Complete Power-Line Narrow Band System for Urban- Wide Communication Gerd Bumiller, Markus Sebeck GmbH Unterschlauersbacher-Hauptstr. 10, D-90613 Großhabersdorf, Germany Phone: +49 9105 9960-51, Fax: +49 9105 9960-19, Email: Gerd.Bumiller@iad-de.com Affiliation: Keywords: Narrow Band PLC s OFDM, OPC, Home Automation, Remote Metering Abstract Typical applications for narrow band PLC are remote metering, value added services for utility companies, intelligent power management, security services and home automation. All these different services have to be integrated into one system and share the same medium. Most utility companies have a communication gateway in their high-voltage medium-voltage transformer stations. From this point hardly any communication lines to the end user are available. In this paper we want to present a complete communication system for low and medium-voltage distribution lines and a communication server to integrate different utility service and home automation software. This system DLC-1000 developed by the company is based on the ASIC with powerful forward error correction and possible data rates up to 100 kbps presented on the ISPLC99 [1], the packet-oriented master-slave network management system optimised for a large number of participants presented on the ISPLC 2000 [2] and a communication server that was developed as service in Windows NT / 2000. Physical Communication In a lot of applications data packets have to be sent to a huge number of receivers. The physical and the medium access layer of the power-line modem is optimised for this task. A transmission packet starts with a preamble for adjusting the automatic gain control and synchronisation. The block is designed to allow the use of the receiver's complete dynamic of 70 db AGC gain without requiring an estimate of the gain and ensure reliable synchronisation without any framing. The total dynamic range of the receiver exceeds 80 db. When any client detects a synchronisation, the transmission is automatically disabled for the duration of one OFDM-Symbol and the danger of collisions is restricted to the period of synchronisation. This characteristics is only used by the network management system in special time slots for auto-logon. The other time, collisions are avoided by resource management of the master. After the preamble an OFDM symbol is transmitted. In combination with the FFT, the cyclic prefix used as guard interval transforms the linear interference from e.g. reflections into factors on the subcarriers. By use of special modulation in combination with powerful forward error correction, the influence of the factors can be ignored and a good channel equalisation is achieved. The forward error correction is implemented by means of convolutional coding and MLSE decoding. A transmission block carries, depending on the mode, between 48 and 159 bytes of data and has, depending on the bandwidth used (4 to 38 khz), a duration between 12.5 and 100 ms. In this application the communication is packet-oriented and every burst has a header with a 32-bit CRC for error detection. The automatic repeat request is part of the network management system. An additional Reed- Solomon code without interleaving over several bursts would not improve the performance. The network management is part of the software, running on the INTEL 8052 compatible MCU integrated into the DLC-2A ASIC. An amplifier module is used to connect the modem directly to the low-voltage grid or to the medium-voltage grid over a coupling unit. As the transformers do not allow direct communication from the low-voltage to the medium-voltage grid, we have nearly independent networks.
Network Management System The network management system is based on a master-slave concept where all modems share the same frequency band and thus a TDMA scheme is applied. The use of different frequency bands allows the coexistence of several logical networks on the same physical network. In order to extend the physical range of the network, two repeater levels are employed and every slave can also be a repeater. There is no direct communication between two slaves and hence the network has a virtual star topology. Slaves are allowed to transmit only on demand of the master. As OFDM performs parallel transmission of bits, decoding can only start after the reception of a complete block. Therefore, responding is not possible within the next TDMA slot but instead an interleaf is specified after which the slave is supposed to respond. Thus, system delays caused by run times of both the physical layer and the layers of the network management software are taken into account. The number of slaves in the network is dynamic and both logon and logoff of the slaves are performed automatically by the master modem. Every modem in the network can be identified by its unique serial number and during logon it is additionally assigned a unique network address by the master which is valid for the period of its logon. If no application requires transmission of data, the master constantly updates its network statistics (logged-on modems, routing tables) by a cyclic poll of all slaves. In the response PDUs to these status polls, the slave provides the master with information on its current status, possible available data and statistical information for routing purposes. The status poll period, i.e. the number of cycles for polling all slaves, depends on the number of slaves in the network and is used as a reference for the computation of the network statistics. The use of repeaters requires the determination of the best routing paths, which is done by the network management software in the master modem. For this purpose, every slave keeps a table with its favourite repeater slaves and in response to status polls forwards the best five entries to the master. In case of telegrams to a single slave, the master has to compare and test eleven different paths including the direct path, 5 paths using one repeater each and another 5 paths using two repeaters. For this purpose, every path is judged by an evaluation figure, computed using the respective response quota, the reception probability of a possibly applied repeater and an additional weighting factor used to prefer short paths. By repeating a block the response quota has changed and therefore, in most cases, the algorithm will use a different path. In case of broadcasts multiple repetitions of the same telegram by the master as well as by certain repeater slaves should provide a good coverage of slaves. In this context, the question arises which repeaters should be used so that only a few repetitions are required. Using the statistical information in its routing table, the master is able to estimate for every slave the probability that it has not yet received the broadcast. This probability is updated after every telegram sent. Thus, the master chooses the path for the next telegram such that the probability for the slave with the currently highest probability is minimised, which also decreases the probabilities of the other slaves. A few telegrams are also transmitted using the direct path. This procedure is repeated until either the probabilities of all slaves have reached a threshold or a maximum number of repetitions is exceeded. This is all done by the master and therefore, higher level applications do not have to know anything about the repeaters. To allow auto logon, the master regularly sends auto logon requests. These are sent as broadcast PDUs using selected repeaters. In order to avoid collisions on the channel as far as possible, the slaves which want to log on each respond after a random delay within a fixed range. If the master detects collisions on the channel, it can reduce the number of slaves which are allowed to answer the request by specifying a few bits of their serial number and thus reduce the probability of collisions. As the random delays can only be multiples of the slot interval, the system can be considered as a slotted ALOHA system during periods of auto logon. Network Structure The resulting network structure for communication is a virtual star on the medium-voltage lines, implemented as a master-slave system, and a second virtual star on the low-voltage lines, also implemented as a master-slave system. In the following figure we give a system overview with hardware and software structure. In this overview every master slave-system is represented by a master and only one slave without any repeater.
Hardware Modem Software Host Computer i.e. NEXUS, MINOS, NAMS PLC, DIANA (virtual Fieldbus Master) Virtual COM Port (Driver) OPC in design Confguration Space / DNS POTS Modem PC VBO1 Connection Server VBO2 RX/TX Hardware VBOn medium-voltage network Master MV M CU CU 20 kv RX/TX Hardware Master Slave external UART Port NMS Master Slave MV NMS-Slave Ports Ports Slave MV + Master LV M S RS232 Master LV external UART Port low-voltage network Slave LV S 230 V Master Slave NMS-Master NMS Slave Ports M = Master S = Slave NMS = Network Management System MV = medium-voltage LV = low-voltage CU = Coupling Unit = Amplifier Module Figure 1: System Overview
Components of the System Both software structure and hardware of the different parts of the system will now be described. The power-line communication of this system is based on the DLC-2A ASIC. The complete network management and control software runs on an INTEL 8052 compatible MCU integrated into DLC-2A ASIC. Based on a multi-tasking real-time operating system, the different layers of the network management software, which are independent from the interface hardware, use several. The application layer is the communication manager between the network management software and the interfaces (cf. Figure 1). Slave Several slaves with different interfaces are specified and also implemented. In some cases the slaves are integrated into other equipment, e.g. a meter, while in other cases it is a separate box with several interfaces. These interfaces can be very simple, such as digital IO, trigger, analog inputs and S0-impulse input for meters, or more complex, such as RS 232, RS485, IEC 1107, M-Bus and internal I²C-bus. On a bus more than one client is possible and the slave works like a gateway or data concentrator. We have built the DLC-100 Module as a universal slave with all these interfaces, integrated power supply, integrated single-phase amplifier module for low-voltage and optionally an external 3-phase amplifier module for low or medium-voltage (cf. Figure 2). Optional 3 phase 1 x RS232/RS485 1 x IEC 1107 TTY/opt. 4 x S0 Impulse 4 x Analog 8 x digital 8 x digital 2 x Input Trigger 4 x Output Relays COM COM Input Input Power Supply Input Output 230V 230V 1 phase Slave LV Power Supply/ PLC Figure 2: Universal Slave In most applications, only a subset of these interfaces are used. Special slaves with only some of the interfaces of this universal slave to reduce costs can easily be realised. The interfaces are logically grouped into ports. Port 0 is reserved for a service interface over power-line for internal administration of the slave. After an automatic logon, the communication server is able to read the version number and the ports existing of the slave and adds this information to his database. In addition to configuration, software download is also possible using this port. All other ports depend on the interfaces realised. The application layer is a multi-port communicator with a universal interface to the ports. Specialised port drivers adapt this interface to the physical medium. If the physical interface is a bus, the port driver is a gateway or data concentrator. Depending on the hardware, the several port drivers can be linked with the application layer and the network management software to a single software application. A power-fail logic saves critical data of the network management software and the port drivers, e.g. meter results, to a flash and automatically restores this data after the power is restored. For integration of the slave and the DLC-2A ASIC into a customer system, a development kit with a library of the network management software is available. The customer has to implement only the controller functions of his system inside the application layer level and additional port driver. A lot of systems, depending on the requirements, do not require a second controller and thus, costs are reduced.
Medium-Voltage Low-Voltage Gateway The MV-LV Gateway is typically installed at transformer stations. Because of the low impedance of the lowvoltage grid at the transformer station, an external 3-phase amplifier module, connected to the low-voltage master, is installed not far from the distribution lines. The low-voltage master of the low-voltage network is a communication board within the DLC-100 Box. The structure of the software of masters and slaves are the same. On the internal MCU, the real-time operating system run and the master of the network management system is implemented using several tasks. For the routing tables of the network management system, the master needs larger memory. The application layer is the communication manager between the ports and the network management software. Port 0 is reserved for internal configuration and, in contrast to the slave, it is possible to access this port from other ports, e.g. an external UART. The communication between medium-voltage slave and low-voltage master is connected by the back-plane of the DLC-100 Box with the serial port of the communication boards. The DLC-100 Box is a waterproof IP54 box with a back-plane for communication and the power supply. Display, relays (3x400V/max.100A) and other devices for special applications also as slave can be provided. Also a 3- phase amplifier module could be included in an DLC-100 Box. This box has space for 3 communication boards. One of the communication boards is the master of the low-voltage grid. The second communication board is the slave of the medium-voltage grid. The slave is connected to a medium-voltage amplifier module and a coupling unit to the medium-voltage network. The software of the medium-voltage slave is nearly identical to the software of the slave. The only difference is the specialised port driver for the external UART and the protocol for the communication with the low-voltage master. The other ports of the slave can be used for control applications or meter reading within the transformer station. In a medium-voltage network more than one medium-voltage low-voltage gateway is possible. Master The medium-voltage master is connected over a medium-voltage amplifier module and a coupling unit to the medium-voltage grid. The communication to the PC is realised directly with RS232 or with a voice-band modem. In another version of the communication ASIC which we also present on ISPLC 2001 as Narrow Band Power-Line Chipset for Telecommunication and Internet [4], Ethernet, ISDN and USB interfaces are available. A possible hardware for the master could be a DLC-100 Box and a communication board. The master version of the network management software for the medium-voltage network and for the lowvoltage network are identical. It is not always necessary to realise the medium- and low-voltage network. If a communication link from the transformer station to the service centre is available, the master is installed directly at the low-voltage network. For the communication with our communication server, application layer software and a special port driver has been realised. The protocol running over the RS 232 or over the voice-band modem is especially developed to meet the requirements of the master with its limited buffer size and the low data rates of voice-band modems (38.4 kbps for duplex communication). We have analysed standard protocols for this interface, but no one has met the requirements. Our protocol is packet-oriented and not classical master-slave protocol. Data initiated from the slaves like e.g. events or the network management like e.g. auto-logon of a new slave has to be supported. Also a data-flow control depending on the slaves is necessary. If the communication to a particular slave is routed over two repeaters and several repetitions are necessary, the data throughput is much lower than to other slaves. If the limited buffers of the master are filled with messages to this particular slave, the other slots of the time frame cannot be used for communication with other slaves and system performance is wasted. The structure of the protocol also supports cascaded systems requiring to address the slave and port on the medium-voltage network and additionally the slave and port on the low-voltage network. For integration of the master and the DLC-2A ASIC into a customer system without PC-based communication server, a development kit with a library of the network management software is available. The customer has to implement only the controller functions of his system inside the application layer level and additional port driver.
Communication Server At a service centre the power-line network has to be connected to several existing management software packages for different services and applications. Even if the structure of the data is packet-oriented, the protocols of existing management software packages often use a point-to-point structure, sometimes in combination with link-oriented protocols for small networks such as M-Bus or IEC 1107. The communication to a bus master is often realised using a voice-band modem. Newer applications in the field of home automation have an OPC interface. The PC-based communication server allows to connect to different management applications and services at the same time. The management applications see one or more virtual serial ports accepting Hayes-modem commands. This virtual serial ports communicate over the DCOM interface with the communication server and therefore, the management applications and the communication server do not have to run on the same computer. The dialled telephone number is an alias name and will be translated by a dynamic name server. Newer applications can be connected to the communication server directly via DCOM interface or via OPC. This applications do not have to run on the same computer either. On the other side several masters, both low- and medium-grid masters, can be connected directly, via a modem or a LAN to one communication server. For all these networks connected to the communication server, one database for the dynamic name server exits. If a new slave is connected to the network, an auto-logon procedure is started by the network management software and the slave will be added automatically to the database. Only the alias name can not generated automatically. The server generates a communication object ( VBOn in figure 1), which defines a virtual connection to a port of a slave, for every connection requested. A dynamic name server converts a number or alias name into one or, in a cascaded system, more power-line address(es). In our case, it is a slave address and port number of the medium-network and a slave address and port number of the low-voltage network. The data packets of the communication objects are scheduled and transmitted via the master and the power-line network to the slaves and vice versa. The connections exist in parallel and the maximum performance of the network can be exploited. It is also possible that different services run independently of each other at the same time on the same network and sometimes even over the same slave. Conclusion Here, we have presented a complex system. A lot of users can be connected and to manage a large and dynamic network on power-line like this is only possible, if routing and logon is supported automatically as implemented in our network management software. The support of very different services is only possible, if the applications of these services run completely independent of each other and of the structure of the network. This is realised with different ports at the slaves and the communication objects of the communication server, with alias names and a dynamic name server. A fair allocation of the resources available is guaranteed by the central resource management of the communication server and masters. All components in this system, from physical layer over protocols to the interfaces, has been optimised to meet the requirements of this system since the start of the development in 1996. It was a long way, but now we are ready. References [1] Manfred Deinzer, Matthias Stöger: Integrated PLC-modem based on OFDM. Proceedings of the 3 rd International Symposium on Power-Line Communications and its s (ISPLC 99) [2] Markus Sebeck, Gerd Bumiller: A Network Management System for Power-Line Communication and its Verification by Simulation. Proceedings of the International Symposium on Power-Line Communications and its s 2000 (ISPLC 2000) [3] Dale E. Rogerson, Inside COM, Microsoft Press 1997 [4] Manfred Deinzer, Gerd Bumiller: Narrow Band Power-Line Chipset for Telecommunication and Internet, ISPLC 2001