'i tion links in form of optical fibers mounted along cess without the necessity of additional cabling.

Transcription

1 ---- RF-MODELS OF THE ELECTRCAL POWER DSTRBUTON GRD Klaus Dostert University (TH) of Karlsruhe nstitute of ndustrial nformation Systems Hertzstr. 16, D Karlsruhe, GERMANY Phone: ; Fax: Emaii: Keywords: Channel Model, Powerline, Telecommunications, Last Mile 1 Abstract With the cessation of the German Telekom's monopoly significant changes are taking place in the field of fixed network telecommunications in central Europe. n this context fast communication links over the so-called,,last mile" are of major interest. The electrical power distribution grid could turn out to be an ideal alternative for the existing standard communication links based on copper wiring, if it were possible to realize data rates in the range from 2.. 1OMbitss. This paper thoroughly examines the physical properties of distribution grids for the intended telecommunication applications. Towards this aim the frequency range up to 20MHz was investigated by extended measurements. n further steps identification and extraction of essential channel parameters towards a channel model was started. First results will be presented. 1. ntroduction rently neither appropriate communication sys- tems, nor regulations concerning frequency distri- The cessation of the i Telekom's monop- bution with respect to electromagnetic compati-, oly at the beginning of 1998 initiated significant bility are available. But work towards solutions,j changes concerning fixed network telecommuni- - - avoears worth while, as using the distribution A A ", cations in central Europe. grids for high-speed communications opens up f Most German power supply utilities (PSU) are completely new fields for telecommunication owners of long-distance high-speed communica- services such as telephony andlor NTERNET ac- 'i tion links in form of optical fibers mounted along cess without the necessity of additional cabling. the high and medium tension lines. During the Therefore considerable effort is currently spent on past, due to the existing mono~ol~~ this system design as well as on finding reliable and could only be partially Now, foundations to set up new spectrum the PSUs are able to offer various telecommuniregulations. cation services, not only for mobile phones but also within fixed networks. This means that true An important series of steps to examine the specom~etition is generallv enabled. cia1 properties of power distribution grids as " communication channels has been performed at At the moment, however, local loop fixed netthe University of Karlsruhe during the past fifteen work access can only be provided over the existmonths El]. First of all profound investigations of ing copper wires, which are still property of the the channel properties were carried out through Telekom. So true competition calls for alternaextensive measurements at various typical chantive, sufficiently fast communication links to nels. Some of the results are presented in the folbridge this so-called,,last mile". n fact, the exlowing, giving an impression of the most imporisting electrical power distribution grid would be tant channel parameters to be regarded for further an ideal candidate, if data rates in the range from steps towards regulations and system design Mbitss could be realized here at reasonable effort. Data rates in that range, however, call for Finally Some conclusions are drawn for suitable 1 RF carriers and bandwidths of several MHz. Cur- Concepts of multiple access systems. A brief dis- t!

2 cussion of the feasibility of single- or multicarrier modulation schemes, which both are of interest for high-speed communications over distribution grids, is presented. 2. Basic Channel Requirements Although data transmission over energy distribution networks has been playing a role for quite a long time, systems operating beyond the CENELEC specifications, i.e. the frequency range from 3 to khz [2], have hardly been considered. Hence, the current use of energy distribution networks is limited to rather low data rates of several kilobits per second, being sufficient for low level applications such as tariff switching or remote meter reading [3]. Recent research results gained during the past year however clearly demonstrate that the frequency range above 200kHz up to 20 MHz and even beyond offers various opportunities for true telecommunication applications [4,5]. The importance of this statement is apparent under the aspect of the new deregulated telecommunication market, especially in Germany. So upgrading power distribution grids to local area networks bridging the last mile can enable true and serious competition within the telecommunication market. As long as the Deutsche Telekom exclusively provides local loop access, the market is in fact not really open. Generally the field of high-speed applications at power lines can be subdivided into telephony (i.e. speech transmission) and pure data transmission. The latter is of major interest for NTERNET access. The requirements of data rate per user, the demand on real time ability, and the allowable channel bit error rate considerably differ for the mentioned applications. Telephony, for example, will need a relatively small data rate for each user, but the rate must be constantly provided over the entire duration of a connection. The average duration of a normal telephone call is generally small compared to typical data transmission connections. For speech transmission, the source coded bit stream can be subdivided into more and less significant bits. Since transmission errors hitting significant parts of the data stream would cause severe disruptions, such parts are carefully channel coded. So the quality of speech can be maintained acceptable, even if the channel bit error rate reaches critical values, e.g. The demand for real time capability is undoubtedly one of the key elements for speech transmission. n case that a data packet is lost, e.g. due to a burst error, a repetition is usually not possible, because an unacceptable delay would be the consequence. Short disruptions can easily be bridged by muting, almost unnoticeably for the user. Longer disruptions, however, in the range of seconds, are intolerable in any case. Data transmission on the other hand has different requirements. For NTERNET access, the amount of data to be transmitted during a session may be moderate most of the time. Sometimes, however, peaks could occur, calling for multiples of the average data rate, e.g. for the download of a picture. So a fixed, equally distributed assignment of channel capacity to all active users cannot be recommended. A flexible assignment of resources giving capacity to the active users depending on their current demands would make more sense. n practice, due to burst-like demands, the maximum data rate for each user can reach several megabits per second. A growing number of users would of course increase the probability of simultaneous demands for high data rates; but as the average duration of a connection is relatively long, generally no limitation for the number of users must be provided. Also for data transmission a low bit error rate is generally imperative. However, lost data packets can be easily repeated, since no real time capabilities are required, and the ~TERNET access is based on packet-oriented transmission protocols (TCPDP). From the typical topology of a power distribution grid in Western Europe a logical network structure can be directly derived: A low voltage network is usually fed by a single transformer, housed in a transformer station. So the assignment of a master control station to each transformer makes sense. Such a master will operate as a bridge between wide area networks and the local loop.

3 3. Channel Properties The properties of a power distribution grid used for high-speed communications considerably differ from other channels such as telephone cable or radio links. Therefore extensive investigations based on numerous measurements in different places had to be carried out [5,6]. Fig. 1 gives an impression of the typical topology of European power distribution grids. Generally up to ten cables are fed by a single transformer, each supplying between 30 and 40 households. So customers are connected to a transformer station. For our investigations a measurement set-up according to Fig. 2 was constructed. Besides the PC or notebook and the spectrum analyzer the remaining components had to be developed and assembled in our university labs. As Fig. 2 illustrates, we finally have an automated measuring system, which can be remotely controlled from the receiver side by transmitting appropriate commands over power line modems operating in the CENELEC A-band. The transmitter section consists of a direct digital synthesis (DDS) system, which is controlled by a microcontroller. The DDS is able to provide continuous sinusoidal outputs from DC to 5OMHz with amplitudes up to 3V. Different kinds of sweeps, varying in speed and frequency span, can be easily programmed as well as frequency hops. The essential part of the receiver equipment is a spectrum analyzer which can be controlled and read out by a personal computer or a notebook. So all tasks of gaining measurement results can be reduced to appropriate programming of a standard PC. 3.1 Propagation loss For typical attenuation measurements, presented in the following, the transmitter was set to sweep mode over a frequency range from 500kHz to 20MHz within a time span of 20s. Due to remote control from the receiver side synchronization of the spectrum analyzer to capture the results of a complete sweep was easy. 3.2 nterference scenario When the transmitter is switched off, everything the spectrum analyzer will receive can be denoted as interference. Due to the mentioned remote control features, it is easily possible to record interference and transmitted signal spectra within short time intervals, gaining the essential data for calculating a signal to noise ratio (SNR), which finally is the important figure to estimate the channel capacity as well as the performance of possible modulation and coding schemes. 3.3 Some measurement results Fig. 3 shows attenuation and noise measurement results gained at a typical earth cable connection with a length of 300m. t can be seen that attenuation generally increases with frequency. Such characteristic low pass behavior has been observed for all investigated earth cable connections, and therefore should be regarded as one of the major properties of distribution grids at least in the frequency range above 500kHz. Furthermore some rather weak but periodic fluctuations can be discerned in Fig. 3. This is an important observation which will lead to interesting channel modeling aspects in the following. Over extended frequency spans the noise floor is well below the received signal, even more than 50dB in the range from 5OOkHz to 6MHz. Up to 17 MHz we still observe a SNR over 20dB, which would be sufficient for excellent error-free communications. At 9 and 12 MHz, however, we register two significant narrow-band peaks within the noise spectrum. A closer look reveals that these peaks are caused by H? broadcast transmitters. With the transmission signal level used for the measurements (i.e. 65dBmV = 1,77V) such kind of interference would not cause serious problems. But due to certain aspects of electromagnetic compatibility (EMC), which must be taken into account for future regulations, the allowable transmission level might be brought down well below lv, significantly lowering the mentioned SNR margin. Fig. 4 gives results for an open wire connection over lkm. n contrast to Fig. 3 strong and frequent fluctuations of attenuation over the fre-

4 quency band are noticed here. Furthermore the low-pass characteristic found for earth cables is no longer clearly visible. Closer examinations reveal that the observed fluctuations must be caused by various echoes in combination with relatively low attenuation. The fact, that notches and peaks appear in close neighborhood, is obviously due to extended wire length (lkm) and low attenuation. The latter is important in this context; otherwise long distance echoes would simply be suppressed. Also the noise spectrum in Fig. 4 is considerably different from Fig. 3. At first glance we notice a high number of narrow-band,,interferersg, which obviously come from broadcast transmitters. The observation is not astonishing, because an open wiring system generally exhibits rather fair antenna capabilities. The above statements as well as the considerations made for Fig. 3 and numerous other measurements [5,6] which cannot be presented here due to limited space, led to the basic idea of an echo model for the powerline channel used for telecommunications in the high frequency range. 4. Channel model ideas n earlier investigations, scattering parameter models were proposed - see e.g. [7]. Fig. 5 gives an impression of the possible set-up of such a model. n practice, however, this attempt proved difficult to handle. The results obtainable with a scattering parameter model were only satisfactory in cases where very detailed knowledge about the network structure and the used cable material were available. This is due to the fact that the complete network topology must be described to characterize the network properties - see Fig. 5. Generally it can be supposed, that such an extended description will contain a lot more information as needed for our goal. Therefore ideas towards a more user-friendly echo model were worked out during the last months. The results available until now are far from complete, but some interesting features can already be presented. 4.1 Echo-based attenuation model Assuming that a transmitted signal with the information bearing signal m(t) and the carrier frequency fo is received together with a series of N echoes, we have the following receiver input signal: n (2) Ai(f) denotes the frequency selective attenuation coefficient belonging to the ith echo signal and q is the corresponding delay of this echo with respect to the direct path. A first simulation result given in Fig. 7 impressively demonstrates, that echo modeling obviously could be a rather simple and efficient method to solve our problem. Starting from the attenuation curve given in Fig. 6, and with the knowledge of some topology details, an approach incorporating three dominant echoes was made with the following delay parameters: As Fig. 7, in comparison with Fig. 6, demonstrates, there is excellent accordance, although only 3 echoes have been considered. After this encouraging result, echo models have be tried on more and different links, leading to the interesting conclusion, that for earth cable modeling generally only a small number of echoes must be considered, in practice 3...5, although there might be up to 40 taps along such a cable, each causing a reflection. This interesting observation is due to the fact, that RF attenuation on earth cables is considerable, causing distant echoes almost to vanish, or at least to be of negligible effect at the receiver. Only dominant reflections in the receiver's closer neighborhood must be taken into account. The above discussion is of significant practical importance, because small numbers of echoes

5 will be easy to handle and parameter identification can be performed with moderate effort. Steps towards this aim are currently under way in our working group at the University of Karlsruhe. Concerning the attenuation coefficients some basic results are already available: Starting from a somewhat different notation as mentioned before, an attenuation coefficient Ai(f) can also be characterized through the cable length e: where a(f) in the exponential function can be split into two parts as follows: Here the contribution ar(f) is mainly caused by skin effect, increasing the resistance of the cable's metal wires, whereas ~ ( f comes ) from dielectric losses which occur within the insulation material between the wires. Since both ar(f) and ~ ( f are ) monotonously growing frequency dependent functions (3) can also be written as The values of K,, c, and Ei were found to be constant for a cable type. E generally is in the range , meaning that we not only observe attenuation from skin effect but also from dielectric losses. For pure skin effect, we would expect ~=1/2. Further investigations will be made to define generally applicable mean values for the parameters K, c and E in (5), making the detailed analysis of different cable types unnecessary. Otherwise insurmountable problems could occur in practice, as often, especially in older supply branches, not all of the different implemented cable types are known, whereas distances needed for estimating echo delays are normally available. 4.2 Remarks on interference Besides attenuation, the quality of a communication link is significantly influenced by interference. The dominant types of interference degrading communications over power lines can be classified as follows: back ground noise (nearly white Gaussian) narrow band (single tone) interference, mostly generated by broadcast radio transmitters impulse noise, either periodical and synchronous with the mains voltage, or stochastic, from various kinds of switching equipment The noise spectrum shown in Fig. 3 is insignificantly stressed by narrow band interference. n the major portions of the spectrum low level background noise is observed, with a spectral amplitude around -28dBmV (= 40pV). This value of course depends on the spectrum analyzer filter bandwidth set to lookhz for this record. So, in order to get the generally applicable figure of a single-sided noise power density No, the following calculation has to be made: The narrow band interference from HF radio transmitters around lomhz can be severe, so that it is advisable to exclude - if possible - such constantly affected portions of the spectrum from telecommunication applications. 5. Channel capacity considerations and EMC discussion Let us now consider a virtual communication system with the following parameters, to get a first impression of the achievable transmission quality at power lines: transmission level: 1 v, data rate: 1 Mbitls carrier frequency: 15MHz When such a system would be applied to a link with characteristics according to Fig. 3, the only interference would be background noise with the above calculated noise power density leading to a ratio of bit energy versus noise power density (ENR) of Eb/N,=18dB at the receiver input, as signal attenuation is around 60dB. This fairly high ENR value would generally make the

6 application of amplitude sensitive and spectrally efficient modulations schemes such as quadrature amplitude modulation (QAM) feasible. But before taking such a step, the following restrictions should be carefully checked: Due to possible EMC problems, upcoming standards might restrict allowable transmission signal amplitudes well below 1V. Numerous earth cable links exceed 300m, exhibiting significantly higher values of attenuation as given in Fig. 3.. n the above calculation the receiver was only disturbed by background noise. n practice narrow band interference and impulse noise cannot generally be excluded. Despite of possible restrictions the example is useful for a clear demonstration of the enormous channel capacity provided by the power distribution grid. mportant further work has to be done to set up reasonable standards and rules for transmission power and use of spectral resources, in order to guarantee electromagnetic compatibility (EMC) on one hand, and on the other to preserve the proposed capabilities of distribution grid for telecommunication purposes. The latter calls for not too restrictive rules, which easily could kill possible innovation potentials given by this truly alternative last mile bridge. Setting up the rules will in fact not be a trivial task, as sophisti- - cated compromises must be found to preserve valuable existing services in the same frequency band, and in parallel give enough room for innovations. 6. Conclusions After cessation of the last telecommunication monopoly structures in Europe, the market is now open for various activities towards true competition. n order to make this possible, alternative cable links to the customer's homes are needed. n fact, plenty of such links exist in form of the electrical power wiring. This paper attempted to provide an answer to the basic question, whether existing distribution grids generally exhibit physical properties, necessary for telecommunication purposes. As a result of various measurements and steps of modeling, it could be pointed out, that data transmission over typical European power distribution grids at rates up to several megabits per second appears realistic. t could be proved, that possibilities for telecommunication purposes exist up to frequencies of 20MHz. Even open overhead wiring does not present crucial obstacles for RF signaling. The properties of different wiring types can be captured by a few parameters, making the proposed echo-based channel model easy to set up and handle with very moderate knowledge of topological and electrical details. Normally interference in the frequency range of interest can be modeled as colored background noise with low power density. n some distinct places of the spectrum, however, broadcast transmitters appear as interferers. Despite of this drawback, large and almost undisturbed gaps can be found, preferably useable for our purposes. Future work will concentrate on measures towards RF isolation of transmission links from other parts of the supply system. On one hand this will be an effective means against EMC problems, and on the other hand transmission power obviously can be lowered in a conditioned network. n any case the benefits of network conditioning will be valuable for the acceptance of new services and for setting up rules and admission regulations for the new communication channels. Currently no system, nor standard hardware is available for telecommunications over the power distribution grid. However, various experimental investigations are under way and have already proved the feasibility of using power lines as high speed data channels. So, undoubtedly there are technical and economical solutions, which will have to be further developed. With the actual state of the art in mind, telecommunication over the power grid appears generally easier than e.g. the management of urban mobile radio channels. To complete channel modeling, some important properties are still missing, such as detailed figures of phase response or group delay. The missing knowledge is e.g. needed for a general answer to the question, whether wide-band modulation techniques with one or very few carriers would be a good choice, or if it would be better to rely on

7 multi-carrier techniques such as orthogonal frequency division multiplexing (OFDM). Furthermore, also questions concerning amplitude sensitive and bandwidth-efficient modulation schemes, such as QAM are expected to be answered by more detailed channel knowledge. 7. References [] Arzberger M., Dostert K., Waldeck T., Zimmermann M.: Fundamental Properties of the Low Voltage Power Distribution Grid. Proceedings of the 1997 nternational Symposium on Power Line Communications and its Applications, Essen (April 1997) [2] EN : Signalling on low voltage electrical installations in the frequency range 3 khz to 148,5 khz, CENELEC, Brussels, [3] Dostert K.: All Digital Spread Spectrum Modems for Power Line Communications. European Transactions on Telecommunications, ETT, Vol. 7, No. 6 (1996) [4] Dostert K.: Telecommunications over the Power Distribution Grid - Possibilities and Limitations. Proceedings of the 1997 nternational Symposium on Power Line Communications and its Applications, Essen (April 1997) 1-9. [5] Dostert K.: Telekommunikation uber Energieverteilnetze - eine Standortbestimmung. Tagungsband der Fachkonferenz,,Powerline - Daten- und Telekommunikation uber das Strornnetz" des nstitute for nternational Research (R), Dusseldorf, June [6] Dostert K., Halldorsson U. Zimmermann M.: Modellierung elektrischer Energieverteilnetze als schnelle Nachrichtenkanale - ein praxisorientierter Ansatz -, Conference Proceedings:,,Power Line Telecommunications - Forschung trifft Markt", EUTELS PTF e.v. Dusseldorf, Nov [7] Threin G.: Datenubertragung uber Niederspannungsnetze rnit Bandspreizverfahren. Fortschritt-Berichte VD, Reihe 10, Nr. 156, VD-Verlag, Dusseldorf, more cables \transformer station Fig. 1: Typical topology of a residential distribution grid in central Europe

8 receiving. e~uivment with remote control facility transmitter - spectrum powerline analyzer 125 MHz < DDS TP system A \ J PC/Notebook - MC - PLC modem modem system Fig.2: Automated measuring system for the transfer function and the interference scenario (100 khz...20 MHz) based on remote control by CENELEC-A-band modems (transmission level: 65dBmV, analyzer bandwidth: 100kHz) Fig. 3: Measurement results for signal and interference levels at a 300m cable link frequency [MHz] *-b

9 (transmission level: 65dBmV, analyzer bandwidth: 100kHz) frequency [MHz] ~-b Fig. 4: Measurement results for signal and interference levels at a looom open wire link - - r1 - Z2 scattering - matrix : [: S= termination [ S port devices 4 Y [s11 [s31 r, [ s41 [% transmitter b EE U receiver Fig. 5: Example of a scattering parameter model application

10 frequency [MHz] Fig. 6: Attenuation measurement results for a 130m earth cable link Fig. 7: Simulation results for the earth cable link from Fig. 6, based on a 3-echo model