A method for eliminating harmonic spurs in a wideband acquisition system based on adaptive recognition

Abstract

The application discloses a wideband acquisition system harmonic spur elimination method based on adaptive identification, and comprises the following steps: S1. in the pre-training process, input signal x i (t) enters the wideband acquisition system, and the output signal itself is accompanied by the generation of harmonic spur and noise, at which time the output acquisition signal is expressed as x i [n]; S2. the amplitudes and phases of the main signal and the second harmonic signal of a plurality of frequency points are fitted, the phase offset of the main signal after being squared and the second harmonic signal is calculated, the phase offset adjustment is carried out by using the phase offset half-phase period compensation technology, the amplitude-frequency adjustment FIR filter, the phase-frequency adjustment frequency domain filter module and the delay compensation missing point number are designed; S3. in the spur elimination process, the input signal x(t) enters the wideband acquisition system, the wideband acquisition system outputs the acquisition signal x[n], and the signal after the spur elimination is obtained The application reduces the implementation difficulty of harmonic spur elimination, and improves the elimination precision.

Description

Technical Field

[0001] This invention relates to a broadband acquisition system, and more particularly to a method for eliminating harmonic spurious emissions in a broadband acquisition system based on adaptive identification. Background Technology

[0002] In broadband acquisition systems, harmonic spurious signals, such as second harmonics, severely impact signal integrity, signal-to-noise ratio, and dynamic range. Existing technologies primarily rely on hardware optimization or digital processing. Hardware optimization traditionally suppresses spurious signals by improving the linearity of analog devices or frequency planning, such as adjusting the phase-locked loop (PLL) reference frequency. However, these methods have several drawbacks. For example, frequency planning requires a pre-set frequency planning table, cannot adapt to dynamic changes in broadband signals in real time, lacks flexibility, and relies on high-precision devices and complex circuits, resulting in high hardware costs. Digital processing typically involves static filter design and open-loop correction algorithms to suppress harmonic spurious signals. However, these methods still suffer from bandwidth limitations, performance degradation in dynamic scenarios, and limited depth of harmonic spurious suppression.

[0003] Currently, research on harmonic spurious cancellation technology in digital processing mainly focuses on static filter design and open-loop correction algorithms. The patent application "CN202010838456.0," titled "A Broadband Signal Generation Device and Method Based on Adaptive Correction," employs a static filter design. However, this method is only applicable to narrowband signals, and since spurious signals are not within the frequency band, it cannot be applied to broadband acquisition. The patent application "CN202411506846.2," titled "A Method and Device for Nonlinear Spurious Suppression of a Broadband Receiver ADC," proposes an adaptive correction algorithm that can identify and eliminate the amplitude and phase of harmonic spurious signals. However, because it requires a complex algorithm for identification with each signal input, it is difficult to implement and has poor real-time performance. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a harmonic spurious cancellation method for a broadband acquisition system based on adaptive identification, which reduces the difficulty of implementing harmonic spurious cancellation and improves the cancellation accuracy.

[0005] The objective of this invention is achieved through the following technical solution: a method for eliminating harmonic spurious signals in a broadband acquisition system based on adaptive identification, comprising the following steps:

[0006] S1. During the pre-training process, the input signal x i (t) When entering the broadband acquisition system, the output signal itself will be accompanied by harmonic spurious signals and noise. At this time, the output acquisition signal is represented as x. i [n];

[0007] S2. Fit the amplitude and phase of the main signal and its second harmonic signal at multiple frequency points using the adaptive RLS algorithm, calculate the phase shift between the squared main signal and the second harmonic signal at each frequency point, use the phase shift half-phase period compensation technology to adjust the phase shift, and design an amplitude-frequency adjustment FIR filter, a phase-frequency adjustment frequency domain filter module, and the number of lost points for delay compensation.

[0008] S3. During spurious signal cancellation, the input signal x(t) enters the broadband acquisition system, and the broadband acquisition system outputs the acquired signal x[n]; after spurious signal cancellation, the signal is obtained.

[0009] The beneficial effects of this invention are: 1) In the broadband signal second harmonic elimination method of this invention, a pre-training and real-time processing module is used, wherein the computational complexity is mainly reflected in the pre-training module. The filter bank information calculated by the pre-training module is put into the real-time processing module during system operation, which greatly improves the real-time performance of the system.

[0010] 2) In the broadband signal second harmonic elimination method of the present invention, the amplitude and phase estimation method of the original signal and spurious signal based on the adaptive method is more accurate than the traditional FFT estimation method. In particular, the accuracy of the estimation of high frequency spurious signals in broadband signals is especially important. The accuracy of its phase information determines the degree of phase alignment between the subsequent second harmonic fitting signal and the real second harmonic signal, and to a large extent determines the suppression depth of the second harmonic of the high frequency signal.

[0011] 3) In the broadband signal second harmonic elimination method of the present invention, compared with the traditional fixed filter design which can only filter out out-of-band spurious signals, the present invention can apply the second harmonic suppression method to broadband signals by accurately calculating the phase shift at different frequency points and designing the filter bank, and also has a good suppression effect on in-band spurious signals.

[0012] 4) Thanks to the accurate estimation of the adaptive method and the technique of dropping sampling points during the half-cycle of group delay, the accuracy of the filter bank is greatly improved and the design difficulty is reduced, enabling the elimination of second harmonics within the 0–18 GHz bandwidth. On the other hand, the improved fitting accuracy for the second harmonic signal can enhance the suppression depth of the second harmonic, resulting in a maximum suppression depth of approximately 35 dB and an average suppression depth of 20 dB at each frequency.

[0013] 5) The broadband signal second harmonic cancellation method of the present invention is not limited to the second harmonic cancellation of broadband signals. Based on the same method, it can be extended to the third or multi-order spurious cancellation, and has strong scalability. Attached Figure Description

[0014] Figure 1 For both pre-training and real-time processing modules;

[0015] Figure 2 This diagram illustrates the data frame positions and the effective data after filtering for non-overlapping frames and frames with two overlapping frames.

[0016] Figure 3 A schematic diagram of the second harmonic of a 1GHz signal and its fitted signal;

[0017] Figure 4 The spectrum diagrams of the original 1GHz signal and the signal after second harmonic elimination are shown.

[0018] Figure 5 A schematic diagram of the second harmonic of a 7GHz signal and its fitted signal;

[0019] Figure 6 The spectrum diagrams of the original 7GHz signal and the signal after second harmonic elimination are shown.

[0020] Figure 7 This is a schematic diagram of the second harmonic suppression depth in the 0-9GHz range. Detailed Implementation

[0021] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following description.

[0022] The technical solution of this invention is divided into a pre-training module and a real-time processing module. For example... Figure 1 As shown, during the pre-training process, the input signal x i (t) When the signal enters the broadband acquisition system, due to the non-ideal characteristics of the system, the output signal itself will be accompanied by harmonic spurious signals and noise. At this time, the output acquisition signal is represented as x. i [n]. For ease of explanation, let's take a single-frequency signal and its second harmonic as an example, assuming... Let Ω be a single-tone signal at input frequency i, where Ω i Its simulated frequency, This is the initial phase. After being acquired by the broadband acquisition system, the sampled signal x... i [n] can be approximated as

[0023]

[0024] Where ω i =Ω i / f s f is the digital angular frequency of the input signal; s The sampling frequency; a i , a' represents the amplitude and phase of the original signal after sampling. i , The amplitude and phase of its second harmonic signal.

[0025] This invention provides an adaptive identification module that utilizes the adaptive recursive least squares (RLS) algorithm to obtain the amplitude and phase information of the main signal and its second harmonic {a}. i 、a' i}, The calculated amplitude and phase information can be used to generate amplitude-frequency adjusted FIR filter coefficients and phase-frequency adjusted frequency filter parameters, thus providing the necessary conditions for the real-time signal processing module.

[0026] This invention utilizes the square of the primary signal to fit its second harmonic signal, and then eliminates the influence of the second harmonic by subtracting it from the primary signal. There is an amplitude difference between the primary and second harmonic signals, and an absolute phase shift exists between the squared primary signal and its second harmonic. Therefore, this invention designs a second harmonic fitting filter bank to compensate for the aforementioned amplitude and phase errors. Specifically, the second harmonic fitting filter bank includes an amplitude-frequency adjusted finite impulse response (FIR) filter and a phase-frequency adjusted frequency domain filter. The amplitude-frequency adjusted FIR filter has a fixed linear delay while changing the amplitude of the original signal; the phase-frequency adjusted frequency domain filter module compensates for the nonlinear absolute phase delay at each frequency point of the broadband acquisition system.

[0027] In a broadband acquisition system, experimental data shows that the phase delay between the original signal and its second harmonic at different frequencies is not the same but remains constant. Frequency sweeping can calculate the phase delay at each frequency point across the entire bandwidth. However, because the swept signal is discrete, the high-frequency portion of the large bandwidth has a high frequency, leading to a large error in the differential operation during fitting. Therefore, using the Infinite Impulse Response (IIR) fitting group delay method to compensate for the phase shift error results in a large error. If FIR is used to compensate for the phase shift, a symmetrical linear phase FIR cannot be used. In the design of this FIR, besides the given points that can be fitted, the remaining frequency points will exhibit large fluctuations, resulting in a similarly large error. In summary, in the phase shift compensation part, this invention adopts a frequency domain filtering technique based on overlapping frequency domain frames, which produces the smallest error compared to IIR and FIR when the amplitude and phase flatness are high. Furthermore, a half-phase periodic compensation technique is proposed in the phase fitting, which can improve the phase shift flatness between each frequency point, reduce the error caused by insufficient frequency resolution in subsequent frequency domain filtering, and improve the fitting accuracy.

[0028] Single-frequency sampling signal x i [n] After passing through the second harmonic fitting filter bank, a squaring operation is performed to fit and reconstruct the second harmonic signal, ultimately resulting in x. i [n] Subtract the second harmonic from the fitted and reconstructed signal to eliminate the second harmonic and obtain the corrected signal x. i [n]. x i The specific expression of [n] is as follows:

[0029]

[0030] Where λ i τ is the amplitude adjustment factor. i This is a delay adjustment factor.

[0031] This invention is achieved through the following specific steps:

[0032] (1) The adaptive RLS algorithm accurately fits and estimates the amplitude and phase of the main and second harmonic signals at each frequency point;

[0033] From formula (1), we can see that four parameters need to be fitted for each frequency point. To facilitate modeling, the cosine function in formula (1) is expanded as follows:

[0034]

[0035] in() T Let a represent the transpose of a vector. i =[cos(ω i n),sin(ω i n),cos(2ω i n),sin(2ω i [n)] is the input vector, For parameter vectors.

[0036] Formula (3) completes the sampling of signal x i [n] modeling, using the RLS algorithm to model w i The specific steps of the RLS algorithm for fitting are as follows:

[0037]

[0038]

[0039] Complete w i , After fitting, it is necessary to restore a. i , and a' i , From formula (3), we can obtain:

[0040] Where w i (1), w i (2), w i (3), w i (4) is a vector w i The 1st, 2nd, 3rd, and 4th elements in the dataset.

[0041] (2) Calculate the phase shift between the squared original signal and the second harmonic signal at each frequency point within the broadband;

[0042] The broadband signal phase offset calculation module will calculate the phase offset between the squared main signal and its second harmonic at each frequency point based on the phase information identified by the adaptive recognition module. In the single-frequency case, assuming the main signal is... Its second harmonic is Let λ i For amplitude adjustment factor, The phase adjustment factor is used to adjust the amplitude and phase of the original signal after passing through a second harmonic fitting filter bank. Subsequently, the original signal is squared to fit the second harmonic signal. During this process, the fitting process needs to satisfy the following equation:

[0043]

[0044] In digital processing, the DC bias λ generated by formula (4) i 2 a i 2 / 2 can be calculated and eliminated through signal averaging, ultimately yielding the phase shift at that frequency.

[0045] (3) Use phase shift half-phase period compensation technology to adjust the passband phase response;

[0046] Based on actual measurement data, the phase shift between the squared signal and its second harmonic signal at each frequency point may exhibit significant fluctuations. The uneven phase shift curve at multiple frequencies can lead to large differences in the phase shift values ​​between adjacent frequency points. Poor flatness of the phase shift curve can result in significant phase shift errors at non-acquisition frequency points in subsequent frequency domain filtering. To reduce this phase shift error, this invention proposes a phase shift half-phase period compensation technique. Since the frequency domain filtering module only considers phase shift correction, without considering amplitude and DC bias, there are...

[0047]

[0048] As can be seen from (5), in phase shift Adding or subtracting half-phase period π to the base phase will result in a second harmonic phase that is added or subtracted by 2π, without affecting the squared phase. Therefore, half-phase period compensation near frequencies with large phase shift fluctuations can improve the flatness of the passband phase shift, reduce the phase shift error of subsequent frequency domain filtering, and improve the suppression depth of the second harmonic signal.

[0049] (4) Design a second harmonic fitting filter bank;

[0050] As shown in formula (4), in order for the original signal to fit its second harmonic signal after squaring, its amplitude and phase must be adjusted. In order to improve the fitting accuracy, the second harmonic fitting filter bank uses an amplitude-frequency adjusted FIR filter and a phase-frequency adjusted frequency domain filter based on frequency domain overlapping frames to adjust the amplitude and phase of the original signal.

[0051] 1) Design of amplitude-frequency adjusted FIR filter

[0052] As shown in formula (4), in order for the squared original signal to have the same amplitude as its second harmonic signal, the following relationship must be satisfied.

[0053]

[0054] The amplitude adjustment coefficient at this frequency point is obtained. After scanning the frequency signal, the amplitude adjustment coefficients across the entire bandwidth can be obtained. Based on the amplitude adjustment coefficients λ at each frequency point... i , The design incorporates compensation coefficients for an amplitude-frequency adjustable FIR filter to ensure that the squared original signal has the same amplitude as its second harmonic signal. Here, the FIR filter employs a symmetrical structure with a fixed group delay, thus not affecting the phase relationship at each frequency point.

[0055] 2) Design of phase frequency adjustment frequency domain filter module

[0056] In practical engineering, continuous transformation of DTFT or IDTFT is not feasible. A common solution is to sample in the frequency domain and convert it into DFT and IDFT for processing. However, sampling in the frequency domain introduces spectral leakage. Furthermore, the actual amount of sampled data may exceed the Mpts level. Directly calculating the spectrum with an extremely large number of points for phase offset adjustment is difficult to implement in practical systems. Therefore, this invention employs a phase-frequency adjustment frequency domain filtering technique based on overlapping frequency domain frames.

[0057] With a fixed number of frequency domain sampling points, frequency shifts at non-integer multiples of the period will cause spectral leakage. This will result in significant fluctuation errors at the beginning and end of the frame signal after IDFT, reducing fitting accuracy. To mitigate this error, this invention employs frequency domain overlapping frame technology to compensate for the phase shift values ​​at each frequency point; for example... Figure 2 As shown, this invention uses two identical filtering modules, each based on the phase offset value at each frequency point. The input data is misaligned and compensated. The outputs of the two filtering modules each retain 50% of the data with smaller errors and are then combined to achieve a high fitting accuracy. This method can also meet the filtering requirements of large data volumes. Furthermore, since frequency domain filtering does not introduce additional delays generated during the design of time domain filters, the filtered signal generally does not have additional data loss due to the filter order.

[0058] (5) Delay compensation and harmonic elimination.

[0059] In the design of the second harmonic fitting filter bank, the amplitude-frequency adjusted FIR filter will generate a fixed time delay, which is (N-1) / 2, where N is the filter order.

[0060] To ensure that the original signal, after passing through the filter bank and being squared, perfectly fits the second harmonic, it is necessary to compensate for the time delay generated by the filter bank. Since the generated time delay is constant, the corresponding number of sampling points can be directly discarded during digital processing to achieve the purpose of time delay compensation.

[0061] The technical effects of this invention will be described in detail below, based on actual engineering tests:

[0062] 1. Test conditions and content:

[0063] Within the 0–18 GHz bandwidth range, since the second harmonic of signals exceeding 9 GHz is not within the frequency band, it can be directly filtered out using a linear FIR low-pass filter. Therefore, the test results of this invention mainly focus on the second harmonic of input signals within the 9 GHz bandwidth range. Using a signal source, a frequency sweep signal is performed in the 0–9 GHz band with steps of 100 MHz, and the signal is sampled by a broadband acquisition system with a sampling rate of 80 GSa / s. This invention employs a broadband signal second harmonic elimination method based on adaptive identification for pre-training the module. After obtaining the specific parameters of the second harmonic fitting filter bank, the parameters are set in the real-time processing module. Subsequently, the real-time elimination processing of the second harmonic signal is verified again using the sampled 0–9 GHz frequency sweep input signal with steps of 10 MHz.

[0064] 2. Test Result Analysis:

[0065] Reference Figure 3 The two signals are the second harmonic signal generated after sampling a single-frequency signal with an input frequency of 1 GHz and the second harmonic signal fitted using the method of this invention. It can be seen that the two signals have a high degree of fit in both phase and amplitude.

[0066] Reference Figure 4 The spectrum diagrams are of the original signal after sampling of a single-frequency signal with an input frequency of 1 GHz and the signal after second harmonic elimination using the method of this invention. It can be seen that the second harmonic suppression at this frequency point can reach about 21 dB.

[0067] Reference Figure 5 The two signals are the second harmonic signal generated after sampling a single-frequency signal with an input frequency of 7 GHz and the second harmonic signal fitted using the method of this invention. It can be seen that the two signals still have a high degree of fit in terms of phase and amplitude.

[0068] Reference Figure 6 The spectrum diagrams are of the original signal after sampling of a single-frequency signal with an input frequency of 7 GHz and the signal after second harmonic elimination using the method of this invention. It can be seen that the second harmonic suppression at this frequency point can reach about 23 dB.

[0069] Reference Figure 7 The value is the second harmonic suppression depth of the signal within an 18GHz bandwidth after using the method of this invention (the second harmonic outside the bandwidth is filtered out by a low-pass filter). It can be seen that the method proposed in this invention is applicable to wide bandwidth signals, with a maximum suppression depth of 35dB and an average suppression depth of 20dB.

[0070] The foregoing description illustrates and describes a preferred embodiment of the present invention. However, as previously stated, it should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the inventive concept described herein through the foregoing teachings or techniques or knowledge in related fields. Any modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for harmonic spurious emission cancellation in a broadband acquisition system based on adaptive identification, characterized in that: Includes the following steps: S1. During the pre-training process, the input signal... When inputting into a broadband acquisition system, the output signal itself will be accompanied by harmonic spurious signals and noise. At this time, the output acquisition signal is represented as... ; S2. Fit the amplitude and phase of the main signal and its second harmonic signal at multiple frequency points using the adaptive RLS algorithm, calculate the phase shift between the squared main signal and the second harmonic signal at each frequency point, use the phase shift half-phase period compensation technology to adjust the phase shift, and design an amplitude-frequency adjustment FIR filter, a phase-frequency adjustment frequency domain filter module, and the number of lost points for delay compensation. S201. The amplitude and phase of the multi-frequency main signal and its second harmonic signal are estimated by fitting the adaptive RLS algorithm; According to step S1, obtain the corresponding values ​​of multiple frequency points i. ; Each frequency point requires fitting 4 parameters. Expanded to: (2) in Represents the transpose of a vector. For the input vector, For parameter vectors; Using the RLS algorithm Perform fitting to complete the process. After fitting, it is necessary to restore the original state. , and , From formula (2), we get: ， ， ， ， ， in , , , For vectors The 1st, 2nd, 3rd, and 4th elements in the dataset; S202. Calculate the phase shift between the squared main signal and the second harmonic signal at each frequency point within the broadband; Based on the identification results of step S201, calculate the phase shift between the squared main signal and its second harmonic at each frequency point: Assume the main signal, i.e. the desired signal corresponding to a single audio point i, is... Its second harmonic is ;set up For amplitude adjustment factor, As the phase adjustment factor, the fitting process needs to satisfy the following equation: (3) In digital processing, the DC bias generated by formula (3) The phase shift at that frequency point is finally obtained by averaging and eliminating the cosine terms on both sides of the equation, making them equal. ; S203. Phase offset adjustment is performed using phase offset half-phase period compensation technology; In phase shift Add / subtract half the phase period based on , get new To achieve phase shift Adjustments; S204. Design a second harmonic fitting filter bank, including an amplitude-frequency adjusted FIR filter and a phase-frequency adjusted frequency domain filter module, and determine the number of lost points for delay compensation; S3. During spurious signal elimination, the input signal The signal enters the broadband acquisition system and is then output by the broadband acquisition system. The signal is obtained after spurious emission reduction. .

2. The method for harmonic spurious elimination in a broadband acquisition system based on adaptive identification according to claim 1, characterized in that: Step S1 includes: Assumption Let i be a single-tone signal at input frequency i, where Its simulated frequency, As the initial phase, after being acquired by the broadband acquisition system, the sampled signal... Approximately expressed as: ； in The digital angular frequency of the input signal; The sampling frequency; , The amplitude and phase of the original signal after sampling; , The amplitude and phase of its second harmonic signal.

3. The method for harmonic spurious emission cancellation in a broadband acquisition system based on adaptive identification according to claim 1, characterized in that: Step S204 includes: Amplitude-frequency adjusted FIR filter design: To ensure that the squared amplitude of the main signal is equal to the amplitude of its second harmonic signal, the following relationship must be satisfied. ； The amplitude adjustment coefficient at this frequency point is obtained. After scanning the frequency signal, the amplitude adjustment coefficient within the full bandwidth is obtained. Based on the amplitude adjustment coefficient of each frequency point The amplitude-frequency adjustment FIR filter compensation coefficient is designed to achieve the goal of making the amplitude of the original signal equal to that of its second harmonic signal after squared. The FIR filter adopts a symmetrical structure, has a fixed group delay, and does not affect the phase relationship of each frequency point. Phase frequency adjustment frequency domain filter module design: Frequency domain filtering techniques using overlapping frequency domain frames are employed to compensate for the phase shift values ​​at each frequency point: two identical filtering modules are used to compensate for the phase shift values ​​at each frequency point. The input data is compensated for for misalignment. The outputs of the two filtering modules are combined by retaining the middle 50% of the data with smaller errors, and finally a higher fitting accuracy is achieved. The number of lost points for delay compensation is determined as: the number of sampling points corresponding to the delay of the FIR filter.

4. The method for harmonic spurious emission cancellation in a broadband acquisition system based on adaptive identification according to claim 1, characterized in that: Step S3 includes: During spurious signal elimination The signal enters the broadband acquisition system and is then output by the broadband acquisition system. ; Collected signals The signals are sequentially fed into the amplitude-frequency adjustment FIR filter and the phase-frequency adjustment frequency domain filter module. The data output from the phase frequency adjustment frequency domain filter module is squared and then time delay compensation is performed: that is, the number of sampling points corresponding to the delay of the FIR filter is directly discarded to achieve the purpose of time delay compensation. Reuse Subtracting the delay-compensated signal yields the spurious-cancelled signal. .

Images