US-12621198-B2 - High-precision multi-phase CFR system and method and use

US12621198B2US 12621198 B2US12621198 B2US 12621198B2US-12621198-B2

Abstract

This application discloses a high-precision multi-phase CFR system and method and use. This application relates to the technical field of PAPR reduction. The CFR system is a single-stage multi-phase structure, which is used for an input signal to perform peak searching and phase recording at a high rate, and to perform peak screening, peak allocation and peak cancellation at a single rate; in the cancellation signal generation, the peak noise after the peak screening is pulse shaped and compensated for the recorded phases at the high rate by multi-phase CPG coefficients; and according to the peak allocation, multiple CPG pulse signals are combined to obtain a final cancellation signal; then a delayed original signal and the final cancellation signal are subtracted to obtain a low-PAPR signal for output.

Inventors

Qiang Gu
Jinbang ZONG

Assignees

CHONGQING WUQI MICROELECTRONICS CO., LTD.
Shanghai Wuqi Microelectronics Co., Ltd.

Dates

Publication Date: 20260505
Application Date: 20240725
Priority Date: 20220127

Claims (2)

1 . A high-precision multi-phase Crest Factor Reduction (CFR) system for performing crest factor reduction on an input signal to obtain a peak-clipped signal for output, the system includes a main circuit, a branch circuit for generating a peak cancellation signal, and an adder/subtractor module; an original input signal is divided into two paths and input into the main circuit and branch circuit respectively; on the main circuit, the original input signal is delayed to generate a main circuit output signal; the branch circuit is provided with an interpolation module, a multi-phase max mag select module, a peak screening module, a peak allocation module, a multi-phase Cancel Pulse Generation (CPG) coefficients generation module and a cancellation signal generation module; the adder/subtractor module subtracts the peak cancellation signal from the main circuit output signal to obtain the peak-clipped signal for output; wherein, the interpolation module is configured to: when the original input signal with a sampling rate of fs is input into the branch circuit, the original input signal is interpolated N times to N*fs sampling rate, fs is an original data sampling rate, and after interpolation, one signal data x (n) becomes N-phase signal data X(Nn), X(Nn+1), . . . . X(Nn+N-1) of different phases, wherein N is an integer greater than or equal to 2, and n, Nn, Nn+1, . . . . Nn+N-1 represent time; the multi-phase max mag select module is configured to: for each signal data x(n), according to the N-phase signal data X(Nn), X(Nn+1), . . . . X(Nn+N-1) obtained after interpolation, compare the magnitudes of these N phases, select a phase with the maximum magnitude from the N phases as a peak point output, and record the phase information of the peak point position; and, form an N-phase peak signal according to the magnitude and phase of the peak point extracted from the N-phase signal data; the peak screening module is configured to: perform two peak screenings in sequence, including an initial peak screening and a secondary peak screening; the initial peak screening is to perform peak detection of the N-phase peak signal by sliding window processing; and the secondary peak screening is peak window screening which selects a maximum peak in the window according to a set window length, after two peak screenings, obtain a maximum peak set which contains multiple maximum peaks; and, store peak noise, which includes: retain magnitude and phase information corresponding to maximum peak points, and set all data at positions other than the maximum peak points to 0 to obtain a noise set, the noise set corresponding to a noise signal; the peak allocation module is configured to: perform filter multiplication allocation on the maximum peaks in the noise set, wherein, a number of multipliers is allocated according to a density of the maximum peaks in the noise set; a filter is used to filter the noise signal to maintain the same spectral characteristics as the original input signal; the multi-phase CPG coefficients generation module is configured to: perform a convolution operation on a single-phase CPG coefficient and a multi-phase fractional delay filter to obtain multi-phase CPG coefficients including multiple groups of CPG coefficients, each group of CPG coefficient corresponds to a phase; the single-phase CPG coefficient is obtained by designing a filter with a same spectrum as the original input signal at the rate fs; the cancellation signal generation module is configured to: for noise of different phases within the noise set, perform a convolution operation on the noise and a CPG coefficient of the corresponding phase to obtain a CPG pulse signal of this phase after delay compensation; and according to peak allocation, the CPG pulse signals of multiple phases are combined to obtain the peak cancellation signal.
2 . A high-precision multi-phase Crest Factor Reduction (CFR) method, comprising: inputting an original input signal; dividing the original input signal into two paths and inputting them into a main circuit and a branch circuit respectively; on the main circuit, delaying and outputting the original input signal; on the branch circuit, processing the original input signal to generate a peak cancellation signal subtracting the peak cancellation signal output by the branch circuit from the delayed original signal output by the main circuit to obtain a peak-clipped signal for output; wherein, generating the peak cancellation signal comprises: step 1, interpolation: when the original input signal with a sampling rate of fs is input into the branch circuit, the original input signal is interpolated N times to N*fs sampling rate, fs is an original data sampling rate, and after interpolation, one signal data x(n) becomes N-phase signal data X(Nn), X(Nn+1), . . . . X(Nn+N-1) of different phases, wherein N is an integer greater than or equal to 2, and n, Nn, Nn+1, Nn+N-1 represent time; step 2, multi-phase max mag selection: for each signal data x (n), according to the N-phase signal data X(Nn), X(Nn+1), . . . . X(Nn+N-1) obtained after interpolation, compare the magnitudes of these N phases, select a phase with the maximum magnitude from the N phases as a peak point output, and record the phase information of the peak point position; and, form an N-phase peak signal according to the magnitude and phase of the peak point extracted from the N-phase signal data; step 3, peak screening: perform two peak screenings in sequence, including an initial peak screening and a secondary peak screening; the initial peak screening is to perform peak detection of the N-phase peak signal by sliding window processing; and the secondary peak screening is peak window screening which selects a maximum peak in the window according to a set window length, after two peak screenings, obtain a maximum peak set which contains multiple maximum peaks; and, store peak noise, which includes: retain magnitude and phase information corresponding to maximum peak points, and set all data at positions other than the maximum peak points to 0 to obtain a noise set, the noise set corresponding to a noise signal; step 4, peak allocation: perform filter multiplication allocation on the maximum peaks in the noise set, wherein, a number of multipliers is allocated according to a density of the maximum peaks in the noise set; a filter is used to filter the noise signal to maintain the same spectral characteristics as the original input signal; step 5, generating multi-phase CPG coefficients: perform a convolution operation on a single-phase CPG coefficient and a multi-phase fractional delay filter to obtain multi-phase CPG coefficients including multiple groups of CPG coefficients, each group of CPG coefficient corresponds to a phase; the single-phase CPG coefficient is obtained by designing a filter with a same spectrum as the original input signal at the rate fs; step 6, delay compensation and cancellation signal generation: for noise of different phases within the noise set, perform a convolution operation on the noise and a CPG coefficient of the corresponding phase to obtain a CPG pulse signal of this phase after delay compensation, the CPG pulse signals of multiple phases are combined to obtain the peak cancellation signal.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS The application claims priority to Chinese patent application No. 202210099158.3, filed on Jan. 27, 2022, the entire contents of which are incorporated herein by reference. TECHNICAL FIELD This application relates to the field of PAPR reduction, and specifically, to a high-precision multi-phase CFR system and method and use. BACKGROUND Compared with a single-carrier system, an OFDM (Orthogonal Frequency Division Multiplex) system has multiple orthogonal subcarriers. If the phases of multiple subcarriers are consistent, the instantaneous power of the superimposed signal will be much greater than the average power of the signal, and the PAPR (Peak to Average Power Ratio, also known as PAR) will increase. If the instantaneous power exceeds the dynamic range of the PA (Power Amplifier), it will lead to nonlinear distortion, destroy the orthogonality of the sub-channels, cause mutual interference, and spectrum leakage will seriously affect adjacent channel signals. For example, in WiFi systems, WiFi6 supports higher-order modulation methods, and the unprocessed PAPR can reach 10 dB or even above 11 dB, which has a significant impact on the linear output of the PA. Therefore, before entering the PA, the PAPR of the modulated signal needs to be appropriately suppressed. In the prior art, CFR is usually used to reduce the PAPR, wherein CFR is the abbreviation of Crest Factor Reduction. Since CFR is a nonlinear process, if limiting is performed directly, it will cause problems such as spectrum leakage and EVM degradation. Therefore, the commonly used methods to reduce PAPR in engineering are the windowing function method (Windowing CFR), Noise Shaping CFR (abbreviated as NS-CFR) and Peak Cancellation CFR (abbreviated as PC-CFR). Among them, the basic principle of the PC-CFR is to use a pulse signal that matches the spectrum characteristics of the original input signal to cancel the peak of the original input signal, thereby achieving the purpose of reducing the PAPR of the signal. Compared with the Windowing CFR and Noise Shaping CFR, the PC-CFR has better peak cancellation performance and controllable resource consumption. When using the PC-CFR scheme for peak cancellation, processing only at a low rate may result in peak leakage or peak regeneration. Therefore, in order to increase the working rate of CFR, multi-stage cascading or multiple interpolation is usually used for peak cancellation at a high rate to obtain better peak cancellation performance. Referring to a typical peak cancellation algorithm shown in FIG. 1, a fractional delay scheme is used to perform peak search and peak cancellation at a single rate; a multi-stage cascade is used, and a fractional delay is performed on the signal between stages, which is equivalent to increasing the sampling rate of the data. And the internal processing of each stage is the same. As an example, FIG. 1 illustrates the use of a four-stage CFR module cascade, including peak cancellation module 1, peak cancellation module 2, peak cancellation module 3 and peak cancellation module 4. Each stage of CFR modules is connected using fractional delays, including fractional delay 1, fractional delay 2 and fractional delay 3. Assuming that the fractional delays are ½ delay, ¼ delay and ½ delay respectively, this is equivalent to increasing the signal to four times the sampling rate, which can reduce leakage peaks and peak regeneration. The peak allocation diagram of the signal after the fractional delay is shown in FIG. 2. It can be seen that a higher peak is generated after the fractional delay. This is because due to the limitation of the sampling rate, the real peak may be hidden at a higher rate. At this time, the single rate cannot see it and cannot eliminate the peak. If it is only processed at a low rate, the peak will still appear after high-rate interpolation, causing peak regeneration. By performing fractional time delay on the signal to remove the peak value of pipeline processing, the purpose of high-speed peak search and peak cancellation can be achieved. At the same time, the prior art also provides a method for reducing PAPR at a high rate at the symbol level. As an example, a method and device for reducing the PAPR of a signal is disclosed in Chinese patent z1200710122825.0, which discloses a method for performing peak cancellation in frequency domain symbols as shown in FIG. 3 and a method for performing peak cancellation in time domain symbols as shown in FIG. 4. By controlling the rotation factor of the frequency domain symbol or the fractional delay of the time domain symbol, and using multi-stage pipeline processing, the purpose of high-rate peak cancellation can be achieved. However, the above-mentioned several existing CFR processing methods achieve the purpose of peak cancellation at a high rate by pipeline processing at a low rate, which results in the process that requires multiple fractional delays or phase transf