A multi-channel high linearity analog-to-digital conversion method and apparatus

Abstract

The present application relates to a kind of multi-channel high linearity analog-digital conversion method and device, belong to digital signal technical field, solve how to realize the low distortion receiving of analog signal and eliminate the distortion problem introduced by conversion device.Method includes: receiving training signal or scrambling signal from digital signal processing module and conversion into analog training signal or analog scrambling signal;Receive analog training signal or receive input analog signal and analog scrambling signal and superimposed as analog input signal;Analog training signal or analog input signal is decomposed and converted into multiple digital training signal or multiple digital input signal;Linear distortion characteristics and nonlinear distortion characteristics are obtained based on multiple digital training signal in training process to generate FIR filter and Volterra filter, and in subsequent real-time correction process, multiple digital input signal is filtered by FIR filter and Volterra filter Process to obtain high linear digital baseband signal.Linear and nonlinear distortion are eliminated by FIR and Volterra filter.

Description

Technical Field

[0001] This invention relates to the field of digital signal technology, and in particular to a multi-channel high linearity analog-to-digital conversion method and apparatus. Background Technology

[0002] Radio frequency (RF) signal receiving systems, with analog-to-digital converters (ADCs) at their core, are an indispensable part of various electromagnetic information devices. Current applications demand increasingly high-performance RF signal receiving systems, making the optimization of designs to achieve higher speed, higher accuracy, and lower power consumption a challenging research problem. In recent years, although the speed and quantization accuracy of ADCs have continuously improved, quantization errors, timing errors, and nonlinearities still hinder the overall performance improvement of RF signal receiving systems, making it difficult to meet the spurious-free dynamic range (SFDR) requirements of high-sensitivity signal reception scenarios.

[0003] Existing methods for nonlinear distortion, spurious suppression, or elimination can only handle certain types of spurious or harmonic, intermodulation, and other nonlinear distortions. If all distorted signals within the operating frequency band cannot be removed, the system's spurious-free dynamic range remains limited by the magnitude of the residual distorted signal, affecting the system's ability to receive low-power signals. Simply combining multiple different types of distortion correction methods would result in excessively high software and hardware resource requirements, significantly reducing the system's practicality. Summary of the Invention

[0004] Based on the above analysis, the embodiments of the present invention aim to provide a multi-channel high linearity analog-to-digital conversion method and apparatus to solve problems such as how to achieve low-distortion reception of analog signals and eliminate signal distortion introduced by the conversion device itself.

[0005] On one hand, embodiments of the present invention provide a multi-channel high-linearity analog-to-digital conversion method, comprising: receiving a training signal or a scrambled signal from a digital signal processing module and converting it into an analog training signal or an analog scrambled signal; receiving the analog training signal, or receiving an input analog signal and the analog scrambled signal and superimposing them into one analog input signal; decomposing the analog training signal or the analog input signal and converting it into multiple digital training signals or multiple digital input signals; and during the training state, obtaining linear distortion characteristics and nonlinear distortion characteristics based on the multiple digital training signals to generate an FIR filter and a Volterra filter, and during the subsequent real-time correction state, filtering the multiple digital input signals through the FIR filter and the Volterra filter to obtain one high-linearity digital baseband signal.

[0006] The beneficial effects of the above technical solution are as follows: The linear and nonlinear distortion characteristics of the internal channels of the receiving device can be obtained by transmitting a single training signal, and these characteristics can be converted into FIR filter and Volterra filter coefficients. By combining channel linearization technology, nonlinear distortion correction technology, and scrambling technology in a hardware and software architecture, the beneficial effects of the above technologies can be achieved simultaneously with a single hardware setup.

[0007] Further improvements to the above method, obtaining linear distortion and nonlinear distortion characteristics based on the multi-channel digital training signals to generate FIR filters and Volterra filters during the training process, further include: when the multi-channel high-linearity analog-to-digital converter is powered on for the first time, closing the external analog signal input channel, and the digital signal processing module entering the training state; generating first and second single-tone signals of different frequencies through the training signal generation module and outputting analog training signals, the analog training signals being output as multi-channel digital signals as training feedback signals via the analog combiner and the multi-channel receiving channels; extracting linear distortion and nonlinear distortion characteristics of each receiving channel based on the training signals and the training feedback signals; generating FIR filter coefficients corresponding to the linear distortion characteristics and Volterra filter coefficients corresponding to the nonlinear distortion characteristics according to the linear distortion characteristics and the nonlinear distortion characteristics, and setting the FIR filter and the Volterra filter based on the FIR filter coefficients and the Volterra filter coefficients, respectively.

[0008] A further improvement to the above method, generating FIR filter coefficients corresponding to the linear distortion characteristics based on the nonlinear distortion characteristics, further includes: expressing the linear distortion characteristics of the j-th receiving channel using the following formula:

[0009] h j =g j (d) / f(d);

[0010] The first FIR filter coefficient v1, corresponding to the linear distortion characteristics of the first receiving channel in the j-channel receiving process, is calculated using the following formula:

[0011]

[0012] The FIR filter coefficients v corresponding to the linear distortion characteristics of the second to nth receiving channels in the j-channel receiver are calculated using the following formula. j :

[0013]

[0014] d = b / Δf + NS / 2;

[0015] M = exp(-2πi·b) T u / f s );

[0016] B = [b1; b2], b = vec(B);

[0017] b1 = F start +iΔQ,i∈0,…,N R -1;

[0018] b2 = F start +iΔQ+ΔP,i∈0,…,N R -1;

[0019] Δf=f s / N S ;

[0020] N R =round((F end -F start ) / ΔR);

[0021] Among them, h j G represents the linear distortion characteristic of the j-th receiving channel; j =G(:,j) represents the spectrum of the training output signal received by the j-th receiving channel; Represents the spectrum. The expression represents the Fourier transform; M represents the linear training matrix; u = [-L / 2, -L / 2+1, ..., L / 2-1] represents a vector; L is the maximum number of taps in the equivalent FIR filter; fs represents the baseband sampling rate; vec(·) represents the matrix vectorization operation; d is a matrix of length 2N. R The vector; b1 represents a single audio point in the spectrum of x1; b2 represents a single audio point in the spectrum of x2; N R Indicates the number of receive channels; F start Indicates the system's initial operating frequency; f s Δf represents the baseband sampling rate; ΔQ and ΔP represent positive integer multiples of Δf; N R This indicates the number of single tones required within the operating frequency band, and round is the rounding operation.

[0022] Further improvements to the above method, generating Volterra filter coefficients corresponding to the nonlinear distortion characteristics based on the nonlinear distortion characteristics, further include: expressing the spectrum g of the j-th receiving channel using the following formula. j The location where the spectral values ​​of the k-th order nonlinear distortion characteristic appear in the table:

[0023]

[0024] The following formula represents the spectral vector of the k-th order nonlinear distortion characteristic of the j-th receiving channel:

[0025]

[0026] The equivalent Volterra filter coefficients corresponding to the k-th order nonlinear distortion characteristics of the j-th receiving channel are calculated using the following formula.

[0027]

[0028]

[0029]

[0030]

[0031]

[0032] w * =N S –w+2;

[0033] in, The training matrix represents the k-th order nonlinear intermodulation distortion of the j-th receiving channel; The equivalent Volterra filter coefficients corresponding to the k-th order nonlinear intermodulation distortion characteristics of the j-th receiving channel are represented; w1 represents a vector of size M×1 and w2 represents a vector of size N×1; vec(·) represents matrix vectorization operation, and unique(·) represents removing duplicate element values ​​from the vector; (·) T This represents the matrix transpose operation; (w) K This indicates that K vectors w are processed sequentially. Operations; Indicates the spectrum g of the j-th receiving channel j In the diagram, the location of the k-th order nonlinear intermodulation distortion characteristic spectrum value corresponding to this receiving channel is shown. This represents the spectral value vector of the k-th order nonlinear distortion characteristic corresponding to the j-th receiving channel; For y j The conjugate signal of (n).

[0034] Further improvements to the above method, in the subsequent real-time correction process, filtering the multiple digital input signals using the FIR filter and the Volterra filter to obtain a high-linearity digital baseband signal further include: after the training state is completed, enabling external analog signal input, and the digital signal processing module entering the real-time correction state; generating a continuous scrambling signal through the digital signal processing device, and combining the continuous scrambling signal with the external analog signal into a single analog signal; decomposing the single analog signal into multiple analog signals with the same or different power and converting the multiple analog signals into multiple digital signals; and linearizing the multiple digital signals according to the linear distortion characteristics and the nonlinear distortion characteristics, then performing spurious frequency point identification and elimination processing, and outputting a high-linearity digital baseband signal.

[0035] Further improvements to the above method include linearizing the multiple digital signals based on the linear distortion characteristics and the nonlinear distortion characteristics, which further comprises: aligning the multiple digital signals through the FIR filter, wherein channels 2 to n have the same linear distortion characteristics as channel 1; subtracting the scrambling signal processed by FIR filter 1 from the aligned multiple digital signals to generate aligned multiple digital signals without scrambling; and then passing the aligned multiple digital signals without scrambling through a K-order Volterra filter to remove the K-order nonlinear distortion signal to generate multiple first digital signals, wherein K is the order of nonlinear distortion present in the operating frequency band, which varies with the frequency band range and the nonlinear characteristics of the device.

[0036] Further improvements to the above method, including spurious frequency point identification and elimination processing and outputting a high linear digital baseband signal, further include: the multi-channel first digital signal is decomposed into frame signals with fixed sample lengths by a frame segmentation module; Fourier FFT processing is performed on each decomposed frame signal to obtain the spectrum of each frame; point-by-point comparison and selection processing is performed on the spectrum of each frame obtained from each channel, and by comparing the magnitude difference of the frequency slot spectrum signal components in each frame spectrum of each channel, it can be identified whether the current signal is an input signal or a spurious signal, and the identified spurious signals are removed one by one to obtain a single-channel spurious-free baseband signal; the single-channel spurious-free baseband signal is processed by inverse Fourier IFFT, and overlapping frames are merged to remove the overlapping part of the signal to generate a continuous high linear time-domain data stream.

[0037] On the other hand, embodiments of the present invention provide a multi-channel high-linearity analog-to-digital converter, comprising: a training and scrambling module for receiving a training signal or a scrambling signal from a digital signal processing module and converting it into an analog training signal or an analog scrambling signal; an analog combiner for receiving the analog training signal, or simultaneously receiving an external analog signal and the analog scrambling signal and superimposing them into one analog input signal; an analog splitter for decomposing the analog training signal or the analog input signal into multiple analog training signals or multiple analog input signals with the same or different power; and multiple receiving channels for converting the multiple analog training signals or the multiple analog input signals into multiple digital training signals or multiple digital input signals; wherein, during the training state, the digital signal processing module obtains linear distortion characteristics and nonlinear distortion characteristics based on the multiple digital training signals to generate an FIR filter and a Volterra filter, and during the subsequent real-time correction state, the multiple digital input signals are filtered by the FIR filter and the Volterra filter to obtain one high-linearity digital baseband signal.

[0038] Based on further improvements to the above-mentioned device, the digital signal processing module includes: a training start-up module, a training signal generation module, a feature extraction module, and a distortion elimination module. The training start-up module is used to close the external analog signal input channel when the multi-channel high-linearity analog-to-digital converter is initially powered on, allowing the digital signal processing module to enter a training state. The training signal generation module is used to generate first and second single-tone signals of different frequencies as the training signal. The training and scrambling module performs digital-to-analog conversion on the first and second single-tone signals of different frequencies and combines them into the analog training signal. The multiple receiving channels are used to receive signals via the analog signal and the scrambling module. The circuit device and the analog splitter device receive the multiple analog training signals and convert them into multiple digital training signals; the feature extraction module is used to extract the linear distortion characteristics and nonlinear distortion characteristics of each receiving channel based on the training signals and the training feedback signals; the distortion cancellation module is used to generate FIR filter coefficients corresponding to the linear distortion characteristics and Volterra filter coefficients corresponding to the nonlinear distortion characteristics, respectively, and set the FIR filter and the Volterra filter based on the FIR filter coefficients and the Volterra filter coefficients, respectively.

[0039] Based on further improvements to the above-mentioned device, the feature extraction module is used to represent the linear distortion characteristics of the j-th receiving channel using the following formula:

[0040] h j =gj (d) / f(d);

[0041] The spectrum g of the j-th receiving channel is expressed by the following formula. j The location where the spectral values ​​of the k-th order nonlinear distortion characteristic appear in the table:

[0042]

[0043] The following formula represents the spectral vector of the k-th order nonlinear distortion characteristic of the j-th receiving channel:

[0044]

[0045] The distortion cancellation module is used to calculate the first FIR filter coefficient v1 corresponding to the linear distortion characteristics of the first receiving channel in the j-channel receiving channels using the following formula:

[0046]

[0047] The FIR filter coefficients v corresponding to the linear distortion characteristics of the second to nth receiving channels in the j-channel receiver are calculated using the following formula. j :

[0048]

[0049] d = b / Δf + N S / 2;

[0050] M = exp(-2πi·b) T u / f s );

[0051] B = [b1; b2], b = vec(B);

[0052] b1 = F start +iΔQ,i∈0,…,N R -1;

[0053] b2 = F start +iΔQ+ΔP,i∈0,…,N R -1;

[0054] Δf=f s / N S ;

[0055] N R =round((F end -F start ) / ΔR);

[0056] Among them, h j G represents the linear distortion characteristic of the j-th receiving channel;j =G(:,j) represents the spectrum of the training output signal received by the j-th receiving channel; Represents the spectrum. The expression represents the Fourier transform; M represents the linear training matrix; u = [-L / 2, -L / 2+1, ..., L / 2-1] represents a vector; L is the maximum number of taps in the equivalent FIR filter; fs represents the baseband sampling rate; vec(·) represents the matrix vectorization operation; d is a matrix of length 2N. R The vector; b1 represents a single audio point in the spectrum of x1; b2 represents a single audio point in the spectrum of x2; N R Indicates the number of receive channels; F start Indicates the system's initial operating frequency; f s Δf represents the baseband sampling rate; ΔQ and ΔP represent positive integer multiples of Δf; N R This indicates the number of single tones required within the operating frequency band; round is the integer operation.

[0057] The equivalent Volterra filter coefficients corresponding to the k-th order nonlinear distortion characteristics of the j-th receiving channel are calculated using the following formula.

[0058]

[0059]

[0060]

[0061]

[0062]

[0063] w * =N S –w+2;

[0064] in, The training matrix represents the k-th order nonlinear intermodulation distortion of the j-th receiving channel; The equivalent Volterra filter coefficients corresponding to the k-th order nonlinear intermodulation distortion characteristics of the j-th receiving channel are represented; w1 represents a vector of size M×1 and w2 represents a vector of size N×1; vec(·) represents matrix vectorization operation, and unique(·) represents removing duplicate element values ​​from the vector; (·) T This represents the matrix transpose operation; (w) K This indicates that K vectors w are processed sequentially. Operations; Indicates the spectrum g of the j-th receiving channel jIn the diagram, the location of the k-th order nonlinear intermodulation distortion characteristic spectrum value corresponding to this receiving channel is shown. This represents the spectral value vector of the k-th order nonlinear distortion characteristic corresponding to the j-th receiving channel; For y j The conjugate signal of (n).

[0065] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0066] 1. By combining channel linearization technology, nonlinear distortion correction technology, spurious emission technology and scrambling technology in a hardware and software architecture, the beneficial effects of the above technologies can be achieved simultaneously with the same hardware settings;

[0067] 2. The linear and nonlinear distortion characteristics of the internal channel of the receiving device can be obtained by transmitting a training signal once, and can be converted into FIR filter and Volterra filter coefficients.

[0068] 3. Distorted signals were eliminated and spurious-free dynamic range was improved using relatively few computing resources;

[0069] 4. By comprehensively utilizing the spectrum information of multiple channels, irregular spurious distortion signals in the received signal are eliminated.

[0070] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0071] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0072] Figure 1 This is a flowchart of a multi-channel high linearity analog-to-digital conversion method according to an embodiment of the present invention;

[0073] Figure 2 This is a schematic diagram of the structural composition of an analog-to-digital converter according to an embodiment of the present invention;

[0074] Figure 3 This is a structural diagram of the internal components of the training and correction module according to an embodiment of the present invention;

[0075] Figure 4 This is a structural diagram of the analog-to-digital converter in training mode according to an embodiment of the present invention;

[0076] Figure 5 This is a structural diagram of the analog-to-digital converter in real-time correction mode according to an embodiment of the present invention;

[0077] Figure 6a and Figure 6b This is a typical scrambled signal spectrum diagram according to an embodiment of the present invention;

[0078] Figure 6c This is an example diagram of the time-domain waveform of the scrambled signal according to an embodiment of the present invention;

[0079] Figure 7a , Figure 7b and Figure 7c These are schematic diagrams of the spectra of the first single-tone signal T1 and the second single-tone signal T2 at the input port and the simulated training signal Tout at the output port of the training and scrambling module according to an embodiment of the present invention during the training process.

[0080] Figure 8a and Figure 8b The effects of linear distortion and nonlinear distortion in the output signal of a single receiving channel according to embodiments of the present invention on the spectrum of the training output signal are respectively described.

[0081] Figure 9 This is a flowchart illustrating the internal processing of a digital signal linearization module according to an embodiment of the present invention.

[0082] Figure 10 This is a flowchart illustrating the internal processing of the spurious frequency identification and elimination module according to an embodiment of the present invention.

[0083] Figure 11 This is a schematic diagram of dividing a continuous time-domain signal into frame signals in the spurious frequency point identification and elimination module according to an embodiment of the present invention;

[0084] Figure 12 This is a schematic diagram of the point-by-point comparison and selection module processing procedure in the spurious frequency point identification and elimination module according to an embodiment of the present invention;

[0085] Figure 13a and Figure 13b The diagrams show the output signal spectrum of a constant-scale analog-to-digital converter and the high-linearity output signal spectrum processed by the multi-channel high-linearity analog-to-digital converter according to an embodiment of the present invention.

[0086] Figure 14 This is a schematic diagram illustrating the main spectral changes of a two-tone signal during real-time correction according to an embodiment of the present invention;

[0087] Figure 15 This diagram illustrates the spectral changes that occur at each step when a two-tone signal is processed by a digital signal linearization module according to an embodiment of the present invention. Detailed Implementation

[0088] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0089] refer to Figure 1 A specific embodiment of the present invention discloses a multi-channel high-linearity analog-to-digital conversion method, comprising: in step S101, receiving a training signal or scrambling signal from a digital signal processing module and converting it into an analog training signal or an analog scrambling signal; in step S102, receiving an analog training signal, or receiving an input analog signal and an analog scrambling signal and superimposing them into one analog input signal; in step S103, decomposing the analog training signal or analog input signal and converting it into multiple digital training signals or multiple digital input signals; and in step S104, during the training state, obtaining linear distortion characteristics and nonlinear distortion characteristics based on the multiple digital training signals to generate an FIR filter and a Volterra filter, and in the subsequent real-time correction state, filtering the multiple digital input signals through the FIR filter and the Volterra filter to obtain one high-linearity digital baseband signal.

[0090] Compared with existing technologies, the multi-channel high-linearity analog-to-digital conversion method provided in this embodiment can obtain the linear and nonlinear distortion characteristics of the internal channels of the receiving device by transmitting a single training signal, and can convert the linear and nonlinear distortion characteristics into FIR filter and Volterra filter coefficients. By combining channel linearization technology, nonlinear distortion correction technology and scrambling technology in a hardware and software architecture, the beneficial effects of the above technologies can be achieved simultaneously with the same hardware settings.

[0091] The following text will refer to Figures 1 to 3 The steps of the multi-channel high linearity analog-to-digital conversion method according to embodiments of the present invention will be described in detail.

[0092] In step S101, a training signal or scrambling signal is received from the digital signal processing module and converted into an analog training signal or an analog scrambling signal. Specifically, the training signal or scrambling signal includes a first single-tone signal and a second single-tone signal of different frequencies. The first single-tone signal is converted from digital to analog to a first analog signal by a digital-to-analog converter DAC1, and the second single-tone signal is converted from digital to analog to a second analog signal by a digital-to-analog converter DAC2. An analog combiner receives the first analog signal and the second analog signal via front-end 1 and front-end 2 respectively, and then combines the first analog signal and the second analog signal into an analog training signal or an analog scrambling signal.

[0093] In step S102, an analog training signal is received, or an input analog signal and an analog scrambling signal are received and superimposed into a single analog input signal. Specifically, during the training state, external analog signal input is turned off, and only the analog training signal is received through the analog combiner. After the training is completed, the system enters the subsequent real-time correction state, and external analog signal input is turned on. The input analog signal and the analog scrambling signal can then be received simultaneously through the analog combiner.

[0094] In step S103, the analog training signal or analog input signal is decomposed and converted into multiple digital training signals or multiple digital input signals. Specifically, the analog training signal or analog input signal is decomposed into analog training signals or analog input signals with the same or different power using an analog splitter.

[0095] In step S104, during the training state, linear distortion characteristics and nonlinear distortion characteristics are obtained based on multiple digital training signals to generate FIR filters and Volterra filters. In the subsequent real-time correction state, multiple digital input signals are filtered by FIR filters and Volterra filters to obtain a single high linearity digital baseband signal.

[0096] Specifically, the process of obtaining linear and nonlinear distortion characteristics based on multiple digital training signals to generate FIR and Volterra filters during the training process further includes: when the multi-channel high-linearity analog-to-digital converter is powered on for the first time, the external analog signal input channel is closed, and the digital signal processing module enters the training state; the training signal generation module generates first and second single-tone signals of different frequencies and outputs analog training signals, which are then output as multiple digital signals as training feedback signals via an analog combiner and multiple receiving channels; the linear and nonlinear distortion characteristics of each receiving channel are extracted based on the training signals and training feedback signals; FIR filter coefficients corresponding to the linear distortion characteristics and Volterra filter coefficients corresponding to the nonlinear distortion characteristics are generated according to the linear and nonlinear distortion characteristics, and the FIR and Volterra filters are set based on the FIR filter coefficients and Volterra filter coefficients, respectively.

[0097] Specifically, generating FIR filter coefficients corresponding to linear distortion characteristics based on nonlinear distortion characteristics further includes: expressing the linear distortion characteristics of the j-th receiving channel by the following formula:

[0098] h j =g j (d) / f(d);

[0099] The first FIR filter coefficient v1, corresponding to the linear distortion characteristics of the first receiving channel in the j-channel receiving process, is calculated using the following formula:

[0100]

[0101] The FIR filter coefficients v corresponding to the linear distortion characteristics of the second to nth receiving channels in the j-channel receiving network are calculated using the following formula. j :

[0102]

[0103] d = b / Δf + N S / 2;

[0104] M = exp(-2πi·b) T u / f s );

[0105] B = [b1; b2], b = vec(B);

[0106] b1 = F start +iΔQ,i∈0,…,N R -1;

[0107] b2 = F start +iΔQ+ΔP,i∈0,…,N R -1;

[0108] Δf=f s / N S ;

[0109] N R =round((F end -F start ) / ΔR);

[0110] Among them, h j G represents the linear distortion characteristic of the j-th receiving channel; j =G(:,j) represents the spectrum of the training output signal received by the j-th receiving channel; Represents the spectrum. The expression represents the Fourier transform; M represents the linear training matrix; u = [-L / 2, -L / 2+1, ..., L / 2-1] represents a vector; L is the maximum number of taps in the equivalent FIR filter; fs represents the baseband sampling rate; vec(·) represents the matrix vectorization operation; d is a matrix of length 2N. R The vector; b1 represents a single audio point in the spectrum of x1; b2 represents a single audio point in the spectrum of x2; N R Indicates the number of receive channels; F start Indicates the system's initial operating frequency; fs Δf represents the baseband sampling rate; ΔQ and ΔP represent positive integer multiples of Δf; N R This indicates the number of single tones required within the operating frequency band, and round is the rounding operation.

[0111] Based on the nonlinear distortion characteristics, the Volterra filter coefficients corresponding to the nonlinear distortion features are generated, which further include: expressed by the following formula for the spectrum g of the j-th receiving channel. j The location where the spectral values ​​of the k-th order nonlinear distortion characteristic appear in the table:

[0112]

[0113] The following formula represents the spectral vector of the k-th order nonlinear distortion characteristic of the j-th receiving channel:

[0114]

[0115] The equivalent Volterra filter coefficients corresponding to the k-th order nonlinear distortion characteristics of the j-th receiving channel are calculated using the following formula.

[0116]

[0117]

[0118]

[0119]

[0120]

[0121] w * =N S –w+2;

[0122] in, The training matrix represents the k-th order nonlinear intermodulation distortion of the j-th receiving channel; The equivalent Volterra filter coefficients corresponding to the k-th order nonlinear intermodulation distortion characteristics of the j-th receiving channel are represented; w1 represents a vector of size M×1 and w2 represents a vector of size N×1; vec(·) represents matrix vectorization operation, and unique(·) represents removing duplicate elements from the vector; (·) T This represents the matrix transpose operation; (w) K This indicates that K vectors w are processed sequentially. Operations; Indicates the spectrum g of the j-th receiving channel j In the diagram, the location of the k-th order nonlinear intermodulation distortion characteristic spectrum value corresponding to this receiving channel is shown. This represents the spectral value vector of the k-th order nonlinear distortion characteristic corresponding to the j-th receiving channel; For y j The conjugate signal of (n).

[0123] In the subsequent real-time correction process, filtering multiple digital signals using FIR and Volterra filters to obtain a single high-linearity digital baseband signal further includes: after the training state is completed, enabling external analog signal input, and the digital signal processing module entering the real-time correction state; generating a continuous scrambling signal through the digital signal processing device, and combining the continuous scrambling signal with the external analog signal into a single analog signal; decomposing the single analog signal into multiple analog signals with the same or different power and converting the multiple analog signals into multiple digital signals; and linearizing the multiple digital signals according to linear and nonlinear distortion characteristics, then performing spurious frequency point identification and elimination processing, and outputting a high-linearity digital baseband signal.

[0124] The following sections provide detailed explanations of linearization processing of multiple digital signals based on linear and nonlinear distortion characteristics, spurious frequency point identification and elimination processing, and output of high linear digital baseband signals.

[0125] The linearization process for multiple digital signals based on linear and nonlinear distortion characteristics further includes: aligning the multiple digital signals through an FIR filter, wherein channels 2 to n have the same linear distortion characteristics as channel 1; subtracting the scrambling signal processed by FIR filter 1 from the aligned multiple digital signals to generate aligned multiple digital signals without scrambling; and then passing the aligned multiple digital signals without scrambling through a K-order Volterra filter to remove the K-order nonlinear distortion signal to generate multiple first digital signals, where K is the order of nonlinear distortion present in the operating frequency band, which varies with the frequency range and the nonlinear characteristics of the device.

[0126] The process of identifying and eliminating spurious frequency points and outputting a high-linearity digital baseband signal further includes: decomposing multiple first digital signals into frame signals with fixed sample lengths through a frame segmentation module; performing Fourier FFT processing on each frame signal to obtain the spectrum of each frame; performing point-by-point comparison and selection processing on the spectrum of each frame obtained from each channel, and identifying whether the current signal is an input signal or a spurious signal by comparing the magnitude difference of the frequency slot spectrum signal components in each frame spectrum of each channel, and removing the identified spurious signals one by one to obtain a single-channel spurious-free baseband signal; performing inverse Fourier IFFT processing on the single-channel spurious-free baseband signal, merging overlapping frames, and removing the overlapping part of the signal to generate a continuous high-linearity time-domain data stream.

[0127] refer to Figure 2 Another specific embodiment of the present invention discloses a multi-channel high-linearity analog-to-digital converter, comprising: a training and scrambling module for receiving training signals or scrambling signals from a digital signal processing module and converting them into analog training signals or analog scrambling signals; an analog combiner for receiving analog training signals, or simultaneously receiving external analog signals and analog scrambling signals and superimposing them into one analog input signal; an analog splitter for decomposing the analog training signals or analog input signals into multiple analog training signals or multiple analog input signals with the same or different power; and multiple receiving channels for converting multiple analog training signals or multiple analog input signals into multiple digital training signals or multiple digital input signals; wherein, during the training state, the digital signal processing module obtains linear distortion characteristics and nonlinear distortion characteristics based on the multiple digital training signals to generate FIR filters and Volterra filters, and during the subsequent real-time correction state, the multiple digital input signals are filtered by the FIR filters and Volterra filters to obtain one high-linearity digital baseband signal.

[0128] Specifically, the digital signal processing module includes: a training start-up module, a training signal generation module, a feature extraction module, and a distortion cancellation module. The training start-up module shuts down external analog signal input channels when the multi-channel high-linearity analog-to-digital converter is first powered on, and the digital signal processing module enters the training state. The training signal generation module generates first and second single-tone signals of different frequencies as training signals. The training and scrambling module performs digital-to-analog conversion on the first and second single-tone signals of different frequencies and combines them into analog training signals. Multiple receiving channels receive multiple analog training signals via analog combiners and analog splitters, and convert these signals into multiple digital training signals. The feature extraction module extracts the linear and nonlinear distortion characteristics of each receiving channel based on the training signals and training feedback signals. The distortion cancellation module generates FIR filter coefficients corresponding to the linear distortion characteristics and Volterra filter coefficients corresponding to the nonlinear distortion characteristics, and sets the FIR and Volterra filters based on these coefficients.

[0129] The feature extraction module includes a linear feature extraction module and a nonlinear feature extraction module. Specifically, the linear feature extraction module is used to represent the linear distortion characteristics of the j-th receiving channel using the following formula:

[0130] h j =g j (d) / f(d);

[0131] The nonlinear distortion location determination module is used to express the spectrum g of the j-th receive channel by the following formula. j The location where the spectral values ​​of the k-th order nonlinear distortion characteristic appear in the table:

[0132]

[0133] The nonlinear feature extraction module is used to represent the k-th order nonlinear distortion characteristic spectrum value vector of the j-th receiving channel using the following formula:

[0134]

[0135] The distortion cancellation module is used to calculate the first FIR filter coefficient v1 corresponding to the linear distortion characteristics of the first receiving channel in the j-channel receiving process using the following formula:

[0136]

[0137] The FIR filter coefficients v corresponding to the linear distortion characteristics of the second to nth receiving channels in the j-channel receiving network are calculated using the following formula. j :

[0138]

[0139] d = b / Δf + N S / 2;

[0140] M = exp(-2πi·b) T u / f s );

[0141] B = [b1; b2], b = vec(B);

[0142] b1 = F start +iΔQ,i∈0,…,N R -1;

[0143] b2 = F start +iΔQ+ΔP,i∈0,…,N R -1;

[0144] Δf=f s / N S ;

[0145] N R =round((F end -F start ) / ΔR);

[0146] Among them, h j G represents the linear distortion characteristic of the j-th receiving channel; j=G(:,j) represents the spectrum of the training output signal received by the j-th receiving channel; Represents the spectrum. The expression represents the Fourier transform; M represents the linear training matrix; u = [-L / 2, -L / 2+1, ..., L / 2-1] represents a vector; L is the maximum number of taps in the equivalent FIR filter; fs represents the baseband sampling rate; vec(·) represents the matrix vectorization operation; d is a matrix of length 2N. R The vector; b1 represents a single audio point in the spectrum of x1; b2 represents a single audio point in the spectrum of x2; N R Indicates the number of receive channels; F start Indicates the system's initial operating frequency; f s Δf represents the baseband sampling rate; ΔQ and ΔP represent positive integer multiples of Δf; N R This indicates the number of single tones required within the operating frequency band, and round is the rounding operation.

[0147] The distortion cancellation module calculates the equivalent Volterra filter coefficients corresponding to the k-th order nonlinear distortion characteristics of the j-th receiving channel using the following formula.

[0148]

[0149]

[0150]

[0151]

[0152]

[0153] w * =N S –w+2;

[0154] in, The training matrix represents the k-th order nonlinear intermodulation distortion of the j-th receiving channel; The equivalent Volterra filter coefficients corresponding to the k-th order nonlinear intermodulation distortion characteristics of the j-th receiving channel are represented; w1 represents a vector of size M×1 and w2 represents a vector of size N×1; vec(·) represents matrix vectorization operation, and unique(·) represents removing duplicate elements from the vector; (·) T This represents the matrix transpose operation; (w) K This indicates that K vectors w are processed sequentially. Operations; Indicates the spectrum g of the j-th receiving channel jIn the diagram, the location of the k-th order nonlinear intermodulation distortion characteristic spectrum value corresponding to this receiving channel is shown. This represents the spectral value vector of the k-th order nonlinear distortion characteristic corresponding to the j-th receiving channel; For y j The conjugate signal of (n).

[0155] In the following text, refer to Figures 2 to 14 The multi-channel high linearity analog-to-digital converter according to embodiments of the present invention will be described in detail by way of specific examples.

[0156] refer to Figure 2 The input analog signal is combined with the output signal from the training and scrambling modules by an analog combiner, then split into two or more analog signals by an analog splitter. These signals then pass through receiving front-ends 1 to n and analog-to-digital converters 1 to n (n is an integer greater than or equal to 2, typically 2, 4, 8, or 16). Finally, they enter the digital signal processing module, which outputs a distortion-free digital baseband signal. The digital processing module can also input training or scrambling signals to the training and scrambling modules for use during the training or scrambling process.

[0157] The ADCs 1 through n in the analog-to-digital converter (ADC) must have two or more sampling clock settings, distributed across different hardware circuit boards. For example, when n is 4 and the system requires a sampling rate of 2490MHz, the sampling rates of ADC1 and ADC2 can be set to 2490MHz, and the sampling rates of ADC3 and ADC4 to 3090MHz. ADC1 and ADC2, along with their clocks, should be located on circuit board 1, while ADC3 and ADC4, along with their clocks, should be located on circuit board 2. The specific circuit designs on circuit boards 1 and 2 can be the same or different. In this case, because the ADC sampling rates and clock frequencies differ between the two circuit boards, all clock-related distortion signals, including clock signal harmonics, intermodulation between the clock and input signals, and clock signal leakage due to hardware design factors, will be located at different positions in the output signal spectrum. These distortions can be subsequently removed using spurious frequency identification and cancellation methods in the digital domain. During training and real-time operation, the settings of each analog signal processing device must remain unchanged.

[0158] like Figure 3 As shown, the digital training or scrambling signal is converted into an analog signal by two independent DACs, then passed through the transmitting front end and finally combined into a single analog signal by an analog combiner. The two independent DAC circuits in the training and scrambling modules can be the same or different, as long as they can provide single-tone transmission capability. The transmitting front end may include a filter, an amplifier, or be omitted.

[0159] Analog combiners are analog devices that combine multiple signals into a single signal, including combiners and couplers. Analog splitters are analog devices that decompose a single signal into multiple signals of the same or different power, including power dividers and couplers. Digital signal processing modules are one or more devices or apparatuses with digital signal processing capabilities, including FPGAs, ASICs, ARM processors, CPUs, and GPUs.

[0160] The function of the receiving front end is to perform processes such as limiting, filtering, and amplifying on the analog signal entering the ADC, so that the processed analog signal matches the performance of the ADC device, allowing the ADC to perform its optimal analog-to-digital conversion capability. Depending on the specific application scenario, it can be a combination of one or more of analog filters, low-noise amplifiers, and limiters, or it can be omitted, allowing the signal to enter the ADC directly after passing through analog splitters.

[0161] ADC1 to n, DAC1 and 2 can be independent devices, or they can be different transmit and receive channels of one or more devices.

[0162] The digital signal processing module has two states: training state and real-time correction state.

[0163] When the analog-to-digital converter is powered on for the first time, external analog signal input is turned off, and the digital signal processing module enters training mode, such as... Figure 4 As shown, the training signal generation module generates single-tone signals 1 and 2 at different frequencies. These signals are then processed by the DUC module and input into the training and scrambling module. The training signals pass through an analog combiner and enter multiple receiving channels. After processing by the DDC module, the corresponding training feedback signals for each channel are obtained. The training and feedback signals at each frequency are then processed by the distortion characteristic extraction module to obtain the linear and nonlinear distortion characteristics of each receiving channel.

[0164] After training is complete, external analog signal input is enabled, and the digital signal processing module enters real-time correction mode, such as... Figure 5 As shown. In real-time correction mode, the digital signal processing device generates a continuous scrambling signal, which is processed by the DUC module and then enters the training and scrambling module, before being combined with the input analog signal. The output signals of each receiving channel enter the digital signal linearization module after passing through the DDC module. This module linearizes the output signals of each channel based on the linear and nonlinear distortion characteristics obtained in the training state, resulting in a single high-linearity digital baseband signal. Simultaneously, this module removes the scrambling signal contained in the receiving channel based on the waveform of the transmitted scrambling signal and the channel's linear distortion characteristics.

[0165] The DUC module is a digital up-converter module, which can be implemented in a digital signal processing device or using the built-in DUC circuitry of some DAC chips. The DDC module is a digital down-converter module, which can be implemented in a digital signal processing device or using the built-in DDC circuitry of some ADC chips. There are several ways to generate scrambling signals, such as... Figure 6a and Figure 6b As shown, its basic characteristic is that the scrambling signal waveform must be randomly and uniformly distributed in the time domain, such as... Figure 6c As shown.

[0166] The steps for training and extracting distortion features are as follows:

[0167] (1) Set waveform parameters: Obtain the system's starting operating frequency F start and termination frequency F end The number of receiving channels is N Rx and baseband sampling rate f s Set the number of samples for the training waveform to N. S The comb spectrum frequency interval is ΔQ, and the two-tone frequency interval is ΔP. ​​Based on this, the required number of single tones within the operating frequency band can be calculated as N. R =round((F end -F start ) / ΔR), where round is the floor operation. Generally, to simplify the complexity of frequency domain separation and improve spectral resolution, the spectral resolution Δf = f s / N S ΔQ and ΔP should be positive integer multiples of Δf. Set a higher N. S A higher ΔQ value and a smaller ΔQ value can help improve training accuracy, but will increase the computational resources required for the training process.

[0168] (2) Design training waveforms: Design them respectively as follows Figure 7a and Figure 7b The spectrum shows training waveforms x1 and x2, both of which are uniform comb spectra. The single-tone point in the spectrum of x1 is b1 = F. start +iΔQ,i∈0,…,N R The single-tone point in the spectrum of -1,x2 is b2=F start +iΔQ+ΔP,i∈0,…,N R -1, define matrix B = [b1; b2], b = vec(B). vec(·) represents matrix vectorization operation. Generates a matrix of length 2N. R Find a vector d such that d = b / Δf + N S / 2.

[0169] (3) Perform training and obtain the output results of each receiving channel: Input training waveforms x1 and x2 from the T1 and T2 input ports of the training and scrambling modules respectively, set x = x1 + x2, and denot the spectrum as . This represents the Fourier transform. T out Port output signal such as Figure 7c As shown. And it receives the output signals y1 to y2 from DDC1 to DDCn. n This is used as the training output. For a single receiving channel, the effect of linear distortion on the spectrum of the training output signal is as follows: Figure 8a As shown, the effect of nonlinear distortion on the spectrum of the training output signal is as follows: Figure 8b As shown.

[0170] (4) Extract the linear distortion characteristics of each receiving channel: generate a generator of size N S ×N Rx Matrix G such that: Represents the Fourier transform, g j =G(:,j) represents the spectrum of the training output signal received by the j-th receiving channel. For the j-th receiving channel, its linear distortion characteristic can be expressed as: h j =g j (d) / f(d).

[0171] (5) Calculate the FIR filter coefficients corresponding to the linear distortion characteristics: Define L as the maximum number of taps of the equivalent FIR filter. Construct the vector u = [-L / 2, -L / 2+1, ..., L / 2-1] and the linear training matrix M = exp(-2πi·b T u / f s The coefficients v1 of FIR filter 1 are calculated by solving the following optimization problem:

[0172]

[0173] Then, by solving the following optimization problem in sequence, calculate the coefficients v corresponding to FIR filters 2 to n. j :

[0174]

[0175] (6) Calculate the position of the nonlinear distortion spectrum value of each receiving channel: For a vector w1 of size M×1 and a vector w2 of size N×1, define the operation:

[0176]

[0177] in,(·) T`vc(·)` represents the matrix transpose operation, `vec(·)` represents the matrix vectorization operation, and `unique(·)` represents removing duplicate elements from a vector. Define `w`. * =N S –w+2, (w) K This indicates that K vectors w are processed sequentially. Operations: The spectrum g of the j-th receiving channel j In the above, the position of the k-th order nonlinear intermodulation distortion characteristic spectrum value corresponding to this receiving channel is:

[0178] (7) Extract the nonlinear distortion characteristics of each receiving channel: The spectrum value vector of the k-th order nonlinear distortion characteristics corresponding to receiving channel j is as follows:

[0179] (8) Calculate the Volterra filter coefficients corresponding to the nonlinear distortion characteristics of each receiving channel: Set the maximum memory depth of the Volterra filter for the receiving channel to M. In order to obtain the k-th order nonlinear intermodulation distortion characteristics of receiving channel j, first construct a filter with size N. S ×M k matrix Make

[0180]

[0181] in, Indicates Fourier transform, For y j The conjugate signal of (n). Then the training matrix corresponding to the k-th order nonlinear intermodulation distortion of the receiving channel j. The equivalent Volterra filter coefficients corresponding to the k-th order nonlinear intermodulation distortion characteristics of the receiving channel j are obtained by solving the following optimization problem.

[0182]

[0183] Where ||·||2 represents the second norm.

[0184] The real-time calibration method for multi-channel high-linearity analog-to-digital converters is as follows:

[0185] (1) When the analog-to-digital converter is in real-time calibration mode, the digital signal processing module continuously generates scrambling signals, which are simultaneously sent to the training and scrambling module via the DUC module. After being combined with the analog signal to be received by the analog combiner, the signals enter the subsequent receiving channel. Simultaneously, the transmitted scrambling signals enter the digital signal linearization processing module, which uses the input source as a reference to remove the scrambling signals from the received baseband signal. The scrambling method can reduce distortion characteristic fluctuations in the analog-to-digital converter, suppress spurious distortion amplitude, and remove most small-amplitude distortion.

[0186] (2) Based on the linear and nonlinear distortion characteristics of the device obtained during the training process, the digital signal linearization processing module receives the baseband signals input from n channels and outputs a high linear digital baseband signal after processing.

[0187] The internal components of a digital signal linearization module are as follows: Figure 9 As shown, it includes the following steps:

[0188] (1) The DDC baseband signal output from each channel is first aligned by an FIR filter. Channels 2 to n will have the same linear distortion characteristics as channel 1. The FIR filter coefficients of the j-th receiving channel are the coefficient vector v calculated above. j ;

[0189] (2) After the scrambled signal is processed by FIR filter 1, it is subtracted from the filter output results of channels 1 to n, thereby removing the scrambled signal from the received signal;

[0190] (3) After removing the scrambling signal, the output signal of each channel is then passed through a K-order Volterra filter to remove the K-order nonlinear distortion signal. Here, K is the order of nonlinear distortion present in the operating frequency band, which varies with the frequency range and the nonlinear characteristics of the device. The K-order Volterra filter coefficients of the j-th receiving channel are the coefficient vector calculated above.

[0191] (4) After removing the nonlinear distortion signal, all channel output signals are processed by the spurious frequency identification and elimination module and merged into one output signal, which is the high linear digital baseband signal finally output by the device.

[0192] The internal components of the spurious frequency identification and cancellation module are as follows: Figure 10 As shown, it includes the following steps:

[0193] (1) The input signals of each channel are first divided into frame signals with fixed sample lengths by the frame segmentation module, and the processing method is as follows: Figure 11 As shown;

[0194] (2) Then perform FFT processing on each frame of the decomposed signal to obtain the spectrum of each frame.

[0195] (3) Perform point-by-point comparison and selection operations on the spectrum of each frame obtained from each channel. The processing procedure is as follows: Figure 12 As shown, for input baseband signals with two or more configurations, the location and magnitude of clock-related spurious signals in the spectrum differ due to different sampling rate settings. Since the hardware of each receiving channel is independent, fixed and random spurious signals generated by hardware generally appear in different locations. By comparing the magnitude differences of the spectral signal components in each frequency slot of each frame of each channel, it is possible to identify whether the signal is an input signal or a spurious signal. After removing the spurious signals one by one, a single-channel spurious-free baseband signal can be obtained. Figure 12 The process for identifying and eliminating baseband spurious frequency points under two sampling rate configurations is presented. Specifically, Figure 12 Sections a through c in the diagram sequentially illustrate the changes in the spectrum for Configuration 1 and Configuration 2 as the spectrum slot shifts from low to high frequency. As the spectrum slot moves, the differences in the spectrum between the different configurations are eliminated, retaining only the common components. Finally, refer to... Figure 12 In part d, all configurations have the same corresponding spectrum, which can be merged into the same spectrum.

[0196] (4) Perform IFFT processing on the final merged single-channel spurious-free baseband signal, merge the overlapping frames, remove the overlapping part of the signal, and obtain a continuous high linear time domain data stream.

[0197] Taking the reception of dual-tone analog signals as an example, a typical analog-to-digital converter (ADC) system, after receiving a dual-tone signal, converts it into a digital baseband signal with the following spectrum: Figure 13a As shown, the spectral components of each distorted signal are marked. In addition to the input two-tone signal (black triangle), the following distortion components also exist:

[0198] a) Linear distortion of the input signal itself;

[0199] b) Second-order, third-order, and higher-order harmonics and intermodulation distortion of the input signal;

[0200] c) Distortion related to the sampling clock;

[0201] d) Distortion resulting from modulation of the input signal with the sampling clock;

[0202] e) Fixed-frequency spurious signals;

[0203] f) Other random frequency spurious signals;

[0204] The presence of these distorted signals significantly reduces the SFDR (Signal Flow Rate) of the analog-to-digital converter system. If the input analog signal also contains small signals, it will be unable to effectively distinguish them.

[0205] refer to Figure 13b The functions of each module or method in the analog-to-digital conversion device of the present invention are as follows:

[0206] • Channel linear distortion matching: Corrects differences caused by varying linear distortion in different receiving channels;

[0207] • Noise scrambling: Reduces distortion fluctuations in the receiving system and suppresses the amplitude of spurious components.

[0208] Eliminate minor stray particles.

[0209] • Nonlinear distortion correction: Eliminates second-order, third-order, and higher-order harmonics and intermodulation distortion components.

[0210] • Spurious identification and elimination: Eliminate distortions related to the sampling clock, distortions generated after the input signal is modulated with the sampling clock, and other residual fixed or random spurious signals;

[0211] After the above processing steps, the spectrum of the resulting high linearity output signal is as follows: Figure 12 As shown in b, the output signal contains only the original input signal, without any other distorted signals, and the noise floor level decreases.

[0212] The main spectral changes of the dual-tone signal during real-time correction are as follows: Figure 14 As shown. Scrambling signal ( Figure 14 b) and the input analog signal ( Figure 14 c) After being combined by an analog combiner, the signal is then split and input to multiple receiving channels. The output signal of a single channel after DDC processing ( Figure 14 d) contains distortion signals introduced by the receiving front-end and ADC. After processing by the digital signal linearization module, a highly linear digital baseband signal is output. Figure 14 (e in the text)

[0213] Figure 15 This demonstrates the spectral changes that occur at each step when a two-tone signal is processed by a digital signal linearization module. The signals entering this module contain various types of distortion introduced by the receiver front-end and ADC. First, an FIR filter processes the input signals of each channel, preserving their linear distortion characteristics once. Then, the introduced scrambling signals are removed, resulting in the obtained spectrum (…). Figure 15 (c) In these output signals, second-order, third-order, and higher-order harmonics and intermodulation distortion are related to the input signals, and therefore occupy the same spectral position in each channel. These are eliminated after processing by the K-order nonlinear distortion correction module. Figure 15(d) Because each ADC has a different sampling clock setting and independent hardware circuitry, clock-related spurious signals, as well as other fixed and random spurious signals, will appear in different locations. After processing by the spurious signal identification and cancellation module, only the same spectral components in each channel will be retained. In the device, this same spectral component is the input signal. This results in a high-linearity baseband output signal.

[0214] The generation of distortion signals in analog-to-digital conversion systems is mostly related to the design of devices and circuits. Due to the numerous factors involved, eliminating such stray signals in practice requires a complex hardware troubleshooting and modification process, which is not only time-consuming and labor-intensive but also makes it difficult to guarantee the final result. This invention, through system architecture design and by combining multiple distortion correction methods and digital correction algorithms, can eliminate most of the distortion signals introduced by the conversion device itself during analog-to-digital conversion at the system level.

[0215] The beneficial effects of the embodiments of the present invention are as follows:

[0216] (1) A hardware and software architecture combining channel linearization technology, nonlinear distortion correction technology, spurious elimination technology and scrambling technology is provided, which can achieve the beneficial effects of the above technologies with the same hardware settings.

[0217] (2) A training waveform design method is provided, which can obtain the linear and nonlinear distortion characteristics of the internal channel of the receiving device by transmitting a training signal once, and can convert them into FIR filter and Volterra filter coefficients.

[0218] (3) A signal processing method is provided that eliminates distorted signals and improves spurious-free dynamic range with less computational resources;

[0219] (4) A multi-channel random spurious signal cancellation method is provided, which eliminates the random spurious distortion signal in the received signal by comprehensively utilizing the spectral information of multiple channels.

[0220] Furthermore, the proposed solution combines channel linearization, nonlinear distortion correction, spurious emission, and scrambling techniques. Even using only some of these techniques can achieve a certain level of distortion reduction. The processing order of the modules within the digital signal linearization module can be adjusted to still achieve the same effect.

[0221] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0222] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-channel high linearity analog-to-digital conversion method, characterized in that, include: Receive training signals or scrambled signals from the digital signal processing module and convert them into analog training signals or analog scrambled signals; Receive the simulated training signal, or receive the input simulated signal and the simulated scrambling signal and superimpose them into a single simulated input signal; The analog training signal or the analog input signal is decomposed and converted into multiple digital training signals or multiple digital input signals; as well as During the training process, linear distortion and nonlinear distortion characteristics are obtained based on the multi-channel digital training signals to generate FIR filters and Volterra filters. In the subsequent real-time correction process, the multi-channel digital input signals are filtered by the FIR filters and the Volterra filters to obtain a high linearity digital baseband signal. The subsequent real-time correction process, in which the multiple digital input signals are filtered by the FIR filter and the Volterra filter to obtain a single high-linearity digital baseband signal, further includes: after the training state is completed, opening the external analog signal input channel, and the digital signal processing module entering the real-time correction state; generating a continuously scrambled signal through the digital signal processing module, and combining the continuously scrambled signal with the external analog signal into a single analog signal; decomposing the single analog signal into multiple analog signals with the same or different power and converting the multiple analog signals into multiple digital signals; and linearizing the multiple digital signals according to the linear distortion characteristics and the nonlinear distortion characteristics, and then removing spurious signals. The process involves frequency point identification and elimination, followed by outputting a high-linearity digital baseband signal. The linearization of the multiple digital signals based on the linear and nonlinear distortion characteristics further includes: channel alignment of the multiple digital signals using the FIR filter, where channels 2 to n have the same linear distortion characteristics as channel 1; subtracting the scrambling signal processed by FIR filter 1 from the aligned multiple digital signals to generate aligned multiple digital signals without scrambling; and then passing the aligned multiple digital signals without scrambling through a K-order Volterra filter to remove K-order nonlinear distortion signals to generate multiple first digital signals, where K is the order of nonlinear distortion present in the operating frequency band, varying with the frequency range and the nonlinear characteristics of the device.

2. The multi-channel high linearity analog-to-digital conversion method according to claim 1, characterized in that, The process of obtaining linear distortion and nonlinear distortion characteristics based on the multi-channel digital training signals during training to generate FIR filters and Volterra filters further includes: When the multi-channel high linearity analog-to-digital converter is powered on for the first time, the external analog signal input channel is closed, and the digital signal processing module enters the training state. The training signal generation module generates first and second single-tone signals of different frequencies. Based on the first and second single-tone signals of different frequencies, digital-to-analog conversion is performed and they are combined into the analog training signal. The analog training signal is then output as multiple digital signals as training feedback signals through an analog combiner and a multiple receiving channel. Based on the training signal and the training feedback signal, the linear distortion characteristics and nonlinear distortion characteristics of each receiving channel are extracted; Based on the linear distortion characteristics and the nonlinear distortion characteristics, FIR filter coefficients corresponding to the linear distortion characteristics and Volterra filter coefficients corresponding to the nonlinear distortion characteristics are generated respectively, and the FIR filter and the Volterra filter are set based on the FIR filter coefficients and the Volterra filter coefficients respectively.

3. The multi-channel high linearity analog-to-digital conversion method according to claim 2, characterized in that, Generating FIR filter coefficients corresponding to the linear distortion characteristics based on the nonlinear distortion characteristics further includes: The following formula represents the first j Linear distortion characteristics of the receiving channel: h j = g j (d) / f(d); Calculate using the following formula j The first FIR filter coefficient v1 corresponding to the linear distortion characteristics of the first receiving channel in the receiving channel: ； The following formula is used to calculate the relationship between the second receiving channel and the j-th receiving channel in the j-channel receiving channel. n The linear distortion characteristics of the receiving channel correspond to the FIR filter coefficients v j : ； d=b / Δf + N S / 2; M=exp(-2πi b T u / f s ); B = [b1;b2], b = vec(B); b1=F start + i ΔQ, i 0,…, N R -1; b2=F start + i ΔQ+ΔP, i 0,…, N R -1; Δf=f s / N S ; N R = round((F end -F start ) / ΔR); Among them, h j Indicates the first j Linear distortion characteristics of the receiving channel; g j =G(:, j ) indicates the first j The spectrum corresponding to the training output signal received by the receiving channel; f= ( x ) represents the spectrum. (·) denotes Fourier transform; M denotes the linear training matrix; u=[-L / 2,-L / 2+1,…,L / 2-1] denotes a vector; L is the maximum number of taps in the equivalent FIR filter; fs denotes the baseband sampling rate; vec(·) denotes matrix vectorization operation, where d is a vector of length 2N. R a vector; b1 represents x 1 represents a single audio point in the spectrum; b2 represents... x 2. Single audio point in the spectrum; N R Indicates the number of receive channels; F start Indicates the system's initial operating frequency; f s Δf represents the baseband sampling rate; ΔQ and ΔP represent positive integer multiples of Δf; N R This indicates the number of single tones required within the operating frequency band, and round is the rounding operation.

4. The multi-channel high linearity analog-to-digital conversion method according to claim 3, characterized in that, Generating Volterra filter coefficients corresponding to the nonlinear distortion characteristics based on the aforementioned nonlinear distortion characteristics further includes: The following formula expresses the... j Receive channel spectrum g j middle, k Location of the spectral values ​​of the first-order nonlinear distortion characteristic: = (d) (k-1) / 2 (d ) (k-1) / 2+1 ; The following formula represents the first j Receive channel k The spectrum value vector of the nonlinear distortion characteristic is: ； The following formula is used to calculate the relationship with the j-th receiving channel. k Equivalent Volterra filter coefficients corresponding to the first-order nonlinear distortion characteristics : ； ； w1 w2=unique(vec(w1·1 1×N +1 M×1 ·w2 T - 1)); (In) K =in In … In; w =N S –w+2; in, Indicates the first j Receive channel k The training matrix corresponding to order-order nonlinear intermodulation distortion; Indicates the first j Receive channel k The equivalent Volterra filter coefficients corresponding to the first-order nonlinear intermodulation distortion characteristics; w 1 represents the sum of a vector of size M×1. w 2 represents a vector of size N×1; vec(·) represents matrix vectorization operation, unique(·) represents removing duplicate elements from the vector; (·) T This represents the matrix transpose operation; (w) K express K vectors w proceed in sequence Operations; Indicates the first j Receive channel spectrum g j In the middle, the corresponding receiving channel k Location of the spectral values ​​of the first-order nonlinear intermodulation distortion characteristics; Indicates the first j Receive channel corresponding k The spectral value vector of nonlinear distortion characteristics; for The conjugate signal.

5. The multi-channel high linearity analog-to-digital conversion method according to claim 1, characterized in that, Further steps include identifying and eliminating spurious frequencies and outputting a high-linearity digital baseband signal, including: The multiple first digital signals are decomposed into frame signals with fixed sample lengths by the frame segmentation module; Perform Fourier FFT processing on each frame of the decomposed signal to obtain the spectrum of each frame; Each frame spectrum obtained from each channel is compared and selected point by point. By comparing the difference in the magnitude of the frequency slot spectrum signal components in each frame spectrum of each channel, it is possible to identify whether the current signal is an input signal or a spurious signal. The identified spurious signals are then removed one by one to obtain a single-channel spurious-free baseband signal. The single-channel spurious-free baseband signal is subjected to inverse Fourier transform (IFFT) processing, and overlapping frames are merged to remove overlapping signals in order to generate a continuous high linear time-domain data stream.

6. A multi-channel high-linearity analog-to-digital converter, characterized in that, The device uses the multi-channel high linearity analog-to-digital conversion method according to any one of claims 1 to 5, comprising: The training and scrambling module is used to receive training signals or scrambling signals from the digital signal processing module and convert them into analog training signals or analog scrambling signals. An analog combiner is used to receive the analog training signal, or to simultaneously receive an external analog signal and the analog scrambling signal and superimpose them into a single analog input signal. Analog splitter, used to decompose the analog training signal or the analog input signal into multiple analog training signals or multiple analog input signals with the same or different power; A multi-channel receiving channel is used to convert the multi-channel analog training signals or the multi-channel analog input signals into multi-channel digital training signals or multi-channel digital input signals; During the training phase, the digital signal processing module obtains linear and nonlinear distortion characteristics based on the multiple digital training signals to generate an FIR filter and a Volterra filter. In the subsequent real-time correction phase, the module filters the multiple digital input signals using the FIR filter and the Volterra filter to obtain a single high-linearity digital baseband signal.

7. The multi-channel high linearity analog-to-digital converter according to claim 6, characterized in that, The digital signal processing module includes: a training start-up module, a training signal generation module, a feature extraction module, and a distortion removal module, wherein... The training start-up module is used to shut down the external analog signal input channel when the multi-channel high linearity analog-to-digital converter is powered on for the first time, and the digital signal processing module enters the training state. The training signal generation module is used to generate first and second single-tone signals of different frequencies as the training signals. The training and scrambling module performs digital-to-analog conversion on the first and second single-tone signals of different frequencies and combines them into the analog training signal. The multiple receiving channel is used to receive the multiple analog training signals sequentially via the analog combiner and the analog splitter, and to convert the multiple analog training signals into multiple digital training signals. The feature extraction module is used to extract the linear distortion characteristics and nonlinear distortion characteristics of each receiving channel based on the training signal and the training feedback signal. The distortion cancellation module is used to generate FIR filter coefficients corresponding to the linear distortion characteristics and Volterra filter coefficients corresponding to the nonlinear distortion characteristics, respectively, and to set the FIR filter and the Volterra filter based on the FIR filter coefficients and the Volterra filter coefficients, respectively.

8. The multi-channel high linearity analog-to-digital converter according to claim 7, characterized in that, The feature extraction module is used to express the first feature using the following formula. j Linear distortion characteristics of the receiving channel: h j = g j (d) / f(d); The following formula expresses the... j Receive channel spectrum g j middle, k Location of the spectral values ​​of the first-order nonlinear distortion characteristic: = (d) (k-1) / 2 (d ) (k-1) / 2+1 ; The following formula represents the first j Receive channel k The spectrum value vector of the nonlinear distortion characteristic is: ； The distortion elimination module is used to calculate the distortion using the following formula. j The first FIR filter coefficient v1 corresponding to the linear distortion characteristics of the first receiving channel in the receiving channel: ； The following formula is used to calculate the relationship between the second receiving channel and the j-th receiving channel in the j-channel receiving channel. n The linear distortion characteristics of the receiving channel correspond to the FIR filter coefficients v j : ； d=b / Δf + N S / 2; M=exp(-2πi b T u / f s ); B = [b1;b2], b = vec(B); b1=F start + i ΔQ, i 0,…, N R -1; b2=F start + i ΔQ+ΔP, i 0,…, N R -1; Δf=f s / N S ; N R = round((F end -F start ) / ΔR); Among them, h j Indicates the first j Linear distortion characteristics of the receiving channel; g j =G(:, j ) indicates the first j The spectrum corresponding to the training output signal received by the receiving channel; f= ( x ) represents the spectrum. (·) denotes Fourier transform; M denotes the linear training matrix; u=[-L / 2,-L / 2+1,…,L / 2-1] denotes a vector; L is the maximum number of taps in the equivalent FIR filter; fs denotes the baseband sampling rate; vec(·) denotes matrix vectorization operation, where d is a vector of length 2N. R a vector; b1 represents x 1 represents a single audio point in the spectrum; b2 represents... x 2. Single audio point in the spectrum; N R Indicates the number of receive channels; F start Indicates the system's initial operating frequency; f s Δf represents the baseband sampling rate; ΔQ and ΔP represent positive integer multiples of Δf; N R This indicates the number of single tones required within the operating frequency band; round is the integer operation. The following formula is used to calculate the relationship with the j-th receiving channel. k Equivalent Volterra filter coefficients corresponding to the first-order nonlinear distortion characteristics : ； ； w1 w2=unique(vec(w1·1 1×N +1 M×1 ·w2 T - 1)); (In) K =in In … In; w =N S –w+2; in, Indicates the first j Receive channel k The training matrix corresponding to order-order nonlinear intermodulation distortion; Indicates the first j Receive channel k The equivalent Volterra filter coefficients corresponding to the first-order nonlinear intermodulation distortion characteristics; w 1 represents the sum of a vector of size M×1. w 2 represents a vector of size N×1; vec(·) represents matrix vectorization operation, unique(·) represents removing duplicate element values ​​from the vector; (·) T This represents the matrix transpose operation; (w) K express K vectors w proceed in sequence Operations; Indicates the first j Receive channel spectrum g j In the middle, the corresponding receiving channel k Location of the spectral values ​​of the first-order nonlinear intermodulation distortion characteristic; Indicates the first j Receive channel corresponding k The spectral value vector of nonlinear distortion characteristics; for The conjugate signal.

Images