Method for allocating data elements in multicarrier applications

To allocate a number of data elements which constitute a data symbol to a set of carriers used for transmission in multicarrier applications, a full capacity step and capacity fine tuning step are executed successively.

In the full capacity step, the individual capacity or maximum amount of data elements that may be allocated to a carrier is determined for each carrier which forms part of the set of carriers. This maximum amount of data elements is then allocated to each carrier in such a way that a full capacity occupation of the carrier set is obtained.

In case of undercapacity, i.e. in case more data elements have to be allocated to the set of carriers, the capacity of the carrier set is enlarged, for example by power boosting, and additional data elements are allocated to the carriers in accordance to a predetermined rule.

In case of overcapacity on the other hand, data bits previously allocated to the set of carriers, are removed from some carriers selected in accordance with another predetermined rule.

<img file="AU5832696A_D0001.tif" />

<img file="AU5832696A_D0002.tif" />

P/00/011 28/5/91

Regulation 3.2

AUSTRALIA

Patents Act 1990 » 0 0 ί Q a

0 · 0

BO O O Go

0 '* «6 0 0 0 o o 0· οασ ooa ao © oo o o

t&gt;U« Q

ORIGINAL

COMPLETE SPECIFICATION

STANDARD PATENT

99 η e οo

DO

0 4 0

Invention Title:

METHOD FOR ALLOCATING DATA ELEMENTS IN MULTICARRIER APPLICATIONS

The following statement is a foil description of this invention, including the best method of performing it known to us:1

<img file="AU5832696A_D0003.tif" />

<img file="AU5832696A_D0004.tif" />

Ο 0 0 ο ο ο ο ο ο ο ο ο ο ο ο ο ο ο ο

Ο 0 • ο ο ο

<img file="AU5832696A_D0005.tif" />

ο

660 6 &gt; ο υοβο

<img file="AU5832696A_D0006.tif" />

This invention relates to a method for allocating a number of data elements, grouped in a packet of data elements called a data symbol and each of said data elements comprising at least one data bit, to a set of carriers to be modulated thereon and to be transmitted via a telecommunication line, a program module for allocating data elements to a set of carriers, an allocation processing unit to perform this method, and a multicarrier modulator including such an allocation processing unit.

Such a method and equipment to perform this method are already known in the art, e.g. from the US Patent 4,679,227, entitled 'Ensemble modem structure for imperfect transmission media' from the inventor Dirk Hughes-Hartogs. Therein, a modem is described which transmits and receives digital data on a set of carriers called an ensemble of carrier frequencies. The modem includes a system for variably allocating data elements or data, and power to the carrier frequencies to be transmitted via a telephone line. In this modem, data elements are allocated to the ensemble of carrier frequencies in a straightforward way. Indeed, as is described on lines 6-11 of column 3 in the above cited us patent, a power allocation system included in the modem first computes the marginal required power to increase the symbol rate on each carrier from n to n + 1 information units. In addition, the system allocates information units or data elements to the carrier that requires the least additional power to increase its symbol rate by one information unit. In other words, according to a predetermined rule - the carrier that requires the least additional power to increase the symbol rate for the modem described in US Patent 4,679,227 the data elements are thus allocated one by one until all data elements which constitute a data symbol are allocated. The predetermined rule is different in other known implementations of multicarrier modems, described in literature, but the basic concept of allocating data elements one by one to thereby build up the carrier occupations straightforwardly remains unchanged in all known solutions.

An object of the present invention is to provide a method, program module and equipment to perform this method of the above known type but wherein the carrier occupations are no longer build up straightforwardly, this method, program module and equipment providing thus an alternative solution for allocating data

<img file="AU5832696A_D0007.tif" />

<img file="AU5832696A_D0008.tif" />

ο ο ο οο

6 Οβ

Ο

Ο ο ο

Ο Ο 8 0 Ο Ο ο ο ο ο ο ο

<img file="AU5832696A_D0009.tif" />

ο ο ο Ο ο δ ο ο ο ο ο ο ο ο ο ο ο

0 Ο ο ο

9 00 0 ‘,... 25 ο ο

0 09

<img file="AU5832696A_D0010.tif" />

elements in multicarrier applications.

According to a first aspect of the invention, there is provided a method for allocating a number of data elements, grouped in a packet of data elements called a data symbol and each of said data elements comprising at least one data bit, to a set of carriers to be modulated thereon and to be transmitted via a telecommunication line, wherein said method includes a first step, the full capacity step, wherein for each carrier in said set of carriers an individual capacity number is determined, said individual capacity number being equal to a maximum amount of data elements that may be allocated to said carrier, and wherein to each said carrier said individual capacity number of data elements is allocated, and a second step, the capacity fine tuning step, wherein in case of undercapacity, i.e. in case said number of data elements grouped in a said data symbol is larger than an overall capacity number, said overall capacity number being equal to the sum of individual capacity numbers of all said carriers, said overall capacity number is enlarged and additional data elements are allocated to said set of carriers in accordance to a predetermined capacity enlarging rule, and wherein in case of overcapacity, i.e. in case said number of data elements grouped in a said data symbol is smaller than said overall capacity number, some of said data elements are removed from carriers in said set of carriers according to a predetermined data removing rule.

According to a second aspect of the invention, there is provided a program module for an allocation processing unit for allocating a number of data elements, grouped in a packet of data elements called a data symbol and each of said data elements comprising at least one data bit, to a set of carriers to be modulated thereon and to be transmitted via a telecommunication line, said program module containing a set of control instructions, wherein said set of control instructions is structured to control a sequence of operations in said allocation processing unit in such a way that in a first phase, the full capacity phase, for each carrier in said set of carriers an individual capacity number is determined, said individual capacity number being equal to a maximum amount of data elements that may be allocated to said carrier, and to each said carrier said individual capacity number of data elements is allocated, and in a second phase, the capacity fine tuning phase, in case of undercapacity, i.e.

I ΐ

.....____Jk10 ο ο ο » e f&gt;

ο ο ο η ο ο α ο β ο ο οο ο ο ο ο

0 0 ο ο ο οο

ΟΟ » • ·· 20 ο οο

Ο Ο 8

8 0 Ο

<img file="AU5832696A_D0011.tif" />

Ο Ο

<img file="AU5832696A_D0012.tif" />

in case said number of data elements grouped in a said data symbol is larger than an overall capacity number, said overall capacity number being equal to the sum of individual capacity numbers of all said carriers, said overall capacity number is enlarged and additional data elements are allocated to said set of carriers in accordance to a predetermined capacity enlarging rule, and in case of overcapacity, i.e. in case said number of data elements grouped in a said data symbol is smaller than said overall capacity number, some of said data elements are removed from carriers in said set of carriers according to a predetermined data removing rule.

According to a third aspect of the invention, there is provided an allocation processing unit provided to calculate a distribution of a number of data bits which constitute a data symbol over a set of carriers, said allocation processing unit being provided with a first input whereto said number is applied and a second input whereto carrier property information is applied, wherein said allocation processing unit includes a memory means a first part of which being provided to store said carrier property information, a second part of which being provided to store carrier requirement information, and a third part of which being provided to store data allocation information, i.e., the amount of said data bits being assigned to each said carrier in said set, a first comparator means, coupled at its first input to an output of said first part of said memory means and at its second input to an output of said second part of said memory means, said first comparator means being provided to compare said carrier property information with said carrier requirement information, to thereby obtain individual carrier capacities for said carriers, and to apply said individual carrier capacities via an output to a processing unit included in said allocation processing unit and coupled at its output to an input of said third part of said memory means, said processing unit being adapted to apply to said third part of said memory means said data allocation information wherein for each said carrier said amount of data bits allocated thereto equals said individual carrier capacity of said carrier, and that said allocation processing unit further includes a second comparator means, coupled at its first input to an output of said third part of said memory means and at its second input to said first allocation processing unit input, said second comparator means being adapted to compare said number of data bits

I

<img file="AU5832696A_D0013.tif" />

<img file="AU5832696A_D0014.tif" />

<img file="AU5832696A_D0015.tif" />

<img file="AU5832696A_D0016.tif" />

which constitute α data symbol with an overall capacity number of said set of carriers, said overall capacity number being equal to a sum of said individual carrier capacities, and to thereby, in a capacity fine tuning step, activate said processing unit to assign additional data elements to said carriers in accordance with a predetermined capacity enlarging rule in case of undercapacity and to remove data elements from said carriers in accordance with a predetermined data removing rule in case of overcapacity.

According to a fourth aspect of the invention, there is provided a multicarrier modulator for modulation of data elements applied to an input thereof on a set of carriers for transmission thereof in a communication network coupled to an output thereof, said modulator including between said input and said output a cascade connection of a mapping unit, an inverse fast fourier transform processing unit, a cyclic prefix adder, a parallel to serial converter and a digital to analog converter, said mapping unit being provided to allocate said data elements to said set of carriers and to thereby generate a frequency domain parallel sequence of data, said inverse fast fourier transform processing unit being included to inverse fast fourier transform said frequency domain parallel sequence of data applied to its input and to thereby generate a time domain parallel sequence of data, said cyclic prefix adder being provided to add a cyclic prefix to said time domain parallel sequence of data to compensate for intersymbol interference due to transmission over transmission lines in said communication network, said parallel to serial converter being adapted to convert said time domain parallel sequence of data into a serial sequence of data which is applied to said digital to analog converter included to transform said serial sequence of data into an analog signal and to supply said analog signal to said output of said modulator, said mapping unit including an allocation processing unit provided to generate a distribution of a number of data bits which constitute a data symbol over a set of carriers, said allocation processing unit being provided with a first input whereto said number is applied and a second input whereto carrier property information is applied, and a data allocation unit, an input of which is coupled to said modulator input and another input of which is coupled to an output of said allocation processing unit, said data allocation unit provided to allocation

I

I t

j

<img file="AU5832696A_D0017.tif" />

<img file="AU5832696A_D0018.tif" />

<img file="AU5832696A_D0019.tif" />

Ο &lt;* 4 * β

• 4 4 4

<img file="AU5832696A_D0020.tif" />

Ο ©4 t

Ο G4 ©4 &lt;14.

4 4 4 processing unit, to said set of carriers, wherein said allocation processing unit includes a memory means a first part of which being provided to store said carrier property information, a second part of which being provided to store carrier requirement information, and a third part of which being provided to store data allocation information, i.e., the amount of said data bits being assigned to each said carrier in said set, a first comparator means, coupled at its first input to an output of said first part of said memory means and at its second input to an output of said second part of said memory means, said first comparator means being provided to compare said carrier property information with said carrier requirement information, to thereby obtain individual carrier capacities for said carriers, and to apply said individual carrier capacities via an output to a processing unit included in said allocation processing unit and coupled at its output to an input of said third part of said memory means, said processing unit being adapted to apply to said third part of said memory means said data allocation information wherein for each said carrier said amount of data bits allocated thereto equals said individual carrier capacity of said carrier, and that said allocation processing unit further includes a second comparator means, coupled at its first input to an output of said third part of said memory means and at its second input to said first allocation processing unit input, said second comparator means being adapted to compare said number of data bits which constitute a data symbol with an overall capacity number of said set of carriers, said overall capacity number being equal to a sum of said individual carrier capacities, and to thereby, in a capacity fine tuning step, activate said processing unit to assign additional data elements to said carriers in accordance with a predetermined capacity enlarging rule in case of undercapacity and to remove data elements from said carriers in accordance with a predetermined data removing rule in case of overcapacity.

Indeed, in the full capacity step, data elements are allocated to the set of carriers without taking into account the number of data bits that actually has to be allocated. Whether the number of data bits that has to be allocated is small or large, the result of the full capacity step remains unaffected and depends only upon the individual capacities of the carriers in the set of carriers. Since for each carrier only its individual capacity for carrying data elements has to be measured and no further

<img file="AU5832696A_D0021.tif" />

<img file="AU5832696A_D0022.tif" />

<img file="AU5832696A_D0023.tif" />

rules or measurements have to be taken into account, the full capacity step is executed quickly. The total processing time for allocating the dala elements equals the sum of the time spent to the full capacity step and the time spent to the capacity fine tuning step which is executed successively. In the capacity fine tuning step the number of allocated data elements is adapted in such a way that the exact number of data bits is allocated to the set of carriers. When additional data elements have to be allocated, i.e. in case of undercapacity, the capacity of the carrier set is enlarged e.g. by power boosting in accordance to a capacity enlarging rule. Capacity is for example enlarged in such a way that minimal power boost is required to allow allocating the additional data elements. When too many data elements become allocated to the carrier set in the full capacity step on the other hand, i.e. in case of overcapacity, some of the allocated data elements are removed. This removal is also based on a predetermined rule, e.g. to maximize the minimum additional noise margin calculated for the carriers, this additional noise margin for one carrier being equal to SNRi - SNRreq, wherein SNRi represents the signal noise ratio measured on a carrier and wherein SNRreq represents the signal noise ratio required to allow allocating to this carrier an integer number of data elements.

Compared to the known straightforward allocation methods, the present method wherein data elements are allocated blindly in the first step till full capacity is reached and wherein the full capacity allocation is modified in a second step, is an alternative allocation method, whose processing time is smaller whenever the number of data elements to be allocated lays in the neighbourhood of the global capacity of the set of carriers.

The individual capacity of a carrier for allocating data elements thereto, can be defined precisely. In a first implementation of the present method, the signal noise ratio measured on a carrier is compared with the required signal noise ratio values to allocate thereto integer numbers of data elements. In a second implementation, these required signal noise ratio values are enlarged by a fixed margin. Such a margin of 6 dB is for example prescribed in the draft American National Standard for Telecommunications on ADSL (Asymmetric Digital Subscriber Line), published by ANSI (American National Standards Institute) in April 1994, paragraph 12.8.3.3, page 100.

<img file="AU5832696A_D0024.tif" />

<img file="AU5832696A_D0025.tif" />

<img file="AU5832696A_D0026.tif" />

*20 • · a • a · • · · a a a a

<img file="AU5832696A_D0027.tif" />

<img file="AU5832696A_D0028.tif" />

An additional preferred feature of the present invention is that in a particular implementation thereof, the full capacity step is performed as described in claim 3. In this way, the individual capacity of each carrier is obtained by measuring the signal noise ratio on this carrier and defining the number of data elements which, when allocated to this carrier, requires a signal noise ratio lower than but as close as possible to the measured signal noise ratio.

Another preferred feature of the present method is that, also in a particular implementation thereof, undercapacity fine tuning is performed. In this way, the global capacity of the set of carriers is enlarged by applying the minimum power boost necessary to allow allocating all data elements to the carriers. In this particular implementation, the power distribution amongst the carriers is supposed to be flat. Therefore, this minimum power boost is applied to all carriers in the set. It is noted however that such a flat power distribution is not necessary when implementing the present method.

Still a further preferred feature of the present allocation method is that, again in a particular implementation thereof, overcapacity fine tuning is performed as in a way that the noise sensitivity of the carriers is minimized by removing data elements from the most noise sensitive carriers in the set. Indeed, a large additional noise margin, calculated as described above, is equivalent to a low noise sensitivity. Since data elements are removed from carriers with least additional noise margin, the minimum additional noise margins amongst the set of carriers is maximized. This renders the data transmission less sensitive for noise.

Yet another preferred feature of the present method is that pseudoovercapacity, caused by undercapacity fine tuning, is eliminated in a pseudoovercapacity fine tuning step. Such a pseudo-overcapacity is due to the fact that additional data elements are allocated to carriers in the undercapacity fine tuning step. These data elements may comprise different amounts of data bits dependent on the carrier whom they are allocated to. For ADSL (Asymmetric Digital Subscriber Line) applications for example, the already mentioned draft Standard excludes the existence of 1 bit constellations. A first data elements allocated to a carrier thus always contains 2 data bits whilst all further data elements allocated to this carrier comprise only 1

<img file="AU5832696A_D0029.tif" />

<img file="AU5832696A_D0030.tif" />

i

I 15

O 99

0.1 ft

0 0

090 t&gt; - C o oo

009 © J0 o©

009 ⁰o^Doo'20

0 0

34

O 0 o oo 0 3 0

0 9

0-300

9 0

B3 O π ¹ !/ ?

I

Oft

0

0009

000 0 , ³⁰ i

data bit. When the last additional data element allocated in the undercapacity fine tuning step contains 2 data bits, whilst only 1 additional data bit had to be allocated, the pseudo-overcapacity occurs. In this pseudo-overcapacity situation, a smaller data element comprising only 1 data bit can be removed from a carrier which is occupied by at least two data elements. To select amongst all carriers occupied by at least two data elements, the carrier where this data bit is removed from, a sequence of substeps similar to the substeps for overcapacity fine tuning described in claim 5 is executed. This implies that the small data element is removed from the carrier with least additional noise margin to thereby make this carrier less noise sensitive.

The above mentioned and other objects and features of the invention will become more apparent and the invention itself will be best understood by referring to the following description of an embodiment taken in conjunction with the accompanying drawings, wherein:

Figure 1 is a block scheme of an embodiment of a Discrete Multi Tone (DMT) modulator according to the present invention;

Figure 2 is a block scheme of an embodiment of an allocation processing unit according to the present invention;

Figure 3 is a flow chart diagram of a particular implementation of the method and program module according to the present invention, illustrating the steps and substeps therein;

Figure 4 includes a table illustrating the contents of the memory means of the allocation processing unit shown in Figure 2 in case 9 data bits have to be distributed over a set of 4 carriers, and further includes a graph illustrating the distribution of data bits over these carriers;

Figure 5 includes a sequence of tables illustrating the contents of the memory means of the allocation processing unit shown in Figure 2 during successive steps of the algorithm shown in Figure 3, in case 10 data bits have to be distributed over a set of 4 carriers, and further includes a sequence of graphs illustrating the evolution of the data bit distribution over these carriers;

Figure 6 includes a sequence of tables il'ustrating the contents of the memory means of the allocation processing unit shown in Figure 2 during successive steps of

<img file="AU5832696A_D0031.tif" />

<img file="AU5832696A_D0032.tif" />

<img file="AU5832696A_D0033.tif" />

co Ο ο On ο ο ο »20 ο ο« 0 0 0 ο ο ο ο ο ο ο α

Ο 0 6

0 Ο

<img file="AU5832696A_D0034.tif" />

ο 0 . ..0.₂₅ ο ο ο ο ο ο ο ο ο a

<img file="AU5832696A_D0035.tif" />

! « 0

U 0 V *» the algorithm shown in Figure 3, in case 6 data bits have to be distributed over a set of 4 carriers, and further includes a sequence of graphs illustrating the evolution of the data bit distribution over these carriers;

Figure 7 includes a sequence of tables illustrating the contents of the memory means of the allocation processing unit shown in Figure 2 during successive steps of the algorithm shown in Figure 3, in case 11 data bits have to be distributed over a set of 4 carriers, and further includes a sequence of graphs illustrating the evolution of the data bit distribution over these carriers; and

Figure 8 is a representation of a 'required SNR per data element'-table used in the method illustrated by the flow chart of Figure 3 to obtain the data bit allocation in the four examples accompanied by Figure 4, Figure 5, Figure ό and Figure 7 respectively.

Referring to Figure 1 and Figure 2, the structure of a Discrete Multi Tone (DMT) modulator MOD, which is a preferred embodiment of the multicarrier modulator according to the present invention will be described in the first part of the description. The basic means included in a Discrete Multi Tone (DMT) modulator MOD as described in the draft ANSI Standard on ADSL are drawn in Figure 1. These basic means and the functions provided thereby will be described in the first paragraphs of this first part. Since the present invention more specifically relates to the mapping unit and the allocation method executed thereby, some paragraphs in addition will be spent on describing a specific embodiment of an allocation processing unit included in such a mapping unit and equipped to perform the method according to the present invention. A detailed block diagram of this allocation processing unit is drawn in Figure 2 whilst the allocation method executed thereby is illustrated by the flow chart of Figure 3.

In the second part of the description, it will be explained how the means included in the allocation processing unit of Figure 2 are controlled to perform the method illustrated in Figure 3. The working of the allocation processing unit will be described by means of 4 examples, each passing through a different sequence of branches in the flow chart of Figure 3. The contents of the memory means MEM in the allocation processing unit APU' of Figure 3 as well as the distribution of data ι

Ο Ο ΰ Ο β Ο ο ο ο ο

4’ Ο Ο

Ο © Ο © Ο Ο ο η a ο

Ο if ο © ο .ν%20 β 9 0 © 0 0 ο ο ο βοο ο ο ο ο ο © ο ο ο ο ο

0©Ο« ο ο ο © Ο

<img file="AU5832696A_D0036.tif" />

elements for the successive steps in these 4 examples are represented in the tables and graphs of Figure 4, Figure 5, Figure 6 and Figure 7 respectively. After having described the successive steps of these 4 examples, all branches of the flow chart of Figure 3 will be passed through.

The Discrete Multi Tone (DMT) modulator MOD in Figure 1 includes between an input DI, the data input, and output MO, the modulator output, the cascade connection of a mapping unit MAP, an inverse fast fourier transform processing unit IFFT, a cyclic prefix adder CPA, a parallel to serial converter PSC, and a digital to analog converter DAC. The mapping unit MAP in this cascade connection includes a data allocation unit DAU and an allocation processing unit APU. The allocation processing unit APU is provided with a first Ml and second NI input and is further equipped with an output O connected to an input of the data allocation unit DASU. The modulator input DI is coupled to another input of the data allocation unit DAU.

According to the draft Standard on ADSL, the modulator MOD modulates data elements grouped in data symbols on a set of carriers having equidistant frequencies, and further applies the modulated carriers via the output MO to a twisted pair telephone line, not shown in Figure 1. To be distributed over the set of carriers, the data elements entering the modulator MOD via DI, are first applied to the data allocation unit DAU which forms part of the mapper MAP. Based on a particular algorithm, the allocation processing unit APU in this mapper MAP calculates a data element distribution. It therefore is provided with carrier property information, the signal noise ratio measured on each carrier, this information being applied via the input Ml, and with information indicating the total amount of data bits comprised by one data symbol, this information being applied via NI. The results of the calculations are supplied to the data allocation unit DAU via output O of the allocation processing unit APU. Upon receipt of these results, the data allocation unit DAU allocates data elements constituting one data symbol to the carriers and decides for each carrier of the set which modulation method has to be executed. The data allocation unit DAU e.g. allocates 2 bits to the first carrier, these 2 bits being modulated on this first carrier via 4 QAM modulation, allocates 4 bits to the second carrier, these 4 bits being modulated on this second carrier via 1 6 QAM modulation and so on. In a

<img file="AU5832696A_D0037.tif" />

<img file="AU5832696A_D0038.tif" />

<img file="AU5832696A_D0039.tif" />

<img file="AU5832696A_D0040.tif" />

ο ο ο ο ο ο ο ο ο η &lt;5 Ο G οο βο ο ο « : Οα ° ^β © Ο Π ® c *Ζϋ e βώ '0 ©ο I Ο Ο4 | οβοο ώ Ο ©

3« ο a ο ο

Ο 3 ο ο

Ο 3 β

3 £&gt;

<img file="AU5832696A_D0041.tif" />

Ο «« &gt; ο β« so οοοο signal plane, each modulated carrier can be represented by a single point, representing the amplitude and phase of the carrier after modulation. Thus, a set of complex numbers represent the modulated carriers and are therefore outputted parallel at the data allocation unit output as a frequency domain parallel sequence of data. This frequency domain parallel sequence of data is converted into a time domain parallel sequence of data by the inverse fast fourier transform processing unit IFFT. If the transmission line would be perfect, i.e. if no intersymbol interference would be caused by the impulse response of the transmission line, the time domain parallel sequences of successive symbols could have been joined into a serial data stream, transformed into an analogue signal and applied to the transmission line. Due to the effective impulse response length of the transmission line however, intersymbol interference occurs. Such intersymbol interference can be compensated by an adaptive filter at the receivers side. In known solutions and also suggested in paragraph 6.10 of the above cited draft Standard, such a digital filter technique at the receivers side is combined with cyclic prefix extension at the transmitters side to obtain sufficient intersymbol interference compensation. The time domain parallel sequence at the output of the inverse fast fourier transform processing unit IFFT is therefore applied to a cyclic prefix adder CPA which, in accordance with paragraph 6.10 on page 44 of the draft ADSL Standard, prepends the last real numbers of the time domain parallel sequence to this time domain parallel sequence to thereby generate an extended time domain parallel sequence of real numbers. The extended time domain parallel sequence is +hen applied to the cascade of parallel to serial converter PSC and digital to analog converter DAC to be successively transformed into a serial digital sequence and analog signal.

It is noted that an allocation processing unit of the above described type can also be provided in the demodulator at the receiver's side. The signal noise ratio on each carrier is then measured e.g. by transmitting in an initial phase, a predetermined signal from the transmitter to the receiver and by analysing this signal in the receiver. The allocation processing unit thereupon calculates the data element distribution, and applies its results via a backward path in the transmission system to the data allocation unit DAU in the modulator at the transmitter's side. In such a

<img file="AU5832696A_D0042.tif" />

i . |

I t

i

<img file="AU5832696A_D0043.tif" />

v

<img file="AU5832696A_D0044.tif" />

<img file="AU5832696A_D0045.tif" />

O Oft ο &gt;

Ο Ο Ο }

Ο &gt;3

UΟ

Ο οο

ΟΙ

Οft οο .V'20

Ο Oft

Ο θ', ο ft

Ο Ο Ο ο

Ο π 1 ΟΠΟ Ο ο ο ο t ft ο ο ο

Γ »”25 system, no data processing unit has to be provided in the modulator at the transmitter's side.

The block scheme of the Discrete Multi Tone (DMT) modulator MOD in Figure 1 will not be described in further detail since such a detailed description is of no importance for the present invention. Further details with respect to ADSL requirements are described in the already mentioned draft ANSI Standard on ADSL whilst specific implementations of Discrete Multi Tone modulators are found in the articles Ά multicarrier El-HDSL Transceiver System with Coded Modulation', written by Peter S. Chow, Naofal Al-Dhahir, John M. Cioffi and John A.C. Bingham and published in the issue Nr. 3 May/June 1 993 of the Journal of European Transactions on Telecommunications and Related Technologies (ETT), pages 257-266, and 'Performance Evaluation of a Multichannel Transceiver System for ADSL and VHDSL Services' from Peter S. Chow et al., published in the issue Nr. 6 August 1991 of the Journal of European Transactions on Telecommunications and Related Technologies (ETT), pages 909-919.

The present invention more specifically relates to the method performed by the allocation processing unit APU in Figure 1. A particular embodiment of such an allocation processing unit APU' is drawn ion Figure 2.

The allocation processing unit APU¹ of Figure 2 includes memory means MEM, a subtraction unit SUB, a summation unit SUM, a first comparator means Cl, a second comparator means C2, a processor PR and a control unit CT. The memory means MEM is subdivided into a signal noise ratio measurement memory MM, a required signal noise ratio memory MR, a data allocation memory MD, a power boost memory MB and an additional noise margin memory MA.

A first input Ml' of the allocation processing unit APU' is connected to an input of the signal noise ratio measurement memory MM. A first output Ol of this memory is connected to a first input Cl 11 of the first comparator means Cl, whilst a second output 02 thereof is connected to a first input SBI1 of the subtraction unit SUB. Similarly, an output 03 of the required signal noise ratio memory MR is connected to a second input Cl 12 of the first comparator means Cl, whilst another output 04 thereof is connected to a second input SBI2 of the subtraction unit SUB. An output

<img file="AU5832696A_D0046.tif" />

<img file="AU5832696A_D0047.tif" />

ο © ο σ ο ο

Ρ J 4 «

Ο

0 0

Ο ϋ Ο

Ο Ο ο 4 ο ο ο ο ο _oV%20

Ο ϋί

Ο Ο 4

CO 0

Ο 4« 0

Ο 0 4 0 Ο &amp;

U ’&gt; &lt;1 Ο

Ο Ο υο α

Οα

Ο υ νβ η *25

<img file="AU5832696A_D0048.tif" />

ίί β

4

04*0

CIO of the first comparator means Cl is connected to a first input Pll of the processor PR. The subtraction unit SUB on the other hand is provided with two outputs, the first SBO1 of which is connected to an input 13 of the additional noise margin memory MA and a second SBO2 of which is connected to an input 12 of the power boost memory MB. The data allocation memory MD is equipped with an input 11 and two outputs, 05 and 06. A connection is made between this input 11 and an output PO of the processor PR. One of the outputs, 05, is coupled to a first input C2I1 of the second comparator means C2 via the summation unit SUM, and the other output 06 is connected to an output O' of the allocation processing unit APU¹. The summation unit SUM thereto is provided with an input SI and output SO. An output 07 of the power boost memory MB and an output of the additional noise margin memory MA are connected to respective second PI2 and third PI3 inputs of the processor PR. Finally, a second input Nl¹ of the allocation processing unit APU¹ serves as a second input C2I2 of the second comparator means C2.

The processor PR, the first Cl and second C2 comparator means, the memory means MEM, the subtraction unit SUB and the summation unit SUM are all controlled by the controller unit CT. In Figure 2, this is indicated by the double arrows which represent control lines between the control unit CT and the other means in the allocation processing unit APU¹. The control lines themselves are not drawn in this figure since this would make the drawing unclear. However, from Figure 2, it is obvious to a person skilled in the art how such control lines should be provided to obtain the working which will be described in the following paragraphs and which is illustrated by the algorithm flow chart of Figure 3.

The flow chart in Figure 3 is build up with different shaped boxes interconnected via horizontal as well as vertical lines. Via the lines a unique tree of successive steps is obtained. The branches of this tree are walked through from top to bottom and from left side to right side of the chart. The actions which have to be executed successively are represented by rectangle boxes. If the left and right side of such a rectangle box are drawn double, this box represents a procedure or cluster of actions. The actions included in this cluster are all connected to the bottom of the rectangle box with double left and right sides. Each hexagonal box indicates that an

<img file="AU5832696A_D0049.tif" />

<img file="AU5832696A_D0050.tif" />

<img file="AU5832696A_D0051.tif" />

<img file="AU5832696A_D0052.tif" />

»« 0

0 0 o o o

0« ¢0

0 o

O 0

0 0

0o

0 0

O 0

9«

0 00

O 0

0 © ο 0 0 o

9 0 0 ο ff 0 o o o

9 0 0 0 0 ©OOO action or plurality of actions has to be executed repeatedly. The condition to determine how many times the actions have to be executed is described within the hexagonal box, whilst the actions that have to be executed are described in boxes connected to the bottom of the hexagonal box. A diamond shaped box on the other hand indicates that one of two actions has to be executed. If a predetermined condition is fulfilled, the action or branch of actions connected to the diamond side marked by Y is executed. If this predetermined condition is not fulfilled, the action or branch of actions connected to the N marked side of the diamond box is executed. The predetermined condition itself is described within the diamond shaped box.

The method whose steps and substeps are represented in the flow chart of Figure 3 and which is performed by the allocation processing unit APU' of Figure 2 is best explained by considering four fictive situa^J’ons wherein a specific number of data bits have to be distributed over a specific set of carriers. Therefore in the following part of the description, 4 examples will be described one after the other. In these examples, data bits have to be allocated to a set of four carriers, fl, f2, f3 and f4. The properties of these carriers are supposed to remain identical in the 4 described examples, which implies that the signal noise ratio values measured for each of the 4 carriers remain fixed. The minimum required signal noise ratio values allowing to allocate to a carrier 2, 3 or 4 data bits are equal to 1 6dB, 20 dB and 23 dB respectively in the present example. The 'required SNR per data element'-table which includes these figures and which is stored in the required signal noise ratio memory MR of the allocation processing unit APU' in Figure 2, is represented in Figure 8. Since the set of carriers and the properties of the carriers in this set remain fixed for the 4 examples, only the number of data bits which has to be allocated is different. Dependent on this number, the steps of the algorithm which have to be executed will be different too.

In the first example, the allocation is performed by only executing the full capacity step of the present method since no overcapacity or undercapacity appears therein. The contents of the memory MEM of the allocation processing unit APU¹ of Figure 2 for this example is shown in the table of Figure 4. A graph attached to this table in Figure 4 illustrates the division of the data bits over the set of carriers. For the

<img file="AU5832696A_D0053.tif" />

<img file="AU5832696A_D0054.tif" />

i

5·

<img file="AU5832696A_D0055.tif" />

β ο ο 0 0ο ο ο ο ο

00 ο ο ο

Ο 0 0 ο« ο » ο οο οο

Ο 0 4 ο οο

ΟΟ

Ο 0 4 ο οο

4 00 ο 04 ο ο ο

Ο © 4 β

Ο ο ο

<img file="AU5832696A_D0056.tif" />

<img file="AU5832696A_D0057.tif" />

4 » Ο 0 βί : ·

<img file="AU5832696A_D0058.tif" />

second example, tables illustrating the contents of the memory MEM and corresponding graphs are drawn in Figure 5. To obtain allocation according to the present invention, in the second example the full capacity step as well as the capacity fine tuning step for undercapacity are executed. In example 3, overcapacity occurs after full capacity allocation. Therefore the capacity fine tuning step dedicated to eliminate this overcapacity is executed in addition to the full capacity step. Tables and graphs accompanying the explanation of example 3 are drawn in Figure 6. Finally, the last example is illustrated in Figure 7. Full capacity allocation and undercapacity fine tuning in this fourth example cause pseudo-overcapacity. Elimination of this pseudo-overcapacity is done in an additional step, called the pseudo-overcapacity fine tuning step. All steps passed through by the four examples will be described in detail in the following paragraphs.

In example 1,9 data bits have to be allocated to 4 carriers, fl, f2, f3 and f4.

To allow allocating 2 bits to a carrier, a signal noise ratio of at least 1 6 dB has to be measured on this carrier. Similarly, 3 or 4 bits may be allocated to a carrier if at least 20 dB or 23 dB is measured for the signal noise ratio respectively. In an initial step, these signal noise ratio values SNRi are thus measured for each of the four carriers fl, f2, f3 and f4. The results of these measurements are applied to the allocation processing unit APU¹ of Figure 2 via its measurement input ΜΓ and additionally are stored in the signal noise ratio measurement memory MM. In the 4 examples that are described, the measured signal noise ratio values equal 17 dB, 25 dB, 22 dB and 14 dB for fl, f2, f3 and f4 respectively. These values are listed in column 2 of the table of Figure 4. The first comparator Cl in Figure 2 compares the minimum required signal noise ratio values stored in Sri to the measured values, and applies the results of this comparison to the processor PR. The measured 17 dB signal noise ratio value on carrier fl for example is compared to the required signal noise ratio values, 1 6 dB, 20 dB and 23 dB. As a result, it is concluded that a maximum of 2 data bits may be allocated to fl. The processor PR then, via its output PO applies a signal to the data allocation memory MD to make it store the figure 2 in its memory location reserved for fl. Ina similar way, it is found that a maximum of 4, 3 and 0 data bits may be allocated to carriers fl, f3 and f4 respectively. These figures all are stored in the data

I

<img file="AU5832696A_D0059.tif" />

Ο Ο Ο ό ·

S ο ο ο ο ο ο &gt; ο ο οβ

060 5 οο οa « 0 2 a

ο 00

5 ο3

Ο Ο Ο Ο

<img file="AU5832696A_D0060.tif" />

ο ο ο ο ο ο ο

9&lt;Ο0 β Ο Ο 6

<img file="AU5832696A_D0061.tif" />

allocation memory MD of Figure 2. The last column of the table in Figure 4 gives an overview of the contents of these memory locations. Additionally, the control unit CT activates the summation unit SUM to calculate the sum of all numbers stored in MD. This sum equals the overall capacity number of the carrier set. In other words, it equals the number of data bits that has to be allocated to the set of carriers to fully occupy this set of carriers if no power boost is applied. The overall capacity in the first example is 9 (2+4 + 3+0). Via the output SO of the summation unit SUM, this overall capacity number is applied to the first input C2I1 of the second comparator means C2I1 which compares this overall capacity number to the number of data bits which is comprised by a data symbol and which thus has to be allocated to the set of carriers. This number of data bits to be allocated enters the allocation processing unit APU' via the input Nl¹ and is applied to the second comparator means C2 via its second input C2I2.

As already mentioned, in the first example 9 data bits constituting one data symbol are to be allocated. The second comparator means C2 therefore informs the control unit CT that there is no positive or negative deficit in data elements after full capacity allocation. The control unit CT hereupon decides that the allocation processing can be terminated. The figures stored in the data allocation memory MD are outputted via O' and applied to a data allocation unit DAU as shown in the mapper MAP of Figure 1. The final data bit distribution for the first example is shown in the graph of Figure 4. The horizontal stripes on each carrier in this graph represent the individual capacity of these carriers, whilst the black filled circles each represent a data bit. As is seen from the graph, each carrier is allocated in its individual capacity number of data bits.

Consider now in a second example the situation wherein 1 0 data bits have to be allocated to the same set of 4 carriers. The full capacity step in this second example is completely identical to the full capaciiy step of the first example and will therefore not be described here. The first table and attached graph in Figure 5 give an overview of the memory contents and data bit distribution after the full capacity allocation step is performed and are nothing but a copy of the table and graph shown in Figure 4. It is clear that to be able to allocate a tenth date bit to the set of carriers , 30

<img file="AU5832696A_D0062.tif" />

<img file="AU5832696A_D0063.tif" />

<img file="AU5832696A_D0064.tif" />

fl, f2, f3 and f4, the overall capacity of this carrier set should be enlarged. The second comparator means C2 after having compared the overall capacity number to the number of data bits that has to be allocated, warns the control unit CT that there is a capacity deficit of one data bit. By this warning, the control unit CT is triggered to make the subtracting unit SUB calculate the power boost Bi necessary to enable allocating an additional data element to each carrier. It therefore subtracts the measured signal noise ratio SNRi from the required signal noise ratio value SNRreq allowing to allocate at least one additional data element thereto. To allocate e.g. an additional data element of 1 data bit to the first carrier fl, a signal noise ratio of 20 dB is required since only 1 7 dB signal noise ratio was measured on this first carrier fl. The required power boost Bi for fl equals thus 3 dB. The figure 3 is therefore stored in the power boost memory location provided for fl. Since f2 already has 4 data bits allocated, its constellation can not be enlarged, which is indicated by the infinitely large required power boost «&gt;. For f3, a power boost Bi of 1 dB is sufficient to allow allocating thereto 4 data bits instead of 3 data bits. Finally, an additional data element of 2 data bits can be allocated to f4 if a power boost Bi of 2 dB is provided.

The power boosts Bi for each of the carriers are thus calculated by the subtraction unit SUB, are stored in the power boost memory MB, and in addition are applied via the input PI2 to the processor PR. The processor PR then determines the minimum power boost that has to be provided to enable allocating an additional data element. From the second table in Figure 5, it follows that this minimal power boost is 1 dB and is found for f3. The processor therefore decides to allocate an additional data element which comprises 1 data bit to f3 and applies a signal to the data allocation memory MD to inform the memory about this decision. The contents of MD is adapted in such a way that it now includes the values listed in the third table of Figure 5. In Figure 2, summation unit SUM calculates the overall capacity number after power boost by 1 dB. This sum equals 10 data bits which in this second example is equal to the number of data bits that has to be allocated, this number being applied to the second comparator C2 via its second input C2I2. The second comparator means C2 tells the control unit CT that no further data bit deficit exists. Under the control of CT, the allocation is terminated by outputting the contents of MD via O'. It is also remarked

<img file="AU5832696A_D0065.tif" />

<img file="AU5832696A_D0066.tif" />

ο OO C &gt;1 β Ο © i

Ο Ο 3 eο

Ο Ο J

Ο β ο _&gt;

ΟJ υο οο 1 β ο .

Ο90 ©a &gt;

οο ο ο ο ο

<img file="AU5832696A_D0067.tif" />

© ©η © Ο · ο ο ο ο ©coo ο - ο ©ο ο

<img file="AU5832696A_D0068.tif" />

that the data allocation unit whom the contents of MD is applied to, further has to be informed about the 1 dB power boost that has to be performed.

An overall power boost of 1 dB is then performed by this data allocation unit but it has to be noted that in an alternative implementation of the present method, this power boost may be applied only to f3 and not to fl, f2 and f4. The final distribution is shown in the third graph of Figure 5. Therein, it is seen that by applying an overall power boost of 1 dB, the individual capacity of f3, represented by the horizontal stripe, has increased whilst the individual capacities for carrying data elements of all other carriers in the set remain unchanged. The tenth data bit is allocated to f3 in such a way that the four carriers again become fully occupied after having performed the full capacity step and undercapacity fine tuning step of the present method.

In example 3, 6 data bits have to be allocated to the carriers fl, f2, f3 and f4. A full capacity step is again executed to determine for each carrier the maximum number of data bits that may be allocated thereto, and to allocate to each carrier its individual capacity number of data bits. This full capacity step and the distribution which is the result thereof are once more completely identical to the full capacity step performed in example 1. The first table and graph in Figure 6 remain unaffected when compared to the table and graph in Figure 4. For further explanations, reference is made to the above paragraphs.

The second comparator means C2 at the end of the full capacity step detects that 9 data bits became allocated to the set of carriers in the full capacity step whilst only 6 data bits had to be allocated. A deficit of -3 data bits or overcapacity of 3 data bits is therefore announced to the control unit CT which triggers the subtraction unit SUB to perform the first substep of the overcapacity fine tuning step. For the four carriers, the subtraction unit SUB calculates the additional noise margin, which is a measure of the noise sensitivity for these carriers. The additional noise margin is calculated by subtracting from the signal noise ratio SNRi measured on a carrier, the required signal noise ratio SNRreq to allocate to this carrier its individual capacity number of data bits. If the additional noise margin of a carrier is small, the carrier is occupied almost completely and has no spare signal noise ratio margin anymore. In ΐ

<img file="AU5832696A_D0069.tif" />

<img file="AU5832696A_D0070.tif" />

<img file="AU5832696A_D0071.tif" />

Ο Ο Ο i &amp;·

Ο β Ο

Ο Ο 0 3Ο

ΟΟ

Ο 9 0 Ο ΟΟ ο ο ο ο ο

0 β&gt;

ο β ο Ο ο 9 9 ο 0 ο ο ο ο ο

0 0 9

Ο 0 9

0 ο ο ο

Ο 9 9 0 0

Ο

Ο ο ο ο ο «

0 9 0 this situation, the respective carrier will be very noise sensitive. The signal noise ratio measured for fl equals 1 7 dB, whilst full capacity occupation of the first carrier fl, i.e. allocation of two data bits thereto, requires 1 6 dB signal noise ratio. The remaining margin or additional noise margin of fl is thus equal to 1 dB. From the second table in Figure 6, it is seen that the additional noise margins ANMi calculated in an analogue way for f2, f3, f4 equal 2 dB, 2 dB and 14 dB respectively. Having the above conclusion that small additional noise margins are equivalent for high noise sensitivities in mind, it is obvious that the additional noise margin ANMi should be as large as possible. If a data element, previously allocated to a carrier, is removed therefrom, the additional noise margin ANMi of this carrier will be enlarged. Thus, in the overcapacity fine tuning step implemented in the described embodiment, data elements are removed from carriers which have the smallest additional noise margins. As already said, the subtraction unit SUB calculates the additional noise margins and in addition applies these additional noise margins to the additional noise margin memory MA to be stored therein. The values stored in MA are then applied to the processor PR which, still under the control of the control unit CT, determines the minimum additional noise margin and which orders the data allocation memory MD to remove a data element from the carrier having this smallest additional noise margin. From the second table in Figure 6, it follows that fl has the smallest additional noise margin. Therefore, a data element comprising two data bits is removed from fl to obtain the distribution drawn in the third graph of Figure 6. Since still 7 data bits are allocated to the set of carriers whilst only 6 data bits should be allocated, the just described procedure of calculating additional noise margins, storing these additional noise margins in the additional noise margin memory MA, determining the minimum additional noise margin and removing a data element from the carrier having this minimal additional noise margin is repeated. The new list of additional noise margins ANMi is shown in the last column of the third table of Figure 6. Carrier f2 and carrier f3 are most noise sensitive and thus are two candidates for data element removal. To f2 however, 4 data bits already became allocated whilst f3 is occupied by 3 data bits and will therefore be a bit less noise sensitive than f2. The processor PR decides to remove a data element comprising 1 i

0 60

4644

9Ο

9Ο

0 0 0 0Ο e 60 ο

Ο 0

440

Ο ® &lt;3 0 0 9

6 6 0 β Ο Ο Ο p Ο οο ο

Ο 040 4

Ο

0044 « 9

0»

<img file="AU5832696A_D0072.tif" />

data bit from the second carrier f2. The final constellation and memory contents are illustrated by the graph and table at the bottom of Figure 6. As a result of the data element removal, the carriers fl and f2 are no longer fully occupied. An overall power reduction cannot be performed since f3 is still completely occupied but in an alternative implementation of the present method, it would be possible to reduce power on fl and f2 without affecting the power allocated to f3 and f4. In the present implementation, an increase of additional noise margin is gained instead of a decrease in power. Reference is made to the last column of the tables in Figure 6 to notice this gain in additional noise margin ANMi.

In the last example, 11 data bits have to be allocated to the carriers fl, f2, f3 and f4. During the full capacity step, 9 data bits are allocated to the carriers and are distributed over these carriers as indicated in the first table of Figure 7. Under the control of the control unit CT in Figure 2, undercapacity fine tuning is provided in the capacity fine tuning step in a way similar to the undercapacity fine tuning performed in example 2 and described in the above paragraphs related thereto. Since the required power boost Bi for allocating an additional data element is minimal for f3, a first additional data element comprising 1 data bit is allocated to f3. As a result, the constellation obtained is equal to the final constellation of example 2. Evidently, the first three tables and graphs in Figure 7 are identical to the tables and graphs in Figure 5.

The second comparator C2 now detects that there is still a deficit of 1 data bit: 10 data bits are allocated, 1 1 data bits have to be allocated. The control unit CT is informed about this deficit and controls the subtraction unit SUB to recalculate the power boosts Bi necessary to add a second additional data element to the carriers fl, f2, f3 and f4. The new power boost values are listed in the last column of the third table in Figure 7. From this list it is seen that a power boost of 2 dB is required to allow allocating a first data element to f4. The processor PR concludes that this is the smallest required power boost and therefore orders MD to allocate an additional data element comprising two data bits to f4. The so obtained distribution is drawn in the fourth graph of Figure 7, whilst the last column of the attached fourth table already contains the required power boost values for allocating a further additional data $ th ’Ii ί

Γ

<img file="AU5832696A_D0073.tif" />

<img file="AU5832696A_D0074.tif" />

<img file="AU5832696A_D0075.tif" />

<img file="AU5832696A_D0076.tif" />

element to the set of carriers. Since no further additional data element has to be allocated, the information in this column and the calculation thereof is superfluous and will not be provided by an intelligent implementation of the present invention.

Instead of undercapacity, the second comparator C2 detects an overcapacity of 1 data bit. This overcapacity is caused by immediately allocating a data element which comprises two data bits to f4. In ADSL (Asymmetric Digital Subscriber Line) applications however, as prescribed in the draft Standard, no 1 bit constellations are allowed. Since allocating 2 data bits to f4 with respect to minimal power boost is a better solution than allocating only 1 data bit to another carrier it is obvious to do so in the undercapacity fine tuning. Because a power boost is requested, the so caused overcapacity thus is a pseudo-overcapacity which results from undercapacity fine tuning and standardization requirements taken into account to build up the preferred embodiment. Upon detecting this pseudo-overcapacity, the control unit CT triggers the means in the allocation processing unit APU¹ to start up a pseudo-overcapacity fine tuning step wherein this pseudo-overcapacity is eliminated. To eliminate the pseudoovercapacity in the concrete situation of the fourth example, 1 data bit has to be removed. Since this bit may not be removed from carriers with 2 bit constellations, only carriers with constellations larger than 2 data bits have to be taken into account. For these carriers, the already described overcapacity fine tuning provides a good solution for eliminating the pseudo-overcapacity. Therefore, subtraction unit SUB in Figure 2 calculates the additional noise margins for the carriers f2 and f3 and applies these additional noise margins to the additional noise margin memory MA to be temporarily stored therein. The additional noise margin for f2 equals 4 dB, the additional noise margin for f3 equals 1 dB. The processor PR searches the minimal additional noise margin amongst f2 and f3 and orders MD to reduce the number of data bits allocated to f3 by 1. The additional noise margins ANMi stored in MA are listed in the last column of the fifth table in Figure 7. The final obtained distribution is drawn in the last graph at the bottom of Figure 7. It is seen from this graph that f3 is no longer completely occupied. By removing 1 data bit from f3, is additional noise margin ANMi is enlarged from 1 dB to 4 dB. This is indicated in the last table of Figure 7. Calculating the additional noise margins ANMi listed in this last table .

:(

<img file="AU5832696A_D0077.tif" />

<img file="AU5832696A_D0078.tif" />

<img file="AU5832696A_D0079.tif" />

<img file="AU5832696A_D0080.tif" />

<img file="AU5832696A_D0081.tif" />

however is useless and will therefore not be done in an intelligent implementation of the present invention. Eleven data elements are now allocated to the set of 4 carriers. The contents of the data allocation memory MD is outputted via output O' in Figure 2. It is noted that a data allocation unit similar to the one drawn in the mapper MAP in Figure 1 has to be provided with the above outputted data allocation information and with additional information indicating that a power boost of 2 dB is necessary to allow allocating the data elements as shown in the final graph of Figure 7.

It is noticed that although this is not stated explicitly in the above description, the number of data elements that constitute a data symbol can be different for successive data symbols that have to be transmitted. Since the allocation processing unit APU' shown in Figure 2 is provided with an input Nl¹ to which the figure is applied which is equal to the number of data bits in a data symbol, the described method and equipment also apply to systems wherein successive data symbols have different lengths. Such a system and a method to modify the length of successive data symbols e.g. is described in the US Patent 5,400,322 entitled 'Updating of bit allocations in a multicarrier modulation transmission system' from the assignee Amati Communications Corp.

Furthermore it is noticed that the present method is also applicable to multicarrier applications wherein data elements according to specific criteria are partitioned into subgroups and wherein the carriers similarly are partitioned into subsets, each subset of carriers being associated to a subgroup of data elements. In such systems, data elements may be allocated only to carriers which form part of the subset associated to the subgroup of data elements where they belong to. To distribute the data elements of a subgroup over the carriers of the associated subset, the present invention also provides an alternative way for the known solutions.

It is further remarked that the present invention is not restricted to the rules and criteria used in the described embodiment to determine the individual capacity of the carriers, to decide which carrier is assigned additional data elements in case of undercapacity, and to decide from which carrier data elements are removed in case of overcapacity or pseudo-overcapacity.

Still a remark is that, although the use of QAM (Quadrature Amplitude jt

Modulation) is mentioned in the above description of the preferred embodiment, it will be obvious to a person skilled in the art that the present allocation method is not restricted to systems wherein QAM modulation is applied, but can be implemented also e.g. in systems with PSK (Phase Shift Keying) modulation or LAM (Linear Amplitude Modulation). As already mentioned before, different ones of these modulation techniques may be applied to different carriers in the set of carriers where data elements are allocated to, and as a result thereof different carriers may be accompanied by different 'required SNR per data element' tables.

Yet it is noticed that although the described embodiment of the modulator is used in ADSL applications, the present method can be implemented in other transmission systems too, e.g. OFDM for coax cable applications.

While the principles of the invention have been described above in connection with specific apparatus, it is to be clearly understood that this description is made only byway of example and not as a limitation on the scope of the invention.

<img file="AU5832696A_D0082.tif" />

o no • Οο ο ο ο a e θο ε eο

Ο ft ο ί Ο 0 Ρο ¹ ο ο ο ο9 ο &lt;?α ίο©t&gt;

! Ο Ο Ο

Ο ο • β Ο4

Ο β οο ο ο ο 1©ο ο 1 ο ο ο ο

0 9 0 , ο 9 ώ ί ο ο ο

<img file="AU5832696A_D0083.tif" />

<img file="AU5832696A_D0084.tif" />

<img file="AU5832696A_D0085.tif" />

<img file="AU5832696A_D0086.tif" />

!

!

O ft ο «οe

Ο-3ΌΦ ο «aο

C ο ο ο α ο »*»0 4

Ο 0Ο

4»Ο

Ο G &gt;i α ο©

Ο 6 &lt;3 α ο a 0*4 ο

Ο Ο«β • 0 Ο Ο ; 0-900 ! 0 U &lt;J 9

Ο 0 0

Ο Ο Q ύ &lt;3 ώ 0^4 ΐ

Ο *#«4 0 0 β C

<img file="AU5832696A_D0087.tif" />