Method for correcting frequency response mismatch error in time-alternating acquisition system

Abstract

This invention discloses a method for correcting frequency response mismatch error in a time-alternating acquisition system. First, the number of channels, the total number of sampling points, and the maximum number of frames of the time-alternating acquisition system are set. Then, logarithmic data is sampled through the time-alternating acquisition system, and the sampled data is divided into master and slave paths. Each frame of data in the master and slave paths is corrected. After the correction is completed, the corrected data of the master and slave paths are combined in time-alternating order. Then, the data of the master and slave paths are selected in order to obtain the final corrected output.

Description

Technical Field

[0001] This invention belongs to the field of time-domain testing technology, and more specifically, relates to a method for correcting frequency response mismatch error in a time-alternating acquisition system. Background Technology

[0002] Due to the current limitations of ADC specifications, a single ADC cannot meet the sampling rate requirements of data acquisition systems in fields such as communications and aerospace. Therefore, building a parallel architecture time-alternating acquisition system becomes an effective method to improve the system's sampling rate. Time-alternating acquisition uses M chips with a sampling rate of f. s The / M ADC increases the sampling rate of the acquisition system to f s This method improves the sampling rate by adjusting the sampling clock phase of M ADCs, causing them to sample the signal at fixed time intervals, and then combining the collected data. However, in practice, the signal enters the M ADCs through different acquisition channels. The inconsistency of the channel physical parameters, coupled with the imperfect sampling timing, results in inconsistent sampling results from each acquisition channel. This inconsistency is essentially caused by the frequency response mismatch between the acquisition channels in the time-interleaved acquisition system. These factors will seriously affect the system's acquisition performance, so it is necessary to correct the frequency response mismatch error.

[0003] Regarding methods for correcting frequency response mismatch errors, after obtaining the corrected frequency response of each acquisition channel, patent CN110557122A uses a cascaded filter bank design method of FIR filter + all-pass filter + fractional delay filter to approximate the corrected frequency response. However, this method requires a large amount of hardware resources and is not suitable for time-alternating acquisition systems with more than four channels. Furthermore, the IIR filter cannot process data in real time in the FPGA implementation structure. Patent CN113114241A designs a frequency response mismatch correction method based on cascaded overlapping FFT, which reduces the resource consumption of hardware implementation and ensures real-time data processing. However, when the number of channels in the time-alternating acquisition system increases to eight or more, its resource consumption is still large, making it difficult to apply to time-alternating acquisition systems with more than eight channels.

[0004] In summary, while current methods for frequency response mismatch error correction in time-alternating acquisition systems ensure real-time data processing and reduce hardware resource consumption, the increase in hardware resource consumption is significant as the number of channels increases, making hardware implementation difficult. Therefore, designing a simpler, more universal, and more resource-efficient method for arbitrary-channel frequency response mismatch error correction is of paramount importance. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for correcting frequency response mismatch error in a time-alternating acquisition system. Based on frequency domain decimation theory, it further reduces hardware resource consumption and data computation, so as to better adapt to multi-channel time-alternating acquisition systems and complete the correction of frequency response mismatch error.

[0006] To achieve the above-mentioned objective, the present invention provides a method for correcting frequency response mismatch error in a time-alternating acquisition system, characterized by comprising the following steps:

[0007] (1) Let the number of channels in the time-alternating acquisition system be M, and the total sampling rate be f. s The number of FFT points in each frame of the correction module is N, and N satisfies that N / M / 4 is an integer. The maximum number of FFT frames during the correction process is K. If the actual acquired data does not satisfy N*K, the total acquired data is made up of N*K by padding with zeros at the ends. Let G be the discrete correction frequency response of the m-th channel of the time-alternating acquisition system. m [n0], n0=0,1,…,N-1, m=0,1,…,M-1;

[0008] (2) Input the signal to be sampled into the time-alternating acquisition system, sample it through the time-alternating acquisition system and combine it in the time alternation order to obtain the sampled data y[n], n=0,1,…,N*K-1;

[0009] (3) The discrete corrected frequency response G m [n0] and delay factor Multiplying them together yields the discrete frequency response Q. m [n0];

[0010]

[0011] (4) Divide y[n] into master-slave path data;

[0012] The main path data input is y[0] to y[N*K-1]; the secondary path data input is controlled by a counter and is y[N / 2] to y[(2K-1)N / 2-1];

[0013] (5) Divide the input data of the master and slave paths into data frames;

[0014] The input data for the master and slave paths is divided into frames with N points each. The data for each frame of the master path is y[0]~y[N-1], y[N]~y[2N-1],…,y[N(K-1)]~y[N*K-1], for a total of K frames; the data for each frame of the slave path is y[N / 2]~y[3N / 2-1], y[3N / 2]~y[5N / 2-1],…,y[(2K-3)N / 2]~y[(2K-1)N / 2-1], for a total of K-1 frames.

[0015] (6) Correct each frame of data in the master-slave path;

[0016] The correction process for each frame of data in the master-slave path is consistent. Assume the data in the kth frame is y. k [n0], then the data correction steps for the k-th frame are as follows:

[0017] (6.1) Calculate y k The mean of [n0] is a k Then from y k Subtract the mean a from [n0] k This removes the DC bias and yields...

[0018]

[0019] (6.2) Regarding Perform an N-point FFT transform to obtain the frequency response Y. k [n0];

[0020] (6.3) Y k [n0] and Q obtained in step (3) m Multiplying [n0] sequentially yields the corrected frequency response.

[0021]

[0022] (6.4) Calculate the spectrum data after M times extraction;

[0023] Will Grouping the data into groups of N / M points, the resulting M groups are represented as follows:

[0024]

[0025] Where n1 = 0, 1, ..., M / N-1, l = 1, 2, ..., M;

[0026] Grouped data The data is accumulated and then multiplied by 1 / M to obtain the spectrum data after M times decimation.

[0027]

[0028] (6.5) will Perform an N / M-point IFFT transform, take the real part of the transformed data, and obtain the corrected data for a single frame, denoted as .

[0029] After correcting each frame of data in the master-slave path according to steps (6.1)-(6.5), the k-th frame data of the m-th channel of the master path is... The correction result is denoted as Data from the m-th channel and k-th frame of the path The calculation result is denoted as

[0030] (7) Merge the data after master-slave path correction;

[0031] (7.1) The M-frame data of the main path M channel after correction in step (6) are combined in alternating time order to obtain

[0032] The process of merging data on the master and slave paths is the same; similarly, the merged data on the slave path can be obtained.

[0033] (7.2) will and Add it to the calculated mean of the data in that frame to obtain and

[0034]

[0035]

[0036] (8) Select and merge the master and slave path data;

[0037] exist Select the first N / 4 points of data, and then alternate from N / 2 points with a period of N / 2 points. and Select the data from the middle N / 2 points of each frame of data, and then select... The last N / 4 points of data are then combined to form the final corrected output y. cali [n];

[0038]

[0039] The objective of this invention is achieved as follows:

[0040] This invention discloses a method for correcting frequency response mismatch errors in a time-alternating acquisition system. First, the length of the FFT within the correction module, the number of FFT frames, and the number of channels in the time-alternating acquisition system are specified, and the total number of sampling points is set to match the FFT length and frame number. The corrected discrete frequency response is delayed to obtain the delayed discrete corrected frequency response. Then, the sampled data is divided into master and slave paths, and after dividing the data into data frames, the mean of the data in each frame is calculated and the DC component is removed. Next, the spectrum of the data is obtained through the master and slave FFT modules. This spectrum is then multiplied sequentially with the delayed discrete corrected frequency response of each channel to obtain the corrected spectrum data for each frame of each channel. This spectrum data is divided into M groups, and the data at corresponding positions in the M groups are accumulated and averaged. This average is then fed into the IFFT output, and the DC component is added to obtain the corrected data for each frame of each channel after M times decimation. Finally, the corrected data of the master and slave paths are combined in time-alternating order, and then the data from the master and slave paths are selected sequentially to obtain the final corrected output.

[0041] Meanwhile, the method for frequency response mismatch error correction in a time-alternating acquisition system of the present invention also has the following beneficial effects:

[0042] (1) By using frequency domain decimation theory, the amount of IFFT calculation in the correction module is reduced. Compared with the original correction method, the number of transformation points of IFFT per channel is reduced to 1 / M of the original. Moreover, the implementation of frequency domain decimation is only an addition calculation, which reduces the implementation difficulty and is more conducive to the hardware implementation of the correction module.

[0043] (2) This invention addresses the correction of data acquired and stitched together from M channels. The correction is performed across M channels, with each channel correcting only the errors it introduces. After correction, the data from that channel is extracted from the corrected data, and the data is stitched together again. The entire correction process is performed in the frequency domain. This invention implements the extraction of corrected data in the frequency domain; that is, the IFFT process only calculates the points that need to be extracted for each channel. This differs from the previous method of first performing IFFT on all data and then extracting in the time domain. This reduces the number of points that need to be calculated during the IFFT process of M correction channels to 1 / M of the original number, thereby reducing hardware resource consumption and data computation. Attached Figure Description

[0044] Figure 1 This is a flowchart of a method for correcting frequency response mismatch error in a time-alternating acquisition system according to the present invention;

[0045] Figure 2 This is a schematic diagram of the frequency domain data grouping after M-fold data extraction;

[0046] Figure 3 This is a schematic diagram of the master-slave path data frame division and data selection;

[0047] Figure 4 This is a schematic diagram of the data after channel correction by a 4-channel time-alternating acquisition system;

[0048] Figure 5 This is a schematic diagram of the spectrum after the data has been extracted at twice the original value. Detailed Implementation

[0049] The specific embodiments of the present invention will now be described with reference to the accompanying drawings to enable those skilled in the art to better understand the invention. It should be particularly noted that in the following description, detailed descriptions of known functions and designs that might obscure the main content of the invention will be omitted here.

[0050] Example

[0051] Figure 1 This is a flowchart of the digital all-pass filter design method based on the hybrid particle swarm optimization algorithm of the present invention.

[0052] In this embodiment, as Figure 1 As shown, the present invention provides a method for correcting frequency response mismatch error in a time-alternating acquisition system, comprising the following steps:

[0053] (1) Assume the number of channels in the time-alternating acquisition system is M = 4, and the total sampling rate is f. s The number of points in each frame of the correction module's FFT is N = 1024, where N is an integer satisfying N / M / 4 = 64. The maximum number of FFT frames during the correction process is K = 10. If the actual acquired data does not meet the requirement of N*K, the total acquired data is padded with zeros at the ends to make it N*K. Let G be the discrete correction frequency response of the m-th channel of the time-alternating acquisition system. m [n0], where n0 = 0, 1, ..., 1023, m = 0, 1, 2, 3;

[0054] (2) Input the signal to be sampled into the time-alternating acquisition system, sample it through the time-alternating acquisition system and combine it in the time-alternating sequence to obtain the sampled data y[n], n=0,1,…,10239;

[0055] (3) The discrete corrected frequency response G m [n0] and delay factor Multiplying them together yields the discrete frequency response Q. m [n0];

[0056]

[0057] (4) Divide y[n] into master and slave path data: the master path data input is y[0] to y

[10239] , and the slave path data input is controlled by a counter to be y

[512] to y

[9728] .

[0058] (5) Divide the input data of the master and slave paths into data frames;

[0059] The input data for the master and slave paths is divided into frames of 1024 points each. The data for each frame of the master path is y[0]~y

[1023] , y

[1024] ~y

[2047] …y9216~y

[10239] , for a total of 10 frames. The data for each frame of the slave path is y

[512] ~y

[1535] , y

[1536] ~y

[2559] …y

[8704] ~y

[9727] , for a total of 9 frames.

[0060] (6) Correct each frame of data in the master-slave path;

[0061] The correction process for each frame of data in the master-slave path is consistent. Assume the data in the kth frame is y. k [n0], then the data correction steps for the k-th frame are as follows:

[0062] (6.1) Calculate y k The mean of [n0] is a k Then from y k Subtract the mean a from [n0] k This removes the DC bias and yields...

[0063]

[0064] (6.2) Regarding Perform an N-point FFT transform to obtain the frequency response Y. k [n0];

[0065] (6.3) Y k [n0] and Q obtained in step (3) m Multiplying [n0] sequentially yields the corrected frequency response.

[0066]

[0067] (6.4) Calculate the spectrum data after M times extraction;

[0068] Will The data is grouped into four sets of 256 points each, and the resulting four sets are as follows:

[0069]

[0070] Where n1 = 0, 1, ... 255, l = 1, 2, 3, 4;

[0071] Grouped data The data is accumulated and then multiplied by 1 / 4 to obtain the spectrum data after a 4x extraction.

[0072]

[0073] After decimating the data by a factor of M, the resulting signal's spectrum is equivalent to first expanding the original signal's spectrum by a factor of M, then shifting it along the ω-axis by a factor of 2πk (k = 1, 2, ..., M-1), reducing the amplitude to 1 / M of the original, and then superimposing the results. This corresponds to the spectrum data after the FFT, as shown below. Figure 2 As shown, the N-point spectrum data is divided into M groups of data in sequence and then the amplitude of each group of data is reduced to 1 / M of the original value before being superimposed.

[0074] (6.5) will Perform a 256-point IFFT transform, and take the real part of the transformed data to obtain the corrected data for a single frame, denoted as .

[0075] After correcting each frame of data in the master-slave path according to steps (6.1)-(6.5), the k-th frame data of the m-th channel of the master path is... The correction result is denoted as Data from the m-th channel and k-th frame of the path The calculation result is denoted as

[0076] (7) Merge the data after master-slave path correction;

[0077] (7.1) The M-frame data of the main path M channel after correction in step (6) are combined in alternating time order to obtain the data.

[0078] The process of merging data on the master and slave paths is the same; similarly, the merged data on the slave path can be obtained.

[0079] (7.2) will and Add it to the calculated mean of the data in that frame to obtain and

[0080]

[0081]

[0082] (8) Select and merge the master and slave path data;

[0083] exist Select the first 256 data points, and then alternate between them in a cycle of 512 points. and Select the middle 512 points of data from each frame of data, and finally select... The last 256 data points are then combined to form the final corrected output y. cali [n];

[0084]

[0085] The master-slave path data selection in the above formula is as follows: Figure 3 As shown, the shaded area represents the selected data from this frame; the first 256 data points of the final corrected output data are selected. The first 256 data points, subsequent data are arranged according to... The order is to select the middle 512 data points, and then select the last one. The last 256 data points.

[0086] Below, we will take a 4-channel time-alternating acquisition system as an example, such as... Figure 4 As shown, the extraction of data after 4-channel calibration is explained. The points to be extracted for each calibration channel are the shaded areas in the figure. It can be seen that the data extraction output of channel 0 is the result of extracting the entire data 4 times from the first data point, channel 1 is the result of extracting the entire data 4 times from the second data point, channel 2 is from the third data point, and channel 3 is from the fourth data point. Corresponding to the M-channel time-alternating acquisition system, the data extraction output of the m-th calibration channel is the result of extracting the entire data M times from the (m+1)-th data point.

[0087] After decimating the data by a factor of M, the resulting signal's spectrum is equivalent to first expanding the original signal's spectrum by a factor of M, then shifting it along the ω-axis by a factor of 2πk (k = 1, 2, ..., M-1), reducing the amplitude to 1 / M of the original value, and then superimposing the results. The time-domain data after FFT corresponds to the coordinates [0, 2π]. Taking a 2x decimation as an example... Figure 5 As shown, the data within [0, 2π] is first extended to multiple periods to obtain... Figure 5 In (a), (b) is obtained by expanding (a) by a factor of 2, and (c) is obtained by shifting (a) by 2π. The amplitude of the data in [0, 2π] in (b) and (c) is reduced to half of its original value, and then they are added together to obtain the spectrum data after the data is extracted by a factor of 2. According to Figure 5 Therefore, the spectrum data after a 2x data extraction is obtained by dividing the original spectrum data into two equal data segments, then adding the corresponding positions and dividing by 2. Correspondingly, the spectrum data after an Mx data extraction is obtained by dividing the N-point spectrum data into M groups of data in sequence, then reducing the amplitude of each group to 1 / M of the original value before superimposing them.

[0088] The M-fold decimation of the aforementioned data starts from the first data point of the entire data segment. However, during time-alternating acquisition system error correction, the starting data point for decimation of the corrected data differs across different correction channels. Therefore, the corrected data needs to be shifted so that the decimation of the corrected data for each channel starts from the first data point. In time-alternating acquisition system error correction, this shifting of the time-domain data corresponds to the FFT-generated spectrum data, which is equivalent to multiplying each spectrum data point by... Where n is the data point number of the spectrum data after FFT, starting from 0, N is the number of FFT points, and m is the number of points that need to be shifted in this correction channel. This step can be integrated into the frequency domain error correction process without consuming additional resources.

[0089] Although the illustrative specific embodiments of the present invention have been described above to enable those skilled in the art to understand the invention, it should be understood that the invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the invention as defined and determined by the appended claims, and all inventions utilizing the concept of the present invention are protected.

Claims

1. A method for correcting frequency response mismatch error in a time-alternating acquisition system, characterized in that, Includes the following steps: (1) Set the number of time-alternating acquisition system channels as M, the total sampling rate as f s , the number of points of each frame of FFT of the correction module as N, and N satisfies N / M / 4 is an integer, the maximum number of FFT frames in the correction process is K; if the actual acquisition data does not satisfy N*K, the total acquisition data quantity is made N*K through the end 0 supplementing mode; set the discrete correction frequency response of the mth channel of the time-alternating acquisition system as G m [n0], n0=0, 1,…, N-1, m=0, 1,…, M-1; (2) Input the signal to be sampled into the time-alternating acquisition system, sample it through the time-alternating acquisition system and combine it in the time alternation order to obtain the sampled data y[n], n=0,1,…,N*K-1; (3) The discrete corrected frequency response G m [n0] and delay factor Multiplying them together yields the discrete frequency response Q. m [n0]; (4) Divide y[n] into master-slave path data; The main path data input is y[0] to y[N*K-1]; the secondary path data input is controlled by a counter and is y[N / 2] to y[(2K-1)N / 2-1]; (5) Divide the input data of the master and slave paths into data frames; The input data for the master and slave paths is divided into frames with N points each. The data for each frame of the master path is y[0]~y[N-1], y[N]~y[2N-1],…,y[N(K-1)]~y[N*K-1], for a total of K frames; the data for each frame of the slave path is y[N / 2]~y[3N / 2-1], y[3N / 2]~y[5N / 2-1],…,y[(2K-3)N / 2]~y[(2K-1)N / 2-1], for a total of K-1 frames. (6) Correct each frame of data in the master-slave path; The correction process for each frame of data in the master-slave path is consistent. Assume the data in the kth frame is y. k [n0], then the data correction steps for the k-th frame are as follows: (6.1) Calculate y k The mean of [n0] is a k Then from y k Subtract the mean a from [n0] k This removes the DC bias and yields... (6.2) Regarding Perform an N-point FFT to obtain the frequency response Y. k [n0]; (6.3) Y k [n0] and Q obtained in step (3) m Multiplying [n0] sequentially yields the corrected frequency response. (6.4) Calculate the spectrum data after M times extraction; Will Grouping the data into groups of N / M points, the resulting M groups are represented as follows: Where n1 = 0, 1, ..., M / N-1, l = 1, 2, ..., M; Grouped data The data is accumulated and then multiplied by 1 / M to obtain the spectrum data after M times decimation. (6.5) will Perform an N / M-point IFFT transform, take the real part of the transformed data, and obtain the corrected data for a single frame, denoted as . After correcting each frame of data in the master-slave path according to steps (6.1)-(6.5), the k-th frame data of the m-th channel of the master path is... The correction result is denoted as Data from the m-th channel and k-th frame of the path The calculation result is denoted as (7) Merge the data after master-slave path correction; (7.1) The M-frame data of the main path M channel after correction in step (6) are combined in alternating time order to obtain the data. The process of merging data on the master and slave paths is the same; similarly, the merged data on the slave path can be obtained. (7.2) Calculation data and (8) Select and merge the master and slave path data; exist Select the first N / 4 points of data, and then alternate from N / 2 points with a period of N / 2 points. and Select the data from the middle N / 2 points of each frame of data, and then select... The last N / 4 points of data are then combined to form the final corrected output y. cali [n];

Images