WO2019027341A1

WO2019027341A1 - Selective tone reservation for papr reduction in wireless communication systems

Info

Publication number: WO2019027341A1
Application number: PCT/RU2017/000572
Authority: WO
Inventors: Andrey Andreevich IVANOV; Dmitry Vladimirovich LAKONTSEV
Original assignee: Autonomous Non-Profit Organization For Higher Education "Skolkovo Institute Of Science And Technology"
Priority date: 2017-08-04
Filing date: 2017-08-04
Publication date: 2019-02-07

Abstract

A method for reducing peak-to-average power ratio (PAPR) in a time domain of Orthogonal Frequency Division Multiplexing (OFDM) signal using selective tone reservation (STR) is provided. The method comprises: distributing a first PAPR reduction signal among unused resource blocks (RBs); distributing a second PAPR reduction signal among occupied RBs; applying non-iterative clipping and filtering to the OFDM signal, modified by the first PAPR reduction signal and the second PAPR reduction signal. The method is fast and has a high performance. STR fully fits in 4G/5G standard requirements and its complexity does not require a huge computational resources.

Description

SELECTIVE TONE RESERVATION FOR PAPR REDUCTION IN WIRELESS

COMMUNICATION SYSTEMS

FIELD OF THE INVENTION

The present invention generally relates to wireless communications. More specifically, it relates to reducing a peak-to-average power ratio (PAPR) for a time domain Orthogonal Frequency Division Multiplexing (OFDM) signal in wireless communication systems.

BACKGROUND OF THE INVENTION

It is widely expected that future mobile communication systems will apply multicarrier OFDM transmission because it allows a high spectral efficiency and has a robustness to channel fading. In recent years, OFDM has been adopted in 4G and 5G communication applications. However, some challenging issues are still unresolved in OFDM systems. One of the issues is the high PAPR, which results in non-linearity in power amplifiers (PA), and causes out of band radiation and in-band distortion. Such distortions could degrade the receiver sensitivity dramatically in the same cell or in a neighbour base station (BS). PAPR is calculated by the equation:

max(|x[k] |²)

PAPR = —

E(|x[k] |²)

k = [lB¾jN]_} where ^x fc] is the signal, represented by N samples; ^ is the sample index; N is the IDFT size; E( ) is the mean operator. Also high values of PAPR result in a low efficient usage of the analog-to-digital converter (ADC) and digital-to-analog converter (DAC) word length at the analog part of a transceiver. This problem requires a PA with a large linear dynamic range or an efficient PAPR reduction technique.

Many PAPR reduction techniques have been investigated over last years.

B. M. Lee, Y. Kim, An Adaptive Clipping and Filtering Technique for PAPR Reduction of OFDM Signals, Circuits, Systems, and Signal Processing, vol. 36, pp. 1335-1349, 2013 and A. Chakrapani, V. Palanisamy, A Novel Clipping and Filtering Algorithm Based on Noise Cancellation for PAPR Reduction in OFDM Systems, The Proceedings of the National Academy of Sciences, Springer, 2014 disclose the method of Clipping and Filtering (C&F). This is the simplest method, in which the out of band radiation generated by clipping is eliminated by subsequent clipping and filtering operations. The filtering operation contributes to a peak regrowth problem, which can be suppressed using iterative techniques. C&F requires inverse discrete Fourier transform (IDFT) operations and the number of iterations leads to the increase in computational complexity and delay. Also clipping noise spectrum occupies full band. Thus, C&F method has a low PAPR reduction capability when harmonics with high modulation order exist in OFDM spectrum.

H. K. Pal, A. K. Singh, PAPR Reduction Technique Using Advanced Peak Windowing Method of OFDM System, International Journal of Soft Computing and Engineering (IJSCE), vol. 3, Issue-2, May 2013 discloses the peak windowing method. This method proposes that it is possible to remove large peaks at the cost of a slight amount of self-interference when large peaks arise infrequently. Peak windowing reduces PAPR at the cost of increasing the performance degradation and out-of-band radiation. Clipping is the simplest case of peak windowing. The technique of peak windowing offers better PAPR reduction with better spectral properties compared with common clipping.

Another method is suggested in A. Kliks, H. Bogucka, Improving Effectiveness of the Active Constellation Extension Method for PAPR Reduction in Generalized Multicarrier Signals, Wireless Personal Communications, Springer, 2011 ; V. Sundeepkumar, S. Anuradha, Adaptive Clipping-Based Active Constellation Extension for PAPR Reduction of OFDM/OQAM Signals, Circuits, Systems, and Signal Processing, Springer Science+Business, 2016; and E. V. Cuteanu, A. Isar, PAPR Reduction of OFDM Signals using Active Constellation Extension and Tone Reservation Hybrid Scheme, AICT: The Eighth Advanced International Conference on Telecommunications^ 12. This method is based on active constellation extension (ACE). Signaling points of the conventional constellation are dynamically moved toward outside of the original constellation in order to reduce the PAPR level of the transmitted signal. The domain for allowed alternative points is chosen so that the signal processing doesn't reduce the constellation's minimum-distance but lowers the PAPR level. However, this process requires many iterative steps with a large number of subcarriers and an additional DFT unit is used. Hence, constellation shaping technique requires a lot of time and computational resources to reach the desired performance.

The huge class of linear methods is represented by approaches like partial transmit sequence (PTS) and selective mapping (SLM). This method are disclosed in C. Duanmu, H. Chen, Reduction of the PAPR in OFDM Systems by Intelligently Applying Both PTS and SLM Algorithms, Wireless Personal Communications, vol. 74 pp. 849-863, Springer, 2014; K. Pachori, A. Mishra, An efficient combinational approach for PAPR reduction in MIMO-OFDM system, Wireless Network, 2016; and S. Khademi, A. Veen, T. Svantesson, Precoding technique for peak-to-average-power-ratio (PAPR) reduction in MIMO OFDM/A systems, IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2012/ The basic idea of the algorithms is to carry out various kinds of randomization of the signal to achieve a lower PAPR. PTS algorithm is intended to carry out the phase optimization processing after the IDFT process, and then select the phase combination, which has the minimum PAPR to transmit, while SLM like algorithms carry out the phase optimization process before the IDFT process, and select the phase combination with the minimum PAPR to transmit. The information overhead of the PTS or the SLM algorithm is small; however, its application is limited since such kind of transformation is not supported by 4G and 5G standards.

Coding techniques use a forward-error-correction code set that excludes the OFDM symbols with a high PAPR, thus reducing the probability of occurrence of a signal with high PAPR. While these schemes reduce PAPR, they also significantly reduce the transmission rate for OFDM systems with large number of subcarriers. Unfortunately, most of such codes don't satisfy 4G/5G requirements.

Another PAPR reduction method is a tone reservation (TR) (see M. M. Hasan, S. S. Singh, An Overview of PAPR Reduction Techniques in OFDM Systems, International Journal of Computer Applications, vol. 60, no. 15, 2012 and M. I. Abdullah, M. Z. Mahmud, M. S. Hossain, M. N. Islam, An Overview of PAPR Reduction Techniques in OFDM Systems, ARPN Journal of Systems and Software, vol. 1, no. 8, 2011. This method uses a small set of tones for PAPR reduction. It can be shown that reserving a small fraction of tones leads to minimization in PAPR using a simple algorithm at the transmitter of the system without any additional complexity at the receiver end. Besides the advantage of no additional distortion, this method also does not need to transmit additional information to the receiver. Because not all subcarriers are used to transmit useful information, this method lowers the data rate of the OFDM-based systems. In order to reduce the computation complexity and to improve the performance, several derivate techniques have been proposed. Most of them are iterative and work until convergence reaches the expected threshold. Hence, calculation time is the limiting factor. Spectrum resource is another limiting factor on its application in 4G/5G technologies since all subcarriers are usually occupied by users (UE) signal.

Therefore, there is a need for a more efficient PAPR reduction algorithm that requires low calculation time.

SUMMARY OF THE INVENTION

To overcome the above-mentioned drawbacks of the prior art a method for selective tone reservation (STR) is proposed. The STR method is a non-iterative version of the TR idea. It is based on minimum mean square error (MMSE) tones reservation for each OFDM symbol separately. STR distributes PAPR reduction signal between selected resource blocks (RBs) in spectrum domain according to allowed error vector magnitudes (EVM) or max allowed noise power for each RB. Moreover, it employs both unused and occupied RBs with limiting PAPR compensation tones power to the desired values.

In one embodiment a method for reducing peak-to-average power ratio (PAPR) in a time domain of OFDM signal is provided, the method comprising: distributing a first PAPR reduction signal among unused resource blocks (RBs); distributing a second PAPR reduction signal among occupied RBs; applying non-iterative clipping and filtering to the OFDM signal modified by the first PAPR reduction signal and the second PAPR reduction signal.

In another embodiment the first PAPR reduction signal has the power, limited according to the inter-cell interference cancellation (ICIC) requirements.

In a further embodiment occupied RBs used for PAPR reduction are modulated according to low-order modulations.

In a still further ambodiment the second PAPR reduction signal has the power limited according to the error vector magnitude (EVM) requirements of low-order modulations.

In another embodiment the PAPR reduction in the OFDM signal is defined according to the following equations:

Z = X - AS

Y = WX

W = peaks (|X|)₅

where Z is the vector of signal with reduced PAPR; X is the initial signal vector; A is the amplitudes vector of PAPR reduction subcarriers; ^{~ p 1 m 1 q}J is the matrix

2nT)iink 2ΠΤ)ίπι 2n¾imN'

T

c =

N m N N

(NxM), consisting of the subcarriers samples ^e ; is the vector, consisting of subcarrier samples; N is the Inverse Discrete Fourier Transform (IDFT) size; M is the number of PAPR reduction subcarriers in current OFDM signal (unused RBs or occupied RBs with low-order modulations); k=[l ...N] is the sample index; m is the subcarrier index ΠΙΒ€€[1ΒΤ>!Ν]. W is the identity matrix, consisting of L non-zero elements "1" corresponding to the max amplitudes of initial signal X; Y is the vector (N*l), consisting of the undesired peaks of X (only L non-zero samples, all other N-L samples are zeros).

In a further embodiment non-iterative clipping and filtering is applied to the entire OFDM signal. In a still further ambodiment the PAPR reduction signal performing non-iterative clipping and filtering has the power, limited according to the EVM requirements of high-order modulations.

In another embodiment, when the OFDM is a dual-band long-term evolution (LTE) signal g[k] consisting of a first LTE signal g[ [k] and a second LTE signal g₂[k], wherein

sW = gi M + ¾W

8i M = *i M

2rmivk

g₂ [k] = x₂[k]e where Xi[k] and x₂[k] are the baseband complex amplitudes, v is the frequency shift between carriers, k=[l ...N] is the sample index;

the PAPR reduction in the OFDM signal is defined according to the following equations:

x = \x₁\ + \x₂\_i

W = peaks

Y_j = WX_j Y₂ = wx₂

¹ H ¹ H

A, = -S^HY, A₉ = -S^HY₇

¹ N ^{1 2} N ²

where W is the identity matrix, consisting of L non-zero elements "1 ", corresponding to the max amplitudes of vector X; Ai, A₂ are the amplitudes of subcarriers; X_{l 5} X₂ are the vectors of baseband complex amplitudes of signals gi[k], g₂[k]; Y_{l s}Y₂ are the vectors, consisting of the undesired peaks of X_{l s} X₂ accordingly (only L non-zero samples, all other N-L samples are zeros); Z\, Z₂ are the vectors of baseband complex amplitudes of signals zj [k], z₂[k]; z[k] is the desired signal with reduced PAPR of g[k].

BRIEF DESCRIPTION OF DRAWINGS

Fig. 1 shows the selective tones reservation scheme.

Fig. 2 shows LTE 10MHz spectrum with the PAPR compensation signal in unused RBs. Fig. 3 shows a time domain LTE symbol: undesired peaks (absolute value) and the PAPR compensation signal in unused RBs. Fig. 4 shows LTE spectrum before and after PAPR compensation by utilizing unused

RBs.

Fig 5 shows CCDF for PAPR reduction using M = 32, 64, 128 and 256 compensation subcarriers in unused RBs.

Fig. 6 shows EVM requirements in LTE DL.

Fig 7 shows an RE power control dynamic range in LTE

Fig. 8 shows a compensation signal distribution in occupied RBs of "Group 1."

Fig. 9 shows CCDF for PAPR reduction using M = 32, 64, 128 and 256 compensation subcarriers in occupied RBs

Fig. 10 shows LTE 10MHz spectrum with PAPR reduction signal

Fig. 11 shows CCDF comparison of joint STR and C&F methods

Fig. 12 shows the dual band STR scheme

Fig. 13 shows time domain OFDM signal (RE part) after STR

Fig. 14 shows CCDF of dual band STR using M=480

DETAILED DESCRIPTION

STR algorithm description

The STR scheme of PAPR reduction is shown in Fig. 1. Its main difference compared to a common TR is the proposed low-complexity non-iterative MMSE algorithm of complex amplitudes calculation. In 4G/5G downlink (DL) the signal consists of occupied RBs which are occupied by user data or control signal and unused RBs where data are not transmitted in current symbol. EVM is different for each RB and depends on the modulation index of transmitted data in case of occupied RB or limited by a fixed value in case of unused RB. That is why common C&F in inefficient by itself.

In 5G information from MAC layer is used to get the modulation index from modulation and coding scheme (MCS) and limit a PAPR compensation tones power according to inter-cell interference cancellation (ICIC) or EVM requirements. In 4G the functional split between baseband unit (BBU) and remote radio unit (RRU) does not allow getting a modulation index directly since data on the transmitter end of RRU is represented by time domain samples after precoding. However, a software DFT with special constellation recognition algorithms on RRU side can be used to detect modulation index for each RB with a low complexity.

Thus, MAC and physical layer (PHY) coordination in DL provides abilities for a PAPR reduction.

STR algorithm includes three steps:

1. Unused RBs reservation according to ICIC requirements; 2. Occupied RBs reservation according to EVM requirements;

3. Non-iterative clipping & filtering according to max modulation index.

In the first step, STR distributes a PAPR reduction signal among unused RBs with limiting the signal power according to ICIC requirements. In the second step, algorithm utilizes occupied RBs to distribute a PAPR reduction signal among them with power according to EVM requirements. And, finally, in the third step, a non-iterative (clipping & filtering) is applied with limiting clipping noise to 201og [3.5/100]=-29dBc level according to required EVM=3.5% of 256QAM. It can be found that step-by-step (1—2→3) a spectrum of the PAPR reduction signal gets higher to provide a better PAPR reduction quality. Further in the paper each step of the presented algorithm will be disclosed in details and provided with related figures.

Unused RBs reservation (step 1)

In 4G/5G systems it is expected that some RBs within a separate cell can be left unused by the resource radio management (RRM) to prevent interference with neighbouring cells (ICIC feature). Such RBs can be utilized by peak reduction signals with limiting their power level to values under the desired value (usually -20dBc to a neighbour cell user level) of fixed level (usually -25dBc to the maximum output power). Null subcarriers (guard bands) in 4G/5G can also be utilized for the same reason. A general MMSE equation of complex amplitudes estimation in unused RBs (or null subcarriers) can be used to suppress the peaks:

Z = X - AS _{j (1)}

S^HY

Y = WX , (3)

W = peaks(|X|)_?

where X is the initial signal vector; Z is the vector of signal with reduced PAPR; A is the amplitudes vector of PAPR reduction subcarriers; ^~~ L ^{p 1 m 1} ^ is the matrix

2IIT)imk Γ 2n¾im 2ΠΤ)ίιτιΝ

T

C_m =

N m N N

BTJ!

(NxM), consisting of the subcarriers samples ^e ; is the vector, consisting of subcarrier samples; N is the Inverse Discrete Fourier Transform (IDFT) size; M is the number of PAPR reduction subcarriers in current OFDM signal (unused RBs or occupied RBs with low-order modulations); k=[l ...N] is the sample index; m is the subcarrier index rnB€€[lB¾]N]. j_{s me} identity matrix, consisting of L non-zero elements "1" corresponding to the max amplitudes of initial signal X; Y is the vector (Nx l), consisting of the undesired peaks of X (only L non-zero samples, all other N-L samples are zeros). Usually L«N and its value depends on IDFT size N, (M/N) ratio and the max allowed compensation signal power, for example, -25dBc in long term evolution (LTE) by default for unused RBs. The compensation signal power have to be limited by RRM according to inter-cell interference coordination requirements.

In practice a wide set of L values can be successfully used in algorithm (1-3) to achieve the desired performance since requirements strongly depend on a PA nonlinearity. A simple empirical table for L calculation was found by intensive simulations and fixed for all schemes to get the best PAPR reduction gain in complementary cumulative distribution function (CCDF) value=10^"5 while this criteria is not unique and can be changed if necessary.

Equation (3) does not require any matrix operations and its complexity is negligible since W is the clipped identity matrix. All subcarrier in OFDM symbol are orthogonal to each other, hence matrices product in (2) can be calculated as:

S^HS=NI, (4)

where I is identity matrix (MxM) . From (2) and (4) can be derived:

¹ H

A = -S^HY

^N (5)

Equation (5) compared to equation (2) doesn't require matrix inverse and division operations since N is the IDFT size and it is represented in the 2^P form, for example 2¹⁰=1024 in LTE 10MHz case. Hence, the division can be implemented by a "shift" operation in binary logic. In practice equation (5) have even a lower complexity and the multipliers could be avoided entirely if a nonlinear quantization is applied under the real and imaginary parts of subcarrier samples in S matrix:

Inmk

N

2mmk

e ^N « P_x (<p)sign[cos(ii?)] + iP₂ (<z>)sign[sin(ii?)] , (6)

where Ρι(φ) and Ρ₂(φ) are represented in 2^"p form, they could have a value in the set 1 1 1 1 1 1

^' 2' 4' 8' 16' 32' 64' ^

to reduce implementation complexity of the matrix A calculation. In equation (6) a sign operations and amplitudes approximation in the 2^_p form is proposed, hence all the multiplication in equation (5) can be replaced by a "shift" and "sum" operations. Also cos(^) and sin( >) can be easily generated by using a well-known low dimension direct digital synthesis (DDS) unit which also has a low complexity. Finally, equation (6) is given by:

2 mk

e ^N « g[cos( )] + /O{sin( )] , (7)

where Q[] is a nonlinear 4-bits quantizer which rounds value to nearest one from set

1 1 1 1 1 1

B ± 1, B ± -, B ± - B ± -, B ±— B ±— B ±— , 0

. Simulations show that such approximation provides enough accuracy for A calculation.

In equation (1) the AS product cannot be simplified by employing formula (7), however,

S consists of subcarrier samples which have an absolute value = 1 , hence a well-known

Coordinate Rotation Digital Computer (CORDIC) algorithm can be used to avoid the multipliers entirely.

In Fig. 2 a LTE 10MHz spectrum is presented with the PAPR compensation signal in unused RBs limited by -25dBc threshold.

In Fig. 3 an absolute value of time domain undesired LTE symbol peaks is shown with the PAPR compensation signal for different L values. It can be found that the compensation sin(x)

signal has a ^x shape because of its limited band.

In Fig. 4 a LTE spectrum is shown before and after PAPR compensation. In can be found that signal power has grown up in unused RBs but doesn't exceed the allowed boundary.

In Fig. 5 a CCDF is calculated for different M values and N=1024 in LTE 10MHz.

Occupied RBs reservation (step 2)

Besides unused RBs a lot of occupied RBs can also be utilized for further PAPR reduction with a higher efficiency. Each RB in DL channel allows an additive noise if final EVM in this RB is less than the specified in standard value. For example, in LTE standard there are restrictions on max EVM for each type of modulation as shown in Fig. 6. The EVM of each E- UTRA carrier for different modulation schemes on DL data channel shall be better than the limits in table.

Information about occupied RBs, MCS index and TX power in DL for each UE per antenna could be available from MAC in 5 G or calculated over a blind constellation recognition in 4G. The RE power control dynamic range is the difference between the power of an RE and the average RE power for a BS at maximum output power for a specified reference condition as p

shown in Fig. 7. It means the UE relative power ^r in DL also makes a sense and PAPR reduction signal power should consider not only modulation order but also meet signal power requirements.

STR challenge scenario: far away UEs with QPSK modulation has a relative power P_r=+4dBc because of long distance (table in Fig. 7). In DL spectrum there are UEs with 256QAM modulation nearby the cell having P_r=0dBc. Therefore, such scenario requires EVM=3.5%/10^{+3dBc 10}=1.4% and a common clipping is inefficient because of ultra-low allowed clipping noise. This is the main STR advantage compared with C&F technique.

Algorithm description

All occupied RBs in DL can be divided into 2 groups according to UEs modulation index:

Group 1 : (QPSK, 16QAM),

Group2: (64QAM, 256QAM).

According to MCS distribution statistics about 80% of all UEs has QPSK or 16QAM modulation scheme, hence all the RBs occupied by Group 1 can be utilized to generate PAPR compensation tones (Fig. 8) by limiting compensation signal power to the P value. For Group 1 the P value is defined as

among all the RBs in Group 1. In general scenarios the power of those tones is limited by P_r— 18dBc (EVM=12.5%).

All the equations for PAPR reduction by using allocated RBs are similar to the equations (1) and (5) derived for unused RBs reservation. The same MMSE estimation of complex amplitudes is used to suppress peaks. The only difference in compensation signal power and location.

In Fig. 9 a CCDF is calculated for different M values and N=1024 in LTE 10MHz (only allocated RBs are reserved).

Joint STR (step 3)

Residual clipping is limited by EVM (3.5%) of high modulation order UEs and a full band can be utilized for extra PAPR reduction. This approach is a common C&F and only efficient on the last stage when PAPR is already reduced significantly by the STR. We only provide the final performance of joint STR algorithm: unused RBs reservation + occupied RBs reservation + final C&F.

In Fig. 10 a LTE 10MHz spectrum with PAPR reduction signal is presented.

In Fig. 11 a CCDF is calculated for joint STR algorithm with M=480 and for C&F method in LTE 10MHz. Proposed STR algorithm outperforms C&F significantly. Dual band STR

Scenario with several single bands allocated in the same overall band of BS happens very often in carrier aggregation of LTE- Advanced since it is a great feature to increase radio resource efficiency and overall data rate of BS. One of the most typical application scenario is a dual band case where one LTE signal g₂[k] is shifted from another one g₂[k] in frequency domain. In such case a sum signal is given by:

g[k]= gi[k]+ g₂[k],

g₂[k] = x₂[k]e^ ,

where x^k] and x₂[k] are the baseband complex amplitudes, v is the frequency shift between carriers. In general, the STR solution under g[k] seems to be the same, however, the frequency shift between carrier- 1 and carrier-2 requires a high sample rate to be used which causes a significant complexity of algorithms (1) and (5). To reduce complexity in dual band scenario an elegant baseband solution can be applied:

x = \x₁\ + \x₂\_t

W = peaks

Y_j = WX_j Y₂ = WX₂

¹ H ¹ H

A, = -S^HY, A, = -S^HY,

¹ N ^{1 2} N ²

2m¾ivk

z[k] = _Zl [k] + z₂[k]e ^N where W is the identity matrix, consisting of L non-zero elements "1 ", corresponding to the max amplitudes of vector X; Ai, A₂ are the amplitudes of subcarriers; X , X₂ are the vectors of baseband complex amplitudes of signals g_\[k], g₂[k]; Yi,Y₂ are the vectors, consisting of the undesired peaks of Xi, X₂ accordingly (only L non-zero samples, all other N-L samples are zeros); Z_\, Z₂ are the vectors of baseband complex amplitudes of signals zj[k], z₂[k]; z[k] is the desired signal with reduced PAPR of g[k]. The main feature of the dual-band STR is included in its property: signals have a high PAPR while ^z W has a low PAPR value. The full scheme realizing the dual-band STR algorithm is shown in Fig. 12.

Demonstrative time domain signal after the dual-band STR compensation is shown in

Fig. 13. CCDF for STR in dual band LTE-A 10MHz scenario (only allocated STR version using M=480) is shown in Fig. 14.

Conclusion

The proposed PAPR reduction technique (STR) is based on the combination of MMSE estimation and TR method. The STR algorithm is applicable for PAPR reduction in wireless communication technologies. In 4G a software DFT can be used on RRU side to identify unused RBs in DL channel and to find high order modulation UEs (blind MCS recognition). This way the MAC data could be eliminated from the algorithm, which makes it applicable for LTE without hardware changes in (BBU-RRU) chain. In 5G information from MAC is available on RRU side. Simulation results shows that STR outperforms all of the exiting PAPR reduction methods in performance (C&F) and in calculation time and resources (ACE, TR) significantly. Finally, proposed STR has the following benefits: PAPR reduction by 4dB in 4G/5G technologies (about 2 dB better compared with a common clipping), has low complexity (doesn't require multipliers), non-iterative solution, without out of band emission, has no performance losses of 64QAM and 256QAM users in DL, fits in 4G/5G.

Claims

1. A method for reducing peak-to-average power ratio (PAPR) in a time domain of Orthogonal Frequency Division Multiplexing (OFDM) signal, comprising:

distributing a first PAPR reduction signal among unused resource blocks (RBs);

distributing a second PAPR reduction signal among occupied RBs;

applying non-iterative clipping and filtering to the OFDM signal, modified by the first PAPR reduction signal and the second PAPR reduction signal.

2. The method according to claim 1, wherein

the first PAPR reduction signal has the power, limited according to the inter-cell interference cancellation (ICIC) requirements.

3. The method according to claim 1, wherein occupied RBs used for PAPR reduction are modulated according to low-order modulations.

4. The method according to claim 1 , wherein

the second PAPR reduction signal has the power, limited according to the error vector magnitude (EVM) requirements of low-order modulations.

5. The method according to claim 1, wherein

Z = X - AS

S^HY

s"s

Y = wx

W = peaks (|x|)

where Z is the vector of signal with reduced PAPR; X is the initial signal vector; A is the amplitudes vector of PAPR reduction subcarriers; ^{δ ~} L P ' ^{m 1 q}J is the matrix

Γ 2n¾im 2Π¾ίπνΝ

T

N ^Cm N N

B¾!

(NxM), consisting of the subcarriers samples ^e is the vector, consisting of subcarrier samples; N is the Inverse Discrete Fourier Transform (IDFT) size; M is the number of PAPR reduction subcarriers in current OFDM signal (unused RBs or occupied RBs with low-order modulations); k=[l ...N] is the sample index; m is the subcarrier index rnB€€[lB¾]N]. j_{s me} identity matrix, consisting of L non-zero elements "1" corresponding to the max amplitudes of initial signal X; Y is the vector (Nx l), consisting of the undesired peaks of X (only L non-zero samples, all other N-L samples are zeros).

6. The method according to claim 1, wherein non-iterative clipping and filtering is applied to the entire OFDM signal.

7. The method according to claim 1, wherein

the PAPR reduction signal performing non-iterative clipping and filtering has the power, limited according to the EVM requirements of high-order modulations.

8. The method according to claim 1, wherein,

when the OFDM is a dual-band long-term evolution (LTE) signal g[k] consisting of a first LTE signal gi[k] and a second LTE signal g₂[k], wherein

g[k] = _gl [k] ₊ g₂[k]

2nf)ivk

g₂[k] = x₂[k]e where xj[k] and x₂[k] are the baseband complex amplitudes, v is the frequency shift between carriers, k=[l ...N] is the sample index;

x = \x₁\ + \x₂\,

W = peaks

^Yi ^{= wx}i Y₂ = wx₂

¹ H ¹ H

A, = -S^HY, A₇ = -S^HY,

1 N N ²

2n¾ivk

z[k] = _Zl [k] + z₂[k]e ^N where W is the identity matrix, consisting of L non-zero elements "1 ", corresponding to the max amplitudes of vector X; A\, A₂ are the amplitudes of subcarriers; Xi, X₂ are the vectors of baseband complex amplitudes of signals gi[k], g₂[k]; Y_l5Y₂ are the vectors, consisting of the undesired peaks of Xi, X₂ accordingly (only L non-zero samples, all other N-L samples are zeros); Z_ls Z₂ are the vectors of baseband complex amplitudes of signals zj[k], z₂[k]; z[k] is the desired signal with reduced PAPR of g[k].