METHOD FOR OPTIMIZING SIGNALS WITH MULTIPLE SUBCARRIERS

I. TECHNICAL FIELD OF THE INVENTION

The invention relates to a Method for coding data in a trans ^¬ mitted signal, wherein the transmitted signal comprises mul ^¬ tiple subcarriers with orthogonal frequencies and to a signal transmitter for transmitting and coding multiple access data in a multiplexed transmitted signal, wherein the transmitted signal comprises multiple subcarriers with orthogonal fre ^¬ quencies .

II. BACKGROUND OF THE INVENTION

The invention relates to the field of cording data in a transmitted signal, wherein the transmitted signal comprises multiple sub-periods with orthogonal frequencies. Data coding methods according to the invention are also known as "Orthogonal Frequency Division Multiplex" (OFDM) schemes, where the signal is composed of a superposition of signals corre ^¬ sponding to a set of sub-carriers, wherein the sub-carriers have mutually orthogonal frequencies. The frequencies are or- thogonal in the sense that the time dependent functions on a support corresponding to the signal duration of the transmit ^¬ ted signal are orthogonal functions in the mathematical sense. The signal duration is therefore inversely propor ^¬ tional to the spacing of neighbouring sub-carriers.

One fundamental problem that is inherent to all orthogonal frequency division multiplex systems is the high Peak-to- Average Power-Ratio (PAPR) . Since the transmitted signal is a superposition of many sinusoidal signals, the time domain samples are approximately gaussian distributed, which gives rise to strong peaks. This is a problem because power ampli ^¬ fiers are linear only within the limited range of input power. Stronger peaks are clipped to the maximum output

power. Hence, either the signal is distorted by the non ^¬ linear clipping operation or the power amplifier must operate with low average power so that even peaks are still within the linear range. This, however, leads to a poor efficiency and over-dimensioned amplifiers. It is therefore important to employ algorithms that reduce the peak to average power ra ^¬ tio .

The second characteristic of orthogonal frequency division multiplex signals is that the Power Spectral Density (PSD) of the transmitted signal decays quite slowly with fT ² outside the allocated frequency range. This is due to the spectral components of the rectangular pulse corresponding to the symbol duration. The Out-of-Band-Power (OBP) therefore inter- feres with systems transmitting on neighbouring frequency bands. Therefore, in order to reduce mutual interference be ^¬ tween two independent systems to an acceptable level, the systems must be separated by a rather large guard interval in frequency domain. Obviously, this guard interval reduces the overall spectral efficiency. In the following, we will use the term Out-of-Band-Power (OBP) for the power that is contained in the power spectral density outside the assigned frequency range. If the Out-of-Band-Power is reduced, then the guard interval between the systems could be reduced and the spectral efficiency could be increased.

Due to the importance of Peak-to-Average-Power-Ratio reduc ^¬ tion, many algorithms have been proposed in literature. Many of these algorithms, however, suffer from the fact that side information needs to be transmitted to the receiver. For in ^¬ stance, one could employ several different interleavers, each of which giving rise to a different time domain signal. Then the signal with the lower Peak-to-Average-Power-Ratio is cho ^¬ sen. The receiver of cause needs to know which interleaver was used. Not only leads the additional transmission of the side information to the decrease of spectral efficiency, but this data is very important because an error in the side in-

formation can lead to a complete loss of the orthogonal fre ^¬ quency division multiplex symbol.

Another Peak-to-Average-Power-Ratio reduction algorithm is proposed by Tellado in his PHD theses (Tellado, "Peak-to-

Average-Power Reduction for Multi-carrier Modulation") . This algorithm does not suffer from the latter drawback. He proposes to reserve a set of sub-carriers for the sole purpose of Peak-to-Average-Power-Ratio reduction: these carriers, called Reserved Tones (RT) , are referred to as reserved sub- carriers in the following. The reserved sub-carriers do not carry any data, but are modulated such that the Peak-to- Average-Power-Ratio of the resulting time signal is mini ^¬ mised. This method yields of course a reduction of spectral efficiency, but since the reserved sub-carriers do not carry any data, the system is not vulnerable to loss of side infor ^¬ mation. In fact, if the channel is slowly varying and the transmitter has channel state information, the reserved sub- carriers could be scheduled on sub-carriers that are in a deep fade, where data transmission would be impossible any ^¬ way. The sub-carrier reservation method has found its way into the WIMAX standard, which shows that this algorithm is indeed of practical interest.

For the problem of Out-of-Band-Power-Reduction, there is also a technique proposed by Brandes (Brandes, Cosovic, Schnell, "Sidelobe Supression in OFDM Systems by Insertion of Cancel ^¬ lation Carriers"), which is based on reserved tones. These tones are sub-carriers, in the aforementioned paper called Cancellation Carriers (CC) are located at the edges of the spectrum and are modulated such that their spectra approxi ^¬ mately cancel the spectra of the data carriers, resulting in the reduction of Out-of-Band-Power of the total orthogonal frequency division multiplex signal.

The above described problems occur in particular in OFDM(A) based Systems. Such Systems are suffering in mobile scenarios uplink connections, by the imperfect time synchronisation of the different allocated chunks. The multitude of mobile ter- minals have slight but notable frequency offsets, which vio ^¬ lates the orthogonal nature of the OFDMA signal at the base station receiver.

Typical examples for OFDM(A) based uplink system concepts are the IEEE 802.16 standard (WiMax) , 802. Hg, or proprietary concepts like Flash-OFDM.

The multitone modulation itself puts a significant require ^¬ ment on the power amplifier design, namely in the need of a back off, which reduces the PA efficiency, or achievable out ^¬ put power. For both problems individual solutions already ex ^¬ ist to reduce the unwanted effects, by using cancellation carriers or reserved tones. However the cancellation carriers are dedicated for a single use to day, to achieve the wanted effect.

OFDM Modulation puts a significant requirement on the power amplifier design, namely in form of a back off, which reduces the PA efficiency, or achievable output power.

Peak to average power ratio (PAPR) is a measure which de ^¬ scribes the required linearity of a transmitter to transmit signals with multi-code or multi-tone characteristic and is also dependent on the used modulation schemes. The aim is to generate a signal w/o significant distortion, therefore avoiding degradations in the Signal-to-Noise-Ratio (SNIR) re ^¬ quired at the receiver.

Therefore, the PAPR describes the needed linearity especially of the power amplifier to achieve a good link performance. The linearity is achieved by introducing a back off from the PA saturation point, commonly known as the IdB compression point, about a certain amount of dB . The magnitude of this back off depends on the used modulation scheme and the multi- code or multi tone characteristic of the system concept. The back off improves the linearity of the PA on the cost of higher power consumption, or a reduced maximum transmission power.

From a mobile terminal point of view, large power back offs are reducing the achievable range, or if the range is main ^¬ tained the PA linearity causes a high power consumption and shorten the battery life time significantly. This effect on increased transmitter design complexity and power demand mounts up in system concepts which are utilizing balanced Multiple Input Multiple Output (MIMO) concepts as e.g. 2X2 MIMO, which requires the presence of two transmitters in the terminal. Complex transmitter designs and high power consump ^¬ tion are duplicated.

Systems with OFDMA multi carrier transmission are suffering on the PAPR effects in the uplink. Especially if mobile ter- minals are being used the effect on the battery life time is negative. There are concepts to reduce the PAPR by means of cancellation carriers which are using reserved tones. Tone reservation dedicates specific subcarriers to be used for something different than the user data transport. Usually, the user data is mapped to a number of subcarriers and is segmented into time intervals, comprising resource block units .

Tone reservation is performed in such a way that additional subcarriers are added to the resource block (or subcarriers of a resource block are being removed from the MAC as carri ^¬ ers bearing user data and getting a reserved status) , where the reserved tones have a modulation which does not depend on user data. The modulation is controlled by the output signal and if the peak power is going to exceed a certain threshold, the modulation of the reserved tones is chosen in such a manner that the PAPR is decreased. Depending on the implemented complexity, PAPR reductions of about 3dB and higher can be achieved, which in terms of power consumption and achievable range is a big step.

The associated disadvantage is that the spectral efficiency is lowered because of the reservation of subcarriers with the only purpose of PAPR reduction. To utilize the result of the PAPR reduction in the cell range, the improved performance is used to increase the average power i.e. utilisation of the amplifier closer to the saturation point. The known method [Brandes, S., Cosovic, I., Schnell, M., "Sidelobe Suppression in OFDM Systems by Insertion of Cancellation Carriers"] to combat these effects is also a tone reservation, however the modulation of the reserved subcarriers is performed in such a manner that the spectral mask gets steeper and the power spill over into adjacent resource blocks is reduced by up to 20 dB . The procedure can be feedback based: using the input data, the resulting output spectrum is processed to derive the parameters for the modulation of the reserved tones. An alternative implementation is a statistical approach produc- ing modulation patterns which are reducing the constructive addition of the tones.

The associated disadvantage of this method is an increase of the PAPR. However, as long as the range is not a point, and spectral efficiency and protection of adjacent bands and re ^¬ source blocks is the objective, the method is very success- ful . The side lobe suppression shows best performance when the cancellation carriers are allocated at the chunks edges as the most outer tones.

Today methods are known to reduce PAPR or to reduce adjacent channel power by means of tone reservation, however exclu ^¬ sively for the one or the other problem, with the result that for each optimisation the position of the cancellation carriers should be different.

III. SUMMARY OF THE INVENTION

The Invention has been made to overcome the above discussed and other drawbacks and obstacles relating to the art.

This may include the problem of finding a method and appara ^¬ tus that allows for both a highly efficient out-of-band power suppression and a Peak-to-Average-Power Reduction.

A further object of the invention is to provide optimisation in the system design, by optimising the system in using a dynamic allocation of cancellation carriers.

According to one aspect of the invention, a method for coding data, in particular multiple access data, in a multiplexed transmitted signal is proposed, wherein the transmitted sig ^¬ nal comprises multiple subcarriers with orthogonal frequen ^¬ cies. The method favourably comprises the step of modulating the signals corresponding to a given subset of the subcarriers of the transmitted signal in order to encode information.

Moreover, the method may comprise the step of modulating the signal corresponding to least one reserved subcarrier depend ^¬ ing on at least one optimization criterion relating to a

quality of the overall transmitted signal. The quality of the signal may be determined depending on the PAPR or on the OBP.

In order to achieve a highly effective joint reduction of en- velope fluctuations (PAPR) and of 0ut-of-Band-Power, it is proposed that the optimization criterion depends on at least one first quantity relating to a peak-to-average-power ratio of the transmitted signal and on at least one second quantity relating to an out-of-band power of the transmitted signal.

According to a further aspect of the invention, the reserved subcarriers are distributed over the available bandwidth of the transmitted signal. In comparison to embodiments where the reserved subcarriers are concentrated on a part of the band, a more effective cancellation may be achieved.

If the subcarriers of the transmitted signal are comprised in a band of subcarriers, wherein at least one subcarrier at each edge of the band of subcarriers is a reserved subcar- rier, a particularly effective Out-of-Band-Power reduction can be achieved. This is due to the fact that the subcarriers on the band edges are especially effective in the sidelobe suppression. In particular if at least two neighbouring subcarriers at an edge of the band are reserved subcarriers, the Out-of-Band-Power (OBP) can be strongly reduced.

A high spectral efficiency can be achieved, if a Quadrature Phase Shift Keying (QPSK) is used for modulating the signals of the subset of subcarriers carrying the information. The QPSK allows transmitting two bits of data per subcarrier and per symbol.

According to a further aspect of the invention, it is proposed that optimization is performed under the constraint that the power allocated to the at least one reserved subcar ^¬ rier is limited to a first preset constant. This avoids a "runaway" of the power allocated to the reserved subcarriers in the optimization.

An explicit choice of the preset constant depending on the total number of subcarriers and/or on the average power allo ^¬ cated to the subcarriers may be avoided if the first preset constant corresponds to the number of reserved subcarriers multiplied with an average power allocated to one of the sub ^¬ carriers of the subset of subcarriers carrying the informa ^¬ tion .

According to a further aspect of the invention, the optimiza ^¬ tion is performed under the constraint that the power allo ^¬ cated one reserved subcarrier is limited to a second preset constant. This avoids a divergence of the optimization method.

An explicit choice of the preset constant depending on the total number of subcarriers and/or on the average power allo ^¬ cated to the subcarriers may be avoided if the second preset constant corresponds to two times an average power allocated to one of the subcarriers of the subset of subcarriers carry ^¬ ing the information.

The optimization can be achieved using numerically well tested algorithms, if the optimization is performed by find- ing an extremum of a weighted sum of the first quantity and the second quantity. Depending on the sign of the optimiza ^¬ tion function, the extremum may be a maximum or a minimum.

If in the weighted sum a weight of the first quantity and a weight of the second quantity are selected depending on at least one parameter relating to a current situation, the method may be easily adapted to the circumstances. If for ex ^¬ ample the guard interval between frequency bands of different systems is particularly large and at the same time, a power amplifier is of poor quality, the weight could be shifted to the first quantity relating to the Peak-to-Average-Power Reduction, since due to the large guard interval, the sidelobe suppression is less essential. On the other hand, if a high

quality power amplifier is applied and at the same time there are many other systems transmitting on neighbouring frequency bands, the weight could be shifted towards the second parame ^¬ ter relating to the Out-of-Band-Power reduction. Parameters describing the circumstances where the method is applied could e.g. be the maximum linear range of the power amplifier or the width of the guard interval.

In a further embodiment of the invention the optimization might also comprise varying the weight of the first quantity and the weight of the second quantity.

In the simplest form, the first quantity relating to the peak-to-average-power ratio is a maximum power of the signal to be transmitted. Alternatively, e.g. an average of several maxima of the transmitted signal might be used.

In a very simple and effective embodiment, the second quan ^¬ tity relating to the out-of-band power of the transmitted signal is determined dependent on a power of the signal to be transmitted in at least one optimization-point-frequency out ^¬ side a band of frequencies assigned to the subcarriers of the transmitted signal. A good Out-of-Band-Power reduction may be achieve, if the second quantity is determined dependent on an average power of at least two frequencies of the transmitted signal .

A further important aspect of the invention is based on the discovery of the inventors that the position of the cancella- tion carriers or reserved tones in the chunk of the available bandwidht is significantly influencing the performance. To achieve best results for the PAPR reduction, positions within the chunk are preferable since PAPR reduction maintains the level of spectral spill over constant by increased usable TX power. However, the PAPR reduction does in this case not result in a tighter spectral mask. The spectral mask can be maintained, but with the difference that a higher output power can be achieved compared to conventional terminals.

In operational scenarios, where range is the crucial issue and not the capacity, the spectral efficiency loss by tone reservation and necessary guard bands due to the PA caused spectral growth is secondary. The primary objective is to re ^¬ duce the PAPR and to enable higher sustained PA output power, not improving the spectral mask.

Another operational scenario is the high capacity scenario, where the range requirements are short, but the spectrum ef ^¬ ficiency shall be optimal. OFDMA in the UL shows here some system inherent effects. The OFDMA signal at the BTS receiver is compiled by a multitude of OFDM signals from various mo ^¬ bile stations. The mobile stations have an imperfect synchro- nisation. Even though they derive their time base from network, delay times, signal jitter, and the limitations of the local oscillators add an error to the signal. These errors of the individual OFDM signals are causing an imperfect demodu ^¬ lation in the BTS receiver, because the signal of adjacent subcarriers would not be perceived as zero at the detector, and a power spill from the adjacent subcarrier is reducing the SINR performance.

This effect is getting more critical when the terminals have different positions in the cell, and the powers of different OFDM signals seen at the receiver have significant power dif ^¬ ferences. In such cases, e.g. on random access channels, packet transmission re-activation, or sub-optimal power control, the power spill over may mask wanted signals. In order to overcome these problems, it is proposed to intro ^¬ duce a dynamic cancellation carrier allocation instead of a predefined allocation of the cancellation carriers or re ^¬ served tones. This Idea applies in particular to systems with OFDM(A) based UL. The dynamic cancellation carrier allocation allows an individual customisation of a terminal connected to a network by means of a base station.

Such a generic concept requires that terminals can support tone reservation for PAPR and/or sideband reduction. The research of the inventors has shown, that the use of can ^¬ cellation carriers aiming to combat both effects simultane- ously is performing better results, as dealing the effects individually or in a sequential process.

According to the needs an algorithm can optimise, i.e. trade off the effects, e.g. depending on the range the terminal is apart from the base station. As described above, this joint optimisation exceeds the solution to switch the cancellation carriers from one purpose to another.

However as previously mentioned, there is no ideal position of cancellation carriers within a chunk which suites best for both purposes. By means of a dedicated signalling, the net ^¬ work respectively the scheduler sent information to a dedi ^¬ cated terminal about the number and positions of cancellation carriers .

If for example the PAPR improvement is not interesting, be ^¬ cause the terminal is close to the base station, the sched ^¬ uler dedicates cancellation carriers to the band edges to achieve optimum side lobe suppression. In the opposite in range limited scenarios, the scheduler select positions of the cancellation carriers in the chunk which allows best back off reduction.

In the operational cases the system benefits from both, de- pending on the effect with dominant weight, the scheduler signals a combination of cancellation carrier positions on chunk edges and in the chunk. The signalling may be performed in the form of a map, addressing tone positions by a number ^¬ ing, and in a second dimension defining the max. usable power per tone.

Accordingly, the invention introduces a dedicated signalling between the network and individual terminal depending on the

link quality, capacity needs and terminal position. It is proposed to implement this by allocating the tones flexible in position and power share, resulting in a fine granularity for trading off the side lobe and power back off performance.

In other words, it is proposed to add the the step of deter ^¬ mining a position of the at least one reserved subcarrier within a bandwidth of the transmitted signal depending on an optimization criterion, which optimization criterion could be used for a separate optimization procedure or for a joint op ^¬ timization procedure including optimizing the signal of the cancellation carriers or reserved tones.

In particular, the position may be determined based on weight factors describing the weight of the first quantity and of the second quantity in the optimization criterion. The thus determined position of the at least one reserved subcarrier may be signalled to the reciever together with the transmit ^¬ ted signal.

Moreover, it is proposed to not only determine the position of the reserved subcarriers dynamically but to also determine a number of reserved subcarriers/cancellation carriers. The different requirements for long-range and high capacity sig- nailing may be taken into account if the latter parameters are determined based on at least a distance between a signal transmitter transmitting the transmitted signal (32) and a station receiving the transmitted signal.

According to a further aspect of the invention, a signal transmitter for transmitting and coding multiple access data in a multiplexed transmitted signal is proposed, wherein the transmitted signal comprises in particular multiple subcarri ^¬ ers with orthogonal frequencies. The signal transmitter may favourably comprise a modulating unit for modulating the sig ^¬ nals corresponding to a given subset of the subcarriers of the transmitted signal in order to encode information. The modulating unit is further suitable for modulating the signal

corresponding to least one reserved subcarrier depending on at least one optimization criterion relating to a quality of the overall transmitted signal.

According to one aspect of the invention, it is proposed that the signal transmitter is provided with an optimizing unit configured to optimize the overall transmitted signal using an optimization criterion depending on at least one first quantity relating to a peak-to-average-power ratio of the transmitted signal and on at least one second quantity relat ^¬ ing to an out-of-band power of the transmitted signal.

IV. BRIEF DESCRIPTION OF THE DRAWINGS

The characterizing features of the invention will become more apparent from the below description of the attached drawings. The drawings relate to a particular embodiment of the inven ^¬ tion which incorporates multiple features in combination. However, the invention is not limited to the combination of features shown in the embodiment. The skilled person will see the advantages of the features of the description, the draw ^¬ ings and the claims taken alone and in any other suitable combination .

Fig. 1 shows schematically an embodiment of a method for cod ^¬ ing multiple access data to be transmitted by use of a multi ^¬ ple access data transmitting system.

Fig. 2 is an illustration of the distribution of data carry- ing subcarriers and reserved subcarriers over an available bandwidth .

Fig. 3 is a graph showing the out-of-band-power suppression and a Peak-to-Average-Power-Reduction versus a trade-off pa- rameter.

Fig. 4 illustrates a trade-off curve achieved with joint op ^¬ timization and corresponding achievable points with separate optimization .

Fig. 5 shows schematically a signal transmitter for transmit ^¬ ting and coding multiple access data.

V. DETAILED DESCRIPTION OF THE EMBODIMENT

Fig. 1 shows schematically an embodiment of a method for cod ^¬ ing data to be transmitted by use of a multiple access data transmitting system. The method is suitable for coding data to be transmitted in a signa, wherein the transmitted signal comprises multiple subcarriers with orthogonal frequencies.

In a first step 10, information 12 is input and the signals corresponding to a given subset of the subcarriers 14 (Fig. 2) of the transmitted signal are modulated in order to encode the information 12.

The keying scheme applied in the first step is Quadrature Phase Shift Keying is used for modulating the signals of the subset of subcarriers 14 carrying the information 12.

In a second step 16, the signal corresponding to least a set of reserved subcarriers 18 is modulated depending on at least one optimization criterion relating to a quality of the overall transmitted signal. The overall signal is composed as a linear superposition of the signals corresponding to the information carrying subcarriers 14 and of the signals corre ^¬ sponding to the reserved subcarriers 18.

The optimization criterion depends on one first quantity re- lating to a peak-to-average-power ratio of the transmitted signal and on one second quantity relating to an out-of-band power of the transmitted signal.

Fig. 2 is an illustration of the distribution of data carrying subcarriers 14 and reserved subcarriers 18 over an avail ^¬ able bandwidth. Due to the symmetry, only the right half of the spectrum is depicted. Please keep in mind that the dif- ferentiation between PAPR reduction tones and out-of-band power only applies to the scenario with two separate optimi ^¬ zation steps. The subcarriers of the transmitted signal are comprised in a band of subcarriers, wherein neighbouring sub ^¬ carriers at an edge of the band are reserved subcarriers 18. The remaining reserved subcarriers 18 are distributed over the available bandwidth of the transmitted signal.

Instead of defining one set of cancellation carriers (at the edges of the spectrum) for the OBP reduction and one set of PAPR reduction carriers (distributed inside the spectrum) , according to one aspect of the invention it is proposed to take only one set of subcarriers 18, in the following called Reserved Tones (RTs) or reserved subcarriers 18, and modulate them in a way that both the out-of-band power and the PAPR are minimized simultaneously.

However, in order to enhance the comprehensibility, the mathematical background of the invention is described in the following using an example where only the BOP is minimized.

Let S _k (f) denote the spectrum of the k-th subcarrier, and

S(f) = ∑S _k (f) the spectrum of the total OFDM symbol. Our aim is then to determine the weights of the reserved subcarriers so that the expression J I S(f) I ² df is minimized, where the in- tegral goes over the frequencies outside the band occupied by our signal.

Since the evaluation of the whole integral is numerically in- feasible, a discrete set ^ of frequencies is selected, where the power is to be minimized. These points should be chosen to lie around the first few sidelobes left and right of the spectrum.

In the following, the variable v = (f — f _c )/Af denotes the nor ^¬ malized frequency, i.e. ^V = O j_ _s the DC frequency bin, and the data carriers are centered at v = ±l,±2,...

The objective function of our optimization problem relating to the sidelobe suppression only then becomes ^IS(V)I ² . veV

In order to derive an expression for ^^ ^v \ let us first con- sider a single unmodulated subcarrier. The discrete time sig ^¬ nal of the k-th tone can be written as

s _k (n)=—exp(2φk/N _c ), -N ≤n≤N _c -l _m

where N _cp i s the length of the cycl ic pref ix . The correspond- ing cont inuous frequency spectrum i s

S _k (v) = Y exp(2τrjnk I N _c ) exp(-2;z/ vnlN _c )

C n=-N _cp

1 N « _rc --1l ∑exp(-2τrj(v - k)n/N _c )

N _C n=-N _rπ

sin(Nx/ 2)

With the Dirichlet function D _N (x) = and N = N _r + N _rn

Nsin(x/2) denoting the total symbol duration, (2) can be written as

With (3) we can calculate the spectrum of the data carriers at the optimization points v eV which will be denoted as b. Furthermore, we can write the spectrum of the CCs at the op- timization points as the linear mapping AX where ^ is a

I V I xR matrix and X is, as above, the vector containing the weights of the reserved tones.

The out-of-band power (OBP) that we wish to minimize is then given as the squared Euclidian norm denoted by Il 11^

OBP(λX+b)=IIλX+bll ₂ . (4)

It suffices to minimize the objective function IIAX +bll ₂ .

If N _cp >0, and hence N _S >N _C , as is usually the case in OFDM systems, the subcarrier spectra become narrower as the symbol duration becomes longer. This means that the zeros of subcar ^¬ rier V ₀ are not at V ₀ +1,V ₀ ±2... any longer but at

V ₀ ±z,v ₀ ±2z...with z = N _c /N _s . In the receiver, the cyclic prefix is removed and orthogonality between the subcarriers is re ^¬ stored.

We mention this fact because when we want to reduce the out- of-band power, we have to differentiate between two different goals:

First, we might want to reduce the power of the transmitted signal within a certain frequency interval. This is the con ^¬ text in which Brandes et al . proposed the cancellation car- rier technique; their goal is to operate the OFDM system as an overlay system which uses a frequency range already occu ^¬ pied by narrowband legacy systems. In order to not interfere with these narrowband systems too heavily, the transmitted signal must contain as little power as possible within the respective frequency band.

We are interested in minimizing the self-interference caused by Doppler spread or bad synchronization. In this scenario, we want to minimize the out-of-band power of the received signal after CP removal.

The difference is that in the first case, we have to take the

CP into account when calculating the spectra, i.e. the pa ^¬ rameters A and b . In the second case, N _cp must be set to zero .

We will now turn to the method where the Out-of-Band-Power reduction and the PAPR is performed simultaneaously . While the term IIλX+bll ₂ may- as described above - be used as second quantity relating to an out-of-band power OBP of the trans- mitted signal, in the below example, the quantity Il CX+ d H ₀₀ is used as a first quantity describing the Peak-to-Average- Power Ratio PAPR of the signal. The function 11...H ₀₀ denotes the maximum component of the vector being the argument of the function. In other words, first quantity Il CX+ d H ₀₀ relating to the peak-to-average-power ratio is a maximum power of the signal to be transmitted.

According to the above example, second quantity HλX+bll ₂ re ^¬ lating to the out-of-band power of the transmitted signal is determined dependent on a power of the signal to be transmit ^¬ ted in several optimization-point-frequencies outside a band of frequencies assigned to the subcarriers of the transmitted signal. An average power of the optimization-point- frequencies is implicitly performed in the matrix multipli- cation.

Therefore, the problem may be formulated as to minimize the vector-valued objective function

By multiplying equation (5) with a two-dimensional weighting vector λ, we transform the objective function into a scalar valued one. Note that norms are convex functions and hence λ ^r f ₀ (X) is a convex function. Its minimum can therefore be found by standard optimization algorithms.

If we solve the unconstrained optimization problem, for some OFDM symbols the RTs or reserved subcarriers will be allo ^¬ cated much more power than the corresponding amount of data carriers. Since an increase of total transmit power is not an option in most systems, the power of the total symbol (in ^¬ cluding the RTs) has to be normalized, which means that the power on the data carriers is reduced if the RTs carry more power than the data carriers. Of course, this is only accept ^¬ able up to a certain degree since the BER increases with de- creasing data carrier power.

Furthermore, if each OFDM symbol is normalized with a differ ^¬ ent normalization factor, we can say that the amplitude of the effective channel varies from symbol to symbol. This is a problem if constellation schemes like QAM, where amplitude carries information, are used, because averaging the channel estimate in time direction is not possible anymore without severely degrading the MSE of the estimation.

We therefore introduce two constraints on X : ensures that the power of each RT does not exceed P _camer , which pre ^¬ vents spectral power peaks. The second constraint IIX|<ζ, _f( lim ^¬ its the total amount of power that is allocated to the RTs.

In other words, the optimization is performed under the constraint that the power allocated to the total number of re ^¬ served subcarriers is limited to a first preset constant P _totcd ' wherein first preset constant P _total corresponds to the number of reserved subcarriers multiplied with an average power allocated to one of the subcarriers of the subset of subcarriers carrying the information.

As a second constraint, the optimization is performed such that that the power allocated one reserved subcarrier is lim- ited to a second preset constant P _carπer . According to one em ^¬ bodiment, second preset constant P _camer corresponds to two times an average power allocated to one of the subcarriers of the subset of subcarriers carrying the information.

The optimization problem can now be formulated as follows:

minimize subject to ⁽⁶⁾ llXII^≤iU,

We stress that the constraint functions are also convex, and hence (6) can be solved by convex optimization algorithms. The optimization is performed by finding an extremum of a weighted sum of the first quantity IICX+dll _∞ relating to the peak-to-average-power ratio and of the second quantity IIλX+bll ₂ relating to the OBP.

The weighting vector is chosen as

With the trade-off parameter μ e [θ,l] we can smoothly adjust the relative weighting of the two optimization criteria: μ close to zero yields a strong PAPR reduction, while μ close to one emphasizes the out-of-band power reduction.

In particular, the weight Kμ of the first quantity and a weight l-μ of the second quantity may be selected depending on at least one parameter relating to a current situation by simply determining the parameter μ in a suitable way.

The constant K =11 d H ₀₀ / Il b H ₂ is strictly speaking not necessary; its only purpose is to scale the μ-axis such that that for the two optimization criteria are approximately equally weighted.

Fig. 3 is a graph showing a first curve 20 describing the

Out-of-Band-Power suppression and a second curve 22 describing the Peak-to-Average-Power-Reduction versus a trade-off

parameter. The graph of Fig. 3 is the result of a numerical simulation which has shown that the joint optimization indeed promises a better performance compared to two separate opti ^¬ mization steps, while providing the system designer with a higher degree of flexibility.

The simulation parameters resulting in the improvements shown in Fig. 3 are as follows:

- FFT-length N ^c = 64 - No cyclic prefix;

- 50 modulated subcarriers on normalized frequencies [±1,±2,...,±25]

- QPSK modulation

- Reserved Tones on carriers L— '— ^■> — ^■> — J . in the main simulation, these carriers are used jointly for both

PAPR reduction and out-of-band power reduction. In the simulations provided for comparison where two separate optimization steps are performed, the carriers L- ^U '- ^υ J are used for PAPR reduction, and the carriers [±24,±25J are used for out-of-band power reduction. The positions of the PAPR reduction carriers are chosen arbitrarily here and may in general be chosen in any suitable way depending on the specific circumstances

- The PAPR is evaluated using a 4-times oversampled time domain signal,

- The out-of-band power is evaluated as JF25 5<\v\<N _c / 2\S(v)\ ² dv.

Note that the spectrum of the discrete-time signal is periodic with period length V ₀ = N _c = 64

Under the assumption that the data carriers are normalized to an average power of 1, we set the constraint constants to P _carner =2 and P _total = R = 8. The first constraint ensures that the reserved tones are not more than 3dB stronger than the data carriers, and the second constraint makes sure that on aver- age, the RTs are not allocated more power than the data car ^¬ riers, which makes a power normalization unnecessary. Accord-

ingly, in the comparison simulation where two separate optimization steps are performed, P _total is set to 4 in both steps.

In the example of Fig. 3, both PAPR reduction and out-of-band power reduction are evaluated by comparison with a reference signal with 50 data carriers. We can observe the smooth trade-off that is achievable by varying μ.

Fig. 4 illustrates a trade-off curve 24 achieved with joint optimization and corresponding achievable points 26, 28 with separate optimization. The reduction achieved by separate op ^¬ timization steps is shown for comparison. The only degree of freedom we have in the latter case of separate optimization steps is the order of the two steps. We see that the "more important" optimization should be performed in the second step, because then the respective algorithms can take the re ^¬ served carriers from the first optimization step into ac ^¬ count. We can furthermore observe that the reduction that can be achieved by separate optimization is far worse than the results of the joint optimization according to the invention.

Fig. 5 shows schematically a signal transmitter 30 for trans ^¬ mitting and coding data in a transmitted signal 32, wherein the transmitted signal 32 comprises multiple subcarriers with orthogonal frequencies (cf. Fig. 2). The signal transmitter 30 may be included in a mobile phone or e.g. in a broadcast system.

The signal transmitter 30 comprises a modulating unit 34 for modulating the signals corresponding to a given subset of the subcarriers of the transmitted signal in order to encode in ^¬ formation and for modulating the signal corresponding to least one reserved subcarrier depending on at least one opti ^¬ mization criterion relating to a quality of the overall transmitted signal 32.

The signal transmitter 30 further comprises an optimizing unit 36 configured to optimize the overall transmitted signal

32 using an optimization criterion depending on at least one first quantity relating to a peak-to-average-power ratio of the transmitted signal 32 and on at least one second quantity relating to an out-of-band power of the transmitted signal 32.

According to further embodiments of the invention, the opti ^¬ mising unit 36 further varies the number and positions of the reserved subcarriers 18 dynamically depending on the instan- taneous requirements, in particular depending on a distance between the sender and the receiver, on the link quality and on the capacity needs .

Moreover, the power share between the subcarriers may be de- termined by the optimising unit and the maximum usable power per reserved tone may be signalled together with the posi ^¬ tions of the reserved tones to the downlink components of the network .

The optimising unit 36 may further be configured to inform a scheduler of the network of the positions and of the number of reserves subcarriers. The optimization may be performed independently for a first group of reserved subcarriers near the edges of the available bandwidth in in order to reduce frequenca spill over and for a second group of reserved sub ^¬ carriers distributed in the chunk of the signal bandwidth.

It is noted that the idea of the dynamical allocation of the reserved subcarriers is not limited to the joint optimization with regard to PAPR reduction and sidelobe suppression but could also be used to optimise the signal with regard to only one of these criteria.

Moreover it is noted that the optimization could also be per- formed by only determining the number of reserved subcarriers, whereas the positions are predetermined given the number of reserved subcarriers. Accordingly, determining whether or not a particular subcarrier is a reserved subcarrier or not

implies the determination of the position of this particular reserved subcarrier.