Passive, Transformerless Coupling Circuitry for Narrow-Band Power-Line Communications

Passive, Transformerless Coupling Circuitry for Narrow-Band Power-Line Communications Mloyiswa P. Sibanda, Petrus A. Janse van Rensburg, and Hendrik C. Ferreira Department of Electrical Engineering Department of Electrical and Electronic Engineering Walter Sisulu University University of Johannesburg P.O. Box 1421, East London, 5200, South Africa P.O. Box 524, Auckland Park, 2006, South Africa Phone: +27-43-702-9262, Fax +27-43-702-9226 Phone +27-11-489-2463, Fax +27-11-489-2357 e-mail: msibanda@wsu.ac.za; pvanren@wsu.ac.za e-mail: hcf@ing.rau.ac.za Abstract The field of power-line communications (PLC) has been maturing and therefore much research is aimed at making PLC technology cheaper. In this paper we investigate passive coupling circuits as an alternative to transformer-capacitor couplers, as the coupling transformer is a relatively expensive component. Various other sub-fields of PLC such as modulation and coding have received a lot of attention in recent years, whereas the sub-fields of coupling, filtering, impedance adaptation and protection have lagged behind. Most electronic circuits, especially filter circuits have been around for years and perhaps their design is sometimes viewed as exhausted. However, the nature of the power-line grid requires careful consideration, as sensitive modems are to be connected to the power-line network, via these coupling circuits. These couplers further have to withstand the strain of relatively large voltages and currents. Hence, sufficient time spent, designing and selecting coupling circuitry elements is essential to avoid unnecessary attenuation as well as possible damage to the more expensive communication circuitry. Keywords Coupling circuits, passive components, filters, impedance adaptation. T I. INTRODUCTION RANSFORMER-CAPACITOR couplers are used extensively in power-line communications as the transformer aids with filtering, protection and impedance adaptation. The design of such a transformer-capacitor coupler was described in [1], while impedance adaptation mechanisms of coupling transformers were investigated in [2]. However, the useful characteristics of the coupling transformer comes at a price. Also, power-line communications as a technology is in tremendous economic competition with other network technologies such as wireless networks and other cable networks. Therefore, costs need to be cut wherever possible, without degrading the performance of the system under investigation. For this reason, this paper investigates the possibility of using a transformerless passive coupling circuit in order to reduce the cost of PLC modems. Filter circuits have the ability to pass or block a certain band of frequencies between a defined lower cut-off frequency and a defined upper cut-off frequency. Hence, to separate the power-line signal at 50 Hz and a communication signal in the kilo-hertz range requires the use of filter circuits. Most passive filter circuits are designed from low-pass filter prototypes, including Butterworth, Chebyshev or Bessel filters. The element values of these prototype filters are often given in electronics and radio-frequency textbooks, e.g. [3], [4]. The Butterworth filter offers a fairly flat pass-band response and a fairly sharp attenuation outside the pass-band, but not as sharp as the Chebyshev. The Chebyshev offers the sharpest attenuation from the cut-off frequency, but carries significant ripple at pass-band frequencies. Thus, for the same order of filter, the Chebyshev filter has a better selectivity than a Butterworth filter. Bessel filters focus on the phase characteristics rather than on the amplitude characteristics. These offer very poor initial stop-band attenuation, but a maximally flat group delay or linear phase characteristics in the filter s pass-band [4]. Butterworth and Chebyshev filters can be constructed using lumped L and C ladder prototypes. Constructing a coupling circuit using only passive L and C components is cheaper than using active components and transformers. This is the reason for using only passive components in this paper. The choice between a Butterworth and a Chebyshev has to do with the amount of ripple at pass-band and the required selectivity of the filter circuit. The Butterworth filter response is more desirable because it provides a flat response in the frequency range of interest. II. ATTENUATION CHARACTERISTICS A typical first step in filter design is to choose the necessary shape factor of the filter circuit. The shape factor quantifies the steepness of the filter s roll-off. The attenuation characteristics for different order filters are often given in filter design textbooks e.g. [3], [4]. Alternatively, for an n- order Butterworth filter, (1) may be used to calculate the attenuation in decibels. A db 2 [ + ( ω )] n = 10log 1 ω (1) C In (1) A db represents the attenuation in decibels, ω is the frequency at which the attenuation is desired, ω C is the cut-off frequency of the filter, and n is the number of elements in the filter circuit. Equation (1) is used for low-pass filter circuits. However, if designing high-pass or band-pass filter circuits the equation is simply transformed by inverting the ratio of 978-1-4244-3790-0/09/$25.00 2009 IEEE 125

ω/ω C for a high-pass filter or using the ratio of bandwidths BW/BW C for a band-pass filter circuit. In power-line communications the coupling filter is required to drastically attenuate the 50-Hz power-line waveform, but freely pass the communication signal. This can be achieved by using either a high-pass filter or a band-pass filter. If the design requires a high-pass filter the frequency ratio ω/ω C in (1) is inverted, giving ω C /ω = f C /f = 2200, for a 110-kHz carrier. Equation (1) then predicts that a single capacitor (first-order) filter would theoretically provide ~67 db attenuation; a two-element second-order LC filter would provide ~150 db attenuation, whereas a three-element thirdorder design would provide ~200 db attenuation to the 50-Hz waveform. Single or double capacitors are often used in power-line communications to block the power waveform and allow communication signals to pass through, see possible arrangement in Fig. 1(a). This arrangement will require a capacitor with a high voltage rating to withstand the peak voltage of the power-line waveform. This voltage strain can be lessened by introducing a second capacitor along the `neutral wire`, the two capacitors now sharing the voltage strain. See Fig. 1(b). The high-pass filter coupling circuit in Fig. 1(c) was extracted from [6]. High-pass filters similar to the one in Fig. 1(c) are common in many PLC applications. The L-shaped high-pass filter is formed by the elements and, as indicated in Fig. 1(c). Without the other circuit elements, the L-shaped filter design has problems, because any dccomponent introduced from the communication device will short-circuit inductor. Hence, the author introduces a capacitive element, to block any dc-component from the communications system end. This decision makes the filter circuit turn similar to T-shaped third-order filter circuits. This improves the filter response, and in addition improves the rolloff figure and attenuation of the 50-Hz power-line waveform. Fig. 1(c) also shows that an inductive element was introduced. The element resonates with the capacitors, transforming what was originally a simple high-pass filter to band-pass filter circuit. The advantage of a band-pass filter is that it allows only frequencies within the band of interest to pass through, blocking all other frequencies. This can help prevent unwanted signals like noise at higher frequencies to pass through. Another advantage is that the voltage strain on the capacitors is lessened when two series capacitors are used, because the two capacitors now share the peak voltage of the power-line waveform. Fig. 1(d) illustrates a T-shaped third-order band-pass filter arrangement. This arrangement allows for more circuit elements, hence less strain in terms of voltage and current rating for individual circuit elements. For this study, a thirdorder band-pass filter is investigated as a possible solution to most of the problems found with first and second-order highpass filters. The use of passive elements also satisfies the universal requirement to minimize costs. (a) Simple first-order capacitor filter. 10uF D1 D2 12uH (b) Dual-capacitor first-order filter. 3.9mH 680nF (c) Improved second-order high-pass filter circuit, extracted from [4]. (d) Proposed third-order band-pass filter circuit. Fig. 1. Filter circuits. III. T- SHAPED VS. Π-SHAPED NETWORKS Fig. 2 illustrates the two possible network topologies for a third-order Butterworth filter. The Pi-network has the component orientation resembling the shape of the Greek lower-case letter pi- π. The T-network has the component orientation resembling the Greek upper-case letter tau- T. As shown in Fig. 2 the two high-pass topologies transform to similar arrangements for the band-pass filter circuits a Pishaped high-pass filter transforms to a Pi-shaped band-pass filter, and similarly for a T-shaped topology. (a) High-pass filter Pi-network topology. LINE NEUTRAL (b) Band-pass filter Pi-network topology. LINE NEUTRAL HIGH PASS FILTER ELEMENTS 126

1 Ω 1 F 1 F 0.5 H 1 Ω (c) High-pass filter T-network topology. Fig. 3. Normalized second-order Butterworth high-pass filter prototype. (d) Band-pass filter T-network topology. Fig. 2 Element arrangements. The problems with the two Pi-network arrangements in Fig. 2(a) and (b) are quite similar to the problems mentioned earlier with the L-shaped high-pass filter circuit. The inductor () in parallel with the terminal power-line impedance will present a very low reactance, drawing huge currents from the power-line. It is further impractical and almost impossible to find inductors that will withstand these required current levels. Also, both Pi-networks present problems, as any dccomponent introduced from the communication device end or the power-line end will saturate and thus short-circuit the inductors or, respectively. Therefore, it is evident that Pi-networks are unusable for power-line communications, and that T-networks should rather be used. IV. LOW-PASS TO HIGH-PASS FILTER TRANSFORMATION Butterworth filter tables, similar to Table I provide element values for an n-order low-pass filter prototype. TABLE I BUTTERWORTH LOW-PASS PROTOTYPE ELEMENT VALUES (RS = RL) n L4 C5 2 1.414 1.414 3 1.000 2.000 1.000 4 0.765 1.848 1.848 0.765 5 0.618 1.618 2.000 1.618 0.618 n C4 L5 Since we are transforming a low-pass to a high-pass filter, each low-pass filter element is replaced with an element of the opposite type. The values of the low-pass filter elements are replaced with a reciprocal value. Fig. 3 is the resulting normalized high-pass filter prototype with a cut-off frequency of 0.159 Hz (or ω = 1 radians/second). In these prototype filters, terminal impedances are normalized to 1 Ω. V. FREQUENCY AND IMPEDANCE SCALING The next step in filter design is to de-normalize and scale all elements to the frequency and impedance of the required design. This transforms the filter prototype circuit to a usable practical circuit. The first transformation scales the normalized high-pass filter cut-off frequency from 0.159 Hz to the required high-pass cut-off frequency, 95 khz. Let us assume that the filter should pass frequencies from the CENELEC Bband (95 khz to 125 khz) upwards. This requirement is achieved by dividing each inductor and capacitor by 2πf C, where f C represents the high-pass filter cut-off frequency in hertz. The second transformation scales both terminal impedances from 1 ohm to the required impedance value. This change in terminal impedances will require the filter circuit elements to match the new terminal impedances. For this study, the terminal impedance is set at 50 ohms as most communication instruments often carry an internal impedance of 50 ohms. For an initial investigation, the output impedance has also been set to 50 ohms to allow for maximum signal power transfer. However, the power-line impedance is a fluctuating parameter in power-line networks. Therefore, performance of the proposed coupler for different impedance levels is discussed in section VIII. This 50-times change in terminal impedance will require 50-times the reactance for the inductors and capacitors in the lumped L and C ladder prototype. This is simply achieved by multiplying all inductors by 50, and dividing all capacitors by 50, as X L =ωl but X C = 1/ωC. The combined transformations can easily be achieved using the equations (2), (3) and (4). For capacitive elements in the series arms, (( 2 f C ) Δ R L ) C ' = C π *. (2) For inductive elements in the parallel arms, ( 2π ) L ' = L * Δ R L f C. (3) C and L represent the high-pass prototype element values, ΔR L represents the change in terminal load impedance, and f C is the high-pass filter cut-off frequency. The resulting transformation for a high-pass filter circuit is shown in Fig. 4. The simulation results are presented in Fig. 4(b). 127

33.51nF ' 41.88uH ' 33.51nF ' (5) and (6). Table II, in Section VII shows the resulting component values, and the resulting band-pass filter circuit is shown in Fig. 5. The simulation results are presented in Fig. 5(b). 7.89nF '' 265.26uH ' 7.89nF '' 265.26uH ' (a) High-pass filter circuit. 212.21nF ' 9.86uH '' (a) Band-pass filter circuit. (b) Simulated high-pass filter response, showing the designed cut-off frequency of 95 khz. Fig. 4 High-pass filter. VI. LOW-PASS TO BAND-PASS FILTER TRANSFORMATION It is possible to convert normalized filter prototypes to band-pass filters. The simplest approach in transforming a low-pass filter circuit directly to a band-pass filter circuit is to replace all capacitors in the series branches with LC seriesresonant branches and all inductors in parallel branches with LC parallel-resonant branches [3], [4]. Using the original normalized low-pass filter element values, equation (2) and (3) can be transform to suit a bandpass filter circuit design. For inductive elements in the series arms, ( 2π * db) L' L * ΔRL BW 3 =. (5) For capacitive elements in the parallel arms, (( 2 * BW db) ΔRL) C' = C π 3 *, (6) where C and L represent the low-pass prototype element values, ΔR L represents the change in terminal load impedance, and BW 3dB is the required band-pass filter bandwidth 30 khz for this design. For inductive elements in the parallel arms, 2 ( * ') L'' = 1 ω 0 C. (7) For capacitive elements in the series arms, 2 ( * ') C' ' = 1 ω 0 L, (8) where C and L represent the new band-pass element values calculated using (5) and (6). Also, ω 0 = 2πf 0 where f 0 represents the center frequency, chosen as 110 khz (within the CENELEC B-band ). Thus the elements C and L are chosen to resonate at 110 khz with elements calculated using (b) Simulated Butterworth band-pass filter response with a center frequency of 110 khz. Fig. 5. Band-pass filter. VII. CHOOSING PRACTICAL COMPONENTS & SPECIFICATIONS Some important checks are necessary when it comes to choosing practical component values. Firstly, C and L component values must be chosen, as close to the design values as is practically possible. At the same time, these chosen elements for each resonant branch must resonate at a frequency as close as (practically) possible to the required resonant frequency, in this case 110 khz. Table II shows a comparison of reactance values and the resonant frequencies for calculated values and some practical components. The first set of closest practical component values already shows a difference in reactance and resonant frequency. TABLE II THEORETICAL VS PRACTICAL COMPONENT VALUES BRANCH SERIES PARALLEL SERIES COMPONENT CALCULATED 7.89nF 265μH 212nF 9.86μH 7.89nF 265μH REACTANCE 183Ω 183Ω 6.82Ω 6.81Ω 183Ω 183Ω f 0 110 khz 110 khz 110 khz PRACTICAL 8.2nF 260μH 220nF 10μH 8.2nF 260μH REACTANCE 176Ω 180Ω 6.6Ω 6.9Ω 176Ω 180Ω f 0 109 khz 107 khz 109 khz AVAILABLE 6.8nF 3*100μH 220nF 8.2μH 6.8nF 3*100μH REACTANCE 213Ω 207Ω 6.6Ω 6.9Ω 213Ω 207Ω f 0 111 khz 118 khz 107 khz To minimize losses, each inductive reactance (X L =ωl) must further theoretically equal the capacitive reactance (X C = 128

1/ωC) at resonance for each series and parallel branch. As shown in Table II, practical component values can alter the designed total reactance, making (X L X C ) larger for the series branch and smaller for the parallel branch. These altered reactance values can cause a significant change in delivered signal power, since the reactance lies between the communication instrument and the receiver. These losses originate from volt-drops across the resultant series reactance (X L X C ) and a higher current loss through the smaller resultant parallel reactance (X L X C ). Further, the voltage and current specifications are very important for each capacitor and inductor, respectively. For the capacitors we can use the voltage divider rule to determine the maximum voltage rating for each capacitor. Take note that in Fig. 5(a), the power waveform voltage is effectively across and needs to be filtered by the various coupling components. Capacitor implements the bulk of the filtering as and are effectively short circuits at 50 Hz therefore requires the highest voltage rating of the capacitors. The recommended voltage rating for is twice the R.M.S. value of the power-line waveform (approximately 60% higher than the peak value). For and a voltage rating equal to the RMS voltage should be sufficient, since these capacitors are each effectively in series with capacitor, sharing the peak voltage to be blocked. Lastly, the inductor current specifications can be calculated using the calculated branch reactance (in Table II), together with the calculated branch voltages. In Fig. 7, three sets of simulation results are shown, indicating the impact of a fluctuating access impedance (modeled here as a single, parallel impedance). In all three cases, the 50 ohm response is used as a benchmark in order to compare performance. (a) VIII. RESULTS To test the practical performance of the band-pass coupling filter, the filter circuit was built using components available at the time. See Table II, section VII. The transfer function of the practical coupling circuit was measured in transmitting mode, with a Goldstar FG2002C function generator to emulate the transmitter and an IFR 2397 spectrum analyzer to emulate a 50-ohm power-line impedance. See Fig. 6. (b) (c) Fig. 7. Simulated band-pass filter response using the available components, see Table II. These graphs show the gradual db-loss as the power-line impedance drops from 50 ohms to (a) 25 ohms, (b) 12.5 ohms and finally (c) 5 ohms. Fig. 8 shows the corresponding measured results, and should be compared with the simulation results of Fig. 7. Fig. 6. Experimental setup to determine the transfer function of the coupler in transmitting mode. (Goldstar FG2002C function generator and IFR 2397 spectrum analyzer.) 129

Attenuation (db) 0 10 20 30 40 (a) 50 ohm vs. 25 ohm 0 Attenuation (db) 10 20 30 40 IX. CONCLUSION Passive coupling filters can play an important role in power-line communications if designed carefully. They present a cheaper solution compared to transformer coupling filters and active coupling filters. It was shown that band-pass coupling filters can be designed to meet the requirements of narrowband PLC. Unfortunately power loss is an unavoidable factor when working with passive components, and their inability to amplify signals is a disadvantage of passive, transformerless coupling filters. It was shown that the preferable topology is T-shaped for power-line coupling filters, and that the Pi-shaped arrangement should never be used. Also, the choice of practical components plays an important role in achieving maximum signal transfer with passive power-line couplers. Performance of a band-pass coupler for CENELEC Bband frequencies was evaluated, which compared favorably with the simulated, expected response. The impact of impedance mismatch between modem and power-line was further explored, showing the necessity of impedance adaptation. The next phase of this study will thus look into synthesizing coupling filters with impedance adaptation circuits, to minimize the losses experienced with variations in power-line impedance. Attenuation (db) (b) 50 ohm vs. 12.5 ohm 0 10 20 30 40 (c) 50 ohm vs. 5 ohm Fig. 8. Measured response of the practical band-pass coupler. Comparing Fig. 8 (a), (b), and (c) with Fig 7. (a), (b), and (c), the measured performance of the practical band-pass coupler is satisfactory. As expected, the practical circuit, composed of real-world components, did incur more losses than the simulated circuit, especially when the impedance mismatch is severe. REFERENCES [1] P. A. Janse van Rensburg, H. C. Ferreira, Design of a bidirectional impedance-adapting transformer coupling circuit for low-voltage power-line communications, IEEE Transactions on Power Delivery, vol. 20, no. 1, January 2005, pp. 64-70. [2] P. A. Janse van Rensburg, H. C. Ferreira, Coupler winding ratio selection for effective narrow-band power-line communications, IEEE Transactions on Power Delivery, vol. 23, no. 1, January 2008, pp. 140-149. [3] Jon B. Hagen, Radio-Frequency Electronics: Circuits and Applications, Cambridge University Press, 1996, pp. 32, 40, 178. [4] Chris Bowick, RF Circuit Design, Howard W. Sams & Co., Inc., Indiana, 1982, pp. 44-65. [5] Signalling on Low-Voltage Electrical Installations in the Frequency Range 3 khz to 148.5 khz, CENELEC Std. EN 50065-1, 1991. [6] O. Hooijen, Aspects of Residential Power Line Communications, Aachen, Germany: Shaker Verlag, 1998, pp. 55, 63. [7] R. M. Vines, H. J. Trussel, K. C. Shuey, J. B. O Neal, Jr., Impedance of the residential power-distribution circuit, IEEE Trans. Electromagn. Compat., vol. EMC-27, pp. 6-12, Feb. 1985. [8] H. C. Ferreira, H. M. Grové, O. Hooijen, and A. J. H. Vink, Power Line Communication (in Wiley Encyclopaedia of Electrical and Electronics Engineering), New York: Wiley, 1999, pp. 706-716. [9] K. Dostert, Powerline Communications, Englewood Cliffs, NJ: Prentice-Hall, 2001, pp. 92, 241, 268. [10] M. Tanaka, High frequency noise power spectrum, impedance and transmission loss of power line in Japan on intrabuilding power line communications, IEEE Trans. Consumer. Electr., vol. CE-34, pp. 321-326, May 1988. 130