WO2010138032A1

WO2010138032A1 - Papr reduction by tone selection

Info

Publication number: WO2010138032A1
Application number: PCT/SE2009/050597
Authority: WO
Inventors: Robert Baldemair; Ali Behravan
Original assignee: Telefonaktiebolaget LM Ericsson AB
Current assignee: Telefonaktiebolaget LM Ericsson AB
Priority date: 2009-05-26
Filing date: 2009-05-26
Publication date: 2010-12-02
Anticipated expiration: 2011-11-26

Abstract

A method and an arrangement (1100) are provided in a communication node for enabling reduction of peaks (302) in transmitted signals. In the method, a sample representing a peak in a signal is selected (202), and a tone frequency, an amplitude and a phase shift are determined (204-208) and are used to create (210) a correction signal which is added (214) to the original signal if it is found (212) to reduce the peak. The procedure is repeated from the operation of selecting a sample representing a peak as long as a stop criterion is not fulfilled (218).

Description

PAPR REDUCTION BY TONE SELECTION

TECHNICAL FIELD

The invention relates to a method and an arrangement in a communication node, in particular to enable the reduction of the peak to average power ratio of transmitted signals using correction signals .

BACKGROUND When a transmit signal comprises power peaks, which are large in relation to the average power of the transmit signal, the signal is said to have a high peak-to-average power ratio (PAPR) . A high PAPR causes either non-efficient use of high power amplifiers when operated in their linear region, or high non- linearity distortion, when operated close to their saturation, i.e. results in degraded performance.

A high PAPR of transmitted signals is a problem, which is typically associated with multicarrier systems, such as OFDM (Orthogonal Frequency Division Multiplexing) based systems. OFDM has been adopted for many high speed wireless data communication standards, such as e-UTRA (Evolved Universal Terrestrial Radio Access) , also referred to as LTE, WiMAX (Worldwide interoperability for microwave access) , DAB (Digital Audio Broadcasting) and DVB (Digital Video Broadcasting) and for wired data communication standards, such as DMT (Discrete MultiTone modulation) .

The simplest and most commonly used method to reduce the PAPR is clipping a signal 100 at a certain power level in the baseband section, which is illustrated in figure 1. However, clipping reduces the signal dynamics, i.e. the PAPR, but also causes an increase in the EVM (Error Vector Magnitude) and the ACPR (Adjacent Channel Power Ratio), i.e. causes an uncontrolled in-band and out-of-band distortion. The received signal must satisfy a certain EVM requirement. Furthermore, after the clipping is performed, sharp filtering is needed in order to satisfy the spectrum masks, i.e. to keep the signal within allowed limits concerning frequency and power. Filters with fast decaying in the transition bands introduce a delay, which may be unacceptably long for many systems . Filtering further consumes part of the cyclic prefix, which is used in OFDM to avoid inter-symbol interference. Thereby, the part of the cyclic prefix which is available for delay spread of the propagation channel is reduced.

Other methods, such as SLM (Selected Mapping) and PTS (Partial Transmit Sequence) require transmission of extra information, also called "side information", to the receiver, and a change in the receiver algorithm is also required for de-mapping the received data.

Further, there are methods based on TI (Tone Injection), and ACE (Active Constellation Extention) , which require knowledge of the de-mapping scheme at the receiver.

PAPR-reduction based on TR (Tone Reservation) either requires solving a complicated optimisation problem, or finding all the correcting subcarriers at once, which may cause peak regeneration .

SUMMARY It would be desirable to obtain an improved procedure for peak-reduction in order to improve amplifier efficiency or reduce amplifier-related distortion. It is an object of the invention to address at least some of the issues outlined above. Further, it is an object of the invention to provide a procedure for enabling peak-reduction in single- and multi- standard radio systems. These objects may be met by a method and apparatus according to the attached independent claims.

According to one aspect, a method is provided in a communication node for enabling reduction of peaks in transmitted signals. In the method, a plurality of discrete time samples is obtained by sampling a signal, which is to be transmitted. One of the samples is then selected to represent a peak in the sampled signal, which it is desirable to reduce. Further, a tone frequency is selected from amongst a set of tone frequencies, which are available for use as correction tones, and an amplitude is determined for the selected tone frequency, such that the determined amplitude satisfies a predefined power limit. A phase shift for the selected tone is also determined, such that the selected tone with the determined phase shift will reduce the peak in the signal, when they are used for creating a correction signal. The correction signal is created based on the selected tone, the determined amplitude and the determined phase shift. Then, a comparison is performed, where the maximum magnitude of the original signal comprising the peak is compared to the maximum magnitude of the combined signal which is formed when the correction signal is added to the signal comprising the peak. If the maximum magnitude of the original signal is the largest, a new "original" signal is set to equal the combined signal, and the selected tone frequency is removed from the set of available tone frequencies. If the compared signal magnitudes are equal or the maximum magnitude of the combined signal is the largest, which means that a new peak has been created, the created correction signal is not used and the selected tone frequency is removed from the set of available tone frequencies. As long as a predetermined stop criterion is not fulfilled, the method will be iterated from the operation of selecting a sample representing a peak. If the predetermined stop criterion is fulfilled, the procedure will end with the last combined signal as output.

In another aspect, an arrangement in a communication node is adapted to enable the reduction of a peak of a transmit signal. The arrangement comprises a sampling unit, which is adapted to obtain a plurality of discrete time samples by sampling a signal, which is to be transmitted. The arrangement further comprises a peak finding unit, which is adapted to select one of the samples, which represents a peak in the sampled signal. The arrangement further comprises a frequency selecting unit, which is adapted to determine which tone frequency to select from a set of tone frequencies, which are available for use as correction tones. The arrangement further comprises an amplitude determining unit, which is adapted for determining an amplitude level for the selected tone k that satisfies a predefined power limit, and a phase shift determining unit, which is adapted to determine a phase shift for a selected tone, so that the selected tone with the determined phase shift will, when used for creating a correction signal, reduce the peak in the signal. The arrangement further comprises a correction signal creating unit, which is adapted to create a correction signal based on the selected tone, the determined phase shift and the determined amplitude. The arrangement further comprises a signal comparison unit, which is adapted to compare the maximum magnitude of the original signal comprising the peak to the maximum magnitude of the combined signal which is formed when the correction signal is added to the signal comprising the peak. The arrangement further comprises a signal setting unit, which is adapted to set a new "original" signal to equal the combined signal if the maximum magnitude of the original signal is the largest. The arrangement further comprises a tone removing unit, which is adapted to remove the selected tone frequency from the set of available tone frequencies when the new signal has been set or when the compared signal magnitudes are equal or the maximum magnitude of the combined signal is the largest, which means that a new peak has been created. The arrangement further comprises an iteration unit, which is adapted to repeat the method from the operation of selecting a sample representing a peak, if a predefined stop criterion is not fulfilled, and otherwise to output the last combined signal.

Different embodiments are possible in the method and arrangement above, of which some example embodiments will be explained in more detail in the detailed description.

The proposed peak-reducing procedure is mainly targeted at OFDM signals, but it can be used for any kind of signal, i.e. for single carrier signals as well as for multicarrier signals, as no property of multicarrier signals is used to find the correcting tones. The procedure can be used in single-standard radio systems or in multi- standard radio systems, in which the signal comprising a peak to be reduced is generated by a radio access technology such as GSM, WCDMA, LTE, or WiMAX or any combination of these radio access technologies .

The proposed procedure may provide optimisation of both tone frequency and phase shift, which allows for the best PAPR- reduction possible with any potential constraints applied to the frequency of the tones or their phase shift.

Further, the procedure has low computational complexity. For example, no FFT operation is needed in finding the correction tones. With a constraint applied to the phase of the correction symbols, the optimum frequency of the tone can be found by a simple comparison of only R real values, and in case of a constraint applied to the frequency of the tone, the optimum phase is found using a simple formula.

Since the tones are added one by one in the proposed procedure, the PAPR of the resulting signal samples can be checked for each correcting tone candidate. This control results in that the tones which are found to create new peaks are discarded, wherefore no peak regeneration will occur.

Further, no knowledge of the transmitter algorithm is needed at the receiver when TI or TR outside the signal bandwidth is used. Therefore, the algorithm may be implemented only at the transmitter side of a communication link, while the receiver side does not need to be changed in any way

BRIEF DESCRIPTION OF THE DRAWINGS

The invention will now be explained in more detail by means of example embodiments and with reference to the accompanying drawings, in which:

-Figure 1 illustrates a signal of which the peaks have been reduced by clipping according to prior art. -Figure 2 is a flow chart illustrating a procedure for peak- reduction according to one example embodiment. -Figure 3 illustrates a transmit signal.

-Figure 4 illustrates a multi -standard system, in which the described embodiments may be used. -Figure 5 illustrates allowed areas for correction tones according to one example embodiment .

-Figure 6 illustrates allowed areas for constellation extension according to one example embodiment. -Figure 7 is a table containing simulation parameters for one example embodiment .

-Figures 8-10 illustrate different simulation results with and without TR (Tone Reservation), i.e. with and without peak- reduction.

-Figure 11 is a schematic view illustrating an arrangement in a communication node according to one example embodiment.

DETAILED DESCRIPTION The invention can be used to enable reduction of the PAPR of a signal. The PAPR can be reduced by using a procedure with a relatively low computational complexity, which only needs to be implemented in the transmitter. The procedure does not require any side information to be sent to the receiver, nor does the receiver need to have any "knowledge" of the procedure. Further, the procedure does not regenerate peaks.

Figure 3 illustrates a signal s 300 with a high PAPR. The peak 302 illustrates a peak with a high power magnitude, which may be desirable to reduce in order to improve the amplifier efficiency or reduce distortion. The signal may for example have been generated in a wireless multicarrier system, such as an LTE system; a wired multi-carrier system, such as a DMT system, or in a multi-standard system. If the signal has been generated in a multi- standard system, it may be a combination of signals generated by different multi-carrier or single- carrier radio access technologies (RATs) . An example multi- standard system involving 3 RATs, in which one embodiment of the described peak-reducing procedure is applied, is illustrated in figure 4. The different RATs can be any multi- carrier or single-carrier RATs, such as WiMAX, LTE, WCDMA, MC- CDMA, CDMA, SC-FDMA, GSM, etc. The baseband signals from different RATs are created and upsampled to a common sampling rate. The signals are then combined, and in order to reduce the PAPR of the combined signal, correction tones are added at the common radio part .

Hereafter, the term "tone" will be used as indicating a signal that is describes as A_k Qxp(j2πf_kt+φ_k) , where A_k is the amplitude of the tone, f_k is the frequency of the tone, and ^. is the phase (or phase shift) of the tone. This signal is normally considered within a limited time interval, O≤t≤T. Also since the operations are done in a discrete-time domain, a tone can

be represented as A_k exp\j lφ_k , where n is the index in

discrete time domain and k is the index in the discrete frequency domain. The expression "tone index" is also used to refer to any of these indices in the discrete frequency domain.

Below, one possible embodiment of the peak-reducing procedure will be described in detail. A generalised embodiment is illustrated in figure 2, which will be described later. Initially, a signal s, such as an N-subcarrier OFDM signal with a high PAPR, is sampled at t=nT/L in order to produce discrete time samples. The nth sample of the signal is described by

s[n] =-_j=∑S_ke ^LN ; n = 0,l,L,LN-l (1)

where Sk; k=0 ,..., K-I is the set of data subcarriers, T is the symbol duration, l/T is the symbol rate and L is the oversampling factor, i.e. LN is the number of signal samples. A set of R possible correcting tone positions can be defined as I^J?=[k₀, ..., k_R-_λ] . The nth sample of the correcting signal can then be defined as:

where C_k=A_k exp (jφ_k) is the complex-valued correcting symbol. The phase shift φ_k can take any value in the range of [0,2π] . In order to reduce a peak as much as possible, it is desirable that the amplitude A_k is as large as possible. Therefore, the amplitude A_k is determined/specified by the maximum allowed power for the kth correction tone, which is specified by the EVM for the in-band tones, see figure 5, and by the spectral mask for the out-of-band tones. Thereby, no further optimisation of the amplitude A_k needs to be done. The amplitude A_k could optionally be set to some smaller value, within the allowed power limit. The operation of determining the amplitude A_k may also involve to set the amplitude A_k to a predetermined value or to verify that an amplitude A_k, which was set during a previous iteration, is valid.

When the correcting signal c[n] is added to the original signal s[n] , the resulting signal can be written as:

Y[n] = s[n] +c[n], ⁽³⁾

and PAPR-reduction may be achieved by finding a correction signal c[n], such that

is minimised, i.e. by finding a correction signal c[n] , which causes the greatest possible reduction of the PAPR of the signal s. Therefore, the correcting symbols C_k=A_k exp ijφ_k) , which causes the greatest possible reduction of the PAPR of the signal s, should be found. In order to locate the largest peak of the signal s [n] , the sample s [n₀] , having the largest instantaneous power, is identified and selected to represent the peak. Alternatively, other samples could be selected to represent the peak, e.g. samples in a limited interval placed around the sample (s) having the largest instantaneous power, or samples depending on this sample in some other way.

In order to reduce said large peak, the frequency and the phase shift of the tone, which minimises the resulting instantaneous power at n₀ should be determined. The optimal tone index k^* and the optimal phase shift φ_k ^*, where "optimal" implies "the one that minimises the resulting instantaneous power at n₀" , can be determined mathematically by solving

k , φ_k A^e _ j l<ψPkk ( 5 )

The frequency of the tone or tone index k and the corresponding phase shift φ_k may be selected from the following predefined sets of frequencies and phase shifts

( 6 ) k e i^R = {£₀,L A-J ; φ_k ≡ [o,2π] -

In one embodiment, the values which the tone frequency or tone index k can take are restricted to a certain, predefined set. One example of such a case is tone-reservation, where only certain tone frequencies are available for PAPR-reduction use. Similarly, in tone-inj ection with a given target EVM the set of available tone indices is restricted, while there is no restriction on the phase shift of each tone. In figure 5, the possible correcting vectors, i.e. the correcting tones which are allowed in order to satisfy the target EVM requirement, are illustrated as small areas around the constellation points of a 16QAM constellation. The target EVM is illustrated as the dashed circumferences around the areas. It can also be seen from the figure that for a given tone index k, the phase shift φjc of the correcting tones can take any value in the interval [0,2π] .

With no constraint applied to the phase shift of the correction tones, the optimum phase shift for tone index k can be determined such that the correction signal c [n] has an opposite phase relative to the signal s [n] at n₀,

2πn_ok

In another embodiment, the values which the phase shift of the tones can take are restricted by a predefined condition. One example of such a case is active constellation extension for OFDM symbols, where the phase shift φ_k has to be within a confined set or range. This will be described further below. For a constant phase constraint, the optimum tone is the tone that has the opposite phase relative to the signal phase at n₀.

Here, the operation of finding the optimal tone index k^* is reduced to a comparison between R real values for a given phase shift φ_k. This means that for all available R tone indices, the expression according to the argument in equation (8) should be found, and the tone index that minimises this argument should be used for tone reduction. Figure 6 illustrates some possible phase values of the correcting vector A_k exp(j'φ^) in a 16 QAM constellation when active constellation extension is used for OFDM symbols. The possible phase values for the correcting vector are shown as dashed lines and shaded areas. In this example, if the kth tone is used, the phase shift can only be O₁ while for tone number k+\ , the phase shift has to be -π /2 in order to maintain good performance. The reason for this is that by allowing other phase shifts, the constellation points would be moved closer to each other, which would make signal detection more difficult at the receiver, since a detected constellation point could thereby more easily be mistaken for an adjacent constellation point.

When the index of the first tone, k, and its corresponding phase shift, φ_k, have been determined, with or without restrictions to the correction tone candidates, new signal samples are formed as

The PAPR of these new signal samples -? [n] may be lower than for the signal samples s [n] , although, it could also be the same or higher due to peak regeneration. This means that even if the large peak, which was identified in s [n] , is reduced in s [n] , a new, possibly even larger peak could have been created in another part of the signal s[n] .

In order to avoid peak regeneration, a test could be performed, which determines whether

i.e. it is determined if a new peak, being larger than the previously identified peak, has been created when the correction signal c [n] has been added to the signal s [n] . When the condition (10) above is not met, that indicates that such a new peak has been created.

If this condition (10) is not met, the tone k should not be used in the current step. If the condition (10) is met, s [n] is set to if [n] , i.e. s [n] = s [n] . In either case, the tone k is removed from the set of available tones, and the procedure is repeated from the step of finding the largest peak of the signal s [n] , if a predefined stop criterion is not fulfilled. If the tone k is not used due to peak regeneration, the procedure can instead be repeated from the operation of determining the optimal tone, if the stop criterion is not fulfilled. When the stop criterion is fulfilled, the procedure will end with the output s [n] .

In the claims and in the description of figure 2, the intermediate sum .τ[n]is not used, and the time discrete signals are written as S_n and C_n.

Figure 2 is a flow chart, which illustrates the procedure steps of one embodiment of using the above described peak- reducing procedure. Initially, a signal s is sampled in step 200. In step 202, a sample representing a peak in the sampled signal is selected. Thereafter, in order to reduce the peak, a frequency index k, a phase shift φ_k for the frequency index k, and an amplitude A_k are determined in the steps 204-206. These determined values are then used in order to create a correction signal C_n in step 210. As described earlier, the steps 204 and 208 can be performed using different formulas depending on if there are restrictions associated with the tones or their phase shifts, or not. Further, the steps 204 and 208 can in some cases be performed in the opposite order. The amplitude can also be determined in other positions than the one illustrated in figure 2.

In order to avoid regeneration of peaks when adding a correction signal C_n to the signal s_n, a comparison test is performed in step 212, where it is determined if the corrected/combined signal s_n+c_n comprises peaks with a magnitude which is larger than the maximum magnitude of the signal S_n If the maximum magnitude of s_n+c_n is smaller than that of s_n, a new signal S_n is set to equal the combined signal s_n+c_n in step 214, i.e. S_n = s_n+c_n. If the maximum magnitude of s_n+c_n is larger than or equal to that of S_n, the correction signal C_n is not used. In either case, the frequency k is removed from the set of available frequencies in step 216. In step 218 it is determined if a stop criterion is fulfilled or not. The stop criterion could for example be one or more of: that all tones available for use as correction tones are used; that some maximum allowed complexity or delay is reached; or that the Peak to Average Power Ratio of the signal S_n has a value which is below a predefined threshold. If the stop criterion is not fulfilled, the procedure is repeated from the operation of selecting a sample representing a peak. It is also possible to have an optional procedure step 220, which determines if it is the same peak which is to be reduced as during the preceding iteration. If the peak to be corrected is the same in step 220 as it was in step 202, the procedure steps could be repeated from step 204 instead of from step 202, since no new sample needs to be selected. The procedure steps could also be repeated from step 204 based on the knowledge that a direct transition from step 212 to step 216 has been performed, i.e. when the signal S_n has not been corrected. If the stop criterion is fulfilled, the procedure is completed and will be ended with the output S_n. One embodiment of the peak-reducing procedure presented above has been applied in a simulated PAPR-reduction in an LTE-DL with a bandwidth of 10 MHz. Since there are no available subcarriers for tone-reservation (TR) within the signal bandwidth in LTE, the correcting tones are placed in the guard band. In this case, the bandwidth of the signal is 9 MHz, and a set of R tones are placed on the two sides of the signal bandwidth. This is a special case of TR, where the correcting tones are localised on the edge of the signal bandwidth. The parameters of the simulated system are displayed in figure 7.

Figure 8 shows a CCDF (Complementary Cumulative Distribution Function) of the PAPR of a signal, with and without correcting tones, for the simulated system with the parameters described in figure 7. As shown in the figure 8, the PAPR of the signal is reduced by almost 2 dB at a PAPR probability of 10^"3, when using a single resource block, i.e. 12 subcarriers, on each side of the signal bandwidth as correcting tones.

Figure 9 shows a CCDF of the instantaneous power of a signal with and without the correcting tones, for the simulated system with the parameters described in figure 7. As mentioned earlier, here the correcting tones are placed at the edge of the signal bandwidth, and thereby, the gain in PAPR-reduction comes at the price of extra correcting tone power at the edge of the signal bandwidth. High power at the edge of the bandwidth is undesired, since it can cause a high out-of-band radiation. However, the power of the correcting tones can be controlled in the above described peak-reducing algorithm by choosing a proper value for A_k in equation (9) . In the simulation, the correcting tones and the data tones are assumed to have equal power. Figure 10 shows the PSD (Power Spectral Density) of the signal with and without correcting tones, for the simulated system with the parameters described in figure 7. The out-of-band radiation is substantially the same for the two cases, and only a small portion of the channel bandwidth is used for PAPR-reduction.

Figure 11 illustrates an arrangement 1100, which is adapted to enable use of one embodiment of the peak-reducing procedure, and which comprises a set of logical units. A sampling unit 1102 samples a signal s. A peak finding unit 1104 selects a sample representing a peak, which is desirable to reduce. A frequency selecting unit 1106 determines which tone frequency k to select amongst a set of possible tones, in order to reduce the peak. An amplitude determining unit 1108 determines an amplitude A_k for the tone k, which amplitude should be kept within certain power limits. A phase shift determining unit 1110 determines a phase shift of the selected tone k, so that the tone k with the phase φ_k will reduce the peak in the sampled signal S_n, when used for creating a correction signal. The frequency selection and the phase shift determination could also in some cases be performed in the opposite order. The amplitude determination may also be performed in other positions than the one illustrated in figure 11. Thereafter, a correction signal c_n is created in the correction signal creating unit 1112, based on the determined frequency k, amplitude A_k and phase shift φ_k. A signal comparison unit 1114 determines if the correction signal C_n regenerates peaks when added to the signal S_n. If the correction signal does not regenerate peaks, a signal setting unit 1116 sets a new signal to equal the corrected signal. Thereafter, the selected tone k is removed from the set of possible tones by a tone removing unit 1118. The iteration unit 1120 determines whether a stop criterion is fulfilled or not, and either repeats the procedure from step 1104, or ends the procedure with the output S_n. The procedure could optionally be repeated from step 1106 if it is the same peak which is to be reduced as during the last iteration, or, if the phase shift should be determined before the tone k, the procedure could be repeated from the step 1110, which then would be placed before step 1106.

It should be noted that figure 11 merely illustrates various functional units in the arrangement 1100 in a logical sense. However, the skilled person is free to implement these functions in practice using any suitable software and hardware means. Thus, the invention is generally not limited to the shown structure of the arrangement 1100.

The performance improvement accomplished by applying the described peak-reducing procedure depends on the number of correction tones as well as their position. The procedure is potentially more effective when using spread correcting tones as compared to when using localised tones, since adjacent tones have substantially the same characteristics, which could be unsuitable for peak-reduction. In the simulations, the tones are considered at the edge of the band, as mentioned previously. The optimal frequencies and phase shifts of a set of tones used for PAPR-reduction of a signal, i.e. the frequencies and phase shifts which maximises the PAPR-reduction, can be found using the described peak-reducing procedure.

In many communication standards, there are constraints on the received signal quality (EVM) , and the out-of-band emission (spectrum masks) . Both of these requirements can be incorporated in the described procedure as the amplitude of the added tones, i.e. the requirements can be fulfilled by determining the amplitude in accordance with these criteria. This is especially useful when the described procedure is applied in a multi- standard communication node, where the different RATs have different spectrum mask and EVM requirements .

The peak-reducing procedure is mainly targeted at OFDM signals, as in the example above, but it can be used for any kind of signal, i.e. for single carrier signals as well as for multicarrier signals, as no property of multicarrier signals is used to find the correcting tones.

As outlined above, the proposed procedure may provide optimisation of both tone frequency and phase shift, which allows for the best PAPR-reduction possible with any potential constraints applied to the frequency of the tones or their phase shift.

Further, no knowledge of the described procedure is needed at the receiver when TI or TR outside the signal bandwidth is used. Therefore, the procedure may be implemented only at the transmitter side of a communication link, while the receiver side does not need to be changed in any way.

The described procedure can be applied to find the correcting tones in both TR and TI methods. In standards such as LTE, where there are no free tones available within the signal bandwidth, TI within the signal bandwidth or TR outside the signal bandwidth can be used. In WiMAX, there are extra tones available within the bandwidth of the signal, and therefore, both TR and TI can be used across the entire frequency axis, as long as the spectrum mask is not violated.

The node(s) in which the described procedure could be applied could for example be an eNB, a repeater node, a terminal node, in a single- or multi-standard radio system, or the corresponding nodes in a wired system.

While the invention has been described with reference to specific example embodiments, the description is in general only intended to illustrate the inventive concept and should not be taken as limiting the scope of the invention. Although the description has mainly described peak reduction of OFDM signals, the scope of the invention is not limited hereto. The invention is generally defined by the following independent claims .

Claims

1. Method in a communication node for enabling reduction of peaks in transmitted signals in a communication system, where the method comprises the following steps:

-obtaining (200) a plurality of discrete time samples by sampling a signal s (300) , to be transmitted, forming a sampled signal S_n,

-selecting (202) a sample s [n₀] representing a peak (302) in the signal s, amongst the discrete time samples s [n] , -determining (204) which tone frequency k to select from a set of tone frequencies available for use as correction tones, and

-determining (206) the amplitude A_k of the tone k, such that the determined amplitude A_k satisfies a predefined power limit,

-determining (208) a phase shift φ_k for the tone k, so that the tone k with the phase φ_k will, when used for creating a correction signal, reduce the peak in the signal s_n

-creating (210) a correction signal C_n based on the tone k, the amplitude^, and the phase ψ_k,

-comparing (212) the maximum magnitude of the combined signal s_n+c_n with the maximum magnitude of the signal S_n, and

- i f max

\^a ; a > 0 , sett ing ( 214 ) a new signal S_n to the combined signal s_n+c_n , i . e .

^Sn = ^Sn ^{+ C}n . -removing (216) the tone frequency k from the set of available tone frequencies, and if a predetermined stop criterion is not fulfilled (218),

-repeating the method steps from the step of selecting a sample s [n₀] representing a peak, otherwise outputting the signal S_n

2. The method according to claim 1, wherein k and φ_k are determined using the formula

2πn_ak k , φ_k } = arg min s[n,] + - LN A_ke jψk k,ψk

wherein LiV is the number of signal samples, s [n₀] is the sample representing the peak, A_k is the amplitude of the correction signal, and A:^" and φ^" indicate optimal values of the tone k and the phase shift φ_k, respectively.

3. The method according to claim 1 or 2 , wherein a constraint is applied to which tone frequencies that can be selected as tone index k and/or a constraint is applied to which values the phase shift φ_k of the tone k can take.

4. The method according to any of the claims 1-3, wherein a constraint is applied to which frequencies that can be selected as tone k, and wherein the phase φ_k is determined using the formula

OJ _LN

wherein the phase shift φ_k can take any value in the range[θ,2;r] .

5. The method according to any of the claims 1-4, wherein a constraint is applied to which values the phase shift φ_k of the tone k can take, and wherein the tone k is determined using the formula

6. The method according to any of the claims 1-5, wherein the correction signal C_n is defined as

Iπnk

where C_k = A_k exp(y<%) ; n = 0,1,...,LN-I ; and LN is the number of signal samples.

7. The method according to any of the preceding claims, wherein the amplitude A_k is defined to be the maximum allowed magnitude for the correction tone k.

8. The method according to any of the preceding claims, where the predefined power limit of the correction signal is derived from an EVM value or from a spectrum mask.

9. The method according to any of the preceding claims, wherein the amplitude A_k of the tone is determined from an EVM criterion if an in-band tone is used, and from a spectral mask criterion if an out-of-band tone is used.

10. The method according to any of the preceding claims, wherein the stop criterion is one or a combination of the following : -all tones available for use as correction tones are used; or

-some maximum allowed complexity or delay is reached; -the Peak to Average Power Ratio of the signal S_n is lower than a predefined threshold.

11. The method according to any of the preceding claims, wherein the method is used within the concept of TR and/or TI.

12. The method according to any of the preceding claims, wherein the signal s is generated by a radio access technology such as GSM, WCDMA, LTE, or WiMAX, or any combination of these radio access technologies in a multi- standard radio system.

13. Arrangement in a communication node (1100), which is adapted to enabling the reduction of a peak of a transmit signal, said arrangement comprising:

-a sampling unit (1102) adapted to obtain a plurality of discrete time samples by sampling a signal s, to be transmitted, forming a sampled signal S_n, -a peak finding unit (1104) adapted to select a sample s [n₀] representing a peak (302) in the signal s, amongst the discrete time samples s [n] ,

-a frequency selecting unit (1106) , adapted to determine which tone frequency k to select from a set of tone frequencies available for use as correction tones, -an amplitude determining unit (1108) , adapted to determine an amplitude A_k of the tone k, such that the determined amplitude A_k satisfies a predefined power limit,

-a phase shift determining unit (1110) , adapted to determine a phase shift φ_k for the tone k, so that the tone k with the phase φ_k will, when used for creating a correction signal, reduce the peak in the signal S_n -a correction signal creating unit (1112) adapted to create a correction signal c_n based on the tone k, the phase shift φ_k, and the amplitude A_k,

-a signal comparison unit (1114) adapted to compare the maximum magnitude of the combined signal s_n+c_n with the maximum magnitude of the signal s_n/

-a signal setting unit (1116) , adapted to set a new signal S_n to equal the combined signal s_n+c_n , i .e . S_n = s_n+c_n , if a max s_^ +c_r <max s,. a>0

-a tone removing unit (1118), adapted to remove the tone frequency k from the set of available tone frequencies, and, -an iteration unit (1120) , adapted to repeat the method steps from the step of selecting a sample s [n₀] representing a peak, if a predefined stop criterion is not fulfilled, and otherwise to output the signal S_n.

14. The arrangement according to claim 13, wherein the frequency selecting unit and the phase shift determining unit are further adapted to determine k and φ_k using the formula

wherein LN is the number of signal samples, s [n₀] is the sample representing the peak, and A^ is the amplitude of the correction signal, and A:^* and φ^" indicate optimal values of the tone k and the phase φ_k, respectively.

15. The arrangement according to claim 13 or 14, wherein the frequency selecting unit and the phase shift determining unit are further adapted to determine k and φ_k when a constraint is applied to which tone frequencies that can be selected as tone index k and/or a constraint is applied to which values the phase shift φ_k of the tone k can take.

16. The arrangement according to any of the claims 13-15, wherein the phase shift determining unit is further adapted to determine φ_k when a constraint is applied to which frequencies that can be selected as tone frequency k, and wherein the phase φ_k is determined using the formula

. _r ., iπnJi OJ _LN

wherein the phase shift ^can take any value in the range[θ,2π] .

17. The arrangement according to any of the claims 13-16, wherein a constraint is applied to which values the phase shift φ_k of the tone k can take, and wherein the tone k is determined using the formula

2πn_ϋk k^* =argmin Zs[n₀] -φ_k- ^■ + π

LN

18. The arrangement according to any of the claims 13-17, wherein the correction signal C_n is defined as

Iπnk c = C₁,e^~™^~ where C_k = A_k exp[jφ_k) ; n = 0,1,...,LN-I ; and LN is the number of signal samples .

19. The arrangement according to any of the claims 13-18, wherein the amplitude A_k is defined to be the maximum allowed magnitude for the correction tone k.

20. The arrangement according to any of the claims 13-19, where the predefined power limit of the correction signal is derived from an EVM value or from a spectrum mask.

21. The arrangement according to any of the claims 13-20, wherein the amplitude A_k of the tone is determined from an EVM criterion if an in-band tone is used, and from a spectral mask criterion if an out-of-band tone is used.

22. The arrangement according to any of the claims 13-21, wherein the stop criterion is one or a combination of the following:

-all tones available for use as correction tones are used; or

23. The arrangement according to any of the claims 13-22, wherein the method is used within the concept of TR and/or TI.

24. The arrangement according to any of the claims 13-23, wherein the signal s is generated by a radio access technology such as GSM, WCDMA, LTE or WiMAX, or any combination of these radio access technologies in a multi- standard radio system.