AU2152699A - Multi-carrier method and device for implementing said method - Google Patents

Abstract

Disclosed is a multi-carrier method for digital data transfer, especially for electrical power distribution networks. According to the invention, N amounts of symbols are assigned to combinations of Id(N) data bits on the basis of the data flow to be transmitted in order to achieve quick and interference-free transfer. Each symbol consists of N signal forms with different frequencies. On the receiver side, a series of 2N digital signal-adapted parallel-working filters are used for simultaneous incoherent optimal reception of all N signal forms used to represent a symbol. A digital numerical value corresponding to the signal form energy is formed as correlation value for each of the N signal forms. An evaluation and scaling of N correlation values is carried out by comparison with a freely chosen threshold. The correlation values forming a symbol are added to obtain the corresponding symbolic value. The corresponding functional units can be advantageously integrated into an ASCI to implement the inventive method.

Description

ABB Patent GmbH Mannheim 10 December 1997 Mp. No. 97/703 PAT 3-Sf/Bt 5 Multicarrier method and device for carrying out the method 10 Description The invention relates to a method and a device for the fast and reliable communication of digital information, especially via electrical power distribution systems, 15 using a number of preferably orthogonal signal forms. In December 1991, European standard EN 50 065, which has the status of a German standard, came into force. This standard regulates the utilization of the 20 frequency band from 3 kHz... 148.5 kHz for signal transmission purposes on electrical power distribution systems. In this standard, the available band is roughly divided into two ranges: the frequency range from 3 kHz ... 95 kHz is reserved for the public 25 utilities where transmitting amplitudes of up to a maximum of 134 dBpV are permissible. The remaining frequency range from 95 kHz ... 148.5 kHz is available to the private user without a licensing requirement but a transmitting amplitude of 116 dBpV must not be 30 exceeded. This rather low transmitting level (less than 1 V) accompanies a narrow available bandwidth. EN 50 065-1 makes high demands on methods and devices for the interference-proof transmission of information via electricity supply lines. Because of the amplitude 35 limitation, transmission methods which can always use the full amplitude have the best chances of success. 47C.

Mp. No. 97/703 - 2 - 10 December 1997 At present, systems with narrow-band or also with wide band, band-spreading modulation, which are described, for example, in printed documents DE-Al-44 23 978, DE-Al-43 23 376, DE-C2-40 01 265, DE-C2-40 01 266, 5 EP-Bl-0 200 016 and EP-B1-0 199 148, are used for transmitting information via electricity supply systems. The method which is probably the simplest method technically and which can always utilize the complete permissible amplitude is FSK (Frequency Shift 10 Keying) . The fact that FSK is simple to implement is certainly a decisive reason for the fact that the majority of systems currently available on the market for transmitting information via electricity systems operates with FSK. The disadvantages of FSK have become 15 apparent incrementally in practical use during numerous field trials and, in accordance with current estimations, it must be said that adequate reliability for services of all types which the public utilities intend to offer via their systems in future will not be 20 achievable by means of FSK. The essential problem of FSK is that the transmission fails completely even when a single carrier frequency is disturbed, either due to selective attenuation which can occur at any time at any point in the time-variant transmission channel, or 25 due to a narrow-band interference source in the form of a television set or of a switched-mode power supply. Band-spreading methods such as, for example, frequency hopping (FH), which allow many types of frequency variation, could provide a remedy here - see, for 30 example, printed document DE-Al-44 23 978. Although the complexity is greater in comparison with an FSK system, it is bearable due to the advances in modern microelectronics. Nevertheless, the practical implementation of currently used FH methods has 35 hitherto not gone beyond a trial stage; in particular, the market has hitherto not provided any operable systems. Although the FH technique is basically capable of overcoming all problems of FSK observed in practice, there are still decisive disadvantages in the method Mp. No. 97/703 - 3 - 10 December 1997 and in the technical implementation which, in the final analysis, are probably the reason why systems with flexible use of a number of carrier frequencies sent successively in time have hitherto only been used hesi 5 tantly. The significant disadvantage when changing from FSK to FH methods is that it is necessary to increase the chipping rate for the same net data rate (compare 10 printed document DE-Al-44 23 978) . With a chipping rate of 1200 s- an FSK system thus achieves, for example, a data rate of 1200 bits/s whilst an FH system with 4 carriers per bit can now only offer 300 bits/s, for example. To achieve the same net data rate with FH, a 15 chipping rate of 4800 s-1 would thus be required. This leads to an increased complexity in the signal generation and processing in every case, but especially in receiver synchronization which would now have to become more accurate by a factor of four. 20 The invention is based on the object of specifying a multicarrier method by means of which the disadvantages listed can be considerably reduced and a fast and interference-resistant transmission can be achieved. In 25 addition, a device for carrying out the method is to be specified. This object is achieved by a symbol-processing multicarrier method having the features specified in 30 Claim 1. Advantageous developments and a device for carrying out the method are specified in the further claims. The invention is described in greater detail by means 35 of the explanation, given below, of advantages of the solution according to the invention, the fundamental concepts and of illustrative embodiments referring to drawing figures, in which: Mp. No. 97/703 - 4 - 10 December 1997 Fig. 1 shows the overall configuration of a modem, Fig. 2 shows a microcomputer system for implementing a number of function blocks from Fig. 1, and Fig. 3 shows a representation for explaining a 5 preferred synchronization method which can be used in the method according to the invention. The advantages of the symbol-processing multicarrier method are, among others, that a constant signal 10 amplitude is always used and thus optimum energy utilization of the channel resources is given whilst adhering to the relevant standards. In the text which follows, a device called a modem (abbreviation for modulator/demodulator) will be used as an example for 15 describing the invention. Such a modem is used for the bidirectional communication of digital information via electrical power distribution systems. The data transmission is in every case resistant to interference on the electricity supply lines and allows reliable 20 fast communications in spite of the adverse transmission characteristics of these lines. Furthermore, simultaneous activity of a number of modems on the same power system is possible without mutual interference (multiple access to the same 25 frequency band). The manufacture and use of an application-specific integrated circuit with mixed analog and digital functions (mixed-signal ASIC) is advantageous for 30 carrying out the method. Thus, the signal generation at the transmitter end and the correlative processing at the receiver end up to the bit decision can take place with high accuracy and 35 exact reproducibility on a digital basis. A virtually unlimited multiplicity of signal forms is available without hardware changes, including the capability of freely selecting various data rates and frequency ranges within wide limits. These digital functions Mp. No. 97/703 - 5 - 10 December 1997 accommodated in the ASIC cannot be implemented in any other way in the current state of the art, not even by using the most elaborate digital programmable signal processors. In addition, the ASIC contains analog 5 functions for amplification and filtering of the received signal and for analog/digital conversion and for digital/analog conversion at the transmitter end, i.e. functional units which are basically not contained in a standard signal processor. The amplification of 10 the received signal is automatically adjusted in such a manner that the analog/digital converter is always operated within a favourable operating range. The invention is based on the following considerations: 15 If N carrier frequencies are used in an FH system, these can be used for forming N! (N faculty = N(N-1) (N 2) ... 1) different combinations which will be called symbols in the text which follows. Thus, 6 symbols can 20 be represented by N = 3 frequencies; N = 5 already provides 120. The 6 symbols with N = 3 would be capable of transmitting [ld(6)] = 2 bits whilst it would be possible to achieve already [ld(120)] = 6 bits with N = 5. In "normal" FH, however, these possibilities are not 25 utilized but only two symbols (combinations) are evaluated which corresponds to information content of 1 bit - see, for example, printed document DE-Al-44 23 978. In normal FH technology, this provides redundancy which, for example, makes a transmission 30 resistant to the failure of individual frequency ranges. The resultant disadvantage is obvious: the transmission speed drops in accordance with the number of carrier frequencies used. The basic concept which led to the present invention avoids this speed 35 disadvantage and, at the same time, unrestrictedly guarantees that the resistance to interference which is inherent in the FH principle is maintained. A few examples will be compared in order to explain this essential aspect of the invention in greater detail: Mp. No. 97/703 - 6 - 10 December 1997 If the three frequencies used are called fi, f 2 and f 3 for N = 3, the following 6 symbols Si... S 6 are obtained which can be utilized in various ways: Symbol No. Frequency FE technique Symbol sequence processing

S

1 fi f 2 f 3 0 00 S2 fi f 3 f 2

-

01

S

3 f 2 fl f 3

-

10

S

4 f 2 f 3 fl - 11

S

5 f 3 fi f 2 1 S 6 f 3 f 2 fi - Table 1: FH technique and symbol processing for N = 3 carrier frequencies 10 In the FH technique column, the high number of unused symbols is immediately obvious which, lastly, supplies the resistance. However, the pure number of unused symbols is not in any way the only measure of the achievable resistance to interference but the 15 determining factor is the number of chips with different frequency in the symbols used for representing information. It can be seen that the symbols Si and S5 marked by 0 and, respectively, 1 differ in the FH technique column in all three chips 20 and can thus be very easily separated in the receiver. Considering the "symbol processing" column, it can be seen that only the symbols Si and S 4 marked with 00 and 11 and 01 and 10 corresponding to S 2 and S 3 differ in all three chips whilst the 00 and 01 or 00 and 10 25 combinations only differ in two chips. The resistance to interference suffers from this fact. If all 6 symbols were to be used for information transmission, Mp. No. 97/703 - 7 - 10 December 1997 there, would only be one different chip in each case as decision criterion; as a result, the transmission would no longer be more reliable than in the case of FSK because the failure of a single frequency would lead to 5 failure. The above table shows that the use of three carrier frequencies obviously does not lead to an advantageous solution. If, in contrast, a sequence of four frequencies is considered which allows 24 symbols but of which only exactly the four differing in all 10 four chips are used for information transmission - see Table 2, Frequency sequence Data symbols fi f 2 f 3 f 4 00 f 2 f 3 f 4 fi 01 f 3 f 4 fi f 2 10 f 4 fi f 2 f 3 11 Table 2: Optimal symbol processing with N = 4 15 the same resistance to interference would be obtained which would be obtained with the standard FH technique but because each one of the symbols shown now represents two data bits, twice the transmission rate 20 is obtained. The principle shown can be used not only for N = 4 but whenever exactly N of the N! possible symbols are used for representing information. In this arrangement, it is advantageous if N is a power of 2 because ld(N) then results -in an integral number of 25 bits. If ld(N) is not integral, only the next smaller integral number can always be used, i.e. the useable net data rate is reduced. After N = 4, N = 8 would then be advantageous again, in which arrangement 4 bits could be transmitted with each of the symbols used, 30 utilizing 8 of the 8! possible symbols. For practical use in power distribution systems, a system based on 4 Mp. No. 97/703 - 8 - 10 December 1997 carrier frequencies will reliably offer adequate resistance to interference in order to guarantee the new services planned by the public utilities with high reliability. As can be seen from Table 2, up to three 5 of the four carrier sequences could fail in boundary cases without this resulting in a bit error. In comparison with FSK where even the failure of a single frequency leads to failure, this is a significant advance. Using normal FH, N 7 carrier frequencies 10 would be necessary and the net data rate would drop to 1/7 of the chipping rate. The method according to the invention provides the same resistance to interference with a ratio of 1/2 of data rate to chipping rate. 15 Practice has shown that with a clever selection of carrier frequencies, the probability that two of these will fail at the same time due to attenuation or inter ference is extremely small. The case that this will occur for three carriers can be virtually eliminated. 20 However, the failure of single carriers is observed relatively frequently. The choice of N = 4 carrier frequencies is certainly advantageous for the practical implementation of the 25 method according to the invention. The illustrative embodiment described in the text which follows is therefore based on N = 4. A largely digital concept with a high degree of 30 integration is of decisive significance for an inexpensive and reliably reproducible implementation. Furthermore, it is advantageous if adaptation to various types of task without changing the hardware is possible and if a large proportion of the analog 35 circuit parts required for the complete construction of a modem can also be monolithically integrated. The embodiment of the invention presented in the text which follows demonstrates a universal advantageous solution.

Mp. No. 97/703 - 9 - 10 December 1997 The starting point of the considerations is a transmission of binary information which is mapped onto frequency sequences (symbols) according to Table 2. During the period T, of a symbol, the transmitting 5 frequency is changed abruptly four times in this example. The larger the N selected, the more resistant to interference can be the transmission of information, with a system complexity, however, which also rises. For the reasons already listed above, N = 4 different 10 and suitably selected frequencies are adequate for most applications of the present invention. A suitable choice means that the frequencies used are distributed in the available transmission band in such a manner that strong selective attenuation or a strong narrow 15 band interference source can never impair two or more frequencies at the same time. To make this choice correctly, adequate background experience is necessary which must be obtained by extensive network investigations (network surveys), field trials and 20 theoretical network modelling based on the measurement results. The frequency band from 3 kHz to 95 kHz, i.e. a bandwidth of B = 92 kHz, is available in the 25 distribution system of a public utility for transmitting information in accordance with the European standard EN 50 065-1. If, for example, it is necessary to transmit binary information with a net data rate (bit rate) rD=1 2 0 0 bits/s by means of the 30 symbol-processing multicarrier method, where N = 4 carrier frequency hops during a symbol period Ts, the frequency hopping rate h = 2400s1. The inverse value T = 1/h of the frequency hopping rate specifies the duration of the time interval during which one of the N 35 frequencies is transmitted in each case. Communication theory says that, within a frequency bandwidth of B = 92 kHz and with a frequency hopping rate h = 2400s~1, a maximum number of FEB/hi = F92000/2400] = 38 signal forms having in each case a frequency offset by 2400 Hz are Mp. No. 97/703 - 10 - 10 December 1997 simultaneously transmitted and can be detected without errors and without mutual interference by receivers which perform correlative signal processing - see also printed document DE-Al-44 23 978. 5 Thus, it would be possible to operate up to F38/41 = 9 modems with symbol-processing multicarrier methods of the type described simultaneously and without mutual interference on an electrical power distribution 10 system, for example between a transformer station and the connected households. The decisive prerequisite for this is not only a highly accurate signal generation but also the accurate insertion of all transmit signals into a global timing pattern and perfect correlative 15 signal processing in the receiver section of each modem. Accurate generation of a transmit signal is not only required with dense channel allocation but also of 20 advantage in the case of a single point-to-point connection. It is also important that all signals are derived from a fixed and highly constant basic frequency and that, when the frequency is changed, there is no phase discontinuity but a continuous phase 25 transition takes place. The frequency, in contrast, must change abruptly, i.e. without transient process. The continuous phase transition is required in every case in order to be able to meet the strict limit values for out-of-band interference in accordance with 30 the standard EN 50 065-1 with supportable filter complexity. As in virtually all high-quality transmission systems, synchronization is basically required with the signal 35 generation and signal processing in a symbol-processing multicarrier system. Synchronization of the received signal with a reference signal existing locally in the receiver is necessary especially in the case of correlative reception. In the case of the symbol- Mp. No. 97/703 - 11 - 10 December 1997 processing multicarrier method, this reference signal comes from a frequency synthesizer. During the transmission, a synchronized frequency synthesizer is also needed, the information-carrying output signals of 5 which are fed into the power system. In systems operating from electricity supply lines, the problem of synchronization can be solved simply and inexpensively with the aid of the alternating system voltage - see, for example, EP-Bl-200 016, EP-Bl-0 199 148 and 10 EP-Bl-0507 087. In a part of this description which is located below, a correlative synchronization which is independent of the power voltage and which is advantageously obtained directly from the symbol processing multicarrier method according to the 15 invention will be presented. The hardware of known transmitting and receiving devices for transmitting information on electrical distribution systems will not allow the use of a 20 symbol-processing multicarrier method. After the method has been described, a device is therefore specified within the context of the present invention, which device includes all essential parts for the synthesis of the transmit signal and for the correlative 25 processing of the received signal for a symbol processing multicarrier method. The new device contains analog and digital sections and, apart from a few components, can be monolithically integrated, standard CMOS technology being adequate for this. A maximum 30 degree of flexibility is achieved with respect to the choice of frequency (position of the frequencies used within the allowable transmission band). The transmitting device in a modem with symbol 35 processing multicarrier method for transmitting digital information via electricity supply lines has the task of accurately generating a multiplicity of signals having relatively closely adjacent frequencies with accurate timing. In this connection, a rapid phase- Mp. No. 97/703 - 12 - 10 December 1997 continuous frequency change must be possible in dependence on the information to be transmitted without transience occurring during the process. Devices constructed with analog technology cannot handle the 5 above requirements and, in addition, are inflexible with respect to frequency changes. The receiving device of a modem must be capable of perfectly isolating signals having adjacent frequencies 10 of unknown phase relationship. For this purpose, matched filtering by correlative means is necessary in a number of branches of the receiver which operate in parallel. The principle of correlation is adequately well known from standard textbooks. An active 15 correlator which generally consists of a multiplier and an integrating unit becomes a matched filter, e.g. for signal forms of chip period Tc, if synchronization is used to ensure that the integrating unit is set to zero at each chip interval end after the value integrated up 20 during the chip period has been sampled for further processing and, if necessary, stored. In analog technology, the complexity of such a circuit would be intolerably high because a separate phase-insensitive quadrature receiver would be required for each carrier 25 frequency - compare also printed document DE-Al-44 23 978. Thus, producing a receiver based on a symbol processing multicarrier method with N = 4 frequencies requires 8 separate correlators. Such structures have hitherto not been achieved which is why no hardware 30 components for implementing multicarrier methods which are resistant to interference to a higher degree are available on the market although there is no dearth of applications for these. 35 The invention provides for the first time a hardware basis for the simple and reproducible manufacture of modems for symbol-processing multicarrier methods and industrial series productions will-make it possible to achieve a continuous drop in costs as the numbers Mp. No. 97/703 - 13 - 10 December 1997 produced increase. The invention is thus suitable for assisting in achieving a breakthrough in the widespread use of interference-resistant data transmission on electricity systems, especially in Europe, where the 5 strict restrictions of EN 50 065-1 are in force. Hitherto, no modems which operate reliably, either with conventional modulation methods or with multicarrier technology for data transmission on electrical distribution systems have been available on the 10 European market. An embodiment of the invention will now be described with reference to Drawing Figures 1 and 2. For reasons of clarity and illustration, the transmission of binary 15 information (i.e. a stream of random successive "H" and "L" bits) having a fixed data rate rD = 1/TB = 1200 bit/s using 4 orthogonal signal forms of different frequency with the chip period Te = TB/2 will be considered. This is therefore a symbol-processing 20 multicarrier system where N 4. The change to other technically meaningful values of N can be carried out by the expert by referring to the text. Incoherent reception is the rule in electrical 25 distribution systems, which is why eight correlators operating in parallel are necessary in the receiver for N = 4. If Tc = T 3 /2, a chipping rate h - 2400s-1 is obtained. In the 9.6 kHz ... 148.8 kHz frequency band, for example, 60 frequencies can be found which, in each 30 case spaced apart by 2400 Hz, form a set of orthogonal frequencies where an integral number of periods in each case fits into each chip period Tc. With a sampling rate of 600 kHz, a maximum of 125 samples are needed for error-free representation. Because of the incoherent 35 reception, 8 reference values are needed for each sample of the received signal, so that a clock frequency of fm = 4.8 MHz is required for outputting the reference signal samples.

Mp. No. 97/703 - 14 - 10 December 1997 In the case of the symbol-processing multicarrier method, an optimum with respect to resistance interference and, at the same time, reliability of manipulation too, can be achieved when the frequencies, 5 over which the user information is distributed, are as far apart as possible. It is then improbable for a number of frequencies to be subject to the same attenuation and/or interference at the same time. For a system with N = 4, for example, the following choice of 10 frequencies is appropriate: fi f2 f 3 f 4 52,800 Hz 62,400 Hz 72,000 Hz 86,400 Hz Table 3: Example of a frequency determination for N = 4 15 Figure 1 shows the overall configuration of a modem, the operation of which will be explained in detail in the text which follows, beginning with the receiving branch. From the power system 1, the received signal 20 passes via a coupler 2 to a band-pass filter 3 which passes the range of the four desired frequencies fi ... f 4 but blocks the remaining frequency range as well as possible. Isolating the mains voltage and other low frequencies is already largely handled by the 25 coupler, which acts as high-pass filter. This is followed by a device for automatic gain control consisting of three operational amplifier stages 4, 5 and 6, the gain of which can be set digitally via an integrated microcontroller system 15. The amplified 30 received signal passes to an analog/digital converter 7 which supplies the digitized samples to the symbol processing system 8, the results of which are accepted and processed further by the integrated microcontroller system 15. Finally, the microcontroller system 15 3 supplies, on the one hand, the data transmitted and Mp. No. 97/703 - 15 - 10 December 1997 received here via a serial interface and, on the other hand, calculates the gain to be set for the three operational amplifier stages 4, 5 and 6. A special feature of this is that two of the three amplifiers are 5 "rapidly adjustable", whereas the third one is adjusted at the most by a factor of 2 from symbol to symbol. The rapidly adjustable amplifier has the task of rapidly responding to abrupt, considerable channel changes whereas the slowly adjustable amplifier is only 10 intended to compensate for slight fluctuations. The control algorithm, which is advantageously implemented in the form of software in the microcontroller system 15, always attempts to set the fast amplifiers to the highest possible values of gain so that a large 15 response margin is available with rapid channel changes. A level estimator which supplies the basis for calculating the amplifier adjustments to the microcontroller system 15 is constructed as averaging circuit in digital hardware and is a component of the 20 symbol-processing system 8. The level estimator receives Z digitized values x(k) of the received signal and uses these to determine the estimated value z 25 X, = E-x(k), 2 L., in each case after a symbol period has elapsed. The ratio between the estimated value XsCh and a nominal 30 value Xo 1 1 supplies the basis for generating the switching commands for the gain adjustment by the microcontroller system 15. The maximum total gain is variable from 1... 4096, each of the three amplifier stages being adjustable to the gain values 1, 2, 4, 8 35 or 16. Another functional unit within the symbol processing system 8 which will be described in detail with reference to Figure 2 in the next section performs the Mp. No. 97/703 - 16 - 10 December 1997 processing of the transmit signal, a digital data stream to be transmitted, which reaches the microcontroller system from a data source, being processed in such a manner that the samples of the 5 signal forms to be transmitted are supplied directly to a digital/analog converter 12. The digital/analog converted transmit signal is filtered in a low-pass recovery filter 11 and, after amplification in a transmission output stage 10, supplied to the power 10 system 1 via the transmit coupler 9. Function blocks 13 and 14 are used for synchronizing the data transmission with the mains alternating voltage. In this process, a highly accurate detection of mains zero transition is carried out in function block 13 and, at the same time, 15 the mains zero transition information is DC-isolated from the power system with the aid of an optocoupler. In block 14, a digital phase-locked loop (PLL) is implemented, with the aid of which high-frequency jitter of any type is removed from the mains zero 20 transmission information so that, finally, a steep edged and stable synchronization signal is supplied to an interrupt input of the microcontroller system 15. The synchronization with the mains AC is not a general optimum solution, which is why an advantageous 25 development of the invention will be described below which implements a correlative type of synchronization and thus achieves perfect synchronization even without the mains alternating voltage. 30 The block diagram shown in Fig. 2 shows an integrated microcomputer system which includes function blocks 7, 8, 12 and 15 from Fig. 1. In the real construction of a complete symbol-processing multicarrier system, it is advantageous to integrate all function units from 35 Figure 1 monolithically in the form of a mixed-signal ASIC with the exception of couplers 2, 9 and of the transmit amplifier 10. At this point, only the section shown in Figure 2 will be considered because it is better suited to explain the essential functions Mp. No. 97/703 - 17 - 10 December 1997 according to the invention than an extensive and thus relatively confusing overall circuit. In Figure 2, a filtered and amplified received signal 5 isolated from the power system first reaches an analog/digital converter 21. Since incoherent reception is a rule in electrical distribution systems as explained above, N = 4 requires eight correlators operating in parallel in the receiver. Because of the 10 incoherent reception, eight reference values are needed per sample of the received signal. With a sampling rate of 600 kHz, a maximum of 125 samples per frequency are needed so that, with a clock frequency of f. = 4.8 MHz, a memory for 2000 samples is required for outputting 15 the reference signal samples. In the text which follows, the digitized received signal is designated by E(iTa), where iTa is the discrete time, with i = 0, 1, 2.... Each of the 20 digitized samples E(iTa) of the received signal is multiplied by eight reference signal samples R(iT 5 +UTa/8), with u 0 ...7, from a signal form or sample memory 214 in a digital multiplier 22. The eight subproducts are then integrated via the signal form 25 period Tc (chip period), i.e. added together digitally in separate accumulators. According to Figure 2, these accumulators are advantageously implemented as follows: an adder 23 can be connected together with eight regis ters 24.. .29, 210, 211 following it, and a switch 212, 30 to form a ring structure. At the beginning of the reception of a signal form, the switch 212 is in position II for the period of eight clock cycles of frequency fm, i.e. the ring structure is opened so that zeros pass to one input of the adder 23 whilst the 35 other input successively receives eight multiplication results E (iTa) -R (iTa+UTa/8) , with U = 0 ...7, from the multiplier 22. The following correlation applies: R(ia+O-Ta/8) / Sine sample fi chip (in-phase component for fi chip) Mp. No. 97/703 - 18 - 10 December 1997 R(iTa+1-Ta/8) / Sine sample f 2 chip (in-phase component for f2 chip) R(iTa+2-Ta/8) / Sine sample f 3 chip (in-phase component for f 3 chip) R(iTa+3-Ta/8) / Sine sample f 4 chip (in-phase component for fa chip) 5 R(iTa+4-T /8) / Cosine sample f, chip (quadrature component for f, chip) R(iTa+5-T,/8) / Cosine sample f 2 chip (quadrature component for f 2 chip) R(iTa+6-T/8) / Cosine sample f 3 chip (quadrature component for f 3 10 chip) R(iTa+7-Ta/8) / Cosine sample f 4 chip (quadrature component for f 4 chip) After 8 clock cycles of frequency fm have passed, 15 registers 24.. .29, 210, 211 contain the following results: Register -> 24 25 26 27 28 29 210 211 Q4Q3 0

Q

1 14 ill d 20 where Ij is the in-phase component and Qj is the quadrature component of the chip having frequency fj, where j=l. .. 4. Because i=0, the following applies at the beginning of 25 a signal form: Mp. No. 97/703 - 19 - 10 December 1997

Q

4 (0) = E(O).R(0+7.T,/8) + 0,

Q

3 (0) = E(O)-R(0+6-Ts/8) + 0,

Q

2 (0) = E(O)-R(O+5.TJ8) + 0,

Q

1 (0) = E(0).R(0+4.TJs) + Qi 14(0) = E(0)-R(0+3-T/8) + 0, I(0) = E(0).R(0+2.TJ8) + 0, 12(0) = E(O)-R(0+i.TJ8) + 0, I1(O) = E(0).R(O+O.T/8) + Q 0 The addition of 0 is obtained because the ring structure is opened by the switch 212. After the above 15 mentioned eight clock cycles of frequency fm have passed, switch 212 is switched to position I, resulting in the ring structure described, in which the contents of register 211 now reach the adder 23. Now arithmetic operations are carried out which accumulate the 20 contents of registers 24.. .29, 210, 211 as follows: Register 24 Q 4 (i)= E(iT.).R iT+ 8 + E E(. R(T. +7 Register25: Qa(i)= E(iT.R iT. +! + EE(T.).R (T.+6 Register 26: Qz(i) = E(iT)-R iT + + )RE(T)R4Ta 8)+ Register 27: Qz(i) = E(iT)- R iT. + + E(4T.)- R(T. + 4. Register28: 14(i)=E(iT)-R iT.+?i + E({T.R (T.+-''Ki Register 29: Is(i) = E(iT. R iT. + + E E({T.)- R (T. + 2. Register 210: Q 2 (i)= E(iT.). R(iTa + ) + E(Ta) R(T +. T. 3T go 8o Register 28: I(i)= E(iT.). R(iT. + 8 )+ Z E(T.). R(T. + 4.0 Mp. No. 97/703 - 20 - 10 December 1997 Assuming that a signal form has the chip period Tc=NATa, the desired eight signal components are obtained in registers 24.. .29, 210, 211 of Figure 2 after i-N clock 5 cycles of frequency fa, i.e. after 8AN clock cycles of frequency fm: 24 25 26 27 28 29 210 211

Q

4 Q 3 Q 2 QI113 2 I For the symbol decision now following, the geometric 10 addition of signal components Ij, Q according to the principle of the quadrature receiver is first required: B 1 =92 2 + + B2 2 +Qi B =12+Q2; B4= 4 +4, resulting in the amounts B 1 .. .B 4 . 15 Although the arithmetic operations of squaring and calculating the square route can be achieved without problems by means of digital hardware, the complexity is considerable, especially with a large dynamic range. 20 In the present invention, therefore, the following approximation is advantageously used: B a max(I ,Q+(I+ min(I,Q,), 25 which supplies equivalent results with much less complexity. In this arrangement, the larger correlation value Ij or Qj of a signal component by amount must be Mp. No. 97/703 - 21 - 10 December 1997 added in each case to the smaller one multiplied by (1/4+1/8) . The simple arithmetic and logic operations necessary for this are implemented in a function block 215, a second switch 216 and a set of absolute-value 5 registers 217. After a chip period has elapsed, switch 212 is changed to position II for exactly eight clock cycles so that the eight correlation values pass into block 215. The four first ones of these (Ii.. .14) are directly shifted on into the set of absolute-value 10 registers 217. Switch 212 is now back in position I so that the correlation values of the next signal form can accumulate in registers 24 ...29, 210, 211. Now the second switch 216 is closed for the further calculation of absolute values Bj in accordance with the above 15 arithmetic rule so that the four values I1... 14 are successively passed back into function block 215 where each of them is compared with the associated quadrature value Q1 ...Q4 still stored in block 215. The larger value in each case is determined and shifted into the 20 set of absolute-value registers 217. The smaller correlation values in each case are now divided by four or, respectively, by eight by being shifted to the right by two or, respectively, three bit 25 positions in function block 215 and the results are added. For the final addition according to the above arithmetic rule, the maximum values stored in the set of absolute-value registers 217 are fed back via the second switch 216 to function block 215 for the purpose 30 of being added to the minimum values scaled in each case by (1/4+1/8), and the results are shifted into the set of absolute-value registers 217 where the four desired absolute values Bj are now available, by means of which the symbol decision can be carried out next. 35 According to Table 2, a symbol decision can take place in each case after four chip intervals have elapsed. For this purpose, it is necessary *to add the absolute values B 1 .. .B 4 from the register set 217 in four Mp. No. 97/703 - 22 - 10 December 1997 successive chip intervals in accordance with the following arrangement and to store the results in the four symbol registers 219.. .222. 5 Chip interval No. 1 2 3 4 Symbol register Data symbol B, + 62 + B3 + B = 222 00

B

2 + B3 + B4 + B 1 = 221 01 B3 + B4 + B1 + 82 -> 220 10 B4 + B, + B2 + B 3 Z :: 219 11 Table 4: Calculation of symbol values The operations needed are carried out with the aid of a multiplexer 225 and an adder 218 interacting with the 10 set of absolute-value registers 217 and the symbol registers 219.. .222. The symbol registers 219.. .222 are assumed to be filled with zeros at the end of the first chip interval. The multiplexer 225 now supplies the four absolute values B1... B4 according to column 1 of 15 table 4 one by one to adder 218, the second input of which, in the meantime, receives zeros. After the second chip interval, the absolute values come from the multiplexer cyclically exchanged according to column 2 of table 4, i.e. B 2 , B 3 , B 4 , B 1 so that, because of the 20 feedback from the symbol registers to the adder 218, the symbol registers now contain the sub-totals according to the first two columns of Table 4 after four addition steps. Continuation of the above considerations supplies the symbol results in registers 25 219.. .222 after the two remaining chip intervals have elapsed. It is now only necessary to determine the maximum of the four results in a decision device 226 in order to determine the associated data bit combination which can then be supplied as received data ED to the 30 data sink. From the results of the. symbol calculation, however, even more information can be obtained which Mp. No. 97/703 - 23 - 10 December 1997 can be utilized advantageously -for a further development of the invention, namely for the autonomous, mains-AC-independent correlative synchronization which will be described in detail in 5 the next section. Before that, the remaining blocks from Figure 2 must still be explained which contain the essential transmit functions. The heart of the transmit hardware is the signal form 10 memory 214 which contains the samples of the transmit signal forms which can largely also be used advantageously as reference signal samples in receiving operation. The data to be transmitted, the transmit data SD, first pass to a function block 213 where they 15 are combined to form frequency sequences in accordance with the symbol formation according to Table 2. In the order predetermined according to Table 2, function block 213 successively addresses and reads out the address areas of the signal form memory 214 in which 20 the samples belonging to the desired frequencies are stored. With a sampling clock frequency of 600 kHz, the address areas would in each case have a maximum length of 125. Overall, a maximum of 500 samples would then have to be stored. To keep down the complexity for the 25 low-pass recovery filter 224 following a digital/ analog converter 223, it is advantageous to increase the clock frequency to 1.2 MHz or 2.4 MHz during the synthesis of the transmit signal SS. This proportionally increases the required memory. Another 30 important detail in designing the transmit side is that there will be no abrupt phase changes at all both during the frequency change within a symbol and at the symbol boundaries, i.e. that phase continuity is ensured. Otherwise, it would not be possible to adhere 35 to the strict regulations with respect to out-of-band interference power according to EN 50 065-1. The description of an exemplary embodiment is now fol lowed in the concluding part of the representation of Mp. No. 97/703 - 24 - 10 December 1997 the invention by a special type of autonomous synchronization which becomes possible as a by-product of the special design of the receiver architecture. 5 In many cases, it is advantageous if the mains AC can be dispensed with as synchronization reference. In some countries of Eastern Europe, where the stability of the mains frequency is much worse than in the interconnected European power system, it is no longer 10 possible to achieve a sufficiently accurate synchronization for chipping rates of approx. 1000 s-1. Or, if an operative communication is required even in the event of a power failure, autonomous receiver synchronization must be possible from the received 15 signal. In general, it is advantageous always to carry out coarse synchronization with the aid of the mains AC and then to compensate for synchronization errors continuously on the basis of the correlative method presented in the text which follows so that, in the 20 final effect, a synchronization is achieved which is so accurate that it would not be possible on the basis of the mains AC alone even in the interconnected European power system, even if it is possible to remove the jitter to a large extent from the synchronization 25 reference, for example with the aid of a digital phase locked loop. In known correlative receivers, autonomous synchroniz ation from the received signal requires, as a rule, 30 more effort than the performance of the correlation itself. In the case of the invention, in contrast, this synchronization can be managed without significant additional effort, using the receiver hardware which is necessary in any case. Because all transmitters and 35 receivers use oscillator crystals for generating the clock as a matter of principle, a sufficiently stable time base is available from the start. It is then ultimately the task of the synchronization device to get the timing pattern existing in the received signal Mp. No. 97/703 - 25 - 10 December 1997 to match that generated locally in the receiver exactly. This complex process will now be explained with reference to Figure 3, which is based on Table 2, and with reference to Table 5. 5 Figure 3 shows the four symbols carrying information, and the associated data bit combinations. Starting with bit combination '00', it can be seen that the symbol belonging to '10' is produced by cyclic right shifting 10 and that belonging to '01' by left shifting. With respect to synchronization, right shifting means that the timing pattern of the local reference in the receiver leads with respect to the received signal, i.e. correction would require a delay. This analogously 15 applies to the left shifting process. If '01' had been selected as the starting point, cyclic right shifting would produce '00' and left shifting would produce '11'. Applying the above consideration to 20 each symbol, it becomes clear that it is possible, by considering the largest and second-largest correlation value alone, to determine whether the receiver timing pattern is leading or lagging. The difference of the absolute values also supplies an indication of the 25 magnitude of the synchronization error, i.e., the greater it is, the more accurate is the synchronization and the less corrective action should be taken. According to the above considerations, the synchronization can be corrected in a simple manner 30 (the starting point is assumed to be the reception of the symbol belonging to '00'): If, for example, a lead with a relatively large difference between the largest ('00') and second-largest ('10') correlation value has been detected, the local reference in the receiver is 35 slightly delayed which can be done digitally quite simply by inserting additional clock pulses. After the subsequent correlation process, the largest and second largest correlation value are examined again. It is now assumed that the symbol belonging to bit combination Mp. No. 97/703 - 26 - 10 December 1997 '11' has been transmitted; for this reason, the largest correlation value is received for this symbol. After the preceding correction of the lead, it is now expected that the second-largest correlation value for 5 '01' will occur and that the difference to the largest has increased, i.e. synchronization has improved. The correction steps described are continued, if necessary with continuous reduction in size, until a lag of the receiver reference is detected instead of a lead. A 10 stable optimum state is reached when, with a large difference between the largest and second-largest correlation value, after each correlation process, i.e. in each case after one symbol period has elapsed, alternating leading and lagging of the receiver 15 reference is detected. In practice, this ideal condition will probably occur quite infrequently since the difference to be examined is subject to fluctuations, even with ideal synchronization, due to interference signals which are always present, which 20 fluctuations cannot be distinguished from synchronization errors. This means that, as a rule, greater corrections (more shifting steps) than would be expected theoretically will occur for reaching the stable state described. 25 Table 5 below illustrates the relationships described with respect to the timing patterns, transmission of the symbol belonging to bit combination '00' being assumed - see top row (transmitted).

IQ,.

Mp. No. 97/703 - 27 - 10 December 1997 transmitted pc f . th n 01 1 I Thsle to smal a pci e vl eaing Atth rcivr n, cnsdralelad of the produced for the transmitted symbol and, on the other hand, contributions being produced in the case of symbols which have not been transmitted. -However, in 10 the synchraion method according to the invention this is the case with only a single symbol, namely the one which is allocated to bit combination the on the case of the other two symbols, no correlation contributions are produced. The vertical dashed 15 rectangles in Table 5 comprise the contributions correlated together at the wrong point due to the timing pattern offset for each of the four chips of a symbol. It can be seen that contributions can only occur in the symbol belonging to '10' - there is no 20 match of frequencies in the case of the other ones '01' and '11'.

Claims (8)

1. Multicarrier method for transmitting digital data, especially via electrical power distribution sys 5 tems, wherein, for the purpose of transmitting information by means of fast continuous-phase fre quency changes and by means of symbol processing, a) from a data stream to be transmitted, combina 10 tions of Id(N) data bits are allocated to a number N of symbols, each symbol of which is composed of N signal forms of different fre quencies, 15 b) at the receiving end, a number 2N of digital matched filters operating in parallel is used for the simultaneous incoherent optimum reception of all N signal forms in each case used for the representation of a symbol, a 20 digital numerical value corresponding to the respective signal form energy being formed as correlation value for each of the N signal forms, 25 c) weighting and scaling of the N correlation values is performed by comparison with a freely selectable threshold, and d) the N correlation values forming a symbol are 30 added together to form the respective symbol value.

2. Method according to Claim 1, characterized in that the threshold used for weighting and scaling is 35 selected as a percentage of the maximum possible autocorrelation value of a signal form and is adaptively set in dependence on received interfe rence and/or useful signal energy. Mp. No. 97/703 - 29 - 10 December 1997

3. Method according to Claim 1 or 2, characterized in that, for the purpose of correlative synchroniz ation, in each case the largest and the second largest symbol value are determined from which 5 both the direction of a timing error of the synchronization (i.e. lead or lag) and the magnitude of this error are determined.

4. Method according to Claim 3, characterized in 10 that, from the direction and magnitude of a phase error, a control information item for a synchronization device is generated which adaptively counteracts the error determined. 15

5. Device for carrying out the method according to one of the preceding claims, characterized in that a) for transmitting and for the incoherent parallel optimum reception of N signal forms of 20 different frequency, a mixed analog/digital integrated circuit (mixed-signal ASIC) exists by means of which the amplified and filtered received signal coupled out of a power system (1), after analog/digital conversion, passes 25 through a digital multiplier (22), the second input of which receives a digital reference signal from a sample memory (214), wherein 2N reference values corresponding to the in-phase and quadrature samples of the N signal forms 30 used are allocated to each received-signal sample, b) a ring structure of an adder (23), a set of 2N registers (24 to 29, 210, 211) and a switch 35 (212) exists by means of which the products produced in the multiplier (22) are accumulated so that, after a signal form period has elapsed, the in-phase components and the quadrature components of N signal forms used Mp. No. 97/703 - 30 - 10 December 1997 for a symbol are accumulated in the 2N registers, c) the switch (212) is also used for subsequently 5 supplying the 2N results from the registers, with opening the ring structure of adder and set of registers, to a geometric addition circuit which forms a digital numerical value corresponding to the absolute value of the 10 respective signal form energy by means of an approximation, d) a further ring structure of an adder (218), a set of N registers (219 to 222) and a 15 multiplexer (225) exists by means of which symbol values are accumulated so that, after a symbol period has elapsed, the respective symbol values which are subsequently supplied to an analysis and decision device (226), which 20 finally supplies the received data, are present in the N registers, and e) an integrated clock generation and control unit (8) exists which, apart from the synthesis of 25 the reference signals in the case of reception, addresses the sample memory (214) during trans mission in such a manner that its output signals form continuous-phase transmit signal forms which, after analog/digital conversion, 30 low-pass filtering and amplification, are coupled into the power system (1).

6. Device according to Claim 5, characterized in that 35 a) for tasks of system control, signal analysis and data transfer, a microprocessor, a microcontroller or a digital circuit (15) of comparable function exists which is a monolithically integrated component of the Mp. No. 97/703 - 31 - 10 December 1997 mixed-signal ASIC which contains the essential components of a modem, b) by means of this digital circuit (15), during 5 the transmitting process, the data bit combination to be transmited can be supplied to the transmitting device and the former first sets the desired data rate and the number N of signal forms of different frequency per symbol 10 by programming the control unit in the mixed signal ASIC.

7. Device according to Claim 5 or 6, characterized in that means (8, 13, 14, 15) exist which have the 15 effect that a) for synchronization of the signal synthesis for transmitting and receiving and for the synchronization of the received-signal 20 processing, the mains AC is used as a global reference signal, the beginning of a data transmission always being coupled to a zero transmission of the mains voltage and a basic timing pattern of 1/6 mains periods being 25 obtained at the 3-phase power system, from which the time bases for various chipping rates such as, for example, 300 s-I, 600 s1, 1200 s-1, 2400 s-1 or 4800 s~1 can be obtained in the European 50 Hz system by integral division, and 30 wherein b) for suppressing jitter during the zero transition detection, a digital phase-locked loop consisting of a digital correlator used 35 for phase detection and a non-linear digital regulator is used so that a very stable synchronization signal freed of short-term fluctuations of any type is-available. Mp. No. 97/703 - 32 - 10 December 1997

8. Device according to one of Claims 5 to 7, charac terized in that a) the required functional groups, with the excep 5 tion of the mains coupler (2, 9), of the zero transmission detector (13) and of the trans mission power amplifier (10) are monolithically integrated in a mixed analog/digital applica tion-specific integrated circuit (ASIC), the 10 memories preferably being constructed as RAM cells, the content of which is loaded via an external read-only memory, e.g. in the form of a ROM, PROM or EPROM, after a circuit reset, and 15 b) the mixed analog/digital application-specific integrated circuit also contains the program memory of the microprocessor, microcontroller or a digital circuit (15) of comparable 20 function in the form of a ROM or EEPROM.