1 Channel Modeling for MV/LV AMI Applications in the Frequency Range < 5kHz Anand Dabak (1), Il Han Kim (), Badri Varadarajan (3), Tarkesh Pande (4) (1) () (4) Texas Instruments E-mail:{dabak, il-han-kim, t-pande}@ti.com, badri.varadarajan@gmail.com Abstract- The medium voltage (MV) to low voltage (LV) crossing is a very important application for power line communications in relation to automated metering infrastructure (AMI) in USA, Japan and also in some parts of Europe. The focus of this paper is on narrow band power line communications (PLC) systems that use frequencies between 1-49 khz. In this context it is important to be able to have a channel model for both MV--MV communication and MV/LV communication. Such a channel model will allow the signal path analysis in terms of the typical attenuation that signals may face under different load conditions. Such an analysis can then further be used to do signal to noise ratio (SNR) calculations for a given MV/LV topology. We propose to use the S-parameters/ABCD parameters for the channel model for the MV/LV case. The reason for using this methodology is that the s-parameters characterize the MV/LV line well and as long as the s-parameters of the different components in the transmit to receive chain are modeled properly, an accurate transmit to receive characteristic of the signal can be obtained. We present measurement of s-parameters for MV/LV transformers and couplers and compare them to measurements in lab. Lastly, these parameters are used to predict the observed MV/MV and MV/LV channel in the field. Index Terms Power-line communications, MV/LV, MV/MV, s-parameters, ABCD parameters I. INTRODUCTION Powerline communications (PLC) is a promising technology that is rapidly gaining traction in smart meter reading (AMR) demand response and SCADA applications. In the US, the FCC permits the use of frequencies less than 5 khz for narrow band PLC. This has resulted in a host of different technologies being developed from single carrier systems such as Lon-Talk (already deployed) to multi-carrier systems such as PRIME, IEEE 191. (in the process of being deployed) for communication purposes. While in Europe and China, each distribution transformer can support tens or hundreds of houses, in the US and Japan, especially in rural areas, each distribution transformer supports only a few houses. Hence, to minimize cost, the data concentrator should reside in the medium voltage (MV) side and it is therefore necessary for signals to cross each distribution transformer to establish communication between smart meters in low voltage (LV) side and the data concentrator. In order to design smart metering networks for these regions properly, it is important to understand the channel and noise characteristics. In this paper we propose a channel model for the MV/LV case using the s-parameters. We present some of the measurements done in the field for MV-MV, MV/LV cases and try to match the simulated channel model against the measurements. The rest of the paper is divided as follows. In Section II a brief background for s-parameters/abcd parameters is given together with their use in calculating the end to end channel response. In section III we compare the s-parameter based model for different components in the MV/LV line such as transformers, couplers and MV cables with the measured voltage transfer function. In section IV we present some field measurements done in the US grid for MV/LV transformers tested under energized conditions which raises several interesting avenues for future research. Finally in section V our conclusions are presented. II. MODELING USING S-PARAMETERS/ABCD PARAMETERS Scattering (S) -parameters and ABCD parameters are tools used to characterize two-port networks as illustrated in Figure1 below. They can also be used to characterize different components in an MV/LV power-line communication channel such as transformers, couplers and cables, thus allowing an end to end voltage transfer function (channel) characterization [1]- [4]. Channel characterization is essentially a three step process: 1) Obtain ABCD parameters for individual components in the MV/LV line ) Obtain the net ABCD matrix by a concatenation of individual ABCD matrices 3) Obtain the end-end voltage transfer function based on the net ABCD parameters. For transformers and couplers, their S-parameters may be
readily measured using a network analyzer. Conversion formulas may then be used to convert the S-parameters to ABCD parameters [5]. S-parameters for -port network ABCD parameters for a -port network Figure1. -port modeling using S/ABCD parameters As an example Figure illustrates a typical MV-LV (or LV-MV) link found in the US: ABCD MV line ABCD MV line1 ABCD txfmr ABCD LV line Txfmr LV MV MV LV ABCD Txfmr Txfmr Z eff Coupler1 ABCD coupler1 Modem V I V 1 I1 Modem1 (home) (home) (home1) Figure A generic MV-LV communication channel In this case, the two port network between modem 1 and modem can be modeled as V 1 Net I1 V I ABCD, MV LV Equation 1 where the net ABCD parameters from the modem at the MV side transmitter to the LV side modem for parameters, V 1, I 1, V, I are given by: Net ABCD, MV LV txfmr After noting that invabcd LV line V I Z eff coupler1 MV line1 where eff Zmod em // Z Home calculation of the voltage transfer function ( V /V1 ). txfmr eff Z, plugging in MV line III. EXPERIMENTAL VALIDATION OF THE DIFFERENT COMPONENTS Equation I into Equation (1) allows for a The three main components that need to be characterized in an MV/LV link are the a) transformer b) coupler and c) MVcable A. TRANSFORMER MODELING S-parameters for distribution transformers used in the USA have been reported in [6]. These S-parameters are converted to ABCD parameters from which corresponding voltage transfer function and equivalent Thevenin impedances are obtained. Next measurements are made on an un-energized 5kVA transformer at the Systems and Application Lab in TI
3 Dallas to experimentally obtain the voltage transfer function and corresponding Thevenin impedance. Figure 3 to Figure 5 illustrate the match between measurements and s-parameter based modeling. Measured Response S-parameter Based Simulated Response Measured versus Simulated Response Figure 3 Comparison of Measured versus S-parameters based LV/MV Magnitude Response for a 5kVA transformer -1 Simulated MV -> LV with various LV side loads LV side open 5 ohm load ohm load - Attenuation (db) -3-4 -5-6 -7 5 1 15 5 3 35 4 45 5 Measured Response S-parameter Based Simulated Response Measured versus Simulated Response Figure 4 Comparison of Measured versus S-parameters based MV/LV Magnitude Response for a 5kVA transformer 5 Real part MV->LV thevenin impedance Imaginary (reactive) part 3.5 x 14 LV->MV thevenin impedance Real part 3 Imaginary (reactive) part.5 Zth (ohm) 15 1 5 Zth, LV->MV(ohm) 1.5 1.5-5 5 1 15 5 3 35 4 45 -.5 5 1 15 5 3 35 4 45 Measured MV-> LV S-parameter based MV->LV Measured LV-> MV S-parameter based LV->MV Figure 5 Measured versus S-parameter based MV- > LV and LV->MV Thevenin Impedance for 5kvA Transformer B. COUPLER MODELING Similar to the transformer, the s-parameter approach was used to match measurements for an MV/LV coupler. Figure 6 illustrates the close agreement between the two different approaches.
4 15 Modem + MV coupler transmission output for various loads 14 Expected Voltage on load (dbuv) 13 1 11 1 9 8 7 Open 4ohm 1ohm 5kVAr cap bank (.7uF) 8nF parasitic cap 5 uh inductor 6 5 1 15 5 3 35 4 45 Freq (khz) Measured Response S-Parameter Based Simulated Response Figure 6 Comparison of Measured versus S-parameters based Magnitude Response for Coupler C. MV-CABLE MODELING In [7], a circuit model is provided for the MV cable which readily allows one to directly determine the corresponding ABCD model. Figure 7 compares the measured attenuation in an MV-MV line and reported in [8] with the predicted attenuation using the ABCD approach for determining the voltage transfer function (V/V1). The R, L, G, C used for this analysis are given by; L = 1.9e-6 H/m, C = 8e-1 F/m, R =.3 Ohm/m, G = 1.5e-6 S/m. Characteristic impedance of above cable Z c R jl G jc R jlg jc / ABCD model: cosh L ABCDcable sinh L / Z Z c sinhl c cosh L Figure 7 Model and Comparison of Measured versus ABCD parameters for an MV-MV line D. AN EXAMPLE ON CONCATENATING DIFFERENT COMPONENTS Lastly, the MV-LV measurement and its comparison to the simulations for a transformer, MV line and a coupler at mile distance is given; Figure 8 The MV-LV measurements at mile for a MV/LV transformer, MV line and a coupler is shown.
5 IV. ENERGIZED TRANSFORMERS The Figure 9 below illustrates an interesting example where impedance measurements where taken on both energized and un-energized transformers at 4 different time snap shots- each a quarter of the AC main cycle. For the energized transformer we find that it exhibits a time varying response that is line to the half AC mains cycle. Un-energized Transformer ( Not connected to MV side) Energized Transformer ( connected to MV side) Figure 9 Time Varying Nature of Energized Transformers This raises a few questions for future areas of research: (1) What may be the principal reason for the transformer impedance and its response variation with half AC mains cycle? Could this effect be captured by understanding the physics of the change of flux in the transformer synched to the AC mains cycle? () Measurements also show that the low voltage side impedance, measured in homes varies as a function of AC mains cycle. Is the change of the impedance on the LV side due to the change of impedance of the transformer which is close to the LV site or it also due to the change of impedance within the home itself? Furthermore which is the dominant component? (3) Most RF simulators using S-parameters currently assume the S-parameters to be stationary with time. However, since the impedance conditions in MV/LV grid are changing with the AC mains, a cyclo-stationary model for the s-parameters is more relevant. In that case, how can current RF simulators be used to simulate the end to end response? One option is to discretize time into multiple bins and then have s-parameters for each of these bins. The s-parameters for the different bins can then be used to simulate the end to end response for the individual bins. V. CONCLUSIONS In this paper we demonstrated that s-parameter/abcd modeling is a viable method for the study of channel characterization in PLC networks. The initial measurement results in most cases match with the predicted results and further characterization for different topologies is on going. Lastly we posed a few open-ended questions for the PLC community REFERENCES [1] Willam C. Black, Nader Badr, High frequency characterization and modeling of distribution transformers, IEEE ISPLC, 1, pages 18-1. [] Eklas Hossain, Sheroz Khan, Ahad Ali, Modeling Low Voltage Power Line as a Data Communication Channel, World academy of science, engineering and technology, 45, 8 [3] Francis Berrysmith A multipath model for powerline communications, [4] IEEE P 191 _Annex_C & N model_v., March 6. [5] Dean Frickey, Conversions between S,Z,Y,h, ABCD, and T parameters which are valid for complex source and load impedances, IEEE Trans. on Microwave Theory and Techniques Vol 4 Feb 1994, pages 5-11 [6] Itron, Itron-solution-S-Parameters-Measurements-for-Distribution Transformer-Model, IEEE 191. contribution document wg-11-69-- PHM5-s-parameters-measurements-for-distribution-transformer-model.docx, 11. [7] Antonio Cataliotti,Alfredo Daidone, Giovanni Tinè, Power Line Communication in Medium Voltage Systems: Characterization of MV Cables, IEEE Trans. on power delivery, vol. 3, no. 4, October 8. [8] TI, L+G, Summary of Channel and Noise Measurements In The FCC Band On A Rural US Grid, IEEE 191. document wg-1-3--phm5- summary-of-channel-and-noise-measurements-in- the-fcc-band-on-a-rural-us-grid.ppt, Nov., 1