A method for suppressing peak-to-average power ratio (PAPR) predistortion of SC-FDMA signals under EVM constraints
By superimposing frequency-domain additive predistortion vectors and sparsity constraints in the symbol domain, and deriving the optimal predistortion symbol vector using Lagrange multipliers, the nonlinear distortion problem in suppressing the peak-to-average power ratio of SC-FDMA signals is solved, achieving efficient signal reconstruction and improved bit error rate.
Patent Information
- Application Number
- CN202611594943.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-15
- Publication Date
- 2026-11-17
AI Technical Summary
In satellite communications, the peak-to-average power ratio (PAPR) of SC-FDMA signals fluctuates significantly under high-order modulation, causing high-power amplifiers to operate nonlinearly, resulting in severe nonlinear distortion and adjacent channel interference. Existing technologies struggle to effectively suppress PAPR while simultaneously considering system bit error rate performance and computational complexity.
By superimposing frequency-domain additive predistortion vectors in the symbol domain, a convex optimization model is established with the goal of minimizing instantaneous power. The optimal predistortion symbol vector is derived through sparsity constraints and Lagrange multipliers. Signal reconstruction is performed using closed-form solutions and linear superposition in the time domain, avoiding complex IFFT domain transformation iterations.
It achieves effective suppression of signal peak-to-average power ratio in high-order modulation scenarios, reduces computational complexity, and improves the communication reliability and bit error rate performance of the system. It is particularly suitable for communication scenarios with stringent computational latency, such as DVB-RCS2.
Smart Images

Figure SMS_74 
Figure SMS_83 
Figure QLYQS_5
Abstract
Description
Technical Field
[0001] This invention provides a method for suppressing the peak-to-average power ratio (PAPR) predistortion of SC-FDMA signals under EVM constraints, belonging to the field of communication technology. Background Technology
[0002] In satellite communications, to improve power efficiency, power amplifiers (PAs) in vehicle-mounted or handheld terminals typically operate near their saturation point. Although the peak-to-average power ratio (PAR) of SC-FDMA is lower than that of traditional OFDM technology, its instantaneous signal power still fluctuates significantly in high-order modulation scenarios such as 64-QAM and 256-QAM. When such signals pass through the high-power amplifiers of satellite transponders or terminals, they can easily force the high-power amplifiers into nonlinear operating regions, leading to severe nonlinear distortion. Nonlinear distortion not only causes in-band distortion, directly reducing the system's bit error rate performance, but also induces out-of-band radiation, causing adjacent channel interference to neighboring users and severely affecting the transmission reliability of satellite links. To avoid nonlinear distortion, amplitude limiting, as the most direct PAR suppression technique, reduces the PAR by limiting the signal amplitude. However, this nonlinear process can suddenly introduce in-band distortion and out-of-band radiation. Although out-of-band radiation can be eliminated through frequency domain filtering techniques, filtering operations often lead to peak regeneration, thus affecting the PAR suppression effect. Therefore, how to further and more efficiently suppress the peak-to-average power ratio of SC-FDMA signals while taking into account both system bit error rate performance and computational complexity remains a key problem to be solved. Summary of the Invention
[0003] To address the shortcomings of existing technologies, this invention presents a predistortion suppression method for SC-FDMA signals under EVM constraints. The method first superimposes a frequency-domain additive predistortion vector onto the original modulation symbols in the symbol domain, establishing a convex optimization model with error vector amplitude budget as the constraint boundary and minimizing instantaneous power as the objective. Then, it performs peak search on the input SC-FDMA time-domain sample signal and calculates the position index of the adjacent modulation constellation points leading to the time-domain peak in the symbol domain. Next, it applies a symbol-domain sparsity constraint to the predistortion vector, making it non-zero only at these two core positions, and expresses the time-domain predistortion signal at the peak position as a weighted superposition of the two symbol-domain distortions. Finally, it introduces Lagrange multipliers and uses KKT conditions to derive and solve for the exact closed-form solution of the optimal predistortion symbol vector, and directly completes the update and reconstruction of the full-length signal through lightweight, fast linear superposition in the time domain. The method of this invention transforms the traditional hard replacement of constellation points into an optimization process with controllable error vector amplitude. It breaks through the performance bottleneck of uncontrollable in-band distortion in traditional amplitude limiting algorithms, which leads to a flat bit error rate at high signal-to-noise ratio. At the same time, it avoids complex IFFT domain transformation iteration by using closed-form solutions and linear superposition in the time domain, reducing the single computation complexity to the linear order. It is particularly suitable for flexibly balancing signal peak-to-average power ratio suppression and system bit error rate performance in high-order modulation scenarios.
[0004] To solve the above problems, the technical solution adopted by the present invention is as follows: A method for suppressing peak-to-average power ratio (PAPR) predistortion in SC-FDMA under EVM constraints includes the following steps: Step A: Obtain the SC-FDMA user time-domain transmit signal vector using the centralized subcarrier mapping LFDMA method. The specific steps are as follows: Step A-1: For the first Data modulation symbol vector for each user Perform DFT precoding to obtain the frequency domain data symbol vector. , is represented as: , in, For user indexing, the system has a total of One user, ; The number of subcarriers allocated to each user; for The normalized DFT matrix of a point is expressed as: ; Step A-2: According to the LFDMA method, convert the frequency domain symbol vector... Map all On each subcarrier, Given the total number of subcarriers in the system, the mapped frequency domain data signal vector is obtained. Represented as: , in User The subcarrier mapping matrix has the following specific form: , in For users The starting subcarrier position, yes An identity matrix of order 1. Yes, the size is A matrix of all zeros; Step A-3: Mapped frequency domain data signal vector through Pointed IDFT transformation back to the time domain yields the user Time-domain transmitted signal vector at the SC-FDMA symbol time , is represented as: , in for Point normalized DFT matrix The conjugate transpose of; Step B: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require the full context and They are respectively used as time-domain transmitted signal vectors for initialization signals. and initialization data modulation symbol vector Input iterative process, for the first... During the iteration process, an optimization model for minimizing the peak-to-average ratio (PAPR) under EVM constraints is established. ,Model The specific expression is as follows: , in, It is the first During each iteration, the optimal additive predistortion sign vector is sought. For iteration rounds, This represents the maximum number of iterations. For the first The time-domain transmitted signal vector during round iteration; For the first The temporal predistortion vector of the round iteration; For the first Peak value of the time-domain transmitted signal during round iteration; for lie in element; peak index Represented as; ; With the The data modulation symbol vector after round iteration; The threshold for EVM constraints; for Located in the index element; and To and The relevant data modulation symbol index positions are represented as follows: , , in, Represents modulo operation; This indicates the floor function; Step C: Optimize the model for minimum peak-to-average ratio under EVM constraints. This simplifies to an optimization model for minimum peak instantaneous power under EVM constraints. ,Model The specific expression is as follows: in, For data modulation symbol vectors In the The average power of each iteration is expressed as: ; Step D: Solve the model To obtain the optimal predistortion symbol vector , is represented as: , in, The sum of energy introduced by the impulse response in the predistorted signal is expressed as: in, , This is the inherent discrete periodic impulse response waveform function of the system; Step E: The peak value after generation is the first... Round iteration time domain transmitted signal vector It is represented as: , in, for The Each time-domain element; until the maximum number of iterations is reached. Terminating the iteration, outputting the result. Time-domain transmitted signal vector of round iteration .
[0005] In step A, the total number of subcarriers in the system The number of subcarriers allocated to each user .
[0006] In step B, the EVM constraint threshold The range of values is .
[0007] In step B, the maximum number of iterations is set to... .
[0008] Beneficial Effects: This invention presents a predistortion suppression method for SC-FDMA signals under EVM constraints. Compared to existing technologies, this invention introduces an adjustable additive predistortion mechanism and establishes a convex optimization model based on error vector amplitude constraints. This transforms the rigid constellation point iterative replacement process in traditional CPS algorithms into a controllable optimization process, solving the problem of uncontrollable in-band distortion in traditional constellation point replacement and amplitude limiting algorithms. It achieves a flexible trade-off between signal peak-to-average ratio suppression and system communication reliability. This invention utilizes the sparsity characteristics of multi-carrier peak formation, performing predistortion control only on the two core adjacent modulation symbols that cause transient peaks in the time domain. It derives the exact closed-form solution of the optimal predistortion symbol vector using the Lagrange multiplier method and KKT conditions, eliminating the need for global optimization calculations across all subcarriers as in traditional optimization algorithms. In the signal reconstruction stage, this method directly utilizes the lightweight linear superposition of time-domain discrete impulse responses to complete signal updates, without performing complex multiple discrete Fourier transform operations. Compared to traditional iterative amplitude limiting filtering methods, it has wider applicability and extremely low latency. During the solution process, by employing a closed-form solution derivation and time-domain superposition strategy, the originally complex global optimization problem was successfully reduced in dimensionality, significantly lowering the computational complexity of a single iteration to the linear order of magnitude. It can quickly reach the performance convergence limit with only a very small number of iterations, making it particularly suitable for SC-FDMA multi-carrier uplink communication scenarios such as DVB-RCS2 that use high-order modulation and have strict requirements for computational latency. Detailed Implementation
[0009] This invention presents a predistortion suppression method for SC-FDMA signals under EVM constraints. The method first superimposes a frequency-domain additive predistortion vector onto the original modulation symbols in the symbol domain, establishing a convex optimization model with error vector amplitude budget as the constraint boundary and minimizing instantaneous power as the objective. Then, it performs a full-range peak search on the input SC-FDMA time-domain sample signal and calculates the position indices of the two constellation points in the symbol domain that cause the time-domain peak. Next, it applies a symbol-domain sparsity constraint to the predistortion vector, making it non-zero only at these two core positions, and expresses the time-domain predistortion signal at the peak position as a weighted superposition of the two symbol-domain distortion values. Finally, it introduces Lagrange multipliers and uses KKT conditions to derive and solve for the exact closed-form solution of the optimal predistortion symbol vector, and directly completes the update and reconstruction of the full-length signal through lightweight, fast linear superposition in the time domain. The method of this invention transforms the traditional hard replacement of constellation points into an optimization process with controllable error vector amplitude, breaking through the problem of uncontrollable in-band distortion in traditional amplitude limiting algorithms. At the same time, it avoids complex IFFT domain transformation iteration by using closed-form solutions and linear superposition in the time domain, reducing the single calculation complexity to the linear order. It is suitable for a flexible balance between signal peak-to-average power ratio suppression and system bit error rate performance in high-order modulation scenarios.
[0010] To better illustrate the method of the present invention, more detailed examples are provided below: This embodiment uses an SC-FDMA system to build a simulation platform, simulating the system's operation under additive white Gaussian noise (AWGN) channel conditions. The specific parameter configurations of the SC-FDMA system used in this embodiment are as follows: ① Total number of subcarriers in the system Number of subcarriers occupied by each user ; Length of the cyclic prefix Subcarrier spacing 15kHz; Maximum number of iterations during iteration ; ② The subcarrier mapping method adopts the centralized subcarrier mapping method LFDMA, that is, each user occupies a continuous subcarrier block. Specifically, user 1 occupies subcarriers 0 to 71, user 2 occupies subcarriers 72 to 173, and so on. ③ Each user's transmitted data uses 64QAM modulation, and the modulation symbols of each user are... After pointwise DFT precoding, it is mapped onto its occupied subcarrier, and then... The time-domain transmitted signal is obtained by point IDFT transformation and adding a cyclic prefix; the transmission channel is an AWGN channel. Based on the above parameter configuration, the specific implementation process of the method of the present invention is as follows: Preparatory steps: First, acquire the SC-FDMA user transmission signal. The specific steps are as follows: Step 1: At the transmitting end, the binary information bitstream of each user terminal sequentially undergoes channel coding and 64QAM modulation to generate a time-domain modulation symbol sequence; the SC-FDMA system performs DFT precoding on the modulation symbols before subcarrier mapping, transforming the time-domain symbols to the frequency domain, and then... After the point DFT operation, the user's frequency domain symbol vector can be represented as: , The number of subcarriers allocated to each user, where, For user indexing, the system has a total of One user, ; For users Data modulation symbols; for The normalized DFT matrix of a point is expressed as: , in, for Point DFT twitch factor, where The imaginary unit; Step 2: Mapped to according to LFDMA method On each subcarrier, the mapped frequency domain signal vector is obtained. Represented as: , in User The subcarrier mapping matrix, under centralized mapping, is a selection matrix composed of a zero matrix and an identity matrix, and its specific form is as follows: , in For users The starting subcarrier position, yes An identity matrix of order 1, used to select user data. Yes, the size is The all-zero matrix corresponds to the subcarrier positions allocated to other users; Step 3: Mapped frequency domain signal vector through Point IDFT transformation back to the time domain yields the first... Time-domain signal of each user at this SC-FDMA symbol time , is represented as: , in for Point normalized DFT matrix The conjugate transpose of; Step A: and They are respectively used as time-domain transmitted signal vectors for initialization signals. and initialization data modulation symbol vector Input iterative process, for the first... During the iteration process, an optimization model for minimizing the peak-to-average ratio (PAPR) under EVM constraints is established. The specific steps are as follows: Step A-1: In the first During the round of iteration, at the 1st The symbol vector of the round iteration The optimal additive predistortion symbol vector is superimposed to obtain the first... Round-by-round iteration of data modulation symbol vector , is represented as: ; At this point, the EVM constraint, used to characterize the degree of power deviation between the transmitted symbol and the ideal constellation point, can be simplified to: Energy and The ratio of total energy is expressed as: , in, For EVM constraint threshold, The second norm of a vector; Step A-2: Apply symbolic field sparsity constraints, peak index Determined by peak search of the current time-domain signal, expressed as: , Based on the physical causes of time-domain peaks, peak indexing is used. Calculation and Related data modulation symbol index positions and It is represented as: , , in, Represents modulo operation; This indicates the floor function; for Apply sparsity control, so that Only and There is a non-zero predistortion correction value on the upper part, and the symbol at the other subcarrier positions is zero, in order to avoid unnecessary in-band interference in the non-peak region, expressed as: ; Step A-3: The objective function is to find This minimizes the peak-to-average power ratio (PAPR) of the newly generated time-domain signal, expressed as: , Therefore, combining the constraints given in steps A-1 and A-2, an optimization model for the minimum peak-to-average ratio under EVM constraints is established, expressed as: ; Step B: Optimize the model for minimum peak-to-average ratio under EVM constraints. This simplifies to an optimization model for minimum peak instantaneous power under EVM constraints. The specific steps are as follows: Step B-1: Because of the introduction Since the energy is usually very weak, it is approximated that the average power remains unchanged before and after the signal update, which is expressed as: , in, For the first The temporal predistortion vector of the round iteration; For data modulation symbol vectors In the Average power of round iteration; Step B-2: Based on the approximation in Step B-1, the problem of minimizing the ratio of peak power to average power in the original objective function can be directly simplified to minimizing the instantaneous peak power, expressed as: , Step B-3: Based on the approximation in step B-1, the number of steps will be... The EVM constraints during round iterations are simplified to: The energy constraint based on the baseline is expressed as: , User labels are omitted to simplify the model. A convex optimization model is established with the objective of minimizing peak instantaneous power and the constraint of total EVM budget, which is expressed as: ; Step C: Reconstruct the Lagrangian function and use the KKT conditions to solve for the optimal predistortion sign vector closed-form solution. The specific steps are as follows: Step C-1: Under the constraint of non-zero frequency points, The time-domain predistorted signal at a given point is expressed as a weighted linear superposition of the discrete periodic impulse responses of the two symbol-domain distortion quantities, as follows: , in, It is the inherent discrete periodic impulse response waveform function of the system, which has a deterministic main lobe and side lobes that decay to both sides. Step C-2: Introduce real Lagrange multipliers Constructing the augmented Lagrangian function for convex optimization problems It is represented as: , in, Impulse response The abbreviation. Based on the KKT optimization conditions, for... Find the partial derivative and set it to zero, let the first... In the next iteration, we have ,in and Let and represent the phase and amplitude of the peak value at the current iteration number, respectively. Expanding the partial derivatives yields the steady-state condition equation, which is expressed as: , ; Step C-3: Introduce the optimal phase condition to make Minimum, and The requirement is an opposite relationship, expressed as: in, To compensate for the loss, substituting this phase condition into the steady-state equation of step D-2, we can obtain... exist and Elements containing offsetting coefficients and , is represented as: , ; Step C-4: The core objective is to minimize PAPR, thus allowing full utilization of the permitted EVM budget. In this case, the complementary relaxation condition is expressed as: Substituting the relaxation condition into the steady-state equation and simplifying, the cancellation coefficient can be obtained. Meanwhile, considering that the domain of the cancellation coefficient in physical reality is... ,like This means that overcompensation leads to a rebound in the negative peak. The actual value can be expressed as: in, , which is the energy introduced by the impulse response in the predistorted signal; Step C-5: Substitute the obtained 𝛼 into the steady-state equation of the optimal phase condition to obtain the optimal predistortion symbol vector. , can be represented as: ; Step D: Obtain and Then, through linear superposition in the time domain, a new time-domain signal vector after suppressing the peak is generated, which is expressed as: , in, for The Each time-domain element; until the maximum number of iterations is reached. Terminating the iteration, outputting the result. Round iteration transmission signal .
Claims
1. A method for suppressing peak-to-average power ratio (PAPR) pre-distortion in SC-FDMA under EVM constraints, characterized in that, Includes the following steps: Step A: Obtain the SC-FDMA user time-domain transmit signal vector using the centralized subcarrier mapping LFDMA method. The specific steps are as follows: Step A-1: For the first Data modulation symbol vector for each user Perform DFT precoding to obtain the frequency domain data symbol vector. , is represented as: , in, For user indexing, the system has a total of One user, ; The number of subcarriers allocated to each user; for The normalized DFT matrix of a point is expressed as: ; Step A-2: According to the LFDMA method, convert the frequency domain symbol vector... Map all On each subcarrier, Given the total number of subcarriers in the system, the mapped frequency domain data signal vector is obtained. Represented as: , in User The subcarrier mapping matrix has the following specific form: , in For users The starting subcarrier position, yes An identity matrix of order 1. Yes, the size is A matrix of all zeros; Step A-3: Mapped frequency domain data signal vector through Pointed IDFT transformation back to the time domain yields the user Time-domain transmitted signal vector at the SC-FDMA symbol time , is represented as: , in for Point normalized DFT matrix The conjugate transpose of; Step B: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require the full context and They are respectively used as time-domain transmitted signal vectors for initialization signals. and initialization data modulation symbol vector Input iterative process, for the first... During the iteration process, an optimization model for minimizing the peak-to-average ratio (PAPR) under EVM constraints is established. ,Model The specific expression is as follows: , in, It is the first During each iteration, the optimal additive predistortion sign vector is sought. For iteration rounds, This represents the maximum number of iterations. For the first The time-domain transmitted signal vector during round iteration; For the first The temporal predistortion vector of the round iteration; For the first Peak value of the time-domain transmitted signal during round iteration; for lie in element; peak index Represented as; ; With the The data modulation symbol vector after round iteration; The threshold for EVM constraints; for Located in the index element; and To and The relevant data modulation symbol index positions are represented as follows: , , in, Represents modulo operation; This indicates the floor function; Step C: Optimize the model for minimum peak-to-average ratio under EVM constraints. This simplifies to an optimization model for minimum peak instantaneous power under EVM constraints. ,Model The specific expression is as follows: in, For data modulation symbol vectors In the The average power of each iteration is expressed as: ; Step D: Solve the model To obtain the optimal predistortion symbol vector , is represented as: , in, The sum of energy introduced by the impulse response in the predistorted signal is expressed as: in, , This is the inherent discrete periodic impulse response waveform function of the system; Step E: The peak value after generation is the first... Round iteration time domain transmitted signal vector It is represented as: , in, for The Each time-domain element; until the maximum number of iterations is reached. Terminating the iteration, outputting the result. Time-domain transmitted signal vector of round iteration .
2. The method for suppressing peak-to-average power ratio (PAPR) predistortion of SC-FDMA signals under EVM constraints according to claim 1, characterized in that, In step A, the total number of subcarriers in the system The number of subcarriers allocated to each user .
3. The method for suppressing peak-to-average power ratio (PAPR) predistortion of SC-FDMA signals under EVM constraints according to claim 1, characterized in that, The EVM constraint threshold mentioned in step B The range of values is .
4. The method for suppressing peak-to-average power ratio (PAPR) predistortion of SC-FDMA signals under EVM constraints according to claim 1, characterized in that, In step B, the maximum number of iterations is set to... .
