CA2180973A1

CA2180973A1 - Method for allocating data elements in multicarrier applications and equipment to perform this method

Info

Publication number: CA2180973A1
Application number: CA002180973A
Authority: CA
Inventors: Jean Francois Van Kerckhove
Original assignee: Jean Francois Van Kerckhove; Alcatel N.V.
Current assignee: Alcatel Lucent NV
Priority date: 1995-07-11
Filing date: 1996-07-10
Publication date: 1997-01-12
Also published as: ES2185681T3; DE69528974T2; EP0753947A1; EP0753947B1; IL118676A0; IL118676A; AU701040B2; US5812599A; AU5832696A; DE69528974D1

Abstract

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.

Description

. ~ 218~973 METHOD FOR ALLOCATING DATA ELEMENTS IN MULTICARRIER
APPLICATIONS AND EQUIIPMENT TO PERFORM THIS METHOD
The present invention relates to a method for allocating data elements to a set of carriers as described in ~:he preamble of claim 1, a program module forallocating data elements to a set of carriers as described in the preamble of claim 7, an allocation l~u~es~ unit to perform this method as described in the preamble of claim 8, and a multicarrier modulator including such an allocation p, oces~,i"g unit as described in the preamble of claim 9.
Such a method and equip~nent to perform this method are already known in the art, e.~. from the US Pafent 4,679,227, entitled 'Ensemble modem structure for impeffect l,d,)s",issiùn media' from the inventor Dirk Hughes-Hartogs. Therein, a modem is cescribed which transmits and receives digital data on a set of carriers called al1 ensemble of carrier frequencies. The modem includes a system for variably allocating data elements or data, and power to the carrier frequencies to be l,d":,",i~ed via a telephone line. In this modem, data elements are allocated to the ensemble of carrier frequencies in a ~l~diyl~rurv.~ald 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 20 computes the marginal required power to increase the symbol rate on each carrier from n to n+1 illfUlllldliull units. In addition, the system allocates illrulllld~io~l units or data elements to the carrier that requires the least additional power to increase its symbol rate by one ill~Ulllldliull unit. In other words, according to a p,t~ ""i"ed rule - the carrier that requires the least additional power to increase the symbol rate for the modem desuibed 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 pl~d~l~llllill~drule is different in other kno~ d~io~s of multicarrier modems, described in literature, but the basic concept of allocating data elements one by . ~ 21~73 one to thereby build up the carrier occupations al,di~,l,lfu,~ardly remains ull(;ll~l~g~d in all known solutions An object of the present invention iâ 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 al~ u,lllrulwardly, this method, program module and equipment providing thus an alternative solution for allocating data elements in m-llticarrier ~ " ,s.
According to the invention, this object is achieved in the method, program module, allocation prucessi"~ unit and multicarrier modulator 1û described in claim 1, claim 7, claim 8, and claim 9 respectively.
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 numb~r of data bits that has to be allocated is small or large, the result of the full c2pacity step remains unaffected and depends only upon the individual capacities of the carriers in the set of carriers. Since for each canrier only its individual capacity for carrying data elements has to be measured and no further rules or measurements have to be taken into account, the full capacity step is execl~ted quickly. The total p, u~eS:~il ,9 time for allocating the data elements e(~uals the sum of the time spent to the full 20 capacity step and the time spent to the capacity fine tuning step which is executed succes~ ly In the cal2acity 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 u, Idercdpdui~y, the capacity of the carrier set is enlarged e.g. by power boostin~3 in acuûl~d"ue to a capacity enlarging rule.
Capacity is for example enlarged in such a way that minimal power boost is required to allow allocating the ~dditional data elements. When too many data elements become allocated to tl1e carrier set in the full capacity step on the other hand, i.e. in case of overcapacity, some of the allocated data elements 30 are removed. This removal is also based on a ~u,,:d~lt,,,,li,,ed rule, e.g. to ~ ~ 218~973 maximize the minimum additional noise margin calculated for the carriers, this additional noise margin for one carrier being equal to SNRi - SNRreq, wherein SNRi l~,olesrc,l~ the signal noise ratio measured on a carrier and wherein SNRreq l~:,ul~s~"l, the signal noise ratio required to allow allocating to this carrier an integer number of data elements.
Compared to the known ~ldigllIrul~dld 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 ,u,ucr~:,i"r,, time is 10 smaller whenever the number of data elements to be allocated lays in the neighbourhood of the global capacity of the set of carriers.
As follows from claim 2, the individual capacity of a carrier for allocating data elements thereto, can be de~ined precisely. In a first i,Il~ dliO~ 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 illl~ul~lllt~llIa~iol1, these required signal noise ratio values are enlarged by a fixed margin. Such a margin of 6 dB is for example ,1., ~s~, iued in the draft American Na~ional Standard for Telecommunlcations onADSL (Asymmetric Digital Sub~scnber Lrne), published by ANSI (Amencan 20 National Standards Institute) in Al~nl 1994, paragraph 12.8.3.3, page 100.
An additional ulldld~ , feature of the present invention is that in a particular i~ e~ l~dliu~ ~ 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 ra~io on this carrier and defining the number of data elements which, ~Ivhen allocated to this carrier, requires a signal noise ratio lower than but as close as possible to the measured signal noise ratio.
Another .,hdld~ , feature of the present method is that, also in a particular illl~ llldliull thereof, ulld_l~d,ud~,iIy fine tuning is performed as30 described in claim 4. In this wa~r, the global capacity of the set of carriers is enlarged by applying the minimul~ power boost necessary to allow allocating all data elements to the carriers In this particular illl,ulenll~llldliOI~, 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 distributic)n is not necessary when illl~ J the present method.
Still a further c~lldldul~liali~ feature of the present allocation method is that, again in a particular illl,UI~ dliUIl thereof, overcapacity fine tuning isperformed as described in claim 5. In this way, the noise sensitivity of the 10 carriers is minimized by removin~ data elements from the most noise sensitivecarriers in the set. Indeed, a large additional noise margin, calculated as described above, is equivalent tc~ 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 ",aA;",i~d. This rendersthe data ~, dl 1~ issiùl1 less sensiti~e for noise.
Yet another ,l Idl d~ lic feature of the present method is that pseudo-overcapacity, caused by u"d~;,~,uauily fine tuning, is e';."i"d~:d in a pseudo-overcapacity fine tuniny step c~s described in claim 6. Such a pseudo-overcapacity is due to the fact tllat additional data elements are allocated to 20 carriers in the u"d~,~,d,ua.;i~y fine tuning step. These data elements may comprise different amounts of data bits ;lept7, Idt~ on the carrier whom they are allocated to. For ADSL (Asymmetric Digital Subscriber Line) ~ 'iùi1s for example, the already ",~,lliu,led draff Standard excludes the existence of 1 bit~" ~ " ~ ,s. A first data elemel1ts allocated to a carrier thus always contains 2 data bits whilst all further data elements allocated to this carrier comprise only 1 data bit. When the last additional data element allocated in the Ul Id 1 uapaui~y fine tuning step contains 2 data bits, whilst only 1 additional data bit had to be allocated, the pseudo-overcapacity occurs. In this pseudo-ovt:,-,d,ua-,ity situation, a smaller data element Cu~ u~iSi~9 only 1 data bit can 30 be removed from a carrier which is occupied by at least two data elements. To ~ ~ 2t80973 select amongst all carriers occupied by at least two data elements the carrier where this data bit is removed flom a sequence of substeps similar to the substeps for overcapacity fine tuning described in claim 5 is executed. This implies that the small data elerrent is removed from the carrier with least additional noise margin to thereby make this carrier less noise sensitive.
The above " ,e, ,liu, ,ed and other objects and features of the invention will become more apparent and the invention itself will be best ulld~ vo~ by referring to the following dt-s._ri,u~iu,~ of an ~Il,bodi,,,_,,L taken in conjunction with the accompanying drawings vlherein:
1û Fig. 1 is a block schemQ ~f an t~ odi",t:"l of a Discrete Multi Tone (DMT) modulator according to the present invention;
Fig. 2 is a block scheme ol an tSIllbo.li",~"l of an allocation p,u~es,;"g unit according to the present inverltion;
Fig. 3 is a flow chart dia3ram of a particular illl,~ llldlitll~ of the method and program module accclrding to the present invention illustrating the steps and substeps therein;
Fis. 4 includes a table illustrating the contents of the memory means of the allocation u~uces~ unit shc)wn in Fig. 2 in case 9 data bits have to be distributed over a set of 4 r~arriers, and further includes a graph illustrating the 2û distribution of data bits over these carriers;
Fig. S includes a sequent~e of tables illustrating the contents of the memory means of the allocatioll p, vct-s~i"g unit shown in Fig. 2 during successive steps of the algorithm shown in Fig. 3 in case 1û 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;
Fig. 6 includes a sequence of tables illustrating the contents of the memory means of the allocatioll p, uct~ ,i"g unit shown in Fig. 2 during successive steps of the algorithm shown in Fig. 3 in case 6 data bits have to be distributed over a set of 4 c~rriers and further includes a sequence of 3û graphs illustrating the evolution of the data bit distribution over these carriers;

218097~

Fig. 7 includes a sequence of tables illustrating the contents of the memory means of the allocation u, ucessil ,9 unit shown in Fig. 2 during successive steps of the algorithm shown in Fig. 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 Fig 8 is a l~ul,:s~,lidliu" of a 'required SNR per data element'-table used in the method illustrated by the flow chart of Fig. 3 to obtain the data bit allocation in the four examples accu",~,a"ied by Fig. 4 Fig. 5 Fig. 6 and Fig. 71 0 respectively.
Referring to Fig. 1 and Fig. 2 the structure of a Discrete Multi Tone (DMT) modulator MOD, which is a preferred ~Illbuvi~ of the multicarrier modulator according to the present invention will be described in the first partof the d~s~ ,liul1. The basic me3ns included in a Discrete Multi Tone (DMT) modulator MOD as described in t~le draft ANSI Standard on ADSL are drawn in Fig. 1. These basic means and thl~ functions provided thereby will be described in the first pdldyldulla of this first part. Since the present invention more aue. iri 'y relates to the mappillg unit and the allocation method executed thereby some pdldyld~l)S in ad~ition will be spent on de~ ly a specihc 20 embodiment of an allocation u~u~essi"y unit included in such a mapping unit and equipped to perForm the m~thod according to the present invention. A
detailed block diagram of this allocation ~luce~ailly unit is drawn in Fig. 2 whilst the allocation method exec~ted thereby is illustrated by the flow chart of Fig. 3 In the second part of the ~ U~ iuliul1 it will be explained how the means included in the allocation p,uceaai"u unit of Fig. 2 are controlled to perform the methûd illustrated in Fig. 3. The working of the allocation ~,luceaai,lg unit will be described by means of 4 examples each passing through a different sequence of branches in the flow chart of Fig. 3 The contents of the memory 30 means MEM In the allocation p,vcessi"g unit APU of Fig. 3 as well as the 21809~3 distribution of data elements for t~le successive steps in these 4 examples are ~ n~DellL~d in the tables and graphs of Fig. 4, Fig. 5, Fig. 6 and Fig. 7 respectively. After having describ~3d the successive steps of these 4 examples, all branches of the flow chart of Fi3. 3 will be passed through.
The Discrete Multi Tone ~DMT) modulator MOD in Fig. 1 includes between an input Dl, the data input, and output MO, the modulator output, the cascade co""e~.ti~", of a mapping unit MAP, an inverse fast fourier transform ~ucessi~g unit IFFT, a cyclic pr~3fix adder CPA, a parallel to serial converter PSC, and a digital to analog corlverter DAC. The mapping unit MAP in this 10 cascade cu""eu~iù~1 includes a data allocation unit DAU and an allocation ~u~ ucesDi"J unit APU. The allocatic)n p, UCeDDil ,g unit APU is provided with a first Ml and second Nl input and is further equipped with an output O cu""e.,~e3d to an input of the data allocation uniit DAU. The modulator input Dl is coupled to another input of the data allocatiorl unit DAU.
According to the draft Standard on ADS~, the modulator MOD modulates data elements grouped in data symbols on a set of carriers having equidistant frequencies, and further applies t~le modulated carriers via the output MO to a twisted pair telephone line, not shown in Fig. 1. To be distributed over the setof canriers, the data elements erltering the modulator MOD via Dl, are first 20 applied to the data allocation uni~: DAU which forms part of the mapper MAP.
Based on a particular algorithm, the allocation lu,u~esDi"g unit APU in this mapper MAP calculates a data ele!ment distribution. It therefore is provided with carrier property ill~UlllldtiUIl, the signal noise ratio measured on each carrier, this illrulllldliol1 being applied via the input Ml, and with illfUlllldliull indicating the total amount of data bits corrlprised by one data symbol, this illfo~ iu~l being applied via Nl. The results of the calculations are supplied to the data allocation unit DAU via ûutput O of the allocation ,u~uc~ssillg 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 the30 set which modulation method has to be executed. The data allocation unit DAU

2~80973 ~, 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 secolnd carrier via 16 aAM modulation and so on.
In a signal plane, each modulated carrier can be ,~p,~se"' i by a single point, ae"li"9 the amplitude and phase of the carrier after modulation. Thus, a set of complex numbers represel1t 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 10 fourier transform ~, u~es:~;l ,9 unit lr-FT~ If the 1, dl lal I ~iss;v~, line would be perfect, i.e. if no intersymbol i"~,r~,~"ce ~,lvould be caused by the impulse response ofthe lldll~lllissio" line, the time domain parallel sequences of successive symbols could have beenioined illtO a serial data stream, t,d",Fu""e i into an analogue signal and applied to the lldll~,llissiull line. Due to the effective impulse response length of the lldl)s,~ ;vll line however, intersymbol i"l~, r~ ce occurs. Such intersymbol i"~, rt , ~ can be Cùl l l,ut~ al~ i by an adaptive filter at the receivers sid3. In known solutions and also s~ est~d in pdldyld~ 6.1û of the above cited draft Standard, such a digital filter techniqueat the receivers side is combined with cyclic prefix extension at the lldllallliL~
2û side to obtain sufficient intersymbol i"l~rFt:,~"ce C~ sd~iu". The time domain parallel sequence at the output of the inverse fast fourier transform ,u, ucesai"g unit iFFT is therefore applied to a cyclic prefix adder CPA which, in accu,dd"ce 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 num~ers The extended time domain parallel sequence is then applied to the cascade of parallel to serial converter PSC and digital to analog converter DAC to be sucu~ssi~ly lld~ru~llled into a serial digital sequence and analog signal.

2I80g73 lt is noted that an allocation ,~" u~sSil ,3 unit of the above described type can also be provided in the demodulator at the receiver's side. The signal noiseratio on each carrier is then measured e.g. by lldll:lllliIlil~g in an initial phase, a ,u,dd~l~""i"ed signal from the l~d"~",ill~r to the receiver and by analysing this signal in the receiver. The allocation ,UIu~55;"~ unit thereupon calculates the data element distribution, and applies its results via a backward path in the lldll~lll;S5iOIl system to the data allocation unit DAU in the modulator at the lldll:lllli~l~l '~ side. In such a systerrl, no data ,~"uces:,i"~ unit has to be provided in the modulator at the lldl l~ side.
The block scheme of the Discrete Multi Tone (DMT) modulator MOD in Fi3. 1 will not be described in furt~ler detail since such a detailed desu,i~,liu,, is of no i,,,,uu, ldl ,~ for the present invention. Further details with respect to ADSL
requirements are described in the already ",~"liu,led draft ANSI Standard on ADSL whilst specific illlUIC:lllt,llldliulls of Discrete Multi Tone modulators are found in the articles 'A multicdrrier E1-HDSL Tldll~ . System with Coded ~ nnll written by Peter S. Chow, Naofal Al-Dhahir, John M. Cioffi and John A.C. Bingham and published in the issue Nr. 3 MaylJune 1993 of the Journal of European T,d"saulio,~s on Telecommunications and Related Te.,lll,olu~i~s (ETT), pages 25~7-266, and 'Pe,ru,l"d"ce Evaluation of a 20 Multichannel Transceiver System lFor ADSL and VHDSL Services' from Peter S.
Chow et al., published in the issue Nr. 6 Au3ust 1991 of the Journal of European Tldll~d~liUII~ on Telecommunications and Related Te~l",olol i-s (ETT), pa3es 909-919.
The present invention more s~euiri~"y relates to the method pe,ru""ed by the allocation ~u~s~i"g unit APU in Fig. 1. A particular er,lL,odi",e"l of such an allocation ~,, uu~ss;, ,~ unit APU' is drawn in Fi3. 2.
The allocation ~J~uce,,i~g unit APU' of Fi3. 2 includes memory means MEM, a subtraction unit SUB, a s~lmmation unit SUM, a first ~,UI~I,UdldlOI meansC1, a second uolll~dlL~ur means C2, a processor PR and a control unit CT.
30 The memory means MEM is subdivided into a si3nal 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 ~,(uces:,i"~ unit APU' is uc~""~_ltld to an input of the signal noise ratio measurement memory MM. A first output 01 of this memory is connected to a first input C111 of the first CUlll~dldlUI means C1, 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 ratiomemory MR is co""e~ed to a second input C112 of the first Cu",~d,L'ù, means 10 C1, whilst another output 04 thereof is cu""e~ d to a second input SBI2 of the subtraction unit SUB. An output C10 of the first ~UllI,UdldlUI means C1 is co""eul~:d to a first input Pl1 of the processor PR. The subtraction unit SUB onthe other hand is provided with two outputs, the first SB01 of which is cu""e,~ d to an input 13 of the additional noise margin memory MA and a second SB02 of which is cu""euI~d 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 cull,,euliul~ 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 C211of the second I~llI,Udl._~l means C2 via the summation unit SUM, and the 20 other output 06 is cù~ e~ d to an output O' of the allocation u,u~,es ,i"~ unit APU'. The summation unit SUM thereto is provided with an input Sl and ou~put SO. An output 07 of the powel- boost memory MB and an output of the additional noise margin memory MA are connected to respective second Pl2 and third Pl3 inputs of the processor PR. Finally, a second input Nl' of the allocation p~uues~ unit APU' s~rves as a second input C212 of the second Colll,udldl~( means C2.
The processor PR, the first C1 and second C2 ~.UIII,Udl~.'UI means, the memory means MEM, the subtraction unit SUB and the summation unit SUM
are all controlled by the controller unit CT. In Fig. 2, this is indicated by the 3û double arrows which represent control lines between the control unit CT and .

the other means in the allocatit~n ~,uces~i"g unit APU. The control lines lvt ~ are not drawn in this fi3ure since this would make the drawing unclear. However from Fig. 2 it is obvious to a person skilled in the art how such control lines should be prl~vided to obtain the working which will be described in the following pdldyl~plls and which is illustrated by the algorithmflow chart of Fig. 3.
The flow chart in Fig. 3 is build up with different shaped boxes r u~ e,~ d 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 10 through from top to bottom and ~rom left side to right side of the chart. Theactions which have to be executed successively are, ~p, ~: e, l~d by rectangle boxes. If the left and right side of such a rectangle box are drawn double this box le,~lt,st",ls a procedure or cl~ster of actions. The actions included in this cluster are all o~ e. L~d to the bottom of the rectangle box with double left and right sides. Each he dyu, Idl box indicates that an action or plurality of actions has to be executed repeatedly. T~e condition to determine how many times the actions have to be executed is described within the l-~ dyUlldl box whilst the actions that have to be executed are described in boxes, ~ e lt, i to the bottom of the l-~dyul-dl box. ~ diamond shaped box on the other hand 20 indicates that one of two actiorls has to be executed. If a pl~d:l_lllli"ed condition is fulfilled the action or branch of actions co""e. le,d to the diamond side marked by Y is executed. If t~lis u,t-dt:~l",i"ed condition is not fulfilled the action or branch of actions cu",~e~ d to the N marked side of the diamond box is executed. The "~ tcll",i"ed condition itself is described within the diamond shaped box.
The method whose steps and substeps are, ~ st, ll~d in the flow chart of Fig. 3 and which is perfommed by the allocation ~,ucessi"g unit APU of Fig.
2 is best explained by cu":,i.le,i"g four fictive situations wherein a specific number of data bits have to bt2 distributed over a specific set of carriers.
30 Therefore in the following part of the dt-s 1 i~ liul, 4 examples will be described one after the other. In these examples, data bits have to be allocated to a set of four carriers, f1, 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 16 dB, 20 dB and 23 dB respectively in the present example. The 'required SNR per data element'-table v,/hich includes these figures and which is stored in the required signal noise ratio memory MR of the allocation, "vces~i"~ unit APU' in Fig. 2, is I ~,u, t~ d in Fig. 8. Since the set 10 of carriers and the properties of tlle carriers in this set remain fixed for the 4 examples, only the number of data bits which has to be allocated is different.
[)ependent on this number, the steps of the algorithm which have to be executed will be different too.
In the first example, the allo~ation is performed by only executing the full capacity step of the present method since no overcapacity or u~,d~,ud,ua~ y appears therein. The contents of t~le memory MEM of the allocation p, uues~il ,9unit APU' of Fig. 2 for this example is shown in the table of Fig. 4. A graph attached to this table in Fig. 4 illustrates the division of the data bits over the set of canriers. For the second example, tables illustrating the contents of the20 memory MEM and co" ~ UI ,~i"g graphs are drawn in Fig. 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 uln~ d~Jd-,ily are executed. In example 3, overcapacity occurs after full capacity allocation.
Therefor the capacity fine tuning step dedicated to eliminate this overcapacity is executed in addition to the full capacity step. Tables and graphs acw",,ud"ying the ~pla"dliu" of example 3 are drawn in Fig. 6. Finally, the last example is illustrated in Fig. 7. Full capacity allocation and u"de",d~,a~,ily fine tuning in this fourth example cause pseudo-overcapacity. Cli,llilldliull ofthis 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 pdl~l~l d,lJI 1:~.
In example 1, 9 data bits h~ve to be allocated to 4 carriers, f1, f2, f3 and f4. To allow allocating 2 bits to a carrier, a signal noise ratio of at least 16 dB
has to be measured on this carri~r. Similarly, 3 or 4 bits may be allocated to acarrier 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 c~rriers f1, f2, f3 and f4. The results of these measurements are applied to the allocation ~,u~,e~si"g unit APU' of Fig. 2 via its measurement input Ml' and a~d;tiu,ldlly are stored in the signal noise ratiomeasurement memory MM. In the 4 examples that are described, the measured signal noise ratio values equal 1 l' dB, 25 dB, 22 dB and 14 dB for f1, f2, ~3 and f4 respectively. These values are listed in column 2 of the table of Fig. 4 The first cu,,,i,a,dlu, C1 in Fig. 2 compares the minimum required signal noise ratio values stored in Sri to the measured values, and applies the results of this ~,u",~arisu,. to the processor PR. The measured 17 dB signal noise ratio value on carrier f1 for example is oompared to the required signal noise ratio values,16 dB, 20 dB and 23 dB. As a rr~sult, it is concluded that a maximum of 2 data bits may be allocated to f1. The l~rocessor PR then, via its output PO applies asignal to the data allocation memory MD to make it store the figure 2 in its memory location reserved for f1. In a similar way, it is found that a maximum of4, 3 and 0 data bits may be allocated to carriers f1, f3 and f4 le~Je-,ti~Oly.
These figures all are stored in the data allocation memory MD of Fig. 2. The last column of the table in Fig. 4 gives an overview of the contents of these memory locations. Additionally, tlle 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 c~rrier set. In other words, it equals the numberof data bits that has to be alloca~ed to the set of carriers to fully occupy this set of carriers if no power boost is al~plied. 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 C211 of the second Cu~ dl~tO~
means C211 which compares thi'; 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. T~lis number of data bits to be allocated enters the allocation ,u~uces~;~lg unit Al'U' via the input Nl' and is applied to the second UU~ dl~tUI means C2 via its second input C212.
As already ~ io~ed, in the first example 9 data bits constituting one data symbol are to be allocated. The second ~olllf~dl~lul means C2 therefore informs the control unit CT that t~lere is no positive or negative deficit in data 1û elements after full capacity alloc~ltion. The control unit CT hereupon decides that the allocation p~uces~ g can be l~""i"dled. 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 Fig. 1. The final data bit distribution for the first example is shown in the graph of Fig. 4. The horizontal stripes on each canrier in this graph represent th~ 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 its individual capacity number of data bits.
Consider now in a second example the situation wherein 1û data bits have to be allocated to the same set of 4 carriers. The full capacity step in this 20 second example is wlllp~_t~ly identical to the full capacity step of the first example and will therefore not be described here. The first table and attached graph in Fig. 5 give an overview of the memory contents and data bit distribution affer the full capacity allocation step is pt:~ ru~ ed and are nothing but a copy of the table and graph shown in Fig. 4. It is clear that to be able to allocate a tenth data bit to the set of carriers f1, f2, f3 and f4, the overall capacity of this carrier set should be enlarged. The second colll~Jdl~lul means C2 after having compared the overall capacity number to the number of data bits that has to be allocated, warrls 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 3û 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 SNF~i 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 f1, a signal noise ratio of 20 dB is reqllired since only 17 dB signal noise ratio wasmeasured on this First carrier f1. The required power boost Bi for f1 equals thus 3 dB. The figure 3 is therefore stored in the power boost memory location provided for f1. Since f2 already ~las 4 data bits allocated, its c~,, Isl~ iol, can not be enlarged, which is indicat~d by the infinitely large required power boost10 ~. For f3, a power boost Bi of 1 dB is sufficient to allow allocating thereto 4 data bits instead of 3 data bits. Firlally, 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 boostsBi for each of the carriers are thus calculated by the subtraction unit SUB, arestored in the power boost memor~ MB, and in addition are applied via the input Pl2 to the processor PR. The F~rocessor PR then cl~ llil,es the minimum power boost that has to be provided to enable allocating an additional data element. From the second table in Fig. 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 r~omprises 1 data bit to f3 and applies a signal 20 to the data allocation memory MD to inform the memory about this decision.
The contents of MD is adapted il1 such a way that it now includes the values listed in the third table of Fig. 5. In Fig. 2, summation unit SUM calculates the overall capacity number after po~Ner 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 cu",i a,dlu, C2 via its second input C212. The second ~,UIll,Uald~Ol means C2 tells the control unit CT
that no further data bit deficit exists. Under the control of CT, the allocation is l~""i"dled by outputting the cont~nts of MD via 0'. It is also remarked that thedata allocation unit whom the contents of MD is applied to, further has to be 30 informed about the 1 dB power boost that has to be performed.

2I8~973 An overall power boost of ~ dB is then performed by this data allocation unit but it has to be noted that in an alternative illl,UI~ llldliVIl of the present method, this power boost may be applied only to f3 and not to f1, f2, and f4.
The final distribution is shown in the third graph of Fig. 5. Therein, it is seen that by applying an overall power boost of 1 dB, the individual capacity of f3, , ~p, ~:se"'~d by the horizontal stripe, has increased whilst the individual capacities for carrying data elerrlents of all other carriers in the set remain ull-,lldll~u,~d. The tenth data bit is allocated to f3 in such a way that the four carriers again become fully occul~ied after having performed the full capacity 10 step and u, ~del~,d,uacily fine tuning step of the present method.
In example 3, 6 data bits ~lave to be allocated to the carriers f1, f2, f3 and f4. A full capacity step is again executed to determine for each carrier themaximum number of data bits tha~ may be allocated thereto, and to allocate to each carrier its individual capacit!/ number of data bits. This full capacity step and the distribution which is the result thereof are once more ,ulll~ ly identical to the full capacity step performed in example 1. The first table and graph in Fig. 6 remain unaffected ~Nhen compared to the table and graph in Fig.

4. For further _~,uldl IdliUI ,~, refererlce is made to the above pdl dyl d,UI 15.
The second culll,udldlol means C2 at the end of the full capacity step 20 detects that 9 data bits became allocdted to the set of carriers in the full capacity step whilst only 6 data bi~:s had to be allocated. A deficit of -3 data bits or overcdpacity of 3 data bits is therefore announced to the control unit CT
which triggers the subtraction urlit SUB to perform the first substep of the overcapacity fine tuning step. Fo~r 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 r~lrlll~Pd by subtracting from the signal noise ratio SNRi measured on a canrier, the required signal noise ratio SNRIeq to allocate to this carrier its individual capacity number of data bits. If th~ additional noise margin of a carrier is small, 30 the carrier is occupied almost cornpletely and has no spare signal noise ratio 218~73 margin anymore. In this situation, the respective carrier will be very noise sensitive. The signal noise ratio measured for f1 equals 17 dB, whilst full capacity occupation of the first carrier f1, i.e. allocation of two data bits thereto, requires 16 dB signal noise ratio. The remaining margin or additional noise margin of f1 is thus equal to 1 dB. From the second table in Fig. 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 rt3spectively. Having the above conclusion hat small additional noise margins é~re equivalent for high noise sensitivities in mind, it is obvious that the additit)nal noise margin ANMi should be as large as10 possible. If a data element, previously allocated to a carrier, is removed therefrom, the additional noise n1argin ANMi of this carrier will be enlarged.
Thus, in the overcapacity fine tuning step i",~,la",~"l~.;l in the described e",~odi",~"~, data elements are removed from carriers which have the smallest additional noise margins. As alre3dy 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 v,/hich, still under the control of the control unit CT, d~ "i"~s the minimum additional noise margin and which orders the data allocation memory MD to remove a data element 20 from the carrier having this small~st additional noise margin. From the second table in Fig. 6, it follows that f1 has he smallest additional noise margin.
Therefore, a data element ~IlIIJl iSi~ ~ two data bits is removed from f1 to obtain the distribution drav,m in the third graph of Fig. 6. Since still 7 data bits are allocated to the set of carriers wllilst only 6 data bits should be allocated, the just described procedure of calc-llating additional noise margins, storing theseadditional noise margins in the additional noise margin memory MA, d,_'~. ",i"i"g the minimum additior~al 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 30 third table of Fig. 6 Carrier f2 and carrier f3 are most noise sensitive and thus 218~973 are two cdll~iddl~:~ for data element removal. To f2 however, 4 data bits already became allocated whilst f3 is occupied by 3 data bits and will thereForebe a bit less noise sensitive tharl f2. The processor PR decides to remove a data element cu,l,,u,isi,,~ 1 datrl bit from the second carrier f2. The final cuns~lldliu" and memory contents are illustrated by the graph and table at the bottom of Fig. 6. As a result of the data element removal, the carriers f1 and f2 are no longer fully occupied. An overall power reduction cannot be pe, ru""ad since f3 is still uolll,u~ut~ly occupied but in an alternative illl~ule:lllalltd~iull of the present method, it would be possible to reduce power on f1 and f2 without 10 affecting the power allocated to f3 and f4. In the present i~l,ul~ lio~, an increase of additional noise margin is gained instead of a decrease in power.
Reference is made to the last col~mn of the tables in Fig. 6 to notice this gainin additional noise margin ANMi.
In the last example, 11 data bits have to be allocated to the carriers f1, f2, f3 and f4. During the full cdpacity step, 9 data bits are allocated to the carriers and are distributed over t~lese canriers as indicated in the first table of Fig. 7. Under the control of the control unit CT in Fig. 2, uln~ wlJd~ily fine tuning is provided in the capacity fine tuning step in a way similar to the url~ ;dpa,,ily fine tuning performed in example 2 and described in the above 20 pdl dyl dl.ll IS related thereto. Since ~:he required power boost Bi for allocating an additional data element is minimal for f3, a first additional data element l,UII I,UI i:~il 19 1 data bit is allocated to f3. As a result, the ~,u, '~ " ' , obtained is equal to the final cu" ' " " ., of e~:ample 2. Evidently, the first three tables and graphs in Fig. 7 are identical to the tables and graphs in Fig. 5.
The second CUIII~Udl~.~JI C2 now detects that there is still a deficit of 1 data bit: 1û data bits are allocat~d, 11 data bits have to be allocated. The control unit CT is informed about ~:his deficit and controls the subtraction unit SUB to re~lrl 1' ' the power boosts Bi necessary to add a second additional data element to the carriers f1~ f2, f3 and f4. The new power boost values are 30 listed in the last column of the third table in Fig. 7. From this list it is seen that a .. . _ _, . , _ , , ,,,,,, , . ,, ,, , ,, , _ _ .. ... .

, 2180973 power boost of 2 dB is required to allow allocating a first data element to f4.
1'he processor PR concludes that this is the smallest required power boost and therefore orders MD to allocate an additional data element co",~., iail ,~ two data bits to f4. The so obtained distriibution is drawn in the fourth graph of Fig. 7, whilst the last column of the atta~hed fourth table already contains the required power boost vaiues for allocating a further additional data element to the set of carriers. Since no further additional data element has to be allocated, the illfulllldlioll in this column and the calculation thereof is superfluous and will not be provided by an intelligent illl~ lldliùll of the present invention.
Instead of ~",de,-,d,uaui~!~, the second Culll,udl~tOI C2 detects an .,,d,uac;ly of 1 data bit. T~lis overcapacity is caused by i"""ed;.A~ly allocating a data element which comprises two data bits to f4. In ADSL
(Asymmetric Digital Subscriber Line) ~ , however, as ,u, ~:,.,, iL.ed in the draft Standard, no 1 bit cu,)s~elldliù":. are allowed. Since allocating 2 data bits to f4 with respect to minimal po~,ver boost is a better solu~ion than allocatingonly 1 data bit to another carrier iit is obvious to do so in the ~" ,de, ud,ua~ity fine tuning. Because a power boost is requested, the so caused overcapacity thus is a pseudo-overcapacity which results from ulldt:lw,ud~iIy fine tuning and ~dllddldi~d~iul~ requirements talken into account to build up the preferred 20 ~",~o.li",~"I. Upon detecting this pseudo-overcapacity, the control unit CT
triggers the means in the allocation ,u,u~,~, ,i"g unit APU' to start up a pseudo-c~ ,apa~ y flne tuning step wherein this pseudo-overcapacity is ~li.llilld~_d.
To eliminate the pseudo-overcal~acity in the concrek situation of the fourth example, 1 data bit has to be removed. Since this bit may not be removed from carriers with 2 bit c~ Ild~iUI ,~, only carriers with cu" '~ " `iù, ,s 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 ~'' llilld~ the pseudo-overcapacity. Therefor, subtraction unit SUB in Fig. 2 calculates the additional noise margins for the carriers f2 and f3 and applies these additional30 noise margins to the additional noise margin memory MA to be ~ Oldlily stored therein. The additional noise margin for f2 equals 4 dB, ~he 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 fiffh table in Fig. 7. The final obtained distribution is drawn in the last 3raph at the bottom of Fig. 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 i~ enlarged from 1 dB to 4 dB. This is indicatedin the last table of Fig. 7. Calculating the additional noise margins ANMi listed 10 in this last table however is useless and will therefore not be done in an intelligent i~ d~iUII 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 0' in Fig. 2. It is noted that a data allocation unit similar to the one drawn in the mapper MAP in Fig. 1 has to be provided with the above outputted data allocation illrulllldliull and with additional illfUlllldlioll indicating that a power boost of 2 dB is necessary to allow allocating the data elements as s~1own in the final graph of Fig. 7.
It is noticed that although this is not stated explicitly in the above des.,,i,u~ioll, the number of data elements that constitute a data symbol can be20 different for successive data syrnbols that have to be ~Idllal~ d. Since the allocation ~,, u~ 55il 19 unit APU ' sllown in Fig. 2 is provided with an input Nl ' to which the figure is applied which is equal to the number of data bits in a data symbol, the described method ar1d equipment also apply to systems wherein successive data symbolâ 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 . " - " ,)s in a multicarrier modulation ns~ ,siol~ system' from the assignee Amati Communications Corp.
Furthermore it is noticed that the present method is also q"~ to multicarrier rr~ wherein data elements according to specific criteria 3û are pdl liliUI ,ed into subgroups and wherein the carriers similarly are pdl liliC~I ,ed .. _ _ _ . _ _ _ , .. _ _ .. .. . . . .. . . . . . .. .. .

into subsets each subset of carriers being associd~ed to a subgroup of data elements. In such systems data elements may be allocated only to carriers which form part of the subset rl ' ' J to the subgroup of data elements where they belong to. To distrib-lte the data elements of a subgroup over the carriers of the Ac5O~ ed subset the present invention also provides an alternative way for the known sol~ltions.
It is further remarked that the present invention is not restricted to the rules and criteria used in the described t~ bo~ l to determine the individual capacity of the carriers, to decide which carrier is assi3ned additional data 10 elements in case of u"d~,~di,a~ily and to decide from which carrier data elements are removed in case of (~vercapacity or pseudo-overcapacity.
Still a remark is that althou3h the use of QAM (Quadrature Amplitude Mo~ nn) is ",e"li~"ed in the al~ove des~ JI I of the preferred el ,Ibodi, "t" ,Iit 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 i"",le",t:"~d also e.~. in systems with PSK (Phase Shift Keying~
modulation or LAM (Linear Amplitude Modulation). As already ,,,e,,liul1ed before different ones of these modulation techniques may be applied to different carriers in the set of carriers where data elements are allocated to 20 and as a result thereof different carriers may be accu,,,~c,,,ied by different 'required SNR per data element' té3bles.
Yet it is noticed that although the described tllll~odilll~lll of the modulator is used in ADSL c~ '; " ,s the present method can be vlt~ d in other lldll~ 2>;Ull systems too e.g. OFDM for coax cable e r ~ ~ 01)s While the principles of the invention have been described above in cu""e~liol1 with specific apparatus it is to be clearly ulld~l~lvod that this d~s( ,i~ is made only by way of example and not as a limitation on the scope'of the invention.

Claims

1. 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, characterized in that 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 individualcapacity 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 accordanceto 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.

2. A method according to claim 1, characterized in that in said full capacity step, a said individual capacity number for a said carrier is obtained by comparing a signal noise ratio value measured on said carrier with required signal noise ratio values for allocating integer numbers of said data elements to said carrier, said required signal noise ratio values being listed in a 'required SNR per data element'-table.

3. A method according to claim 1, characterized in that said full capacity step comprises for each said carrier a first substep wherein a signal noise ratio value (SNRi) is measured on said carrier, a second substep wherein said measured signal noise ratio value (SNRi) is compared with required signal noise ratio values (SNRreq) for allocating integer numbers of said data elements to said carriers a third substep wherein said individual capacity number is determined as being an integer number whose associated said required signal noise ratio value (SNRreq) is lower than said measured signal noise ratio value (SNRi) but larger than or equal to all said required signal noise ratio values (SNRreq) which are lower than said measured signal noise ratio value (SNRi) and a fourth substep wherein said individual capacity number of said data elements is allocated to said carrier.

4. A method according to claim 1 characterized in that said capacity fine tuning step in case of undercapacity comprises a first substep wherein for each said carrier a required power boost (Bi) is calculated said required power boost (Bi) being equal to SNRreq' - SNRi wherein SNRreq' represents a required signal noise ratio value which allows to allocate an additional data element to said carrier and wherein SNRi represents a signal noise ratio value measured on said carrier a second substep wherein in said set of carriers said carrier whose said required power boost (Bi) is minimal is determined, a third substep wherein a said additional data element is allocated to said carrier withminimal said required power boost (Bi) and a fourth substep wherein an overall power boost equal to said minimal required power boost (Bi) is applied to each said carrier which forms part of said set of carriers said first second third and fourth substeps being repeated until said undercapacity is eliminated.

5. A method according to claim 1 characterized in that said capacity fine tuning step in case of overcapacity comprises a first substep wherein for each said carrier an additional noise margin (ANMi) is calculated said additional noise margin (ANMi) being equal to SNRi - SNRreq wherein SNRi represents a signal noise ratio value measured on said carrier and wherein SNRreq represents a required signal noise ratio value to allow allocating to said carrier its individual capacity number of data elements, a second substep wherein in said set of carriers said carrier whose additional noise margin (ANMi) is minimal is determined and a third substep wherein a said data element allocated previously to said carrier with minimal said additional noise margin (ANMi) is removed therefrom, said first, second and third substep being repeated until said overcapacity is eliminated.

6. A method according to claim 1, characterized in that said method further includes a pseudo-overcapacity fine tuning step executed in case a last said additional data element allocated to a said carrier in said capacity fine tuning step in case of undercapacity comprises more said data bits than necessary to eliminate said undercapacity, said pseudo-overcapacity step comprising a first substep wherein for each said carrier occupied by data elements which comprise less data bits than said last allocated data element an additional noise margin (ANMi) is calculated, said additional noise margin being equal to SNRi - SNRreq, wherein SNRi represents a signal noise ratio value measured on said carrier, and wherein SNRreq represents a required signal noise ratio value to allow allocating thereto the number of data elementspreviously allocated thereto in said full capacity step and said capacity fine tuning step, a second substep wherein amongst said carriers occupied by data elements which comprise less data bits than said last allocated data element, said carrier whose additional noise margin (ANMi) is minimal is determined, and a third substep wherein a said data element which comprises less data bits is removed from said carrier with minimal said additional noise margin (ANMi).

7. A program module for an allocation processing unit (APU') 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, characterized in that said set of control instructions is structured to control a sequence of operations in said allocation processing unit (APU') 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. 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 accordanceto a predetermined capacity enlarging rule, and in case of overcapacity, i.e. incase 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.

8. An allocation processing unit (APU) 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 (APU) being provided with a first input (NI') whereto said number is applied and a second input (MI') whereto carrier property information is applied, characterized in that said allocation processing unit (APU') includes a memory means (MEM) a first part (MM) of which being provided to store said carrier property information, a second part (MR) of which being provided to store carrier requirement information, and a third part (MD) 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 (C1), coupled at its first input (C1I1) to an output (O1)of said first part (MM) of said memory means (MEM) and at its second input (C1I2) to an output (O2) of said second part (MR) of said memory means (MEM), said first comparator means (C1) 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 (C1O) to a processing unit (PR) included in said allocation processing unit (APU') and coupled at its output (PO) to an input (I1) of said third part (MD) of said memory means (MEM) said processing unit (PR) being adapted to apply to said third part (MD) of said memory means (MEM) 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 (APU') further includes a second comparator means (C2) coupled at its first input (C2I1) to an output (O5) of said third part (MD) of said memory means (MEM) and at its second input (C2I2) to said first allocation processing unit input (NI') said second comparator means (C2) 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 saidprocessing unit (PR) 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 undercapacity.

9. A multicarrier modulator (MOD) for modulation of data elements applied to an input (DI) thereof on a set of carriers for transmission thereof in a communication network coupled to an output (MO) thereof said modulator (MOD) including between said input (DI) and said output (MO) a 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) said mapping unit (MAP) 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 (IFFT) 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 (CPA) 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 (PSC) 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 (MO) of said modulator (MOD), said mapping unit (MAP) including an allocation processing unit (APU) 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 (APU) being provided with a first input (NI) whereto said number is applied and a second input (MI) whereto carrier property information is applied, and a data allocation unit (DAU), an input of which is coupled to said modulator input (DI) and another input of which is coupled to an output (O) of said allocation processing unit (APU), said data allocation unit (DAU) provided to allocate said data elements, based on said distribution generated by said allocation processing unit (APU), to said set of carriers, characterized in thatsaid allocation processing unit (APU) 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 memorymeans 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 property 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 (APU) 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 (APU) 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 (NI) 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 undercapacity.