance shown in Fig. 3.
10
0

AWGN Channel DQPSK DBPSK ROBO

BER

In order to correct for errors from both sources, a twodimensional interleaving scheme is applied where one dimension is interleaving the bits in time, and the second dimension is interleaving in frequency domain. This technique prevents frequency-dependent fading and impulsive noise from disrupting the data. Following the modulator an IFFT is applied to the data. The IFFT block takes a 256-point IFFT of the data and generates 256 time-domain samples that are then pre-pended by 30 samples of cyclic prefix. In order to reduce the out-ofband emission and the spectral side lobes, Raised Cosine shaping is applied to both data and frame control symbols. III. SYSTEM SIMULATION

10

-1

10

-2

10

-3

10

-4

10

-5

10

-6

-6

-4

-2

0 2 SNR (dB)

In order to get a reference point for the performance of a G3-PLC system, a full C fixed-point simulation was developed. The parameters used for the simulation are listed in Table I.
TABLE I. SYSTEM PARAMETERS Value 400 256 1.56 8 30 7.5 13 252

Figure 3. BER performance of the system

IV.

FIELD TRIAL RESULTS

Parameter Sampling frequency (kHz) Number of FFT points Carrier spacing (kHz) Number of overlapped samples Number of cyclic Prefix samples Number of symbols in Preamble Number of FCH symbols Max Number of Data symbols

Several prototypes of the G3-PLC modems were developed, installed and tested on the ERDF powerline network in France. The objective was to establish communication between a master device (concentrator) and slave devices (attached to the meters). Different test configurations were used to evaluate the performance of the G3-PLC system. Fig. 4 illustrates different segments of a typical power distribution network. The network includes a concentration of homes connected to low voltage (LV) powerlines, which are in turn connected to the medium voltage (MV) portion of the network through a transformer substation. While the data concentrators can reside on either the LV lines or the MV lines, the power station where power generators reside is usually located on the far end of the MV lines.

The channel was simulated by a low-pass filter with 3-dB cutoff frequency of 92 kHz followed by a white noise generator that added noise to the filter output. The maximum theoretical physical layer data rates for all three modes of operation are shown in Table II. The SNR of the channel determines which of the three modes gets selected.
TABLE II. Modulation ROBUST DBPSK DQPSK
PHY LAYER MAXIMUM THEORETICAL DATA RATE

LV MV

LV

RS Convolution Repetition Data Rate (bps) Parity Code Rate 8 16 16 1/2 1/2 1/2 4 1 1 5592 20009 34160 Figure 4. Typical distribution network

Fig. 3 shows the bit error rate (BER) for the overall system corresponding to the data rates in Table II. At a BER of 10-4 it

A. LV to LV Results The LV-LV configuration is one of the most widely used configurations on a powerline network. A transformer

315

substation converts the 20 kV MV to 220 V and distributes it to several homes. In this test, performed in a rural area in central France, the master device was installed on the LV side of a transformer substation and several slaves were connected to meters residing on the LV side at different distances from the transformer. Fig. 5 shows one of the test sites with a distance of 200 m between the master and slave. This specific connection exhibited a strong attenuation of about 30 dB between the two modems which is most likely due to the fact that the power cables between these two points were not all made from the same material but rather some sections were made from copper while other sections were made from aluminum. This resulted in impedance mismatches and hence signal reflections between the different sections.

TABLE III. Test Configuration LV to LV MV to MV MV to LV

MAC LAYER DATA RATE FOR TYPICAL CONFIGURATIONS Master to Slave Slave to Master Data Rate Data Rate Mod FER Mod FER (bps) (bps) 17280 17280 8640 DQPSK 0% DQPSK 0% DBPSK* 1% 17280 17280 2680 DQPSK DQPSK 0% 0%

ROBUST 10%

* For some tests ROBUST modulation had to be used

Fig.6 shows the power spectral density (PSD) of the received signal captured on the slave side of a typical channel on which the tests were performed. It is obvious from this figure that the spectrum of the transmitted OFDM signal (which resides between 36 kHz and 91 kHz) exhibits significant frequency-dependent attenuation and narrowband interference at several frequencies. One particular interference that consistently appeared in our LV powerline field tests was located around 84 kHz. The source of that interference was unknown to us. Yet, in spite of this strong in-band interference, the G3-PLC modems were able to communicate successfully due to their ability to disable transmission on sub-carriers residing close to this interference.
5

200

S
Power Spectrum Magnitude (dB)

-5

-10

-15

Figure 5. LV to LV test setup

-20

The G3-PLC modems used in these tests were able to achieve 2-way communication using DQPSK modulation. The achieved data rate and frame error rate (FER) is shown in Table III. It should be noted that the data rate shown, includes the MAC layer overhead (response time, ACK frame, and CSMA overhead). Moreover, for the prototypes used in the field trials, a DSP processor was used to implement the physical layer. This resulted in an inter-frame gap of about 8 msec due to processing delay, which in turn reduced the effective data rate.

-25

-30

-35

-40

-45

0.2

0.4

0.6

0.8

1 1.2 Frequency (Hz)

1.4

1.6

1.8 x 10

2
5

Figure 6. PSD of an LV to LV channel (4 branches, 200 m)

Slave Point (S) Capacitive coupling on MV network

6,4 km
(No repeaters)
Figure 7. MV to MV test setup

Master Point (M) Capacitive coupling on MV network

316

B. MV to MV results In this test data was injected to, and received from an aerial MV cable using a capacitive MV coupler. Fig. 7 shows the test setup. The master and slave devices were installed 6.4 km apart. The G3-PLC modems were able to achieve 2-way communication at a data rate of 1.728 kbps using DQPSK modulation as shown in Table III. Fig. 8 shows the power spectral density for one of the MV to MV channels. An interesting phenomenon observed in this test was the existence of a very strong interference (120 dBV) at about 21 kHz as shown in the figure. This frequency turned out to be used for Maritime transmission, and the long MV powerline acted as a perfect low-frequency antenna for that particular frequency. A fifth order analog filter was placed in front of the G3-PLC modems in order to reduce this 21 kHz interference by around 40 dB, resulting in the G3-PLC modems being able to communicate in normal DQPSK mode. The spectrum in Fig. 8 also shows an out-of-band interference around 162 kHz which turned out to be a long wave (LW) radio signal generated by France Inter Radio station. Finally, Fig. 9 shows the spectrum of the noise that was residing on the MV lines of this test.
20

C. MV to LV and LV to MV Results In this test a master device (concentrator) was installed on the MV network and two slave devices (meters) were connected to the LV sides of two aerial MV/LV transformers. In order for the slave devices to communicate with the master device the signal had to "cross" the transformer, a phenomenon that is not easily achieved due to the fact that the MV-LV transformer introduces significant distortion as well as severe frequency-dependent attenuation of over 40 dB on the OFDM signal. It is important to note that crossing the transformer is very valuable for the case of isolated meters, where a small overhead transformer substation is used to power a few homes. In such a case, a concentrator would normally be required for each MV-LV transformer. Hence, for the configuration shown in Fig. 10, three expensive concentrators would be required to establish communication. However, by taking advantage of the G3-PLC modems ability to cross transformers, two of the concentrators can be eliminated, allowing the meters residing on the LV side to directly communicate with the concentrator residing on the MV side. It is worth noting that for the test setup shown in Fig. 10, the G3-PLC modems were not always able to achieve reliable 2-way communication in normal DQPSK or DBPSK mode. In the master to slave direction (MV-LV), reliable communication was rarely achieved in DQPSK, and most of the time we had to switch to DBPSK or ROBUST mode to achieve the desired reliability. However, in the slave to master direction (LV to MV), only ROBUST mode was able to consistently achieve reliable communication between the master and slave devices. The results for both directions are summarized in Table III. Fig. 11 and Fig. 12 show the PSD of the MV-LV and LVMV channel used in this test, including the effect of the transformer. From these figures it appears that the transformer frequency response is different between MV-LV and LV-MV direction. Our results show that MV-LV direction allows a much wider portion of the OFDM spectrum to pass through the transformer compared to the LV-MV direction.

10

Power Spectrum Magnitude (dB)

-10

-20

-30

-40

0.2

0.4

0.6

0.8

1 1.2 Frequency (Hz)

1.4

1.6

1.8 x 10

2
5

Figure 8. PSD of an MV to MV channel (6.4 km)

(M)

2 km 1,4 km
(S) (S)

20

10

Power Spectrum Magnitude (dB)

-10

Figure 10. MV to LV test setup


-20

-30

-40

0.2

0.4

0.6

0.8

1 1.2 Frequency (Hz)

1.4

1.6

1.8 x 10

2
5

Figure 9. Noise captured on MV side

One possible reason for the asymmetric behavior of the transformers is that these transformers are known as isolated step-down transformers which convert a network to a Y network. For the network, residing on the MV side, there is no single point common to all phases. However, the Y network residing on the LV side is connected to a common point, called a neutral or star point. As a result, the transformers exhibit higher isolation from Y to direction than the opposite

317

direction. We are currently investigating the asymmetric behavior of the MV to LV transformers to analyze its effect on the data rate.
80

70 Power Spectrum Magnitude (dB)

60

of a 20kV 220V transformer that interfaced an MV powerline to an LV powerline. The data showed that the ROBUST mode was essential to establishing reliable communication among channels that include MV-LV transformers. Finally, it was noted that provisions are being added to the G3-PLC specification to support FCC operating frequency band, which extends up to 480 kHz, resulting in higher data rates. ACKNOWLEDGMENT The authors would like to express their great appreciation for Gerard Etheve and Fabien Schneider from the ERDF R&D division, Aline Pajot and Gaizka Alberdi from the EDF R&D division, for their great technical support and expertise during the field trials. The authors would also like to thank the Gateways and Energy Efficiency Department at Sagem Communications for developing the G3-PLC application and adaptation layers used in the meters and data concentrators. Finally, the authors would like to thank the Application Group at Maxim France, and in particular, Aurelie Maras for her technical support. REFERENCES
[1] M. Zimmermann, K. Dostert, A multipath model for the power line channel, IEEE Trans. Commun., vol. 50, no. 4, pp.553-559, April 2002. [2] M. Tlich, A. Zeddam, F. Moulin, and F. Gauthier, "Indoor power-line communications channel characterization up to 100 MHz - part I: Oneparameter deterministic model," IEEE Trans. on Power Delivery, Vol. 23, No. 3, pp. 1392-1401, July 2008. [3] S. Galli, "A simplified model for the indoor power line channel," IEEE Intl. Symposium on Power Line Communications and Its Applications, Mar 29 - April 1 2009, pp. 13-19. [4] H. Meng, S. Chen, Y. Guan, C. L. Law, P. L. So, E. Gunawan, and T. T. Lie, "Modeling of Transfer Characteristics for the Broadband Powerline Communication Channel," IEEE Trans. on Power Delivery, Vol. 19, No. 3, pp. 1057-1064, 2004. [5] G. Mathias, M. Rapp, and K. Dostert, "Power line channel characteristics and their effect on communication system design," IEEE Communications Magazine, pp. 78-86, April 2004. [6] "Signalling on low-voltage electrical installations in the frequency range 3 kHz to 148.5 kHz part 1: general requirements, frequency bands and electromagnetic disturbances," CENELEC EN-50065-1, Belgium, July 2001. [7] "Title 47 CFR part 15 subpart B," Federal Communications Commission, 1998. [8] "HomePlug A2 psecification: Medium interface specification," Home Plug Powerline Alliance, Oct 2001. [9] "IEEE P1901 draft standard for broadband over power line networks: Medium access control and physical layer specifications," IEEE P1901 working group, July 2009. [10] PRIME Project, "PRIME Technology white paper: PHY, MAC and convergence layers," version 1.0, Iberdrola, July 2008. [11] S. Bradbury, HomePlug Command & Control Overview White Paper, http://www.homeplug.org/tech/whitepapers/HomePlug_CC1_White_Pap er.pdf, September 2008. [12] "G3-PLC physical layer specification," Electricite Reseau Distribution France, Aug. 2009.

50

40

30

20

10

0.2

0.4

0.6

0.8 1 1.2 Frequency (Hz)

1.4

1.6

1.8 x 10

2
5

Figure 11. PSD of a typical MV to LV channel (captured on LV side)

100

90 Power Spectrum Magnitude (dB)

80

70

60

50

40

0.2

0.4

0.6

0.8 1 1.2 Frequency (Hz)

1.4

1.6

1.8 x 10

2
5

Figure 12. PSD of a typical LV to MV channel (captured on MV side)

V.

CONCLUSION

An overview of a new physical layer specification, called G3-PLC, was presented. The specification includes a ROBUST mode that significantly reduces bit error rate by using 3-bit redundancy and two layers of error correction. Simulation results showing the expected performance of a typical G3-PLC-compliant modem were also presented, followed by field trial results showing performance of several G3-PLC modems operating on actual MV and LV powerline channels. Included in these results was data showing the performance of the modems when residing on opposite sides

318

You might also like