Multi-channel error spectrum cancellation method for digital bandwidth interleaved sampling system correction

Abstract

The application discloses a multi-channel error spectrum cancellation method for digital bandwidth interleaved sampling system correction, and belongs to the field of wideband signal processing and high-performance digitization sampling. The application realizes the method as follows: determining an error spectrum generation interval in a working frequency band coupling mapping table; determining all affected frequency intervals corresponding to each error spectrum generation interval according to a preset tolerance threshold; based on linear regression of the distribution of the two types of intervals, establishing a frequency mapping relationship between error spectrum frequency components and original input frequency components; based on Hilbert transform, carrying out correlation calculation to obtain the amplitude and phase mapping relationship between the error spectrum frequency components and the original input frequency components in each affected interval, and recording the frequency mapping relationship and the amplitude and phase mapping relationship in the working frequency band coupling mapping table; completing frequency conversion mapping and amplitude and phase compensation according to the working frequency band coupling mapping table recording the frequency, amplitude and phase mapping relationship, reconstructing the error spectrum components, eliminating the reconstructed error spectrum components from the original output signal, and realizing error spectrum cancellation.

Description

Technical Field

[0001] This invention relates to a correction method for a digital bandwidth interleaved sampling system, and to an error spectrum cancellation method for a digital bandwidth interleaved sampling system based on multi-channel information interaction, belonging to the fields of broadband signal processing and high-performance digital sampling technology. Background Technology

[0002] In modern broadband signal detection systems, to address the challenge posed by the sudden increase in instantaneous bandwidth to analog-to-digital converters (ADCs), a digital bandwidth interleaved sampling system (FIADC) based on the generalized sampling theorem is often employed. This architecture divides the broadband signal into multiple sub-bands using analog filter banks and utilizes a low-speed ADC array for parallel acquisition to alleviate the pressure on individual circuits. However, due to manufacturing limitations, the non-ideal frequency response characteristics of analog filter banks (such as finite stopband suppression) can lead to severe error spectrum interference. Specifically, each sub-channel introduces cross-channel leakage components when acquiring the target signal. These components enter the digital domain during the sampling process, forming complex error spectrum components. Furthermore, after subsequent digital reconstruction operations, the complexity of aliasing components further increases, severely weakening the system's time-frequency accuracy and effective observation bandwidth.

[0003] Traditional digital bandwidth interleaved sampling system correction schemes are mostly based on single-channel digital compensation logic. These methods typically acquire the frequency response of each frequency point of the system through offline frequency sweeping, construct error curves, and use fixed finite impulse response (FIR) filters for amplitude and phase correction. While this scheme has good industrial feasibility, its underlying logic assumes only a linear one-to-one mapping relationship between the system input and output. In scenarios with significant cross-channel aliasing, the output component of a certain channel no longer depends solely on the input component at that frequency, but rather exhibits the result of nonlinear coupling of multiple frequency input components. Traditional one-dimensional compensation methods struggle to establish cross-channel correlation models and cannot fundamentally eliminate the aliasing error spectrum introduced by insufficient filter bank suppression.

[0004] Therefore, there is currently a lack of a method to model cross-channel aliasing caused by the non-ideality of analog filters and to achieve online real-time correction without prior information. How to utilize system identification theory and multi-channel information interaction mechanisms to construct a compensation framework capable of extracting and canceling error spectral components in real time, thereby achieving online, real-time, and effective error spectral elimination, has become a key problem that urgently needs to be solved to improve the dynamic performance of ultra-wideband sampling systems. Summary of the Invention

[0005] To address the error spectrum suppression problem in digital bandwidth interleaved sampling systems under non-ideal analog filtering conditions, this invention aims to provide a multi-channel error spectrum cancellation method for correcting digital bandwidth interleaved sampling systems. Based on offline system identification to establish a global frequency mapping relationship, and combined with an online multi-channel information interaction mechanism, the method adaptively extracts and cancels cross-channel aliasing components to achieve error spectrum suppression in the digital bandwidth interleaved sampling system, thereby improving the spurious-free dynamic range and signal-to-noise ratio performance of the system. This invention is applicable to high-dynamic, high-bandwidth sampling applications such as broadband receivers, communication monitoring, electronic reconnaissance, and spectrum monitoring.

[0006] The objective of this invention is achieved through the following technical solution.

[0007] The multi-channel error spectrum cancellation method for correcting digital bandwidth interleaved sampling systems disclosed in this invention includes the following steps:

[0008] S1. By coarse-precision frequency sweeping, the frequency range in which the performance degradation is caused by the error spectrum in the digital bandwidth interleaved sampling system to be corrected is obtained, and then the error spectrum generation range in the working frequency band coupling mapping table is determined.

[0009] Construct a high-precision single-tone excitation signal sequence, and set the complete operating frequency band of the digital bandwidth interleaved sampling system as follows: Within this frequency band, at preset low frequency intervals Step sampling is performed to generate a series of single-tone continuous wave signals. .in, For frequency sweep index, For the first The center frequency of each sweep point For the amplitude of the excitation signal, Initial phase. Frequency step interval. The selection of frequency step intervals needs to meet the requirements for detailed characterization of the transition band of analog filters. The value should not exceed 1% of the sub-ADC bandwidth.

[0010] The generated single-tone excitation signal The digital bandwidth interleaved sampling system to be calibrated is injected, and after analog processing and digital reconstruction within the system, the discrete output sequence of the system corresponding to each frequency sweep input is obtained in the digital domain. .in, Indicates the index of the time sampling point. This represents the total number of samples captured in a single acquisition. This sequence reflects the overall response characteristics of the system at a specific input frequency, including the main signal components and various error spectral components introduced by system non-ideals.

[0011] Spurious-free dynamic range (SFDR) characteristics at each frequency point were extracted using spectral analysis. For each swept frequency point... The time-domain output sequence implement Point-wise discrete Fourier transform (DFT) yields the frequency domain representation sequence. :

[0012] (1)

[0013] In the formula: Output sequence in the time domain index, Frequency domain representation sequence .

[0014] Calculate the power spectral density at each frequency sweep point according to equation (1). The power of the main component of the signal is located using a peak search algorithm. and its corresponding frequency index Excluding the main signal spectral line and its adjacent spectral lines. Then, identify the maximum error spectral component power of the remaining frequency bands within the full bandwidth. The system's power at a given frequency is calculated based on the formula for the maximum error spectral component power. The normalized error suppression ratio at that point, i.e. :

[0015] (2)

[0016] An adaptive thresholding method is used to construct the frequency-SFDR response curve and extract the error-sensitive region. The calculation results corresponding to all frequency sweep points are then aggregated and mapped to construct the system's full-band performance response characteristic curve. The local mean baseline for this curve is calculated using a sliding window. Therefore, an adaptive decision threshold is defined. ,in This is a preset performance degradation tolerance factor. The factor that satisfies the condition is extracted through logical comparison operations. The frequency set was used to identify regions of performance degradation with significant dips in the curve, which were defined as the error spectral response frequency band intervals. ,in Indicates the identified first A set of consecutive error frequency bands. The frequency range in which the cross-channel aliasing effect is most significant in the system was recorded.

[0017] S2. By finely scanning the frequency, the system obtains the output error spectrum response characteristics of the input frequency components within all error spectrum generation intervals. Based on the preset tolerance threshold, all affected frequency intervals corresponding to each error spectrum generation interval are determined. Based on the linear regression of the two types of interval distributions, the frequency mapping relationship between the error spectrum frequency components and the original input frequency components is established.

[0018] For the error spectral response frequency band range identified in step S1 Perform refined frequency sweep analysis. Within the interval Internal step frequency Generate high-density monotone input signal ,in for Frequency sweep index within the range, This corresponds to the input excitation frequency. The series of signals is injected into a digital bandwidth interleaved sampling system to obtain the digitally reconstructed output sequence. For each output sequence, high-resolution power spectral density estimation is performed to obtain the spectral function. :

[0019] (3)

[0020] In equation (3), For windowing functions, For sequence index, Let f be the sequence length, and f be the sequence of spectral functions. Index. In the spectral function Mid-position principal component power point and their corresponding frequencies ,in In an ideal state, it should equal... .

[0021] Perform error peak filtering based on a tolerance parameter. An adjustable, manually set tolerance parameter is used. Its unit is dBc. Traversal spectral function Identify the set of error spectral peaks for all local maxima points outside the main component frequency region. ,in This serves as the index for the error spectrum peaks. For each element in the set, calculate its power difference relative to the principal component. If the judgment condition is met. < Then the frequency point The error spectrum frequency is identified as needing to be eliminated. Due to the hardware characteristics of the digital bandwidth interleaved sampling system, this error spectrum frequency... Located in Within different physical channel frequency bands.

[0022] Construct a frequency mapping function and the set of affected intervals. As the input frequency... exist Internal smoothing variation, recording the frequency of the selected error spectrum The motion trajectory is used to establish a mapping vector from the input frequency to the error spectrum frequency. By scanning all frequency points Cluster analysis was performed on the mapped vectors to identify the set of all frequency intervals within the system's operating frequency band that were affected by the error spectrum, resulting in performance degradation. , where u is the index of the affected interval. For each affected interval The error spectrum generation interval is generated by inversely tracing the mapping vector. .

[0023] This characterizes the algebraic coupling relationship between the generating interval and the affected interval. For each pair of mapping intervals... The frequency transformation slope was determined by linear regression. Due to the presence of variable sampling rate components and multi-stage mixing structures within the system, the mapping relationship exhibits either a forward or inverse mapping: when When > 0, it satisfies A positive growth relationship; when When < 0, a frequency reversal characteristic is exhibited, meaning the high-frequency band of the generated interval corresponds to the low-frequency band of the affected interval. Ultimately, a frequency band containing any... With its Corresponding frequency mapping parameter pair parameter pair set .

[0024] S3. Based on the Hilbert transform, perform correlation calculations to obtain the amplitude and phase mapping relationship of the error spectrum frequency component relative to the original input frequency component in each affected interval. Record the frequency mapping relationship together with the error spectrum frequency component in the working frequency band coupling mapping table.

[0025] Extract the system output sequence obtained in step S2 And perform signal component separation based on the established frequency coupling relationship. For the frequency sweep index... The principal component frequency below With error spectrum frequency Two sets of narrowband high-performance digital filters were designed respectively. and Output sequence By inputting the above filters respectively, time-domain signals containing only the principal components are obtained. With time-domain signals containing only specific error spectral components The filtering process is expressed as follows:

[0026] (4)

[0027] In equation (4), * denotes discrete convolution operation. This is an index for the time-domain signal sequence.

[0028] Perform signal spectrum shifting and alignment based on frequency mapping principles. This is done based on the frequency deviation between the principal components and error spectrum components in the mapping vector established in step S1.2. Construct a real-coefficient frequency conversion reference signal ,in The system sampling rate. The principal component signal... and Spectrum shifting is achieved by performing point-by-point multiplication operations to generate an intermediate signal containing a mixing copy. Then a low-pass filter was used. By filtering out irrelevant spectral copies generated after the shift, a reference principal component signal with the same center frequency as the error spectral component is obtained. .

[0029] Calculate the amplitude gain and phase mismatch parameters of the error spectral components relative to the reference principal component signal. Construct the model using the Hilbert Transform. and The analytic signal forms are respectively expressed as: and Define the amplitude gain coefficient. The root mean square ratio of the two signals within the observation window:

[0030] (5)

[0031] In equation (5), To parse the signal sequence index, The sequence length is given.

[0032] Meanwhile, the phase mismatch angle is obtained through the cross-correlation phase estimation method. :

[0033] (6)

[0034] In equation (6), Indicates taking the complex argument. Indicates complex conjugation. To parse the signal sequence index, For sequence length, and The definition is the same as in equation (5). This parameter is for It accurately characterizes the distortion features of the error spectrum generated by the coupling of the input signal in the amplitude and phase dimensions.

[0035] Perfecting the parameter pair set Coupling Mapping Table with Operating Frequency Band The parameters will be paired. Composed amplitude and phase error sequence As an attribute field, it is added to the corresponding parameter pair collection. In the middle. Generate a set of intervals based on the error. Affected frequency frequency range set Parameter pair set Output a complete working frequency band coupling mapping table This table records in detail all affected frequency ranges. and its source error spectrum generation interval Boundary frequency, mapping coefficient With mapping bias And the amplitude and phase error sequence pairs that evolve with frequency for each mapping pair This provides complete prior identification data for error compensation signal reconstruction in the subsequent online calibration process.

[0036] S4. For each affected frequency range, extract the approximate input frequency component from the output signal, complete the frequency conversion mapping and amplitude-phase compensation based on the working frequency band coupling mapping table that records the frequency, amplitude and phase mapping relationship, reconstruct the error spectrum component, remove the reconstructed error spectrum component from the original output signal, and realize error spectrum cancellation.

[0037] Extracting the working frequency band coupling mapping table constructed in the offline stage based on the working frequency band coupling mapping table. The topology information is used to initialize the online cancellation network. This is applied to the original signal to be corrected output by the system in real time. Identify the set of all frequency intervals affected by the error spectrum within its full bandwidth. For each defined affected frequency range ,from Retrieve the corresponding error spectrum generation interval set in the middle. ,in This indicates the total number of error spectrum sources pointing to the affected interval.

[0038] Secondly, precise extraction of the source signal of the error spectrum is performed. This is done for each error spectrum generation interval. Design a corresponding high-performance bandpass digital filter Its passband frequencies are strictly aligned The boundary. The original signal... Input this filter bank to obtain the clean main signal components in each source interval. :

[0039] (7)

[0040] In equation (7), * denotes discrete convolution operation. This step indexes the time-domain signal sequence. It ensures that the reference signal used in subsequent cancellation reconstruction contains only the source frequency components that generate the specific error spectrum, thus avoiding interference from out-of-band noise in the correction process.

[0041] Error compensation signals are reconstructed based on coupling relationships. Frequency mapping coefficients recorded in With bias coefficient Calculate the center frequency of the generated interval to the center frequency of its generated error spectrum frequency offset Constructing a real-time frequency conversion operator For the extracted source signal Perform mixing and utilize a low-pass filter. Suppress high-frequency copies of the mixed signal to obtain the pre-aligned signal. . use The amplitude-phase mismatch vector stored in The real-coefficient compensated filter is obtained by fitting using the least squares method. By passing the pre-aligned signal through this filter, dynamic compensation for amplitude gain and phase mismatch is achieved, generating an estimated error spectrum component. :

[0042] (8)

[0043] In equation (8), * denotes discrete convolution operation. This is an index for the time-domain signal sequence.

[0044] The separation and elimination of all reconstruction error spectral components from the original output signal is accomplished directly using linear subtraction. For each affected frequency range... , and its corresponding all estimation error spectral components Accumulate the signals to construct a composite error mask signal. The original signal Subtracting the composite error mask signals corresponding to all affected intervals yields the canceled correction signal. :

[0045] (9)

[0046] In equation (9), * denotes discrete convolution operation. For time-domain signal sequence index, for affected interval. index, Affected range The total number.

[0047] Beneficial effects:

[0048] 1. This invention discloses a method for characterizing the error spectrum generation mechanism based on a working frequency band coupling mapping table. By performing frequency sweep analysis on the digital bandwidth interleaved system to be corrected, a working frequency band coupling mapping table can be established to characterize the numerical model of the error spectrum generation mechanism. This table accurately characterizes the error spectrum generation law under a given tolerance threshold. Compared with traditional error characterization methods based on analytical models, this invention is easier to implement in engineering and is more suitable for nonlinear error correction of digital bandwidth interleaved sampling systems.

[0049] 2. The present invention discloses an online error spectrum reconstruction method based on an offline extraction system working frequency band coupling mapping table. By utilizing the mapping table parameters obtained in the offline stage, the error spectrum source components are extracted from the original output without needing to know the external input signal in the online stage. Frequency, amplitude, and phase compensation are performed according to the mapping table parameters, thereby realizing real-time blind reconstruction of error spectrum components under online conditions and ensuring reliable correction of nonlinear errors in the digital bandwidth interleaved sampling system caused by imperfect analog filters.

[0050] 3. The present invention discloses a digital bandwidth interleaved sampling system correction method based on error spectrum reconstruction, which can realize the reconstruction and compensation of error spectrum introduced by the imperfection of any analog filter. Compared with the traditional digital bandwidth interleaved sampling system correction method based on synthetic filter, the present invention can avoid the dependence on the assumption of subband bandwidth consistency and realize the effective correction of DBI-ADC system with non-uniform channel division. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a structural diagram of the digital bandwidth interleaving system to be corrected.

[0053] Figure 2 This is a flowchart of the offline system identification process.

[0054] Figure 3 This is a flowchart of the online system calibration process based on online error spectrum cancellation.

[0055] Figure 4 is a comparison of the power spectrum and spurious-free dynamic range performance of the system output signal under single-tone input at typical frequencies before and after correction.

[0056] Figure 5 It is a graph showing the frequency-spurious-free dynamic range of the system before and after correction.

[0057] The annotations in the attached figures are explained as follows:

[0058] Figure 4 shows the power spectrum and spurious-free dynamic range performance of the digital bandwidth interleaved sampling system before and after output correction. Figure 4(a) shows the power spectrum and spurious-free dynamic range performance of the digital bandwidth interleaved sampling system before (top) and after (bottom) output correction for a 21.7 GHz single-tone cosine input signal; Figure 4(b) shows the power spectrum and spurious-free dynamic range performance of the digital bandwidth interleaved sampling system before (top) and after (bottom) output correction for a 25.5 GHz single-tone cosine input signal; Figure 4(c) shows the power spectrum and spurious-free dynamic range performance of the digital bandwidth interleaved sampling system before (top) and after (bottom) output correction for a 39.7 GHz single-tone cosine input signal. Detailed Implementation

[0059] To better illustrate the purpose and advantages of the present invention, the invention will be further described below in conjunction with the accompanying drawings and examples.

[0060] Example 1:

[0061] This embodiment discloses a multi-channel error spectrum collaborative cancellation method for correcting a digital bandwidth interleaved sampling system, the structure of which is shown in the attached figure. Figure 1 As shown. This embodiment uses a typical digital bandwidth interleaved parallel sampling simulation system with M = 3 channels as the system to be calibrated. The frequency ranges of the three channels are 0~18.5GHz, 18.5GHz~30GHz, and 30GHz~45GHz, respectively. The system achieves a total observation bandwidth of 0~45GHz with a single-channel ADC sampling rate of 40GHz, far exceeding the 20GHz synchronous sampling bandwidth limited by the Nyquist sampling theorem. The analog domain simulation sampling rate is 240GHz, and the digital domain simulation sampling rate is 120GHz. The specific implementation steps are as follows:

[0062] S1. Error spectrum distribution characteristics and response frequency band identification:

[0063] First, a high-precision single-tone excitation signal sequence is constructed. The system operating frequency band is set to [missing information]. Within this frequency band, frequency intervals are used. Perform step-by-step scanning to generate a monotone sequence. The frequency sweep index is... Each monotone sequence contains 65,536 sampling points at an analog domain simulation sampling rate of 240 GHz.

[0064] The above signal is injected into the system, and an output sequence with a length of 32,768 sampling points is obtained at the digital reconstruction end. .right implement Point Fast Fourier Transform yields the frequency domain sequence. Calculate the power spectral density. The principal component power was obtained through the search. Exclude the area adjacent to the main signal. Then, the power of the maximum error spectral component is extracted. .

[0065] According to the formula The performance metrics for each frequency point were calculated, as shown in the attached figure. Figure 2 The frequency-SFDR response curve shown Set an adaptive threshold tolerance factor. Using the sliding window mean benchmark Determine the decision threshold By identifying the concave regions in the curve below a threshold, the following set of error spectral response frequency bands is obtained. :

[0066]

[0067] S2. Cross-channel frequency mapping and coupling correlation modeling:

[0068] against Each internal frequency point, with finer stepping. Generate single-tone input. (See attached image) Figure 3 As shown (the figure should display the power spectrum including the principal component and multiple error peaks), in the output power spectrum In the middle, set the tolerance parameters .

[0069] Identify all that meet the requirements Error spectrum peak frequency With input frequency The smooth shift of the error spectrum peaks is recorded, and their frequency trajectories are mapped using a linear fit. Determine the set of affected frequency intervals Its source generates interval sets (and the set of frequency bands for error spectrum response) The mapping relationship between each pair of corresponding frequency intervals in the (consistent) model. For the three pairs of intervals appearing in this embodiment, the following mapping relationships exist:

[0070]

[0071] It can be seen that the mapping coefficients in the three equations All values ​​are -1, meaning that all three pairs of intervals exhibit frequency reversal relationships caused by the variable sampling and frequency conversion links of the digital bandwidth interleaved sampling system.

[0072] S3. Gain response and phase mismatch characterization:

[0073] Using the finely swept frequency output signal sequence obtained in S2 Based on the frequency interval mapping relationship stored in the mapping table, for each pair of source-affected frequency intervals, design two sets of 256th order bandpass FIR digital filters corresponding to all swept frequency points. and Extract the principal component signals respectively. With error spectral component signal .

[0074] Calculate the frequency deviation between the principal component signal and the error spectrum component signal based on the frequency interval mapping relationship. For example, for a mapping relationship of interval pairs, That's equivalent to 40GHz. Constructing a frequency conversion signal. .Will Multiply it and pass it through a low-pass filter After processing and eliminating the mixed copies, a reference signal shifted to the error spectrum frequency is obtained. The analytic signals of both signals are obtained through Hilbert transform. and .

[0075] Calculate the amplitude gain coefficient for each set of analytic signals. Phase mismatch angle These parameters are then encapsulated together with the frequency mapping parameters of S2 to construct a working frequency band coupling mapping table. This table provides complete error cancellation guidance data for subsequent online calibration.

[0076] The complete process included in steps S1, S2, and S3 is attached. Figure 2 As shown.

[0077] S4. Cross-channel error spectrum collaborative cancellation guided by working frequency band coupling mapping table:

[0078] After completing the aforementioned offline system identification and obtaining the working frequency band coupling mapping table containing frequency mapping and amplitude-phase mismatch information, Then, at the simulation sampling rate An online calibration and cancellation model was built for the digital domain of the 120GHz system. The specific calibration process is as follows:

[0079] First, initialize the online cancellation module. According to... The logic recorded in the middle identifies the set of affected frequency intervals. For each set of mapping relationships, this embodiment pre-constructs... =A 256th-order FIR digital filter bank. Among them, the bandpass filter... It is used to extract error source signals, and its passband range strictly corresponds to the error spectrum generation interval. Frequency boundaries; low-pass filter It features a flat passband response and high stopband attenuation, making it suitable for spurious suppression after mixing.

[0080] Secondly, real-time branch extraction of the error source signal is performed. This involves extracting the original discrete sequence to be corrected captured by the system. Input parallel filtering architecture. The raw signal is passed through a bandpass filter. Processing and separating the source component signal that generates the specific error spectrum. :

[0081]

[0082] Among them, index The value range is determined by the simulation step size. This step enables the precise removal of characteristic frequency components that cause cross-channel aliasing in the full-band signal.

[0083] Subsequently, spectrum shifting and alignment were performed based on frequency mapping rules. Frequency offset parameters recorded in Generate a real coefficient cosine reference sequence for mixing. The extracted source component signal A dot product operation is performed with the reference sequence to perform a spectrum shift. The generated signal is then passed through a low-pass filter. Filter out the generated spectral copies to obtain the pre-aligned signal. The frequency spectrum of this signal completely overlaps with the aliasing error spectrum of the original signal.

[0084] Then, fine reconstruction compensation of amplitude and phase characteristics is performed. This is based on the frequency-dependent amplitude and phase mismatch vector obtained in the offline stage. In this embodiment, a set of compensation filter coefficients is fitted using the least squares criterion. The pre-alignment signal By inputting this compensation filter and adjusting the amplitude gain and phase delay of the signal, a high-fidelity estimation error spectrum component is generated. :

[0085]

[0086] Finally, a coordinated subtraction cancellation is performed. This will target the same affected interval. All reconstructed The estimated error spectral components are accumulated in the time domain to obtain the composite error mask signal. The mask signal is subtracted from the original signal after delay alignment to obtain the corrected system output signal. :

[0087]

[0088] in This is the group delay compensation value introduced by the cascading of each filter during the simulation process.

[0089] Step S4 includes the online system calibration process based on online error spectrum cancellation, as shown in the appendix. Figure 3 As shown.

[0090] After correction, for the principal components of the input signal at each frequency point within the error generation interval, the power of the error spectrum signal generated by the system in the output is significantly reduced. For the three error spectrum generation frequency intervals obtained in step S2, which exist in the digital bandwidth interleaved sampling system to be corrected in the embodiment, the embodiment selects one typical frequency point as an example for each interval. Specifically, the system input is set to a single-tone cosine signal at this frequency point, and the output signal power spectrum and spurious-free dynamic range performance before (top) and after (bottom) correction are plotted to demonstrate the error spectrum suppression effect of the present invention on the digital bandwidth interleaved system. It can be seen that after correction, the error spectrum peaks at the three typical frequency points are significantly reduced, and the spurious-free dynamic range is also significantly improved.

[0091] The frequency-spurious-free dynamic range performance curves of the system before and after correction across the entire operating frequency band are attached. Figure 5 As shown, comparing the spurious-free dynamic range curves of the system before and after correction, it can be seen that this embodiment significantly reduces the error spectral power caused by cross-channel aliasing within the system's operating frequency band. The performance degradation of the spurious-free dynamic range caused by the three error spectrum generation frequency ranges obtained in step S2 is eliminated, and the lower limit of the system's spurious-free dynamic range is improved by more than 12dB.

[0092] The above detailed description further illustrates the purpose, technical solution, and beneficial effects of the invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-channel error spectrum cancellation method for correction of digital bandwidth interleaved sampling systems, characterized in that: Includes the following steps: S1. By coarse-precision frequency sweeping, the frequency range in which the performance degradation is caused by the error spectrum in the digital bandwidth interleaved sampling system to be corrected is obtained, and then the error spectrum generation range in the working frequency band coupling mapping table is determined. S2. By fine-grained frequency sweeping, the system obtains the output error spectrum response characteristics of the input frequency components within all error spectrum generation intervals. Based on the preset tolerance threshold, all affected frequency intervals corresponding to each error spectrum generation interval are determined. Based on the linear regression of the two types of interval distributions, the frequency mapping relationship between the error spectrum frequency components and the original input frequency components is established. S3. Based on the Hilbert transform, perform correlation calculations to obtain the amplitude and phase mapping relationship of the error spectrum frequency component relative to the original input frequency component in each affected interval, and record it together with the frequency mapping relationship in the working frequency band coupling mapping table. S4. For each affected frequency range, extract the approximate input frequency component from the output signal, and complete the frequency conversion mapping and amplitude-phase compensation according to the working frequency band coupling mapping table that records the frequency, amplitude and phase mapping relationship. Reconstruct the error spectrum component, remove the reconstructed error spectrum component from the original output signal, and realize error spectrum cancellation.

2. The multi-channel error spectrum cancellation method for correction of a digital bandwidth interleaved sampling system as described in claim 1, characterized in that: The implementation method of S1 is as follows: Construct a high-precision single-tone excitation signal sequence, and set the complete operating frequency band of the digital bandwidth interleaved sampling system as follows: Within this frequency band, at preset low frequency intervals Step sampling is performed to generate a series of single-tone continuous wave signals. ;in, For frequency sweep index, For the first The center frequency of each sweep point For the amplitude of the excitation signal, Initial phase; frequency step interval The selection must meet the requirements for detailed characterization of the transition band of the analog filter; The generated single-tone excitation signal The digital bandwidth interleaved sampling system to be calibrated is injected, and after analog processing and digital reconstruction within the system, the discrete output sequence of the system corresponding to each frequency sweep input is obtained in the digital domain. ;in, Indicates the index of the time sampling point. This represents the total number of samples captured in a single acquisition; the sequence reflects the overall response characteristics of the system at a specific input frequency, including the main signal components and various error spectrum components introduced by the non-idealities of the system. Spurious-free dynamic range (SFDR) characteristics at each frequency point were extracted using spectral analysis; for each swept frequency point... The time-domain output sequence implement Point-wise discrete Fourier transform (DFT) yields the frequency domain representation sequence. : (1) In the formula: Output sequence in the time domain index, Frequency domain representation sequence ; Calculate the power spectral density at each frequency sweep point according to equation (1). The power of the main component of the signal is located using a peak search algorithm. and its corresponding frequency index Excluding the main signal spectral line and its adjacent spectral lines. Then, identify the maximum error spectral component power of the remaining frequency bands within the full bandwidth. The system's frequency is calculated based on the formula for the maximum error spectral component power. The normalized error suppression ratio at that point, i.e. : (2) Constructing frequency using adaptive thresholding method The system's response curves are plotted and the error-sensitive intervals are extracted. The calculation results corresponding to all frequency sweep points are then aggregated and mapped to construct the system's full-band performance response characteristic curves. The local mean benchmark for this curve is calculated using a sliding window. Therefore, an adaptive decision threshold is defined. ,in The preset performance degradation tolerance factor is used; the performance is extracted through logical comparison operations. The frequency set was used to identify regions of performance degradation with significant dips in the curve, which were defined as the error spectral response frequency band intervals. ,in Indicates the identified first A continuous interval of error frequency bands; this interval set The frequency range in which cross-channel aliasing is most pronounced in the recording system.

3. The multi-channel error spectrum cancellation method for correction of a digital bandwidth interleaved sampling system as described in claim 2, characterized in that: The implementation method of S2 is as follows: For the error spectral response frequency band range identified in step S1 Perform refined frequency sweep analysis; within the interval Internal step frequency Generate high-density monotone input signal ,in for Frequency sweep index within the range, The corresponding input excitation frequency is used; this series of signals is injected into a digital bandwidth interleaved sampling system to obtain the digitally reconstructed output sequence. For each set of output sequences, high-resolution power spectral density estimation is performed to obtain the spectral function. : (1) In equation (3), For windowing functions, For sequence index, For sequence length, Spectral function sequence Index; in spectral function Mid-position principal component power point and their corresponding frequencies ,in In an ideal state, it should equal... ; Perform error peak screening based on tolerance parameters; Preset an adjustable human tolerance parameter Its unit is dBc; ergodic spectral function Identify the set of error spectral peaks for all local maxima points outside the main component frequency region. ,in Use the error spectral peak index; for each element in the set, calculate its power difference relative to the principal component. If the judgment conditions are met Then the frequency point The error spectrum frequency is determined to be the one to be eliminated; due to the hardware characteristics of the digital bandwidth interleaved sampling system, this error spectrum frequency... Located in Within different physical channel frequency bands; Construct the frequency mapping function and the set of affected intervals; as the input frequency... exist Internal smoothing variation, recording the frequency of the selected error spectrum The motion trajectory is used to establish a mapping vector from the input frequency to the error spectrum frequency. By performing cluster analysis on the mapping vectors of all frequency sweep points i, the set of all frequency intervals within the system's operating frequency band that are affected by the error spectrum and thus suffer from performance degradation is identified. , where u is the index of the affected interval; For each affected interval The error spectrum generation interval is generated by inversely tracing the mapping vector. ; Characterize the algebraic coupling relationship between the generating interval and the affected interval; for each pair of mapping intervals The frequency transformation slope was determined by linear regression. Due to the presence of variable sampling rate components and multi-stage mixing structures within the system, the mapping relationship exhibits either a forward or inverse mapping: when When > 0, it satisfies A positive growth relationship; when When <0, it exhibits frequency reversal characteristics, meaning the high-frequency segment of the generated interval corresponds to the low-frequency segment of the affected interval; obtaining a frequency range containing arbitrary... With its Corresponding frequency mapping parameter pair parameter pair set .

4. The multi-channel error spectrum cancellation method for correction of a digital bandwidth interleaved sampling system as described in claim 3, characterized in that: The S3 implementation method is as follows: Extract the system output sequence obtained in step S2 And perform signal component separation based on the established frequency coupling relationship; for the frequency sweep index The principal component frequency below With error spectrum frequency Two sets of narrowband high-performance digital filters were designed respectively. and ; Output sequence By inputting the above filters respectively, time-domain signals containing only the principal components are obtained. With time-domain signals containing only specific error spectral components ; The filtering process is expressed as follows: (1) In equation (4), * denotes discrete convolution operation. For indexing time-domain signal sequences; Perform signal spectrum shifting and alignment based on frequency mapping rules; based on the frequency deviation between the principal component and the error spectrum component in the mapping vector established in step S1.

2. Construct a real-coefficient frequency conversion reference signal ,in The system sampling rate; the principal component signal and Spectrum shifting is achieved by performing point-by-point multiplication operations to generate an intermediate signal containing a mixing copy. Then a low-pass filter was used. By filtering out irrelevant spectral copies generated after the shift, a reference principal component signal with the same center frequency as the error spectral component is obtained. ; Calculate the amplitude gain and phase mismatch parameters of the error spectral components relative to the reference principal component signal; construct the model using the Hilbert Transform. and The analytic signal forms are respectively expressed as: and Define the amplitude gain coefficient. The root mean square ratio of the two signals within the observation window: (2) In equation (5), To parse the signal sequence index, The sequence length; Meanwhile, the phase mismatch angle is obtained through the cross-correlation phase estimation method. : (3) In equation (6), Indicates taking the complex argument. Indicates complex conjugation. To parse the signal sequence index, For sequence length, and The definition is the same as in equation (5); this parameter is for Accurately characterize the distortion features of the error spectrum generated by the coupling of the input signal in the amplitude and phase dimensions; Perfecting the parameter pair set Coupling Mapping Table with Operating Frequency Band ; will be determined by parameter pairs Composed amplitude and phase error sequence As an attribute field, it is added to the corresponding parameter pair collection. In the middle; generate a set of intervals based on the error. Affected frequency frequency range set Parameter pair set Output a complete working frequency band coupling mapping table This table records in detail all affected frequency ranges. and its source error spectrum generation interval Boundary frequency, mapping coefficient With mapping bias And the amplitude and phase error sequence pairs that evolve with frequency for each mapping pair This provides prior identification data for error compensation signal reconstruction in the subsequent online calibration process.

5. The multi-channel error spectrum cancellation method for correction of a digital bandwidth interleaved sampling system as described in claim 4, characterized in that: The S4 implementation method is as follows: Extracting the working frequency band coupling mapping table constructed in the offline stage based on the working frequency band coupling mapping table. The topology information is used to initialize the online cancellation network; the original signal to be corrected is output in real time by the system. Identify the set of all frequency intervals affected by the error spectrum within its full bandwidth. For each specific affected frequency range ,from Retrieve the corresponding error spectrum generation interval set in the middle. ,in This indicates the total number of error spectrum sources pointing to the affected interval; Secondly, the source signal of the error spectrum is accurately extracted; for each error spectrum generation interval... Design a corresponding high-performance bandpass digital filter Its passband frequencies are strictly aligned The boundary; the original signal Input this filter bank to obtain the clean main signal components in each source interval. : (1) In equation (7), * denotes discrete convolution operation. For indexing time-domain signal sequences; Reconstruct the error compensation signal based on the coupling relationship; according to Frequency mapping coefficients recorded in With bias coefficient Calculate the center frequency of the generated interval to the center frequency of its generated error spectrum frequency offset Constructing a real-time frequency conversion operator For the extracted source signal Perform mixing and utilize a low-pass filter. Suppress high-frequency copies of the mixed signal to obtain the pre-aligned signal. ;use The amplitude-phase mismatch vector stored in The real coefficient compensated filter is obtained by fitting. ; By passing the pre-aligned signal through this filter, dynamic compensation for amplitude gain and phase mismatch is achieved, generating an estimated error spectrum component. : (2) In equation (8), * denotes discrete convolution operation. For indexing time-domain signal sequences; By directly using linear subtraction, all reconstruction error spectral components are separated and eliminated from the original output signal, resulting in the canceled correction signal.

6. The multi-channel error spectrum cancellation method for correction of a digital bandwidth interleaved sampling system as described in claim 7, characterized in that: In S4, a linear subtraction operation is used to separate and eliminate all reconstructed error spectral components from the original output signal, resulting in the canceled corrected signal. The specific implementation method is as follows: For each affected frequency range , and its corresponding all estimation error spectral components Accumulate the signals to construct a composite error mask signal. ; the original signal Subtracting the composite error mask signals corresponding to all affected intervals yields the canceled correction signal. : (1) In equation (9), * denotes discrete convolution operation. For time-domain signal sequence index, for affected interval. index, Affected range The total number.

7. The multi-channel error spectrum cancellation method for correction of a digital bandwidth interleaved sampling system as described in claim 1 or 2, characterized in that: In S1, Frequency step interval The value should not exceed 1% of the sub-ADC bandwidth.

8. The multi-channel error spectrum cancellation method for correction of a digital bandwidth interleaved sampling system as described in claim 7, characterized in that: In S4, The real-coefficient compensated filter is obtained by least-squares fitting. .

Images