System for the transmission of data by means of multiple transmission channels with a protective band between each channel

DATA TRANSFER SYSTEM

System for the transmission of data using a large number (N) of transmission channels formed by spaced carrier frequencies on a line (10) between a central point (1) and a large number of subscribers (2, 3, connected to the line (10) at different junction points) ..., 8, 9), with a transmitting and receiving part (11) arranged at the central point (1) and transmitting and receiving parts (18 , 19), the transmission parts each having a multiplex unit (50) for dividing the data stream to be transmitted into data blocks, coding units (21, 22, 23, 24), an inverse Fourier transformation filter bank (30, 31, 32, 33) formed from several identical prototype filters ( 20) and the receiving parts each comprise a Fourier transform filter bank (11) formed from a plurality of identical prototype filters (55, 56, 57, 58), decoder units (41, 42, 43, 44) and a demultiplexer unit (50), and wherein the processing of the data stream to be transmitted within the inverse Fourier transform filter bank and the Fourier transform filter bank takes place with a block length M, and wherein the block length M is greater than the number of transmission channels N.

Fl6.1

DVR 0078018

AT 408 595 B

The invention relates to a system for the transmission of data using a plurality (N) of transmission channels formed by spaced carrier frequencies on a line between a central point and a plurality of subscribers connected to the line at different junction points with a transmitting and receiving part arranged at the central point and Send and receive parts assigned to the participants, wherein the transmitting parts each have a multiplex unit for dividing the data stream to be transmitted into data blocks, coding units, an inverse Fourier transform filter bank formed from several identical prototype filters and the receiving parts each having a Fourier transform filter bank formed from several identical prototype filters, decoder units and a demultiplexer unit include, and wherein the processing of the data stream to be sent takes place within the inverse Fourier transform filter bank and the Fourier transform filter bank with a block length (M).

Attempts have been made for some time to enable data transmission between a central point and several connections, which exist at different distances from this, on an already existing line connected to the central point.

A particularly important application for this are telematics systems, which enable the transmission of data via power supply lines. For this purpose, a central point is arranged, for example in the area of ​​a transformer, from which data is sent via the existing mains power line to the power connections connected to it. The power connections can exist in households, commercial operations or any other pantograph. By means of the subscriber devices connected to the power connections, data can be received from the central point and data can be sent to it.

The applicability of the invention is not limited to power line networks but is also given for other existing line networks.

Many of the previously known transmission systems only work with relatively low data rates and are therefore not suitable for modern applications with a high data throughput, such as video on demand, Internet, Voice over IP.

Problems with the use of the power grid as a transmission medium arise mainly from the disturbances occurring on the power lines and the different distances of the participants from the central point as well as in the constantly changing loads that are switched on and off at the individual power connections and are therefore strong Cause impedance changes.

The different transit times between the central point and the individual participants prevent the data output by the individual participants from being sent synchronously to the central point and thus also the use of frequency multiplexing methods, since the Fast Fourier transformation used is based on parallel block processing of the in based on the information transmitted by several carriers formed by channels, which can only take place synchronously.

The object of the invention is therefore to specify a data transmission system of the type mentioned at the outset which enables interference-free transmission even when network conditions change.

Another object of the invention is to specify a data transmission system with which a transmission of data between a central point and differently distant, mutually unsynchronized subscribers is possible by means of frequency division multiplexing.

According to the invention, this is achieved in that the block length M is greater than the number of transmission channels N.

Since the orthogonal position of the individual carriers cannot be maintained due to the different distance of the participants from the central point, increasing the block length increases the blocking attenuation between neighboring transmission channels, so that the interference on neighboring transmission channels is reduced to such an extent that interference-free block processing by Fourier -Transformation can be made.

In a further embodiment of the invention it can be provided that the block length is equal to twice the number of transmission channels.

According to a variant of the invention it can be provided that a filter, preferably a low-pass filter, is provided parallel to the output or the input of each transmitting / receiving part,

AT 408 595 B whereby the constant changes in impedance at the connection points of the line or in the area of ​​the central point can be compensated.

The invention is preferably used for data transmission in power networks, it being possible to use an existing power line of a power network as the line for data transmission.

In a further development of the invention, the prototype filters can be designed as low-pass filters.

An alternative embodiment of the invention can consist in that the prototype filter is designed as a 48-band filter.

In order to be able to make the best possible use of the individual available transmission channels, a communication channel formed from a relatively small number of carrier frequencies can be provided according to a further embodiment of the invention for the cyclical query of the demand for the amount of data to be transmitted from the subscribers to the central point.

The invention also relates to a method for transmitting information using a data transmission system according to the invention.

According to the invention, a corresponding number of carrier frequencies is assigned to each of the participants as a function of the determined transmission requirements of the participants.

This allows a dynamic adaptation of the carrier frequency occupancy for the individual subscribers, whereby, in contrast to a rigid distribution of the carrier frequencies to the subscribers, a better use of the available channels can be achieved.

In a further embodiment of the invention, the data rate of the information to be transmitted can be selected for each carrier frequency as a function of the level of the carrier frequency.

In this way it is possible to divide the entire amount of data to be transmitted into the available carrier frequencies in such a way that the signal-to-noise ratio can be optimized for the transmitted information. Since low carrier frequencies are less susceptible to interference than higher carrier frequencies, it will generally be advantageous to occupy the lower frequency range of the carrier frequencies with a higher data rate than the upper frequency range.

The invention is explained in detail below with reference to the exemplary embodiments shown in the accompanying drawings. It shows

1 shows a schematic representation of an embodiment of the data transmission system according to the invention;

FIG. 2 shows a block diagram of the central point according to FIG. 1;

FIG. 3 shows a block diagram of the transmitting part of the central point according to FIG. 1;

FIG. 4 shows a block diagram of the IFFT filter of the central point according to FIG. 1;

5 shows a schematic representation of the transfer function of a band filter;

6 and 7 each show a schematic representation of the spectral separation of several channels;

FIG. 8 shows a block diagram of a receiving part of the central point according to FIG. 1;

FIG. 9 shows a block diagram of the FFT filter of the central point according to FIG. 1;

10A shows a representation of a possible frequency division;

1 OB shows the transfer function of different prototype filters;

11 shows a schematic representation of a further embodiment of the data transmission system according to the invention;

12 shows a block diagram for calculating an equalizer;

13 shows a diagram of the amount of the transfer functions between the central point and various participants;

14 shows a diagram of the impulse responses that occur between the central point and various participants.

15 shows a diagram of the voltage values ​​to be maintained for the CENELEC A frequency band;

FIGS. 16 and 17 show a diagram of the magnitude frequency response of the filters;

18 to 20 each show a diagram of the signal-to-noise ratio achieved in the Q channels on the central side;

FIG. 21 shows a diagram of the signal-to-noise ratio achieved in the I channels on the central side; FIG.

AT 408 595 B

22 and 23 each show a diagram of the signal-to-noise ratio achieved in the Q channels on the central side;

FIG. 24 shows a diagram of the signal-to-noise ratio achieved in the I channels on the subscriber side;

25 shows a diagram of the signal-to-noise ratio achieved in the Q channels on the central side;

26 shows a diagram of the mean signal-to-noise ratio over all Q channels on the central side;

27 shows a diagram of the signal-to-noise ratio achieved in the I channels on the subscriber side and

28 shows a diagram of the mean signal-to-noise ratio over all I channels on the subscriber side.

1 shows a system for transmitting data between a central point 1 and a large number of subscribers 2, 3, 4 and 5 connected to a line 10, which subscribers are connected to the line 10 via feed lines at different junction points. The data separation takes place by means of frequency division multiplex data traffic with a large number of N transmission channels formed by equidistant carrier frequencies.

In the exemplary embodiment shown, the line is an already existing power line 10 of a power network, which is used for data traffic.

FIG. 2 shows the individual functional units of the data transmission system shown in FIG. The central point 1 contains a transformer 12 which converts a high voltage applied to its primary side into a low voltage, for example 220V. The subscribers 2, 3 shown in FIG. 2 are each connected to the line 10 via a filter 14, 15 and a counter 16, 17. The subscriber's power supply connection is established via the connecting lines, not shown, from the counters 16, 17, while the data traffic with the central point 1 takes place via a respective transmitting / receiving unit 18, 19 which are connected in parallel to the input of the filters 14, 15 . In the same way, a filter 13 is connected between the transformer 12 and the connections of a transmitting / receiving part 11 of the central station 1. Due to load changes or Changes in the type of power consumer causes each participant on the line 10 constant changes in impedance, which have a negative effect on the quality of the data transmission. The filters 13, 14, and 15 are provided to shield these impedance changes as much as possible and thus to ensure that conditions on the line remain as constant as possible. The filters 13, 14 and 15 are therefore designed as low-pass filters which block in the data transmission frequency range.

The data access to the transmitting / receiving unit 11 can take place via a wide variety of interfaces, for example PCM according to ITU G.703, ATM, 10 or 100 Base T or the like.

The transmitting part of the transmitting / receiving part 11 is shown in Figure 3 in the form of a block diagram. The incoming serial data stream is multiplexed in a multiplex unit 25 and divided in such a way that one data block with the block length M is fed into several parallel branches, each branch being assigned to a carrier frequency. In all branches, N identical coding units 21, 22, 23, 24 are arranged, which code the data blocks according to the modulation method used in each case. The coded data blocks are fed to an inverse Fourier transform filter bank 20 and processed therein in blocks.

4 shows a possible processing of the data blocks in detail. The number of carrier frequencies and thus the available channels is N. Within each frequency band, two separate data streams can be transmitted in the in-phase (I) and in the quadrature-phase channel (Q). A real-valued pulse amplitude modulation is transmitted via each of these channels. For this purpose, the data blocks are divided into in-phase data blocks c '[n] and quadrature data blocks c^Q[n] divided.

An interpolation unit 26, 27, 28, 29 and a prototype filter 30, 31, 32, 33 and a mixing unit 34, 35, 36, 37 are provided for each in-phase and quadrature branches. The prototype filters 30, 31, 32, 33 are identical apart from a mutual phase shift.

Each information stream c [n] to be sent is first interpolated in the interpolation units 26, 27, 28, 29 with a factor of 2M and then filtered with a prototype filter g [m] 30, 31, 32, 33. The filtered signals are then passed through in the mixing units 34, 35, 36, 37

Modulation with (cos (~ ^ (k + or sm ^ (k + ^) m) k = 0 ..... N-1, N brought into bandpass position and superimposed in a summation unit 38 on the transmission signal, whereupon this on the Line 10 to

AT 408 595 B is sent to participants 2, 3, 4, 5. The distance between two carrier frequencies is—. A real-valued pulse amplitude N.

Transmit modulation.

The prototype filters 30, 31, 32, 33 are low-pass filters and have a cut-off frequency θ₉ = π / 2Ν. Because the channel spacing is π / Ν, only the stop bands of the individual channels overlap. Due to the different transit times between the subscribers and the central point, the orthogonality of the prototype filters that is customary for a frequency multiplex method cannot be maintained. Instead, the latter must have as high a blocking attenuation as possible, because because the orthogonality of neighboring channels is no longer present, they suddenly show a disruptive effect.

If the block length M and the number of channels N are chosen to be the same, the transmission system works with maximum spectral efficiency. However, this choice has the disadvantage that a root 2N band filter at π / 2Ν only has an attenuation of 3 dB, which circumstance is evident from FIGS. At this frequency, however, the filters 30, 31, 32, 33 should already block, otherwise channel interference will occur.

According to the invention it is therefore provided that the block length M is greater than the number of transmission channels.

With M> N, a filter can be realized which at π / 2Μ has an attenuation of 3 dB and at π / 2Ν has already reached a high blocking attenuation, as can be seen from FIG. The larger M is selected in comparison to N, the easier it is to meet the required blocking attenuation with the prototype filters g [m]. At the same time, however, the spectral efficiency is also reduced.

10B shows the transfer functions A, B of two adjacent prototype filters for M = N and, in contrast to this, the transfer functions A ', B' for M> N. It can be clearly seen that in the first-mentioned case there is an overlap of adjacent channels and interference is possible if the orthogonality is no longer applicable, while in the last-mentioned case there can be no disruptive overlap.

The block length M is preferably selected to be twice the number of carrier frequencies N (M = 2N).

In the exemplary embodiment shown, the in-phase and quadrature channels provide a total of 2N channels, so that a preferred block length of M = 4N results.

Not all of the 2N real channels can be used for a downlink connection from the central point to the individual subscribers because the subscribers also need channels for the upstream connection. Which individual channels are assigned to which participants depends on the respective needs. Basically, however, it is certainly advantageous to assign channels with low frequencies to participants who are further away and to supply the nearby participants with the higher-frequency carriers, because the cable attenuation increases at high frequencies and therefore the cable attenuation occurring for the more distant participants can be approximately compensated . Fig.lOA shows a possible form of the channel division.

For the symmetrical case of the same data rates in the upstream and downstream direction, two variants can be implemented:

1) Both the in-phase and the quadrature channel of a frequency band are only used for upstream or downstream data traffic. The transmission in the other direction takes place in a different frequency band.

2) Within a frequency band, the upstream and downstream data traffic are assigned to the in-phase and quadrature channels.

The signal sent by the subscribers on the line 10 is received by the receiving part of the transmitting / receiving part 11, as shown in FIG. 8 in the form of a block diagram. The received signal is fed to a Fourier transform filter bank 40, subsequently to equalizers 41, 42, 43, 44 and decoders 45, 46, 47 and 48, then reassembled in the correct order in a demultiplexer unit 50 and via an interface to one Subscriber terminal forwarded.

FIG. 9 shows the Fourier transform filter bank 40 according to FIG. 8 in detail with equalizers 41, 42, 43, 44, the received signal being split into 2M lines via mixing units 51, 52, 53, 54 and demodulated separately. The filter bank h [m] 55, 56, 57, 58 consists of identical,

AT 408 595 B frequency-shifted low-pass filters.

It has been found to be advantageous if the combination of g [m] and h [m] results in a 2M band filter.

The calculation of h [m] and g [m] is possible as follows, for example. First, a filter p [m] = {h * g} [m] is calculated. p [m] should be a 2M band filter with a cutoff limit of π / 2Ν. Furthermore p [m] = p [-m] should apply.

The calculation can be done with the own filter method. With the help of the spectral factorization, p [m] can be split into a minimum and maximum phase component. The minimum phase part corresponds to g [m], the maximum phase part corresponds to h [m]. The symmetrical choice of p [m] also means that h [m] = g [-m].

In a practical implementation, however, each branch is not formed by a single filter on the transmitter and receiver side, but rather by a filter bank with several identical, phase-shifted prototype filters.

The received signal is a superposition of the send signals from all participants. Because the individual participants are arranged at different locations at different distances from the central point, the transmission links from the individual participants to the central point have different impulse responses and transit times.

For this reason, it does not make sense to use an orthonormal set for the transmission and reception filters g [m] and h [m], because this is destroyed by the different impulse responses.

After filtering with the prototype h [m] 55, 56, 57, 58, sampling is carried out by means of sampling units 59, 60, 61, 62 with the period 2M / L.

If L = 1 is selected, the following equalizers 41, 42, 43, 44 have the same clock rate as the transmission signals c_k'[n] and c_k^Q[n] k = 0.1, ... N-1.

The equalizers work in a symbol cycle. If L is chosen to be greater than 1, eg 2 or 4, the equalizers work with twice or four times the symbol clock. In this case, the filtered signal must still be sampled with L. An equalizer with a higher clock rate is useful, for example, if the sampling is not carried out with exactly the correct phase.

If the receiver is implemented efficiently, each individual branch is not filtered with h [m], but efficient filter bank structures are used. With these structures, however, there is only access to the signal after scanning with 2M / L. As mentioned above, the transmission signals of the individual participants are distorted by different impulse responses with different transit times. If the individual participants are not synchronized, the individual components of the received signal naturally have different phase positions. Sampling with 2M supplies a signal in the symbol cycle, which can, however, be phase-shifted. For this reason, it makes sense to use an equalizer with a higher clock rate.

For the transmitting and receiving parts of the individual subscribers 2, 3, 4, 5 arranged at different locations, the same applies as for the transmitting and receiving part 11 of the central point 1. Because only a small number of the channels are provided for a specific subscriber are, it only has to process these. It can therefore make sense to implement only one of the branches of the transmitting and receiving part 11 for the individual subscribers.

The calculation of the equalizer is based on the block diagram shown in FIG. 12, the calculation of the equalizer for the in-phase channel of the j-th carrier being shown. The interpolation units 81, 82, 83, 84 and the low-pass filters 85, 86, 87, 88 of the transmitting part and filters 89, 90, 91, 92 corresponding to the impulse responses are provided on the transmission side.

The equalizer calculation for the individual carriers as well as for the I and Q channels can be carried out independently of one another. For the Q channels, the equalizer is calculated in the same way as for the I channels.

The top branch in FIG. 12 represents the flow of the desired signal. The filter g ', [m] is the cosine-modulated prototype filter.

<img file="AT408595B_D0001.tif" />

m-

<img file="AT408595B_D0002.tif" />

AT 408 595 B g^Qj [m] = g [m] sin

<img file="AT408595B_D0003.tif" />

The filter c'j [m] is the impulse response of the in-phase channel from that subscriber to the central point 1 who is transmitting on channel j. The components of the jth Q channel and the I and Q channels of the carriers i = 0, 1 ...... N-1, i ^ j are then added to this desired signal. Each of these individual components is then filtered with a filter c'i [m] or c^Qj [m], which corresponds to the impulse response of the corresponding transmission channel, is folded. With the exception of the first component mentioned, all other components act as noise. Receiving a message from a distant participant can be problematic because this is already much more attenuated than the transmission signal from a nearby participant. This naturally occurs in a different frequency band and is strongly attenuated by the prototype filter h [m]. For this reason, a high blocking attenuation of h [m] is required.

This applies in the event that the equalizer is calculated in the central point 1. If this is to be done on the participant side, the model can also be used except for small modifications.

The filters c'j [m] and c^Qi [m], i = 0, 1 ...... N-1, tej no longer contain the impulse responses of the

Transmission links from the subscriber to the central point but those channel impulse responses that occur between the subscriber who is in band i, kj and the subscriber under consideration who receives in band j. The subscriber who transmits in band i / j sends his data to the control center, but because all subscribers are connected via a transmission medium, this signal in band j naturally acts as a disturbance.

A problem can arise between two neighboring subscribers who are located at a great distance from the control center. The transmission link between the central point and the subscriber can already be strongly attenuated due to the great distance, while the transmission link from subscriber to subscriber has only weak attenuation due to the short distance from one another. In this case, a high blocking attenuation of the prototype filter h (m) 55, 56, 57, 58 is decisive.

The calculation model described above can be used for any allocation of carriers to subscribers. If, for example, the upstream connection is implemented through the in-phase and quadrature channels of one carrier and the downstream connection through the I and Q channels of another carrier, the two filters are c'i [m] and c^Qi [m], i = 1,2 ...... N-1 ident.

However, if upstream and downstream connections are implemented in the in-phase or quadrature channel of a carrier, different filters result.

Viewed from the central point, data is sent to the individual participants on the quadrature channels if the above assumption is made. The effect on the in-phase channels is described by the echo equalization, which is achieved by means of the filter c^Qj [m], i = 1, 2 ..... N-1 is modeled. The filters c'i [m], i = 1,2 ..... N-1 contain the impulse responses of the individual participants

Headquarters.

This overlay is filtered with h'jfm] on the receiver side. This filter is the cosine modulated prototype filter h [m], h^Qj [m], which occurs during the equalization of the Q channels, is the sine-modulated prototype filter.

h'jfm] = h [m] cos

<img file="AT408595B_D0004.tif" />

h% [m] = h [m] sin

<img file="AT408595B_D0005.tif" />

The filtered signal is then sampled with the period - ^ -. In the case L = 1, the following equalizer works with the symbol rate, L&gt; 1 corresponds to an equalizer that operates with a rate higher than the symbol rate.

Let the equalizer f'j [m] have K coefficients which are combined in the column vector ή. The following system of linear equations can then be used for the equalizer coefficients

AT 408 595 B can be written

<td>f / V-1 x</td><td></td>

<td></td><td></td>

<td>_k /=0 '</td><td> &gt;</td>

The quantities σ², and σ² i = 1, 2, ..., N-1, are the mean powers of the symbol currents ^ci Cj c'i [m] and cp [m]. The L_HxL_h&gt; Matrix R_n is the autocorrelation matrix of the noise process n [m]. L._His the length of the prototype filter h [m].

The following applies to the matrices D? · 'And Dp /

[DPf]_kJ = [Dff]_k2MJ with resp.

<img file="AT408595B_D0006.tif" />

<img file="AT408595B_D0007.tif" />

<img file="AT408595B_D0008.tif" />

<img file="AT408595B_D0009.tif" />

<img file="AT408595B_D0010.tif" />

<img file="AT408595B_D0011.tif" />

<img file="AT408595B_D0012.tif" />

<img file="AT408595B_D0013.tif" />

<img file="AT408595B_D0014.tif" />

<img file="AT408595B_D0015.tif" />

2M

<img file="AT408595B_D0016.tif" />

d? / [L_d -1]

<img file="AT408595B_D0017.tif" />

<img file="AT408595B_D0018.tif" />

<img file="AT408595B_D0019.tif" />

d? / [L_d- \\ J

AT 408 595 B d! L [m] and dQI [m] are the impulse responses bj d = {g'i * c'i and

The matrix Hj is defined as follows:

^r H^!j [O]

Ä &lt;[1]

2M

6'[0] ^[1]

<img file="AT408595B_D0020.tif" />

h '[0] h &lt;[l]

<img file="AT408595B_D0021.tif" />

The column vector is sj

The cell vector e_{Previous year} is at the point v_y One, otherwise zero. v, · is the shift of the equalized signal cjfw-vj compared to cj [w].

The exemplary embodiment of the data transmission system according to the invention shown in FIG. 11 is examined below. The central point 1 is with a total of eight participants 2, 3, ...,

8, 9 connected via line 10, from which several branch lines 72, 73 ....... 78, 79 lead to subscribers 2, 3 ...... 8, 9. The continuous line 10 is designed, for example, as an AWG 24 cable and the branch lines 72, 73, ..., 78, 79 are designed, for example, as an AWG 26 cable.

In the exemplary embodiment shown according to FIG. 11, the distances between the individual connection points on the line 10 and the branch lines 72, 73,..., 78, 79 are entered in meters. Both the input and output impedance of all participants is assumed to be 50 Ω. Branch lines that are not terminated represent idle line ends 71 ​​and 78. 13 shows the transmission functions up to 10 MHz occurring between the central point 1 and the participants 2, 3, ..., 8, 9.

Fig. 14 shows the impulse responses between the central point and the participants 2,

3, ..... 8, 9 occur. In contrast to the transfer functions of Fig. 13, in which only the

Line 10 and the branch lines 72, 73 ... 78, 79 themselves have been taken into account, the analog low-pass filters in the transmitting and receiving parts are already taken into account in the illustration in FIG. The cutoff frequency is 95 kHz.

For example, the CENELEC A frequency band is selected for transmission. The voltage values ​​to be observed for this are given in Fig. 15. According to the measuring instructions, these voltage values ​​must be measured at a load resistance of 50 Ω. If an average voltage value of 125 dB pV is assumed, this corresponds to a power density of 18 dBm at 50 Ω.

A total of sixteen carriers are used for data transmission between the eight subscribers 2, 3... 8, 9 of the examined embodiment, so that two carriers are available for each subscriber. This results in a bandwidth of per carrier

AT 408 595 B

95kHz-9kHz = 5.3 &amp; kHz

An additional guard band is taken into account so that the bandwidth per participant is selected to be 5.2 kHz.

The examined exemplary embodiment does not use any additional modulation after the Fourier transformation and superposition of the individual signals and thus works in the baseband.

The lowest two frequency bands occupy the frequency ranges from 0 to 5.2 kHz and from 5.2 kHz to 10.4 kHz and may therefore not be used for data transmission within the CENELEC A band. It must therefore be provided 18 carriers, of which the bottom two are not controlled. In order to facilitate the analog filtering required both in the receiver and in the transmitter, which will not be discussed in greater detail here, two more carriers are provided above the carriers used, which are also not modulated. With this measure, the demands on the slope of the analog low-pass filter are reduced.

The sum of the carrier frequencies is thus twenty, sixteen of which are actually used for data transmission. Two supports each are provided below and above the supports used. The top frequency band occupies a frequency range from 19-5.2 kHz to 20 · 5.2 kHz = 98.8 kHz - 104 kHz. The upper frequency 104 kHz corresponds to half the Nyquist frequency. This results for the sampling period

2.104½ = 4.8 / zs

The sixteen available frequency bands must now be divided between the eight participants. It is advantageous to supply subscribers who are further away from the central point with the lower frequency bands in order to compensate for the increased cable attenuation. The following table shows the allocation of the frequency bands to the participants.

<td>Attendees</td><td>Frequency band [kHz]</td>

<td> 1</td><td> 16,17</td>

<td> 2</td><td> 14,15</td>

<td> 3</td><td> 12,13</td>

<td> 4</td><td> 10,11</td>

<td> 5</td><td> 8,9</td>

<td> 6</td><td> 6,7</td>

<td> 7</td><td> 4,5</td>

<td> 8</td><td> 2,3</td>

Each participant 2, 3 ...... 8, 9 communicates with the central point 1 via two frequency bands. Two variants were examined.

In the first variant, a first complete frequency band with in-phase and quadrature channels is used for the upstream connection and a second complete frequency band for the downstream connection. Thus, with this method, for example, the frequency bands with an even index can be used for the downstream and those with an odd index for the upstream. This method is referred to below as "even / odd".

In the second variant, the in-phase channel of two frequency bands is used for the downstream connection and the quadrature channels are used for the upstream connection. This method is referred to below as "IQ".

Two possible embodiments of prototype filters are examined. The first embodiment is formed by an ordinary low-pass filter, while for the second embodiment a 48-band filter is designed for {g * h} [m], the minimum-phase component g [m] and the maximum-phase component h [m] through spectral factorization. is assigned. As already mentioned above, the selection Μ&gt; N significantly reduces the requirements for the filter.

N - 20 is the number of channels in the system, while M = 48/2 = 24.

AT 408 595 B

Both the ordinary low-pass filter and the 48-band filter can be calculated using the eigenfilter method. The blocking limit is selected in both cases with - = -. The Filter2N 40 length is 511.

Fig. 16 and Fig. 17 show the magnitude frequency responses of the filters g₀[m], gi [m] and g₂[m] for the low-pass filter (Fig. 16) and the 48-band filter (Fig. 17).

18 to 20 show the channel signal-to-noise ratios achieved which occur in the receiving section of the central point with different equalizer lengths. The 48-band filter is used as the prototype filter. The I / Q method is used and the upstream data is thus transmitted in the Q channels. Because the signal-to-noise ratio is observed centrally, the signal-to-noise ratios of the Q channels are entered in FIGS. The central point transmits on the l-channels, the signal-to-noise ratio within the l-channels is therefore irrelevant. The equalizer lengths in Figures 18 to 20 are K = 16, K = 24 and K = 32. Within each diagram of FIGS. 18 to 20, equalizers of the same length with different oversampling factors are compared.

For Figs. 18 to 24, white noise with a noise power density of N is added to the transmission signal_O = -30 dBm added. It can be seen that an increase in the filter length has no significant influence on the signal-to-noise ratios that can be achieved. On the other hand, an oversampled equalizer does lead to significant improvements.

In FIGS. 21 and 22, the signal-to-noise ratio is also observed at the central point, but the even / odd method is used. The 48-band filter is again used as the prototype filter. The equalizer length is K = 24. It can be clearly seen that the in-phase / quadrature method delivers significantly better results.

If the normal low-pass filter is used as the prototype filter, the result shown in FIG. 23 is obtained. The equalizer length is again K = 24. The upstream connection is carried in the Q channels of all frequency bands. As can be clearly seen, significantly better results can be obtained when using a 48-band filter.

FIG. 24 shows the signal-to-noise ratio within the I channels on the subscriber side. The I / Q method is used to separate upstream and downstream, i.e. the downstream is transmitted in the I channels. The 48-band filter is used as a prototype filter. The equalizer length is K = 24.

All of the simulations shown in FIGS. 18 through 24 are N noise power density with white noise_O = -30 dBm. In FIGS. 25 to 28, the achievable signal interference levels are shown as a function of the noise power density N._O shown. The simulation on which FIGS. 25 and 26 are based again uses the 48-band filter and the I / Q method as the prototype filter. The equalizer length is K = 24, the oversampling factor L = 4. Fig. 25 shows the signal-to-noise ratio of the individual Q channels at the central-side point. In the top with N₀= 0 * is valid in addition to N₀= 0 also means that only the channel from which the signal-to-noise ratio is currently being calculated is transmitting. All other channels are not modulated at this point. It is clear that it is not the white noise but the channel interference that is the main source of interference.

Fig. 26 shows the mean signal-to-noise ratio over all channels as a function of N._O represent.

FIGS. 27 and 28 show the results obtained in the same simulation on the participant side. Here, too, it can be clearly seen that the channel interference is the main disturbance.

Another embodiment of the invention, not shown, is that a communication channel formed from a relatively small number of carrier frequencies is provided for cyclical interrogation of the demand for the amount of data to be transmitted from the subscribers to the central point. The transmission capacity incurred for each subscriber in the upstream and downstream direction is continuously determined via this channel and the division of the number of carrier frequencies for each subscriber is determined from this.

A corresponding number of carrier frequencies is thus assigned to each of the participants as a function of the determined transmission requirements of the participants. The data rate of the information to be transmitted for each carrier frequency can also be selected as a function of the level of the carrier frequency. Such a selection is particularly useful for increasing the signal transmission quality. Lower carrier frequencies are generally less susceptible to interference than higher carrier frequencies, which is why it makes sense to occupy the lower frequency range of the carrier frequencies with a higher bit rate and towards the upper frequency range of the carrier frequencies11

AT 408 595 B zen to lower the data rate.