阅读说明：本技术 一种基于正频率频带复信号的mwc系统传感矩阵系数校准方法 (MWC system sensing matrix coefficient calibration method based on positive frequency band complex signal ) 是由 刘素娟 吕逍遥 于 2021-07-29 设计创作，主要内容包括：一种基于正频率频带复信号的MWC系统传感矩阵系数校准方法,属于高速模拟信息转换技术领域。MWC系统硬件带有的例如器件非线性、延迟和噪声等各种非理想因素的影响,改变了输入输出之间的关系,只有利用校准方法校准出实际传感矩阵中的系数才能成功重构信号。由于MWC系统将信号的奈奎斯特频率范围划分为若干个宽度相等的子带,所以可以设计一种只在正频率范围内的每两个相邻子带内各存在一个宽度小于一半子带宽度频带的测试信号来进行传感矩阵系数的校准,该种测试信号可以充分利用子带内的空间,一个测试信号即可校准得到每个通道的两组传感矩阵系数。(A method for calibrating a coefficient of a sensing matrix of an MWC system based on a positive frequency band complex signal belongs to the technical field of high-speed analog information conversion. The MWC system hardware has the influence of various non-ideal factors such as device nonlinearity, delay and noise, the relation between input and output is changed, and the signal can be successfully reconstructed only by calibrating coefficients in an actual sensing matrix by using a calibration method. Because the MWC system divides the Nyquist frequency range of the signal into a plurality of sub-bands with equal width, a test signal with a width less than half of the sub-band width frequency band respectively exists in every two adjacent sub-bands in the positive frequency range to calibrate the sensing matrix coefficients, the test signal can fully utilize the space in the sub-bands, and one test signal can be calibrated to obtain two groups of sensing matrix coefficients of each channel.)

1. A method for calibrating a coefficient of a sensing matrix of an MWC system based on a positive frequency band complex signal is characterized by comprising the following steps: designing a frequency band complex signal with frequency spectrum content only in a positive frequency sub-band as a test signal, wherein the frequency band of each test signal is respectively positioned at different sides of every two adjacent sub-bands, and the width of each frequency band is set to be less than 50% of the width of the sub-band; then, the designed test signals are divided into a real part and an imaginary part and respectively added into an MWC system, two groups of sampling sequences output by each channel are respectively synthesized into a group of complex sequences, Fast Fourier Transform (FFT) is carried out on the obtained complex sampling sequences to obtain FFT sequences, the FFT sequences are multiplied by corresponding points of a filter compensation sequence, the compensated FFT sequences corresponding to each frequency band are intercepted and are subjected to corresponding point division with the FFT sequences at the corresponding frequency band positions of the test signals, the average value is obtained, and the obtained average value is conjugated according to the conjugate relation of the sensing matrix coefficients, so that two groups of sensing matrix coefficients of each channel can be calibrated; before calibration, the frequency domain response of the low-pass filter in the actual circuit is measured, the compensation sequence of the filter is calculated, and then the actual low-pass filter is compensated into an ideal low-pass filter in the digital domain.

2. The method according to claim 1, characterized by the following specific steps:

the method comprises the following steps: collecting or setting various parameters of the MWC system: the Nyquist frequency is set to f_nyqThe number of physical channels is m, and the passband range of the ideal low-pass filter is set to [ -f [ ]_s/2，f_s/2]The sampling rate of the ADC is thenThe number of sampling points is respectively set to f_sAnd N, the period of the pseudo-random sequence is set to T_pSetting the number of high and low level changes in each period as M, wherein M is set as odd number and (M +1)/4 is integer, and satisfying f_s＝f_p＝1/T_pWherein f is_pDividing the width of a sub-band for the MWC system; from the above parameters, the MWC input signal has a spectrum inAre divided equally into M sub-bands in the range, each sub-band having a width f_p(ii) a Adding a sequence number to each sub-band, wherein the sequence number of each sub-band from negative frequency to positive frequency is-M₀，…，-1，0，1，…，M₀Wherein M is₀Determining the number of the test signals needing to be designed to be (M + 1)/4;

step two: measuring bilateral frequency response of the low-pass filter by using a frequency point scanning method, and discretizing the bilateral frequency response into a sequence H [ k ] of N points according to an arrangement form of FFT results (a positive frequency point value is on the left, a negative frequency point value is on the right), wherein the value of k is 0 to N-1; according to the relation of B [ k ] ═ Ik ]/Hk ], calculating a filter compensation sequence B [ k ], wherein Ik is an FFT sequence of an ideal low-pass filter;

step three: design test signal FFT sequence F with length of MN_r[n]Wherein, subscript R represents the serial number of the test signal, R takes the value of 1, 2., R, wherein, R ═ M +1)/4 is the total number of the test signal, and n takes the value range of 0 to MN-1; designed frequency bandwidth of B₀(B₀＝0.375f_p) Then it is in FFT sequence F_r[n]The equivalent point number in isHalf of the sub-band width equivalent point number isNumber of interval equivalent points isThe above calculation symbolsRepresents rounding down; then there is an FFT sequence F_r[n]The composition rule of (1) is: first, let the FFT sequence F be 1₁[n]N of (2)_GDot to Nth_G+N_B-1 points all having a value of 1, and letting the FFT sequence F₁[n]To (1) aPoint to the firstThe values of the points are all 1, and finally, the FFT sequence F is ordered₁[n]The values of other position points are all 0; then when r is more than or equal to 2, the r-1 th FFT sequence F_r-1[n]Circularly right-shifting 2N points to obtain an FFT sequence F_r[n]；

Step four: carrying out inverse Fourier transform on FFT sequences of R test signals constructed according to the rule to obtain a time domain sequence x of the test signals_r[n]Respectively taking the real part and the imaginary part of the obtained time domain sequence to obtain a real part sequence alpha_r[n]And imaginary sequence beta_r[n]；

Step five: the real part sequence alpha is transmitted by a signal generator_r[n]The generated test signal is input into an MWC system, and an output sequence a of each channel is acquired_r，i[k]Wherein the value range of k is 0 to N-1, i represents the serial number of each channel and the value of i is 1 to m; then the imaginary part sequence beta is processed by a signal generator_r[n]The generated test signal is input into an MWC system, and an output sequence b of each channel is acquired_r，i[k]Wherein the value range of k is 0 to N-1, and i represents that the value of the serial number of each channel is 1 to m; according to y_r，i[k]＝a_r，i[k]+b_r，i[k]J a relational construct sequence y_r，i[k]Wherein j is an imaginary unit;

step six: for the sequence y_r，i[k]Fast Fourier transform is carried out to obtain FFT sequence Y of the signal_r，i[k]Then will beY_r，i[k]And compensation filter sequence B k]Multiplying corresponding points to obtain compensated FFT sequence Y_r，i[k]；

Step seven: intercepting the sequence Y according to the spectrum shifting relation of the MWC system_r，i[k]N of (2)_GDot to Nth_G+N_B-1 point value constitutes a length of N_BSequence F of_r，i，OUTLAnd then truncating the sequence Y_r，i[k]To (1) aPoint to the firstThe value of a point constitutes a length of N_BSequence F of_r，i，OUTH；

Step eight: because the frequency band amplitudes of the test signals are all 1, the coefficient of the sensing matrix of the MWC system can be calculated according to the following simplified formula:

wherein the subscript l₁2r-2, the operator "SUM". represents the SUM of all elements of the vector, and the superscript "represents the conjugate of the complex number;

wherein the subscript l₂The operator "SUM" ("SUM") denotes vector calculation 2r-1The superscript "", indicates the conjugate of the complex number; and after all the coefficients are solved, the calibration of the sensing matrix C of the MWC system is completed.

Technical Field

The invention relates to a method for calibrating a coefficient of a sensing matrix of a Modulated Wideband Converter (MWC) based on a positive frequency band complex signal, and belongs to the technical field of high-speed analog information conversion.

Background

MWC is an analog information converter system based on the compressive sensing theory. The MWC is composed of a plurality of channels parallel random demodulation structures, each channel includes a mixer, an Analog low-pass filter, an ADC, and other devices, and the MWC performs mixing with a periodically varying high-frequency signal that randomly jumps between high and low levels, i.e., a pseudo-random sequence, then performs low-pass filtering, and finally performs low-speed sampling, compresses the high-frequency Analog signal, and performs sampling with a relatively low-speed Analog-to-Digital Converter (ADC). The MWC can be considered as a sampling structure, which can complete the sampling of the input signal at a sampling frequency far lower than the nyquist frequency of the signal under the condition that the input signal spectrum is sparse and unknown, and can perfectly reconstruct the original input signal waveform by using the sampling information and the reconstruction algorithm. The signal reconstruction is to solve a compressed sensing problem Y ═ AX by using a reconstruction algorithm, where Y is a matrix formed by a sampling sequence of outputs of the MWC system, a is a theoretical sensing matrix representing a relationship between inputs and outputs of the MWC system, and X represents an original input signal to be solved.

However, due to the influence of various non-ideal factors such as device nonlinearity, delay and noise carried by MWC system hardware, the relationship between input and output becomes Y ═ CX, the coefficient in the actual sensing matrix C greatly deviates from the theoretical value in a, and if signal reconstruction is directly performed by using the theoretical sensing matrix a, signal reconstruction fails or even signals cannot be reconstructed. Therefore, the coefficients in the actual sensing matrix need to be calibrated by using a calibration method, and the original signals can be accurately reconstructed only by performing signal reconstruction by using the calibrated sensing matrix C.

The currently proposed calibration method uses test signal types as follows: the frequency point signal of a single sine or cosine signal, a plurality of frequency point signals superposed by a plurality of sine or cosine signals and a single frequency band signal. The test signals are real signals, the positive frequency axis and the negative frequency axis of the frequency spectrum of the test signals have frequency spectrum content and are symmetrical about a zero frequency axis, and when the test signals are used for calibrating the MWC system sensing matrix, half of space in a sub-band divided by the MWC after low-pass filtering is wasted, namely, the contribution of the half of space to the coefficient calibration of the sensing matrix is zero. The invention aims to design a positive frequency band test signal to fully utilize the space in a sub-band to calibrate the coefficient of an MWC sensing matrix.

Disclosure of Invention

The invention aims to provide a method for calibrating the coefficient of a sensing matrix of an MWC system based on a positive frequency band test signal, the test signal can fully utilize the space in a sub-band to calibrate the coefficient of the sensing matrix, and the number of the coefficient of the sensing matrix calibrated by one test signal is twice of that of the original double-side frequency single-band test signal.

The invention is realized by adopting the following technical scheme:

since the MWC system divides the nyquist frequency range of the signal into several sub-bands with equal width, a band complex signal in which the spectral content exists only in the positive frequency sub-band can be designed as the test signal, and the frequency band of each test signal is respectively located at different sides of each two adjacent sub-bands, and the width of each frequency band is set to 37.5% of the sub-band width (the ratio is variable, but needs to be less than 50%). Then adding the real part and the imaginary part of the designed test signal into an MWC system respectively, synthesizing two groups of sampling sequences output by each channel into a group of complex sequences respectively, performing Fast Fourier Transform (FFT) on the obtained complex sampling sequences to obtain FFT sequences, multiplying the FFT sequences with corresponding points of a filter compensation sequence, intercepting the compensated FFT sequences corresponding to each frequency band position and the FFT sequences corresponding to the frequency band position of the test signal, performing corresponding point division and averaging, and conjugating the obtained average according to the conjugation relation of the sensing matrix coefficients to calibrate two groups of sensing matrix coefficients of each channel. In addition, it should be noted that before calibration, the frequency domain response of the low-pass filter in the actual circuit is measured and the compensation sequence of the filter is calculated, and then the actual low-pass filter is compensated to be the ideal low-pass filter in the digital domain.

The method comprises the following specific steps:

the method comprises the following steps: collecting or setting various parameters of the MWC system: the Nyquist frequency is set to f_nyqThe number of physical channels is m, and the passband range of the ideal low-pass filter is set to [ -f [ ]_s/2，f_s/2]Then the sampling rate and the number of sampling points of the ADC are set to f_sAnd N, the period of the pseudo-random sequence is set to T_pSetting the number of high and low level changes in each period as M, wherein M is set as odd number and (M +1)/4 is integer, and satisfying f_s＝f_p＝1/T_pWherein f is_pThe width of the sub-bands is divided for the MWC system. From the above parameters, the MWC input signal has a spectrum inAre divided equally into M sub-bands in the range, each sub-band having a width f_p. Adding a sequence number to each sub-band, wherein the sequence number of each sub-band from negative frequency to positive frequency is-M₀,…,-1,0,1,…,M₀Wherein M is₀And determining that the number of test signals required to be designed is (M + 1)/4.

Step two: and measuring bilateral frequency response of the low-pass filter by using a frequency point scanning method, and discretizing the bilateral frequency response into a sequence H [ k ] of N points according to the arrangement form of FFT results (positive frequency point values are on the left, negative frequency point values are on the right), wherein the value of k is 0 to N-1. And obtaining a filter compensation sequence B [ k ] according to the relation of B [ k ] ═ I [ k ]/H [ k ], wherein I [ k ] is an FFT sequence of an ideal low-pass filter.

Step three: design test signal FFT sequence F with length of MN_r[n]Wherein, the subscript R represents the serial number of the test signal, R takes the value of 1,2, …, R, wherein, R ═ M +1)/4 is the total number of the test signal, and n takes the value range of 0 to MN-1. Designed frequency bandwidth of B₀(B₀＝0.375f_p) Then it is in FFT sequence F_r[n]The equivalent point number in isHalf of the sub-band width equivalent point number isNumber of interval equivalent points isThe above calculation symbolsIndicating a rounding down. Then there is an FFT sequence F_r[n]The composition rule of (1) is: first, let the FFT sequence F be 1₁[n]N of (2)_GDot to Nth_G+N_B-1 points all having a value of 1, and letting the FFT sequence F₁[n]To (1) aPoint to the firstThe values of the points are all 1, and finally, the FFT sequence F is ordered₁[n]The values of other position points are all 0; then when r is more than or equal to 2, the r-1 th FFT sequence F_r-1[n]Circularly right-shifting 2N points to obtain an FFT sequence F_r[n]。

Step four: carrying out inverse Fourier transform on FFT sequences of R test signals constructed according to the rule to obtain a time domain sequence x of the test signals_r[n]Respectively taking the real part and the imaginary part of the obtained time domain sequence to obtain a real part sequence alpha_r[n]And imaginary sequence beta_r[n]。

Step five: the real part sequence alpha is transmitted by a signal generator_r[n]The generated test signal is input into an MWC system, and an output sequence a of each channel is acquired_r,i[k]Wherein the value range of k is 0 to N-1, i represents the serial number of each channel and the value of i is 1 to m; then the imaginary part sequence beta is processed by a signal generator_r[n]The generated test signal is input into an MWC system, and an output sequence b of each channel is acquired_r,i[k]Wherein the value range of k is 0 to N-1, and i represents that the serial number of each channel is 1 to m. According to y_r,i[k]＝a_r,i[k]+b_r,i[k]J a relational construct sequence y_r,i[k]Where j is an imaginary unit.

Step six: for the sequence y_r,i[k]Fast Fourier transform is carried out to obtain FFT sequence Y of the signal_r,i[k]Then, Y is added_r,i[k]And compensation filter sequence B k]Multiplying corresponding points to obtain a compensated FFT sequence Y'_r,i[k]。

Step seven: intercepting sequence Y 'according to spectrum shifting relation of MWC system'_r,i[k]N of (2)_GDot to Nth_G+N_B-1 point value constitutes a length of N_BSequence F of_r,i,OUTLAnd then truncating sequence Y'_r,i[k]To (1) aPoint to the firstThe value of a point constitutes a length of N_BSequence F of_r,i,OUTH。

Step eight: because the frequency band amplitudes of the test signals are all 1, the coefficient of the sensing matrix of the MWC system can be calculated according to the following simplified formula:

wherein the subscript l₁The operator "SUM". cndot.2 r-2 denotes the SUM of all elements of the vector, and the superscript "denotes the conjugate of the complex number.

Wherein the subscript l₂The operator "SUM". cndot.2 r-1 denotes the SUM of all elements of the vector, and the superscript "denotes the conjugate of the complex number. And after all the coefficients are solved, the calibration of the sensing matrix C of the MWC system is completed.

The invention has the beneficial effects that: the invention uses the positive frequency band complex signal as the calibration test signal of the MWC system, the sensor matrix coefficient obtained by calibration is the average value in the sub-band partial range, the calibrated sensor matrix coefficient has more reliability, and meanwhile, the calibration method fully uses the space in the sub-band, reduces the number of the test signals and improves the calibration efficiency.

Drawings

FIG. 1 is a diagram of a MWC system architecture;

FIG. 2 is a schematic diagram of MWC spectral partitioning and test signal spectra;

FIG. 3(a) is a time domain waveform of an original signal;

FIG. 3(b) is a time domain waveform of a signal reconstructed without the method of the present invention;

FIG. 3(c) is a time domain waveform of a reconstructed signal calibrated by the method of the present invention;

FIG. 4(a) is the original signal spectrum;

FIG. 4(b) is a diagram of a signal spectrum reconstructed without the method of the present invention;

fig. 4(c) shows the spectrum of the reconstructed signal after calibration by the method of the present invention.

Detailed Description

The following describes in detail an embodiment of the present invention with reference to the drawings and an example.

Fig. 1 is a schematic diagram of a modulated broadband converter (MWC) system. Taking the number of channels m-25 as an example, each channel is composed of a mixer, an analog low-pass filter and a low-speed analog-to-digital converter (ADC). The signal x (t) enters each channel simultaneously after passing through the beam splitter, and is associated with the pseudo-random sequence p_i(t) after mixing, sampling and outputting a sampling sequence y by filtering_i[k]Setting the Nyquist frequency f of the signal_nyq1GHz, pseudorandom sequence period T_p＝7.5×10^-8s, the number of times M of high and low level changes in each period is 75, that is, the number of divided sub-bands is 75, and the width f of each sub-band is_p13.33MHz, the passband of the analog low pass filter is set to 6.67MHz, 6.67MHz]If the ADC sampling rate is also set to 13.33MHz, and the sampling sequence length is set to N ═ 325, then the input signal length is known to be M × N ═ 24375.

FIG. 2 is a diagram of the spectral division and the test signal spectrum where there are two bands, one for eachWidth B of frequency band₀5MHz, corresponding to a length N in the FFT sequence_B121, the FFT sequence value in which the frequency band is located is set to 1.

According to the first step, various parameters of the MWC system are firstly collected and set, in this example, the channel number m is 25, and the nyquist frequency f is_nyq1GHz, pseudorandom sequence period T_p＝7.5×10^-8s, the number of times M of high and low level changes in each period is 75, that is, the number of divided sub-bands is 75, and the width f of each sub-band is_p13.33MHz, a sequence number is added to each sub-band, the sequence number of each sub-band from negative frequency to positive frequency is given by-M₀,…,-1,0,1,…,M₀Wherein M is₀The number R of required test signals is determined to be 19, 37. The passband range of the low-pass filter is set to be [ -6.67MHz, 6.67MHz]If the ADC sampling rate is also set to 13.33MHz, and the sampling sequence length is set to N ═ 325, then the input signal length is known to be M × N ═ 24375.

According to the second step, measuring the bilateral frequency response of the low-pass filter by a frequency point scanning method, and discretizing the bilateral frequency response into a 325-point sequence H [ k ] according to the arrangement form of FFT results (positive frequency point values are on the left, negative frequency point values are on the right), wherein the value of k is 0 to N-1. And obtaining a filter compensation sequence B [ k ] according to the relation of B [ k ] = I [ k ]/H [ k ], wherein I [ k ] is an FFT sequence of an ideal low-pass filter with the amplitude value of 1 in the cut-off frequency range and the amplitude value of 0 at the rest positions.

Designing a test signal FFT sequence F with the length of 24375 according to the third step_r[n]Wherein, the subscript R represents the serial number of the test signal, R has a value of 1,2, …, R, wherein R ═ (75+1)/4 ═ 19 is the total number of the test signals, and n has a value ranging from 0 to 24374. Designed frequency bandwidth of B₀＝0.375f_pAt 5MHz, it is in FFT sequence F_r[n]The equivalent point number in isHalf of the sub-band width equivalent point number isNumber of interval equivalent points isThe above calculation symbolsIndicating a rounding down. Then there is an FFT sequence F_r[n]The composition rule of (1) is: first, let the FFT sequence F be 1₁[n]The 20 th point to the 140 th point of the FFT sequence are all 1, and the FFT sequence F is made₁[n]The values from the 182 th point to the 302 th point are all 1, and finally, the FFT sequence F is ordered₁[n]The values of other position points are all 0; then when r is more than or equal to 2, the r-1 th FFT sequence F_r-1[n]Circularly right shifting 650 points to obtain an FFT sequence F_r[n]。

According to the fourth step, the FFT sequences of the 19 test signals constructed according to the rule are subjected to inverse Fourier transform to obtain a time domain sequence x of the test signal_r[n]Respectively taking the real part and the imaginary part of the obtained time domain sequence to obtain a real part sequence alpha_r[n]And imaginary sequence beta_r[n]。

According to step five, the real part sequence alpha is transmitted by a signal generator_r[n]The generated test signal is input into an MWC system, and an output sequence a of each channel is acquired_r,i[k]Wherein the value range of k is 0 to 324, i represents the serial number of each channel and the value of i is 1 to 25; then the imaginary part sequence beta is processed by a signal generator_r[n]The generated test signal is input into an MWC system, and an output sequence b of each channel is acquired_r,i[k]Where k ranges from 0 to 324 and i represents the number of each channel from 1 to 25. According to y_r,i[k]＝a_r,i[k]+b_r,i[k]J a relational construct sequence y_r,i[k]Where j is an imaginary unit.

According to the sixth step, for the sequence y_r,i[k]FFT conversion is carried out to obtain the FFT sequence Y_r,i[k]Then, Y is added_r,i[k]And compensation filter sequence B k]Multiplying corresponding points to obtain a compensated FFT sequence Y'_r,i[k]。

According to the seventh step, intercepting the sequence Y 'according to the spectrum shifting relation of the MWC system'_r,i[k]20 th to 140 thThe values of the points constitute a sequence F of length 121_r,i,OUTLAnd then truncating sequence Y'_r,i[k]From 182 th point to 302 th point of (a) form a length-121 sequence F_r,i,OUTH。

According to the eighth step, since the frequency band amplitudes of the test signals are all 1, the coefficient of the sensing matrix of the MWC system can be calculated according to the following simplified formula:

wherein the subscript l₁The operator "SUM". cndot.2 r-2 denotes the SUM of all elements of the vector, and the superscript "denotes the conjugate of the complex number.

Wherein the subscript l₂The operator "SUM". cndot.2 r-1 denotes the SUM of all elements of the vector, and the superscript "denotes the conjugate of the complex number. And after all the coefficients are solved, the calibration of the sensing matrix C of the MWC system is completed.

After the calibration is finished, the signal reconstruction test needs to be carried out on the calibration matrix, the evaluation index is the signal reconstruction signal-to-noise ratio,for the reconstructed signal, x is the input original signal,representing the square of the vector two norm. The original signal is set to have carrier frequency center at 100MHz, 200MHz and 300MHzThe frequency band signals, each frequency bandwidth is 15Mhz, and the signal-to-noise ratio of the input signal is 60 dB. As shown in fig. 3(a), fig. 3(b) reconstructs a time domain waveform of a signal by using an uncalibrated matrix, and fig. 3(c) reconstructs a time domain waveform of a signal by using a calibrated matrix. Fig. 4(a) is the frequency domain response of the original signal, fig. 4(b) is the frequency domain response of the reconstructed signal using the uncalibrated matrix, and fig. 4(c) is the frequency domain response of the reconstructed signal using the calibrated matrix. The result is a signal to noise ratio of 56dB reconstructed using the calibration matrix.