Broadband noise generation method based on predistortion calibration

Abstract

The invention discloses a broadband noise generation method based on predistortion calibration, and relates to the technical field of communication, and the method comprises the steps: S1, setting a flatness threshold value, and screening out an original noise signal needing to be calibrated; s2, generating an inverse filter coefficient and a filter suppression coefficient for the original noise signal; s3, performing frequency response compensation by using an inverse filter coefficient, performing band-pass filtering by using a filter suppression coefficient, performing down-conversion to a baseband, and reducing a sampling rate; and S4, comparing the processed signal with a flatness threshold value, and if the processed signal does not exceed the flatness threshold value, indicating that signal calibration is completed, otherwise, entering calibration again. By introducing a pre-distortion calibration mechanism, the inverse filter coefficient is calculated in real time to compensate the system frequency response, the spectrum distortion caused by a fixed filter is effectively eliminated, and the method has the advantages of realizing dynamic and adaptive filter adjustment, effectively reducing the manual participation frequency and improving the continuous operation capability and the test efficiency of the system.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a method for generating broadband noise based on predistortion calibration. Background Art

[0002] In the fields of wireless communication testing, radar system calibration, and electronic countermeasures, broadband noise signals are the core benchmark for evaluating device performance. However, existing noise generation technologies have some limitations. For example, in application scenarios such as noise injection in communication base station transmitter testing, radar receiver sensitivity calibration, and verification of the anti-interference performance of electronic warfare equipment, the nonlinear distortion of the hardware system causes the noise signal to attenuate or distort in the high-frequency band, making it impossible to provide an accurate reference signal for device performance evaluation. Traditional methods typically use filters with fixed parameters to generate noise, which cannot adapt to dynamic environmental changes and lead to spectral distortion. When faced with the nonlinear characteristics of the hardware system, this method cannot adjust the noise signal in real time, resulting in the noise signal's spectrum being unable to meet test requirements and the parameters being unable to be adjusted dynamically in real time.

[0003] Currently, common noise generation methods use filters with fixed parameters, such as the single-filter noise generation method. This method is characterized by fixed filter parameters and an inability to compensate for dynamic nonlinear distortion. Because the nonlinear characteristics of hardware systems vary with environmental and operating condition parameters, fixed-parameter filters cannot adjust in real time to accommodate these changes. This results in significant distortion in the output noise signal, attenuating or distorting high-frequency signals, making the noise signal's spectrum uneven and further causing spectral distortion, making it unable to meet the requirements of high-precision testing. Furthermore, existing calibration technologies have a low level of automation, resulting in inefficiency caused by manual intervention and an inability to respond to system changes in real time. In complex test environments, manual calibration cannot adjust noise generation parameters in a timely manner, impacting test progress and accuracy. Summary of the Invention

[0004] The present invention provides a broadband noise generation method based on predistortion calibration, which solves the problem in the prior art of noise generation based on filters that the noise signal is easily distorted due to the lack of dynamic adjustment of the fixed parameters of the filter.

[0005] The present invention is achieved through the following technical solutions: A method for generating broadband noise based on predistortion calibration, the method comprising: Step S1: Generate an original noise signal, collect noise spectrum data of the original noise signal and mark it as first spectrum data, set a flatness threshold for the first spectrum data, input the original noise signal into the calibration process when the first spectrum data exceeds the flatness threshold, and output the original noise signal when the first spectrum data does not exceed the flatness threshold; Step S2: Smoothing the original noise signal entering the calibration process, performing predistortion parameter calculation on the processed original noise signal, generating inverse filter coefficients for compensating the frequency response, constructing a bandpass filter with frequency band constraints, constructing a Hamming window function for the bandpass filter, and generating a filter suppression coefficient for limiting the noise frequency band; Step S3: Using the inverse filter coefficients to perform frequency response compensation on the original noise signal, and using the filter suppression coefficients to perform bandpass filtering, down-converting the original noise signal after bandpass filtering to baseband through a digitally controlled oscillator and reducing the sampling rate, outputting the processed original noise signal and marking it as a calibration noise signal; Step S4: Collect the noise spectrum data of the calibration noise signal and mark it as the second spectrum data, compare the second spectrum data with the flatness threshold, and re-input the calibration noise signal into the calibration process when the second spectrum data exceeds the flatness threshold. When it does not exceed the flatness threshold, output the calibration noise signal to indicate that the signal calibration is completed.

[0006] Existing noise signal calibration processes typically utilize filters with fixed parameters, which cannot be adjusted in real time to accommodate these changes. This results in significant distortion in the output noise signal, attenuating or distorting high-frequency signals, making the noise signal spectrum uneven and further causing spectral distortion, making it unable to meet the requirements of high-precision testing. Based on this, the present invention provides a broadband noise generation method based on predistortion calibration, addressing the problem of noise signal distortion caused by the lack of dynamic adjustment of fixed filter parameters in existing filter-based noise generation processes.

[0007] Furthermore, the process of calculating the predistortion parameters of the original noise signal to generate the inverse filter coefficients includes: The frequency response curve of the original noise signal is collected, the target frequency response is preset, and a frequency response compensation function representing the inverse system response is constructed according to the frequency response curve. The frequency response compensation function is converted from the frequency domain response to the time domain impulse response through inverse FFT transformation, and the FIR filter coefficients generated after the conversion are set as the inverse filter coefficients.

[0008] Furthermore, the frequency domain least squares fitting method is used to construct the frequency response compensation function, and the construction process includes: The actual frequency response of the current system is collected at discrete frequency points, and a calculation formula for the time domain inverse filter coefficient is constructed using the target frequency response and the actual frequency response, and the calculation formula is converted into a frequency domain transformation matrix containing complex exponentials; a least squares objective function formula is constructed based on the time domain inverse filter coefficient, and the least squares objective function formula is converted into an error matrix formula containing a frequency domain transformation matrix formula; the FIR filter coefficients in the error matrix formula are calculated through the least squares solution, and the frequency response function generated by the FIR filter coefficients in the error matrix formula through FFT conversion is set as the frequency response compensation function.

[0009] Furthermore, the frequency domain transformation matrix is decomposed by singular values to obtain characteristic frequency components, the characteristic frequency components with the highest frequency values are selected as adjustment parameters and added to the error matrix to solve the FIR filter coefficients, and a singular value amplitude threshold is set for the singular values.

[0010] Furthermore, the process of constructing a window function using a Hamming window to generate a filter suppression coefficient includes: The Hamming window function is generated according to the length of the bandpass filter; the ideal impulse response of the bandpass filter is windowed using the Hamming window function; the frequency response function generated by the FFT transformation of the filter time domain coefficients generated after the windowing process is set as the filter suppression coefficient.

[0011] Furthermore, the functional form of the Hamming window function is set to: Let the Hamming window function be ω, let the number of sample points in the Hamming window function sequence be n, let the length of the bandpass filter be M; let the sidelobe adjustment coefficient be α, Then the functional form of the Hamming window function ω is set to: .

[0012] Furthermore, the sidelobe adjustment coefficient α of the window function is dynamically adjusted according to the passband bandwidth and out-of-band suppression index of the filter target frequency response.

[0013] Furthermore, the down-conversion process includes: Set the down-conversion target frequency, multiply the original noise signal after bandpass filtering by the output of the numerically controlled oscillator, and use orthogonal IQ modulation to perform down-conversion to baseband. Use the CORDIC algorithm for the baseband signal after down-conversion to baseband, mix the two components generated by orthogonal IQ modulation along the target frequency, and reduce the sampling rate of the baseband signal generated after mixing and mark it as the calibration noise signal.

[0014] Furthermore, a low-pass filter and a target sampling rate are set, and the low-pass filter sets the passband bandwidth according to the target sampling rate to meet the Nyquist sampling condition; the baseband signal generated after the mixing process is passed through the low-pass filter to eliminate the mixing image spectrum, and the baseband signal after eliminating the mixing image spectrum is marked as a calibration noise signal after the sampling rate is reduced.

[0015] Furthermore, a decimation operation is performed on the baseband signal after the mixing image spectrum is eliminated to reduce the sampling rate, and the decimation rate is the ratio of the original sampling rate of the original noise signal to the target sampling rate.

[0016] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1. By introducing a pre-distortion calibration mechanism, the inverse filter coefficients are calculated in real time to compensate for the system frequency response, enabling the noise spectrum to dynamically adapt to changes in system characteristics and effectively eliminating the spectrum distortion caused by fixed filters; 2. If the compensated spectrum still does not meet the requirements, the pre-distortion parameters can be iterated to achieve dynamic and adaptive filter adjustment, avoiding human intervention and improving calibration accuracy and automation level; 3. Automatically judge the quality of the noise spectrum and automatically adjust the filter coefficient, effectively reducing the frequency of manual participation and improving the system's continuous operation capability and test efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings: Figure 1 It is a flowchart of the present invention. DETAILED DESCRIPTION

[0018] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with examples and drawings. The exemplary embodiments of the present invention and their descriptions are only used to explain the present invention and are not intended to limit the present invention.

[0019] Example 1 like Figure 1 As shown, this embodiment is a method for generating broadband noise based on predistortion calibration, the method comprising: Step S1: Generate an original noise signal, collect noise spectrum data of the original noise signal and mark it as first spectrum data, set a flatness threshold for the first spectrum data, input the original noise signal into the calibration process when the first spectrum data exceeds the flatness threshold, and output the original noise signal when the first spectrum data does not exceed the flatness threshold; Step S2: Smoothing the original noise signal entering the calibration process, performing predistortion parameter calculation on the processed original noise signal, generating inverse filter coefficients for compensating the frequency response, constructing a bandpass filter with frequency band constraints, constructing a Hamming window function for the bandpass filter, and generating a filter suppression coefficient for limiting the noise frequency band; Step S3: Using the inverse filter coefficients to perform frequency response compensation on the original noise signal, and using the filter suppression coefficients to perform bandpass filtering, down-converting the original noise signal after bandpass filtering to baseband through a digitally controlled oscillator and reducing the sampling rate, outputting the processed original noise signal and marking it as a calibration noise signal; Step S4: Collect the noise spectrum data of the calibration noise signal and mark it as the second spectrum data, compare the second spectrum data with the flatness threshold, and re-input the calibration noise signal into the calibration process when the second spectrum data exceeds the flatness threshold. When it does not exceed the flatness threshold, output the calibration noise signal to indicate that the signal calibration is completed.

[0020] The original noise signal refers to the initial broadband noise signal without any preprocessing, compensation, or filtering correction. It is directly generated by the noise source and is used as the input signal for subsequent spectrum evaluation and calibration. The noise spectrum data of the original noise signal refers to the spectrum amplitude distribution data obtained after frequency domain analysis of the original generated noise signal. It describes the energy distribution of the noise signal at each frequency point. In specific implementations, it is typically a frequency-amplitude sequence and can be expressed in the form of a spectrum graph, spectrum density graph, etc. The flatness threshold is a tolerance range or limit value set after quantitatively measuring the spectral flatness. When the spectral flatness index of the actual noise signal exceeds this threshold, it is considered to be excessively distorted and requires calibration. In specific implementations, it can be set based on historical data or empirical rules. The inverse filter coefficients are used for frequency response compensation to offset frequency response distortion in the signal chain. If certain frequency bands of the system are severely attenuated or over-enhanced, the inverse filter reversely enhances the attenuated bands and suppresses the over-enhanced bands, so that the overall output approaches an ideal flat spectrum response. This is equivalent to inverting a nonlinear system to make its output close to the ideal state. For example, if the system's frequency response in the high frequency band is 0.5, representing a 50% attenuation, the inverse filter can compensate by setting a gain of 2 in the corresponding frequency band. The compensated noise signal will then present a compensated spectrum after passing through the system. Bandpass filtering using the filter's rejection coefficient controls the filter's ability to suppress non-target frequency bands, ensuring that the final noise signal remains within the set frequency range while suppressing energy leakage in other undesirable frequency bands. Down-converting the original noise signal to baseband using a numerically controlled oscillator (NCO) involves moving the high-frequency noise signal to a lower-frequency baseband for processing. The sampling rate is then appropriately reduced to minimize resource overhead while preserving key signal information, resulting in a more efficient output of a high-quality calibrated noise signal. Down-conversion involves shifting the entire spectrum of a high-frequency signal to a lower-frequency region. After bandpass filtering, the original noise signal remains at a relatively high frequency. The NCO is then used to mix it with a local oscillator signal, shifting its center frequency to near baseband 0Hz. Down-conversion reduces the signal's frequency range, and the required sampling rate is correspondingly lowered. This downsampling process typically employs low-pass anti-aliasing filtering and decimation. Whether to continue calibration is determined by judging whether the spectrum of the noise signal after calibration meets the flatness standard; if not, the compensation process is repeated; if it meets the standard, the calibration is completed and the result is output.

[0021] Furthermore, as a feasible implementation method, the process of calculating the predistortion parameters of the original noise signal to generate the inverse filter coefficients includes: The frequency response curve of the original noise signal is collected, the target frequency response is preset, and a frequency response compensation function representing the inverse system response is constructed according to the frequency response curve. The frequency response compensation function is converted from the frequency domain response to the time domain impulse response through inverse FFT transformation, and the FIR filter coefficients generated after the conversion are set as the inverse filter coefficients.

[0022] The original noise signal is sent into the system, and its frequency domain response is measured, and an actual frequency response function curve is obtained, which reflects the non-ideal characteristics of the system such as spectrum attenuation and distortion. An ideal target frequency response is set here, for example, a response with a flat spectrum and constant amplitude is required, which represents the expected standard to be achieved. The frequency response compensation function indicates that by applying this compensation to the input signal, the frequency distortion caused by the system can be offset. In a specific implementation, the frequency response compensation function can be constructed as the ratio of the preset target frequency response to the frequency response curve. The time domain impulse response is the impulse response of the FIR filter. This process converts the frequency domain compensation into a filter form that can be realized in the time domain. The generated FIR filter is called an inverse filter, and its coefficients are used for subsequent pre-distortion processing of the noise signal. By explicitly constructing a compensation function in the form of an ideal and actual frequency response ratio, correction can be directly performed on the frequency distortion area in the system response.

[0023] Furthermore, as a feasible implementation method, a frequency domain least squares fitting method is used to construct a frequency response compensation function, and the construction process includes: The actual frequency response of the current system is collected at discrete frequency points, and a calculation formula for the time domain inverse filter coefficient is constructed using the target frequency response and the actual frequency response, and the calculation formula is converted into a frequency domain transformation matrix containing complex exponentials; a least squares objective function formula is constructed based on the time domain inverse filter coefficient, and the least squares objective function formula is converted into an error matrix formula containing a frequency domain transformation matrix formula; the FIR filter coefficients in the error matrix formula are calculated through the least squares solution, and the frequency response function generated by the FIR filter coefficients in the error matrix formula through FFT conversion is set as the frequency response compensation function.

[0024] The actual gain and phase response data of the system are obtained at a set of selected frequency points for subsequent comparison and compensation with the target response. In actual implementation, the frequency response cannot be accurately obtained in the continuous frequency domain, so a set of discrete frequency points will be selected, and for each frequency point, the actual frequency response is calculated by measuring the ratio of the system output to the input. In order to compensate for the frequency distortion of the system, the response values of the actual system to these frequencies are collected at multiple specific frequency points for comparison with the target values, and a frequency compensation model or filter is constructed accordingly. The calculation formula of the time domain inverse filter coefficient means that by comparing the target frequency response with the actual frequency response, a calculation formula of the inverse filter is constructed, and it is converted into a frequency domain matrix form composed of complex exponentials, so that the FIR filter coefficients in the time domain can be calculated by the least squares method or other numerical methods, and finally used to compensate the frequency response of the noise signal. The least squares objective function is constructed based on the time domain inverse filter coefficients. In specific implementation, the process can be set as follows: first, set an FIR filter, whose coefficients (i.e., impulse response) are the time domain inverse filter coefficients that we want to optimize. After the filter acts on the original signal, its frequency response should be as close to the ideal response as possible. Construct a least squares objective function to measure the error of this approximation. Convert the least squares objective function into an error matrix containing a frequency domain transformation matrix, perform a DFT transformation on the filter impulse response, and construct a frequency domain transformation matrix. This matrix converts the sum of squares of the frequency response errors at the frequency points into a linear least squares form between the frequency domain transformation matrix and the time domain filter coefficients, so that the optimal FIR filter can be efficiently solved using linear algebra methods. As a feasible implementation method, the least squares objective function can be set as: , where k represents the discrete frequency point number, H i represents the target frequency response, H h represents the actual frequency response, f k represents the frequency value of the kth discrete frequency point, h represents the FIR filter coefficient (i.e., the impulse response of the FIR filter), and J(h) represents the least squares objective function.

[0025] Furthermore, as a feasible implementation method, the frequency domain transformation matrix is decomposed by singular values to obtain characteristic frequency components, the characteristic frequency components with the highest frequency values are selected as adjustment parameters and added to the error matrix to solve the FIR filter coefficients, and a singular value amplitude threshold is set for the singular values.

[0026] The characteristic frequency components refer to the frequency structural components represented by the left-hand singular vectors representing the principal directions extracted by singular value decomposition (SVD). These directions represent the main trends in the system's frequency-domain response in the spectral space. Each left-hand singular vector corresponds to a frequency characteristic component. The larger the singular value, the greater the contribution of that component to the frequency-domain variation. These principal components can be considered the frequency structural patterns where the system's energy is most concentrated in the frequency-domain response. In this embodiment, SVD is used to reveal which frequency components dominate the system's response in the spectral space and identify the high-energy frequency directions representing the principal components, providing a basis for optimizing FIR filter design. When designing an FIR filter based on the least squares method, SVD extracts the strongest principal frequency response (i.e., the eigenvalue corresponding to the largest singular value) from the frequency-domain matrix. This is then incorporated into the filter error model as a guiding or adjustment parameter, ensuring that the resulting FIR filter coefficients better align with the system's principal frequency characteristics, improving the accuracy and stability of frequency response compensation. The singular values set a singular value amplitude threshold to filter out frequency noise with near-zero amplitude in the system response.

[0027] Example 2 In this embodiment, the process of using the Hamming window to construct a window function to generate a filter suppression coefficient includes: The Hamming window function is generated according to the length of the bandpass filter; the ideal impulse response of the bandpass filter is windowed using the Hamming window function; the frequency response function generated by the FFT transformation of the filter time domain coefficients generated after the windowing process is set as the filter suppression coefficient.

[0028] In bandpass filter design, to make the ideal but infinitely long impulse response feasible in a practical system, a Hamming window function is used to perform windowing. This smoothes and truncates the ideal impulse response to generate time-domain coefficients of finite length and gradual edge transitions, thereby improving the filter's realizability and suppressing spectral leakage. The finite-length time-domain coefficients obtained by windowing the ideal bandpass filter impulse response are then subjected to an FFT transform to obtain their frequency response. This frequency response function is then used as a suppression coefficient for subsequent modulation of the noise signal spectrum, thereby controlling the frequency range and spectral flatness of the noise signal.

[0029] Furthermore, as a feasible implementation method, the functional form of the Hamming window function is set to: Let the Hamming window function be ω, let the number of sample points in the Hamming window function sequence be n, let the length of the bandpass filter be M; let the sidelobe adjustment coefficient be α, Then the functional form of the Hamming window function ω is set to: .

[0030] When constructing the bandpass filter's window function, a high-order cosine term (4πn / (M-1)) is introduced on top of the standard Hamming window to further control its frequency-domain sidelobe characteristics. The sidelobe adjustment coefficient α of this term is used to adjust the window function's frequency-domain sidelobe height, width, or steepness, making the filter's frequency response more controllable and adaptive. If the sidelobe adjustment coefficient α is greater than 0, an additional suppression notch is created in certain frequency ranges, enhancing out-of-band attenuation. If the sidelobe adjustment coefficient α is less than 0, the mainlobe edge is widened, improving passband flatness. Adjusting the sidelobe adjustment coefficient α can enhance suppression capability and adjust the mainlobe width. By introducing the additional high-order cosine term α∙cos(4πn / (M-1)), the Hamming window function enhances the ability to adjust the frequency response sidelobes. As a window function parameter, the sidelobe adjustment coefficient α can be flexibly set according to filter performance requirements, thereby controlling the out-of-band suppression level and passband edge shape in the frequency response, improving the frequency-domain shaping capability and adaptability of the bandpass filter. In particular, the sidelobe adjustment coefficient α of the window function is dynamically adjusted based on the passband bandwidth and out-of-band suppression index of the filter's target frequency response. It is clarified that the setting method of the sidelobe adjustment coefficient α in the window function is a dynamic adaptive adjustment mechanism; the sidelobe adjustment coefficient α is not a fixed constant, but is dynamically set according to the passband bandwidth requirements and out-of-band suppression performance index of the target filter, thereby achieving precise control of the window function shape, and giving the generated filter higher spectral shaping capabilities and environmental adaptability.

[0031] Example 3 In this embodiment, the down-conversion process includes: Set the down-conversion target frequency, multiply the original noise signal after bandpass filtering by the output of the numerically controlled oscillator, and use orthogonal IQ modulation to perform down-conversion to baseband. Use the CORDIC algorithm for the baseband signal after down-conversion to baseband, mix the two components generated by orthogonal IQ modulation along the target frequency, and reduce the sampling rate of the baseband signal generated after mixing and mark it as the calibration noise signal.

[0032] The original signal is multiplied by the output waveform of the digitally controlled oscillator, which is equivalent to shifting the signal spectrum from the center frequency to 0Hz in the baseband. Orthogonal IQ modulation is used to mix the input signal with the cosine and sine waves respectively to form two components, forming a complex baseband signal. This complex signal can fully represent the amplitude and phase information of the original high-frequency signal, facilitating frequency domain analysis and further filtering, sampling rate adjustment and other processing. The orthogonal IQ modulation method can completely preserve the phase and amplitude characteristics of the signal, providing a standardized complex baseband input for subsequent spectrum compensation and calibration processing. The CORDIC algorithm is an efficient algorithm based on iterative rotation. Using CORDIC to further rotate the IQ baseband signal is equivalent to performing a fine-tuning frequency compensation operation. The orthogonal IQ components refer to the product of the original signal and the cosine local oscillator and the product of the original signal and the sine local oscillator. Mixing along the target frequency means applying an equivalent rotation angle to the IQ complex signal through the CORDIC algorithm to achieve spectral offset or compensation, which is used to compensate for NCO frequency error or frequency fine-tuning during the down-conversion process. It is equivalent to fine-tuning the mixing in the frequency domain to align its spectrum more accurately with the target frequency. The mixed signal is then downsampled to reduce the data bandwidth, ultimately generating and outputting a calibrated noise signal with a flat spectrum and optimized sampling rate.

[0033] Furthermore, as a feasible implementation method, a low-pass filter and a target sampling rate are set, and the low-pass filter sets the passband bandwidth according to the target sampling rate to meet the Nyquist sampling condition; the baseband signal generated after the mixing process is passed through the low-pass filter to eliminate the mixing image spectrum, and the baseband signal after eliminating the mixing image spectrum is marked as a calibration noise signal after the sampling rate is reduced.

[0034] The low-pass filter is used to limit the frequency range of the baseband signal, and the target sampling rate is used to define the target sampling rate. The Nyquist sampling condition ensures that the sampled signal is free of aliasing. This condition causes the low-pass filter to remove frequency components above the target frequency before downsampling, effectively removing image signals outside the main frequency band after mixing, thereby creating a clean spectrum for subsequent decimation and downsampling. The mixing image spectrum refers to symmetrical frequency components in non-target frequency bands introduced during the mixing or downconversion process due to frequency folding or local oscillator (LO) image leakage. As a feasible implementation, the baseband signal, after eliminating the mixing image spectrum, is subjected to a decimation operation to reduce the sampling rate. The decimation rate is the ratio of the original sampling rate of the original noise signal to the target sampling rate. The decimation rate in the decimation operation is set to the ratio between the original noise signal sampling rate and the target sampling rate. This is used to downsample the filtered baseband signal, ensuring that the sampling process meets the Nyquist criterion and achieving data compression, thereby improving signal processing efficiency and system resource utilization.

[0035] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A broadband noise generation method based on predistortion calibration, characterized in that: The method includes: Step S1: Generate an original noise signal, collect noise spectrum data of the original noise signal and mark it as first spectrum data, set a flatness threshold for the first spectrum data, input the original noise signal into the calibration process when the first spectrum data exceeds the flatness threshold, and output the original noise signal when the first spectrum data does not exceed the flatness threshold; Step S2: Smoothing the original noise signal entering the calibration process, performing predistortion parameter calculation on the processed original noise signal, generating inverse filter coefficients for compensating the frequency response, constructing a bandpass filter with frequency band constraints, constructing a Hamming window function for the bandpass filter, and generating a filter suppression coefficient for limiting the noise frequency band; Step S3: Using the inverse filter coefficients to perform frequency response compensation on the original noise signal, and using the filter suppression coefficients to perform bandpass filtering, down-converting the original noise signal after bandpass filtering to baseband through a digitally controlled oscillator and reducing the sampling rate, outputting the processed original noise signal and marking it as a calibration noise signal; Step S4: Collect the noise spectrum data of the calibration noise signal and mark it as the second spectrum data, compare the second spectrum data with the flatness threshold, and re-input the calibration noise signal into the calibration process when the second spectrum data exceeds the flatness threshold. When it does not exceed the flatness threshold, output the calibration noise signal to indicate that the signal calibration is completed.

2. The method for generating broadband noise based on predistortion calibration according to claim 1, wherein: The process of calculating the predistortion parameters of the original noise signal to generate the inverse filter coefficients includes: The frequency response curve of the original noise signal is collected, the target frequency response is preset, and a frequency response compensation function representing the inverse system response is constructed according to the frequency response curve. The frequency response compensation function is converted from the frequency domain response to the time domain impulse response through inverse FFT transformation, and the FIR filter coefficients generated after the conversion are set as the inverse filter coefficients.

3. The method for generating broadband noise based on predistortion calibration according to claim 2, wherein: The frequency domain least squares fitting method is used to construct the frequency response compensation function. The construction process includes: The actual frequency response of the current system is collected at discrete frequency points, and a calculation formula for the time domain inverse filter coefficient is constructed using the target frequency response and the actual frequency response, and the calculation formula is converted into a frequency domain transformation matrix containing complex exponentials; a least squares objective function formula is constructed based on the time domain inverse filter coefficient, and the least squares objective function formula is converted into an error matrix formula containing a frequency domain transformation matrix formula; the FIR filter coefficients in the error matrix formula are calculated through the least squares solution, and the frequency response function generated by the FIR filter coefficients in the error matrix formula through FFT conversion is set as the frequency response compensation function.

4. The method for generating broadband noise based on predistortion calibration according to claim 3, wherein: The frequency domain transformation matrix is decomposed by singular values to obtain characteristic frequency components, the characteristic frequency components with the highest frequency values are selected as adjustment parameters and added to the error matrix to solve the FIR filter coefficients, and a singular value amplitude threshold is set for the singular values.

5. The method for generating broadband noise based on predistortion calibration according to claim 1, wherein: The process of using the Hamming window to construct a window function to generate a filter suppression coefficient includes: The Hamming window function is generated according to the length of the bandpass filter; the ideal impulse response of the bandpass filter is windowed using the Hamming window function; the frequency response function generated by the FFT transformation of the filter time domain coefficients generated after the windowing process is set as the filter suppression coefficient.

6. The method for generating broadband noise based on predistortion calibration according to claim 5, characterized in that: The functional form of the Hamming window function is set to: Let the Hamming window function be ω, let the number of sample points in the Hamming window function sequence be n, let the length of the bandpass filter be M; let the sidelobe adjustment coefficient be α, Then the functional form of the Hamming window function ω is set to: .

7. The method for generating broadband noise based on predistortion calibration according to claim 6, wherein: The sidelobe adjustment coefficient α of the window function is dynamically adjusted according to the passband bandwidth and out-of-band suppression index of the filter target frequency response.

8. The method for generating broadband noise based on predistortion calibration according to claim 1, wherein: The down-conversion process includes: Set the down-conversion target frequency, multiply the original noise signal after bandpass filtering by the output of the numerically controlled oscillator, and use orthogonal IQ modulation to perform down-conversion to baseband. Use the CORDIC algorithm for the baseband signal after down-conversion to baseband, mix the two components generated by orthogonal IQ modulation along the target frequency, and reduce the sampling rate of the baseband signal generated after mixing and mark it as the calibration noise signal.

9. The method for generating broadband noise based on predistortion calibration according to claim 8, wherein: A low-pass filter and a target sampling rate are set, wherein the low-pass filter sets a passband bandwidth according to the target sampling rate to meet the Nyquist sampling condition; a baseband signal generated after the mixing process is passed through the low-pass filter to eliminate the mixing image spectrum; the baseband signal after the mixing image spectrum is eliminated is subjected to a reduced sampling rate and marked as a calibration noise signal.

10. The method for generating broadband noise based on predistortion calibration according to claim 9, characterized in that: A decimation operation is performed on the baseband signal after the mixing image spectrum is eliminated to reduce the sampling rate, where the decimation rate is the ratio of the original sampling rate of the original noise signal to the target sampling rate.

Images