WO2007066056A1

WO2007066056A1 - Method for energy assignment in sub-bands of a multicarrier modulation communication system

Info

Publication number: WO2007066056A1
Application number: PCT/FR2006/051323
Authority: WO
Inventors: Mohamed Tlich
Original assignee: France Telecom
Priority date: 2005-12-08
Filing date: 2006-12-08
Publication date: 2007-06-14
Also published as: FR2894750A1

Abstract

The invention relates to a method for assigning a totality of available energy (E) in the sub-bands (SBI) of a multicarrier modulation communication system wherein each sub-bands (SBI) is defined by a noise margin Γi and a standard signal/noise ratio (RSBNi). The inventive method consists in classifying the sub-bands (SBi) according to the raising values of the parameter yi =Γi/RSBNi, in carrying out a sequence of iterations each of which comprises operations consisting in assigning a energy Ein = 2.yi[yn+1/yi]2 - yI to the sub-band (SBi), wherein i=1,..., n, n is an iteration row and an operator [x]2 being defined by [x]2=2[log2(x)], wherein [x] designates the integer part of x, in calculating the total energy Formula (I) assigned to the row n iteration *, in comparing the totality of assigned energy (En) with the total energy (E), in stopping the iteration sequence at the row n* such as En* = E et En*+1 > E, in assigning to each sub-band (SBI), with i = 1 n* energy Ei,: Ei=2.Yi,[Yn*+1/Yi,]2 - Yi, and number of bits: bin* = 1 + [Iog2(Yn*+1 /Yi)]. The invention can be used for multicarrier modulation multi-channel connections.

Description

METHOD FOR ALLOCATING ENERGY IN SUB-BANDS OF A MULTI-CARRIER MODULATED COMMUNICATION SYSTEM

The present invention relates to a method for allocating total energy available in sub-bands of a multicarrier modulation communication system.

The invention finds a particularly advantageous application in the field of multichannel links with multicarrier modulation, such as xDSL, PLT or Wireless links.

It will be recalled that in communication systems with multicarrier modulation, in particular discrete multi-carrier modulation DMT (“Discrete MultiTone”), the bandwidth used is divided into a plurality of sub-bands SBu each sub-band being the subject of modulation by independent carrier. The number of bits and the transmission energy allocated to a sub-band depend on the constraints imposed on the transmission line in terms of attenuation, stationary noise, etc., and therefore vary from sub-band to other.

On the other hand, it is known that a multichannel link can transport a certain number of flows, each flow being associated with a channel of the link. For example, an ADSL link is provided to transmit seven streams distributed in seven different channels, namely four unidirectional high speed channels, and three bidirectional low speed channels. The transport channels of the link must therefore be chosen according to the service corresponding to the flow to be transported considered.

It is therefore understood that, in multichannel link transmission systems like ADSL, several services can be offered simultaneously with the constraint that the sum of the energies to be transmitted in the SBi sub-bands does not exceed the total energy E configured for binding. Conventionally, the distribution of the energies in the sub-bands SBi is carried out by allocating to each sub-band, by means of binary allocation algorithms of RA (“Rate Adaptive”) type, an energy E, which can be emitted in this sub-band, taking into account the signal-to-noise ratio of the transmission line for the sub-band considered and also taking into account a certain noise margin. The greater the noise margin, the more robust the transmission with respect to the various disturbances that may occur on the line, impulsive disturbances in particular, and the better the quality of QoS service.

An often used RA type allocation algorithm is known as Hughes-Hartogs.

This algorithm is based on the principle of an allocation of bits in the sub-bands ordered according to the smallest quota of additional energy necessary. In other words, the allocation of the bits to be transmitted is done iteratively, bit by bit and with each allocation of a bit, the sub-band for which this allocation requires the most additional energy quota is sought. low.

It is recalled that for a sub-band SB ₁ defined by a normalized signal-to-noise ratio RSBN, and a noise margin X _x , the relationship between the number b, of bits to be transmitted in the sub-band and the energy Ej of necessary transmission is given by Shannon's formula:

b, = Iog ₂ (1 + E ₁ -RSBN ₁ Zr ₁ ) The transmission of the first bit in the subband SB ₁ requires an energy y, verifying the equality resulting from the Shannon formula above:

1 = Iog ₂ (1 + Y _L RSBN / Γ,)

let γ _s -! 7RS BN,

In general, the transmission of the k ^lth bit of all b, = k bits of the sub-band SB, requires an additional energy quota in this sub-band of ΔE, ¹³ 'given by:

AE ₁ ¹ "≈ 2 ^bM γ, To illustrate the Hughes-Hartogs algorithm, we take the example of a DMT modulation transmission with eight sub-bands defined by the following energy quotas Y ₁ :

Yi = I ₁ O γ ₂ = 1, 1 Ya = 1, 1 Y ₄ = 1, 3 Ys = 5.5 γ _e = 6.5 γ ₇ = 10.2 γ _δ = 40.0

In FIG. 1 is given a table showing in this particular case the procedure for allocating bits in the subbands SB ₁ . At each iteration n is allocated a bit in the sub-band for which the additional energy quota ΔE, ¹³¹ is the smallest.

We see in this example that the first bit is allocated to the last sub-band i = 8 to ia n ~ 32 ^th iteration.

Of course, this procedure according to the Hugh-Hartogs algorithm stops when the total energy for all of the sub-bands has reached the maximum available energy E.

It is noted that this known algorithm requires a high number of iterations, this being due to the fact that a single bit is allocated to each iteration.

It is to remedy this situation that the invention proposes a method for allocating total available energy E in sub-bands SB ₁ of a communication system with multicarrier modulation, each sub-band SB ₁ being defined. by a noise margin T ₁ and a normalized signal-to-noise ratio RSBN, remarkable in that said method comprises the following steps:

- classify the sub-bands SB, according to the increasing values of the parameter Y ₁ = FVRSBN ₁ ,

- carry out a series of iterations, each iteration comprising the operations consisting in:

^* allocate an energy E ^ι _n ≈ 2.Ys [Y _{n +} IZyJ ₂ - Yi to the sub-band SB, with i = 1, ... n, n being the rank of the iteration, and the operator JXk being defined by [x] ₂ = 2 ^[ÎO9 2 ^{(x) l} where Ix] denotes the integer part of x,

n n

^* calculate a total energy E _n = 2, χγ, [γ _n + i / γι] 2 -∑Yi allocated to the iteration

ι = 1 ι = 1

of rank n,

* compare the total energy allocated E _n to the total energy E, - stop the sequence of iterations at rank n such that E _n * ≤ E and E _n * + i> E ₁

- allocate to each sub-band SB ₁ with i = 1 ,, .., n the energy E ₁ :

This method is based on a bit allocation mode consisting in each iteration n to be allocated to each sub-band SB, with i = 1, .. ,, n a number of bits b ' _n given by:

b ' _n = 1 + [! θg ₂ (Yn ₊ i / Yr)]

The table in FIG. 2 shows that this bit allocation procedure is much faster than the previous one since the same filling level of the sub-bands is reached in 7 iterations instead of 31. This result is obtained from the fact that several bits can be allocated simultaneously during the same iteration.

The above relation coupled to Shannon's formula then leads to the energy E ^ι _n = 2.γ, [γ _n + i / γ _f ] 2 - Yi allocated to each sub-band SBi at each iteration of rank n . We deduce the corresponding total necessary energy:

n n

E _n = 2.IY ₁ [Y _{n + 1} ZYi] ₂ - IY ₁ (1)

The comparison of this value with the total available energy E makes it possible to determine the rank n * of the iteration for which the energy E _n * is just less than E, and to deduce therefrom the final energy allocation of the sub -bands with the formula E ₁ = 2.γ | γ _n * + 1 / γ _ι ] 2 - y.

Note, however, that this process requires the calculation of expression (1) above, the first sum of which is particularly complex to implement since each of its terms must be recalculated at each iteration.

Also, to simplify these calculations, the invention proposes a method according to which, [x] z being approached by xΛ / 2, the energy allocated to each sub-band SBj is given by:

E ₁ ≈ V2.γV _{+ i} - Y.

Indeed, one can use the following approximation: for all real x, on average the whole part [x] is worth x - 0.5.

In this case, the quantity [x] _{2 is} worth on average xΛ / 2, and the formula (1) giving the total energy allocated to the iteration of rank n is written:

not

E _n = V2n y _{n +} i -∑Yι

1 = 1

It is noted that this formula is much easier to calculate than the previous one, hence a saving in calculation time adding to the gain already achieved on the number of iterations to be performed.

As before, the comparison of the energy E _n thus obtained with the available energy E leads to the determination of the number n * of iterations to be performed for which the energy E _n * is just less than E.

Knowing n *, we can calculate the parameter γV + i giving exactly E _n * = E:

YV ₊₁ = (E + ∑γ,) / n ^* V2

1 = 1

We then deduce very simply the energy distribution by sub-band i:

as announced above. The number of bits per sub-band i is:

It is possible to refine this model by a better estimation of [xfe using a method which, according to the invention, is remarkable in that, [xfe being approached by x / S / 2.α,

- we estimate a first value of α by:

not*

α = V2 / (n ^* γv + i) ∑γ _f [γV ₊ i / γ,] ₂

ι = 1

where n ^* and γV + i are calculated as before,

not

- we compare the total energy E _n = V2n.α.γ _n + i -∑γ, allocated to the rank iteration n to the total energy E, and we deduce a new value n ^* 'such that E _n * -≤ E and E _n - + i> E, and a new value γV + i such that E _n * • = E ,

- we estimate a new value α ^! by :

- the estimates of α are stopped if the difference between two successive estimated values of α is less than a given value, or at the end of a given number of estimates.

We now write [x] = x - 0.5 + β, which gives [x] ₂ ≈ xΛ / 2.α with α =

2 ^β .

In this context, the energy E ' _n allocated by sub-band SB, to the iteration of rank n is given by:

E ^f _n = V2, α. γ _{π +} i - Y, and the total energy E _n necessary for the iteration of rank n is worth:

not

E _n = V2n.α.γ _{n + 1} -∑ψ _{ (2)

) = 1

To find the best couple (n ^* , γV + 0 in the sense of the previous process, it is necessary to obtain a satisfactory estimate of the parameter α.

For this, we first determine n * and γV ₊ i in the hypothesis α = 1, as described above.

We then write the equality between the approximate total energy EV:

not*

EV = V2n ^* .α.γV ₊₁ -∑Y, and the exact energy E _n *:

n ^* n *

E _n * = 2.XYS [YVWY ₁ I ₂ - IY,

i = 1 j = 1

We thus obtain a first value of α:

not*

α = V2 / (n ^* .γV ₊ i) ∑γ, ÏγWYi32

i = 1

From this value of α, we calculate the total energy allocated E _n by the relation (2) and we compare the available energy E to deduce a new value n * 'for which E _n *> is just less than E. We also deduce a new value γV ₊ i which gives the equality between E _n - and E. We then estimate a new value of α, namely α ':

not*'

α '= V2 / (n ^* ' .γV ₊ i) ∑Y. [γV + i / γJ ₂ (3) The iteration on the values of α continues thus until the difference between two successive values is less than a given value. The limit of iterations can also be set by a maximum number of estimates.

A variant of the process which has just been described consists, after having calculated α by:

n ^*

α = V2 / (n ^* .γV _{+ i} ) ∑YilγV ₊ i / γ.] 2.

i = 1

not*

to calculate the total allocated energy E _n * = V2n ^* .α.γV + i -∑γ, and compare it to E.

i≈1

According to the result of the comparison, the value of n ^* is decremented or incremented to obtain a new value n * 'such that E _n * ¹ is just less than E, As before, we calculate a new value γV + i giving the exact equality of energies.

A new value α 'of α is calculated by relation (3), the iterations on α ending when the values obtained are sufficiently stable or after a given number of estimates.

The invention also relates to a computer program product comprising program code instructions for executing the steps of the method according to the invention when said program is executed on a computer. Similarly, the invention relates to a support for said program product.

The invention further relates to a distribution module in the sub-bands of a transmitter of a multi-channel link, remarkable in that it comprises means for implementing the method according to the invention.

Finally, the invention relates to a transmitter of a multichannel link, remarkable in that it comprises the distribution module according to the invention.

Claims

1. Method for allocating a total available energy E in sub-bands SB ₁ of a communication system with multicarrier modulation, each sub-band SB, being defined by a noise margin T ₁ and a signal to normalized noise RSBN ,, said method comprising a step of classifying said sub-bands SB, according to the increasing values of the parameter Y _J ≈ π / RSBN _s ,

characterized in that said method further comprises the following steps:

- carry out a series of iterations, each iteration comprising Its operations consisting in:

* allocate an energy E ' _n = 2.γ, [γ _n + i / γ ₃ ] 2 - Yi to the sub-band SB, with i = 1, ..., n, n being the rank of the iteration , and the operator [x] ₂ being defined by

[x] ₂ = 2 ^[log 2 ^{(x) l} where [x] denotes the integer part of x,

n n

^* calculate a total energy E _n = 2.∑Yi [γn + i / γj] 2 -∑Yt allocated to the iteration i ₌ 1, = 1

of rank n,

* compare the total energy allocated E _n to the total energy E,

- stop the sequence of iterations at rank n ^* such that E _n * ≤ E and En * + i> E,

- allocate to each sub-band SB, with i - 1, ..., n ^* the energy E ₁ :

and the number of bits:

b'n * = 1 + [Iog ₂ (γ _n * ₊ i / y,)] 2. Method according to claim 1, characterized in that, [x] ₂ being approached by X / Λ / 2, the energy allocated to each subband SB ₁ is:

Er ^ Y ₁ [YWYJ ₂ - y, and the number of bits per subband SB ₁ is: b ' _n * = 1 + [log ₂ (γV ₊ i / Yi)]

not*

with γV ₊ i = (E + ∑γ ₍ ) / n ^* V2

ι = 1

3. Method according to claim 1, characterized in that, [x] ₂ being approached by xΛ / 2.α,

- we estimate a first value of α by:

n ^*

α = V2 / (n * .γV ₊ i) ∑γ _t [γV ₊ i / γ,] ₂

i = 1

where n ^* and γV + i are calculated according to claim 2,

- we compare the total energy E _n ≈, allocated to the row iteration

n to the total energy E ₁ and we deduce a new value n ^* 'such that E _n * <E and

E,

- we estimate a new value α 'by:

not*'

α '= V2 / (n ^* ' .γV + i) ∑γ, [γV ₊ i / γj2

₁ = 1

4. Method according to claim 1, characterized in that, [x] ₂ being approached by xΛ / 2.α,

- we estimate a value of α by:

not*

α≈V2 / (n ^* .γV ₊ i) ∑Y. [γWγ.] 2

where n ^* and γV + i are calculated according to claim 2,

not*

- we compare the total energy E _n * = V2n * .α.γV + i -∑γ, to the total energy E,

ι = 1

- we decrement or increment n * to obtain a new value n * 'such that E _n - -≤ E and E _n - + i> E,

- we calculate a new value y V + i such that E _n * = E ₁

- we estimate a new value α 'by:

5. Computer program product comprising program code instructions for executing the steps of the method according to any one of claims 1 to 4 when said program is executed on a computer,

6. Support for the computer program product according to claim 5,

7. A distribution module in the sub-bands of a transmitter of a multichannel link, characterized in that it comprises means for implementing the method according to any one of claims 1 to 4.

8. Transmitter of a multichannel link, characterized in that it comprises the distribution module according to claim 7.