A Real-Time Dynamic Multipath Identification Method for Wireless Frequency Synchronization

Abstract

This invention discloses a real-time dynamic multipath identification method for wireless frequency synchronization, comprising: receiving a signal superimposed from N propagation paths from a slave node, down-converting the signal to obtain a baseband signal; performing pulse compression using a structurally underconstrained least-squares pulse compression filter, detecting N maxima points to determine the propagation delay of the N propagation paths; performing pulse compression on the baseband signal using a spectrum-corrected matched filter, and performing frequency domain weighting on the result to suppress sidelobes, thereby obtaining a second compression result; constructing a system of linear equations and solving for the amplitude and additional phase of the N propagation paths; using the path with the largest amplitude or the smallest delay as the direct wave signal, extracting its propagation delay and additional phase, and adjusting the slave node's crystal oscillator frequency accordingly. This invention's real-time dynamic multipath identification method for wireless frequency synchronization identifies dynamic multipaths in real time, reconstructs the direct wave, and has high propagation delay resolution and good noise immunity.

Description

Technical Field

[0001] This invention belongs to the field of frequency synchronization technology in distributed coherent wireless systems, and specifically relates to a dynamic multipath real-time identification method for wireless frequency synchronization. Background Technology

[0002] In distributed coherent systems, high-precision frequency synchronization is a core technology for ensuring efficient and reliable system operation. Wireless frequency synchronization methods based on custom protocols have gradually become a major research trend in this field due to their advantages such as high accuracy, flexible deployment, adaptability to node mobility, and wide applicability. However, dynamic multipath effects caused by environmental changes severely affect the accuracy and stability of wireless frequency synchronization methods based on custom protocols, posing a significant challenge to related research.

[0003] Currently, research on dynamic multipath interference under wireless frequency synchronization is still in its early stages both domestically and internationally. The vast majority of wireless frequency synchronization mechanisms based on custom protocols do not consider multipath factors, with only a few proposing methods primarily focused on suppressing multipath effects. These methods can only reduce the impact of dynamic multipath effects to a certain extent, resulting in significant accuracy loss and making them unsuitable for high-precision wireless frequency synchronization. Summary of the Invention

[0004] This invention provides a dynamic multipath real-time identification method for wireless frequency synchronization to solve the aforementioned technical problems, specifically adopting the following technical solution:

[0005] A dynamic multipath real-time identification method for wireless frequency synchronization includes the following steps:

[0006] S1: The master node sends a wireless frequency synchronization signal in period T0;

[0007] S2: Receive the signal r from the node after it has traveled through N propagation paths and been superimposed. N (t), and for r N (t) performs down-conversion to obtain the baseband signal y N (t);

[0008] S3: Using a structurally underconstrained least-squares pulse compression filter to apply y N (t) Perform pulse compression to obtain the first compression result, and detect the N maxima in the first compression result to determine the propagation delay τ1…τ of the N propagation paths. N ;

[0009] S4: First, use a spectrum-corrected matched filter to apply the filter to y. N (t) Perform pulse compression, and then perform frequency domain weighting on the result to suppress sidelobes, thereby obtaining the second compression result;

[0010] S5: Based on the determined propagation delay τ1…τ N Extract the complex values ​​at the corresponding positions from the second compression result as constant vectors, construct a system of linear equations and solve for the amplitude and additional phase of N propagation paths;

[0011] S6: Using the path with the largest amplitude or the smallest delay as the direct wave signal, extract its propagation delay and additional phase, and adjust the slave node crystal oscillator frequency accordingly to achieve high-precision frequency synchronization with the master node.

[0012] Furthermore, the structurally underconstrained least-squares pulse compression filter is generated from an ideal pulse compression matrix U, and the matrix is ​​obtained by setting the elements of several adjacent rows before and after the l-th row corresponding to the main lobe of U to zero. by As filter coefficients. Further, the frequency domain weighting adopts any one of the following: Hamming window, 3:1 cone ratio window, cosine square window, cosine cubic window, or cosine fourth power window.

[0013] Furthermore, in step S6, the path with the largest amplitude or the smallest propagation delay is used as the direct wave signal.

[0014] Furthermore, the position of the first path signal is detected by setting an amplitude threshold, and convolution operations are performed only on Q sampling points on the left and right sides of that position, where Q is a preset integer.

[0015] Furthermore, the period T0 satisfies T0≥τ max , τ max To maximize the identifiable propagation delay, in order to avoid overlap of compression results between adjacent periodic pulses.

[0016] Furthermore, the cascaded order of the spectrum correction and frequency domain weighting is to perform spectrum correction first and then frequency domain weighting.

[0017] Furthermore, the latency resolution is 5ns.

[0018] Furthermore, the additional phase estimation error of the direct wave signal is less than 0.02°.

[0019] Furthermore, the wireless frequency synchronization signal is a linear frequency modulation (LFM) reference signal.

[0020] The advantages of this invention are that it provides a real-time dynamic multipath identification method for wireless frequency synchronization, which identifies dynamic multipath in real time, restores the direct wave, and has high propagation delay resolution and good noise resistance. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application 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 only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating a dynamic multipath real-time identification method for wireless frequency synchronization according to this application.

[0023] Figure 2 This is another flowchart illustrating a dynamic multipath real-time identification method for wireless frequency synchronization according to this application.

[0024] Figure 3 This is a schematic diagram of the low-latency hardware logic design framework of this application;

[0025] Figure 4 This is a schematic diagram of another low-latency hardware logic design framework for this application. Detailed Implementation

[0026] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0027] Furthermore, the terms "first" and "second" in the specification, claims, and drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those described herein. At the same time, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. Unless otherwise expressly specified and limited, the terms "set," "arranged," "installed," "connected," and "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to the internal connection of two components or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this case based on the specific circumstances and in conjunction with the prior art.

[0028] Distributed coherent systems that use custom protocols for wireless frequency synchronization often employ a master-slave architecture. The master node uses a high-precision, high-stability crystal oscillator as the standard for frequency synchronization across the entire coherent system. The remaining slave nodes use their own programmable crystal oscillators and periodically perform frequency calibration with the master node according to the custom protocol to ensure the frequency synchronization accuracy of the entire coherent system over a long period.

[0029] Taking a linear frequency modulated (LFM) periodic pulse signal as a wireless frequency synchronization signal as an example. Let the LFM reference signal be:

[0030]

[0031] Where A is the amplitude of the LFM signal, T is the duration of the LFM signal, and B is the bandwidth of the LFM signal. It can be remembered as:

[0032]

[0033] If the frequency of the LFM signal is the modulation frequency, then the LFM reference signal can be simply referred to as:

[0034]

[0035] During wireless frequency synchronization, all nodes are kept stationary.

[0036] The master node up-converts the LFM reference signal and sends it to each slave node. The transmitted signal is as follows:

[0037]

[0038] Where f0 is the carrier frequency, The phase deviation introduced by the up-conversion.

[0039] In the case of a single path, suppose a certain slave node receives the following signal:

[0040]

[0041] Where Δt is the propagation delay from the autonomous node to the slave node, it can be further expressed as:

[0042]

[0043] d is the distance from the master node to the slave node, and v is the signal propagation speed. The slave node down-converts the received signal, and the down-converted signal is:

[0044]

[0045] Where f0′ is the carrier frequency calculated by the slave node based on its own crystal oscillator frequency. The phase deviation introduced by the down-conversion. Let the crystal frequency of the master node be F, and the crystal frequency of the slave node be F′, with the frequency deviation being:

[0046] ΔF=F′-F

[0047] Then f0 and f0′ have the following relationship:

[0048]

[0049] Substituting y(t) into the equation, we get:

[0050]

[0051] Let the master node send the LFM reference signal with a period of T0, where T0 satisfies T0>T and T0f0 is an integer, then we have:

[0052]

[0053] r(t+T0)=x(t-Δt+T0)=x(t-Δt)=r(t)

[0054]

[0055] Therefore, it can be remembered as:

[0056]

[0057] And there are:

[0058] y(t+T0,t c )=y(t,t c )

[0059] say This is the additional phase of y(t).

[0060] To compress y(t), taking the most common matched filtering as an example, a matched filter is constructed using the LFM reference signal to perform linear filtering on y(t), resulting in the following pulse-compressed signal:

[0061]

[0062] Let the sampling rate of the master node be F. s The sampling rate of this node is F. s ′, and there is:

[0063]

[0064] The discretized pulse compression signal is then:

[0065]

[0066] when:

[0067]

[0068] When (round() means rounding), the magnitude of z[n] reaches its maximum value, which is called the pulse compression peak. The phase at this point is:

[0069]

[0070] It is easy to observe that the position and phase of the highest point of pulse compression correspond one-to-one with the propagation delay Δt of the original LFM reference signal and the additional phase of the signal y(t) after downconversion to the slave node.

[0071] Let the slave node extract the signal with a period of T0′ (i.e., the master node signal transmission period calculated by the slave node based on its own crystal oscillator frequency) and perform pulse compression, and we have:

[0072] T0F=T0′F′

[0073] Since all nodes remain stationary, the distance d between the master node and the slave node remains constant, i.e., Δt remains constant. Therefore, the slave node can be determined based on the position n of the highest point of pulse compression. m and phase φ m The variation Δn during adjacent weeks and The frequency of its own crystal oscillator is adjusted to synchronize with the master node, thereby achieving time synchronization with the master node. The specific crystal oscillator frequency adjustment scheme is not the focus of this application; please refer to relevant literature.

[0074] This application primarily considers the impact of dynamic multipath effects caused by environmental changes on a wireless frequency synchronization method based on a custom protocol. Since the equivalent velocity of environmental changes is generally small, the resulting Doppler frequency offset is also small, and its impact on LFM signal pulse compression results is negligible. Therefore, the Doppler effect will not be considered in the modeling of the following problems.

[0075] Considering multipath effects, suppose that at a certain slave node receiver, there are N LFM signals with different amplitudes, different propagation delays, and different phases (reflection may introduce unknown phase changes), but with the same time width T, the same bandwidth B, the same modulation frequency K, and the same carrier frequency f0, which are interleaved and superimposed:

[0076]

[0077] Among them, A i Let Δt be the amplitude of the i-th LFM signal. i Let be the propagation delay of the i-th LFM signal. Let be the phase of the i-th LFM signal. The signal after down-conversion of the received signal by the slave node is:

[0078]

[0079] make:

[0080]

[0081] Then we have:

[0082]

[0083] The goal of the problem is to solve in real time at the slave node for the propagation delay and additional phase of the direct wave signal (corresponding to the position and phase of the highest point of pulse compression in the direct wave signal). Depending on the conditions, the LFM signal with the largest amplitude (i.e., the smallest attenuation coefficient) or the LFM signal with the smallest propagation delay can generally be considered the direct wave signal. Based on this, the goal of the problem can be further transformed into solving for the amplitude A of all N LFM signals. i Propagation delay Δt i and additional phase φ i .

[0084] There are two common techniques for pulse compression of LFM signals: matched filtering and least squares. However, both have their own drawbacks when applied to multipath identification for radio frequency synchronization. This article will first introduce how to apply these two methods to multipath identification for radio frequency synchronization, and then introduce a joint multipath identification algorithm that can circumvent the drawbacks of both methods.

[0085] When the T·B of the LFM reference signal s(t) is greater than 1 (meaning that TB is much greater than 1), performing a Fourier transform on it yields an approximate estimate of its spectrum:

[0086]

[0087] Similarly, we can obtain y N (t) Approximate estimation of the spectrum:

[0088]

[0089] Consider the transfer function of the matched filter as follows:

[0090]

[0091] Where k is a proportionality constant and t0 is the sampling time. Without loss of generality, let k = 1 and t0 = 0, then we have:

[0092]

[0093] Therefore, the spectrum of the signal after matched filtering (i.e., pulse compression) can be obtained as follows:

[0094]

[0095] For Z N (f) Performing an inverse Fourier transform yields the matched-filtered signal as follows:

[0096]

[0097] If we let:

[0098]

[0099] Then we have:

[0100]

[0101] Under certain conditions, when:

[0102] t=Δt i

[0103] z N The modulus of (t) reaches its maximum value. z can be detected at this slave node. N The N pulse compression maxima in (t) correspond one-to-one with the propagation delays of the original N LFM signals on the time axis. Furthermore, the following system of linear equations can be solved to obtain the amplitudes and additional phases of the original N LFM signals:

[0104]

[0105] in, Let N be the complex unknowns of the linear system of equations, and Δt i and z N (Δt i It can be obtained by detecting the maximum point of pulse compression.

[0106] As can be seen from the above, the key to the problem lies in accurately obtaining Δt by detecting the maximum point of pulse compression. i and z N (Δt i This requires that the matched-filtered LFM signal ideally have a very narrow main lobe and no side lobes. The ideal result would be the impulse function δ(t), but this is not the case in reality. In fact, the main lobe width of the matched-filtered LFM signal is... Worse still, it also has high range sidelobes. Sidelobes of matched-filtered LFM signals with larger amplitudes can significantly influence the main lobe of matched-filtered LFM signals with smaller amplitudes, making it impossible to correctly obtain Δt by detecting the pulse compression maxima.i and z N (Δt i Therefore, sidelobe suppression is required for the matched filtering results of the LFM signal. Two common methods are introduced below: frequency domain weighting and spectral correction.

[0107] The frequency domain weighting method involves cascading a sidelobe suppression filter with a certain tapered frequency response after the matched filter. The general form of a commonly used weighting function is:

[0108]

[0109] When L = 0.08 and m = 2, W(f) is a Hamming weighted function; when L = 0.33 and m = 2, W(f) is a 3:1 cone ratio weighted function; when L = 0 and m = 2, 3, 4, W(f) are cosine square, cosine cube, and cosine fourth power weighted functions, respectively.

[0110] Taking the weighted function when m=2 as an example, we can obtain z N The frequency domain weighted result of (t) is:

[0111]

[0112] It is easy to see that d N The modulus of (t) is still at t = Δt i It reaches a maximum value at a point where the phase of the maximum point remains unchanged.

[0113] Cascading a sidelobe suppression filter after a matched filter is essentially a mismatch correction. While reducing sidelobes and increasing the main-to-sidelobe ratio, it also causes main lobe broadening and signal-to-noise ratio loss. The performance of common weighting functions can be found in relevant literature.

[0114] Theoretically, the range sidelobes generated by matched filtering are closely related to spectral edge jumps and in-band ripples. Frequency domain weighting aims to smooth spectral edge jumps, but it cannot suppress Fresnel ripples within the spectral band. For LFM signals with a large time-bandwidth product (T·B), their amplitude spectrum has an approximately rectangular characteristic; while for LFM signals with a small time-bandwidth product, their amplitude spectrum typically has large Fresnel ripples, requiring spectral correction methods to further suppress sidelobes.

[0115] Let the actual spectrum of the LFM reference signal used for matched filtering be... Then y N The actual spectrum of (t) is:

[0116]

[0117] The actual transfer function of the matched filter is:

[0118]

[0119] To correct the amplitude spectrum of the LFM signal to a rectangular shape, a new filter needs to be designed, with the following transfer function:

[0120]

[0121] in, The spectrum is an ideal rectangular shape. Therefore, the corrected spectrum y can be obtained. N (t) Signal z after pulse compression N The actual spectrum of (t) is:

[0122]

[0123] Note that the spectrum of the pulse compression result after spectral correction is an ideal rectangular spectrum, while the spectrum of the original pulse compression result is only approximately an ideal rectangular spectrum, and Fresnel ripples still exist within the band.

[0124] Spectrum correction can remove Fresnel ripples within the LFM signal's spectral band, further suppressing sidelobes in the matched-filtered LFM signal. However, neither frequency-domain weighting nor spectrum correction can narrow the main lobe of the matched-filtered LFM signal; in fact, frequency-domain weighting may even widen the main lobe to some extent. Therefore, matched filtering has low propagation delay resolution when applied to wireless frequency synchronization multipath identification, only achieving this resolution when Δt... i Only when the pairwise differences are large can Δt be accurately obtained by detecting the maximum point of pulse compression. i and z N (Δt i This is the biggest drawback of the method.

[0125] For the least squares method, we directly consider the digital domain and represent the pulse compression process of the LFM signal as a matrix-vector multiplication:

[0126] p = Uh

[0127] Where p is the (l+m-1)×1 dimensional pulse compression result, h is the l×1 dimensional filter weight, and U is the (l+m-1)×1 dimensional matrix composed of the m-point sample shifts of the LFM reference signal s(t), i.e. (where l≥m):

[0128]

[0129] Let the ideal pulse compression result (extremely narrow main lobe, no side lobes) be:

[0130]

[0131] Where 1 is located in the l-th row. Let the sum of the squares of the differences between the ideal pulse compression result and the actual pulse compression result be:

[0132] E(h) = (p - Uh) T (p-Uh)

[0133] The goal is to find an h such that E(h) is minimized, therefore let:

[0134]

[0135] We can obtain:

[0136] h=(U T U) -1 U T p

[0137] It can be considered that h is the optimal pulse compression filter in the least squares sense (i.e., linear sense).

[0138] When using the filter obtained from the above equation to pulse compress the LFM reference signal, two relatively high side lobes will be generated on both sides of the main lobe. To solve this problem, a matrix can be used. Replacing matrix U, we obtain a new pulse compression filter:

[0139]

[0140] Among them, matrix This is obtained by setting the first and last few rows of the l-th row of matrix U to zero. Essentially, this is a form of "structural under-constraint" that better concentrates the main lobe energy by reducing the fitting dimension of the main lobe shape. This effectively reduces the values ​​of higher side lobes and enhances the filter's noise immunity, but it also slightly broadens the main lobe. With the matrix... Increasing the number of rows with zeros improves both the main lobe-to-side lobe ratio and noise immunity, while also increasing the main lobe width ratio. Furthermore, increasing the filter length *l* can further suppress sidelobes and enhance the filter's noise immunity.

[0141] Clearly, the filter can be used at the node end. Identify the propagation delay of N LFM signals; and due to the filter It will not affect the additional phase, therefore it can be used based on the LFM reference signal. The results are then used to further determine the amplitude and additional phase of the N LFM signals.

[0142] Use filters The main lobe width of the obtained LFM signal pulse compression result is less than that of the matched filter result. Furthermore, it exhibits a high main-to-sidelobe ratio, thus providing high propagation delay resolution. However, due to the filter... It is data-driven, directly constructed from sampling of an ideal LFM reference signal, therefore its adaptability to noise is poor. For N noisy LFM signals, although this method can accurately obtain Δt by detecting the pulse compression maxima, it is not ideal for this purpose. i However, the additional phase estimate obtained by further solving has a large error, which seriously affects the accuracy of wireless frequency synchronization.

[0143] Based on the above discussion, this application proposes a dynamic multipath real-time identification method for wireless frequency synchronization, such as... Figure 1 As shown, the dynamic multipath real-time identification method for wireless frequency synchronization uses a joint multipath identification algorithm based on least squares and matched filtering to decouple the work of identifying the propagation delay of N LFM signals from solving the amplitude and additional phase of the N LFM signals. Specifically, from the node end, a structurally underconstrained least squares method is first used to analyze the down-converted signal y. N (t) Perform pulse compression, detect the position of each pulse compression maxima on the time axis, and obtain the propagation delay of N LFM signals; based on this, further process the signal y N (t) Perform pulse compression with spectral correction matched filtering, then perform frequency domain weighting to suppress sidelobes, and then use the identified propagation delay to construct a system of linear equations to solve for the amplitude and additional phase of N LFM signals, and finally restore the propagation delay and additional phase of the direct wave signal.

[0144] Specifically, such as Figure 2 As shown, the dynamic multipath real-time identification method for wireless frequency synchronization includes the following steps:

[0145] S1: The master node transmits a radio frequency synchronization signal with a period of T0. In this application, the radio frequency synchronization signal is a linear frequency modulation (LFM) reference signal. It is understood that other signals, such as NLFM signals (Non-Linear Frequency Modulation), phase-coded signals, etc., can also be used as radio frequency synchronization signals for real-time identification of dynamic multipath.

[0146] S2: Receive the signal r from the node after it has traveled through N propagation paths and been superimposed. N (t), and for r N (t) performs down-conversion to obtain the baseband signal y N (t).

[0147] S3: Using a structurally underconstrained least-squares pulse compression filter to apply y N (t) Perform pulse compression to obtain the first compression result, and detect the N maxima in the first compression result to determine the propagation delay τ1…τ of the N propagation paths. N .

[0148] S4: First, use a spectrum-corrected matched filter to apply the filter to y. N (t) is pulse compressed, and the result is then frequency-domain weighted to suppress sidelobes, thus obtaining the second compression result.

[0149] S5: Based on the determined propagation delay τ1…τ N The complex values ​​at the corresponding positions are extracted from the second compression result as constant vectors, and a system of linear equations is constructed to solve for the amplitude and additional phase of N propagation paths.

[0150] S6: Using the path with the largest amplitude or the smallest delay as the direct wave signal, extract its propagation delay and additional phase, and adjust the slave node crystal oscillator frequency accordingly to achieve high-precision frequency synchronization with the master node.

[0151] The joint multipath identification algorithm based on least squares and matched filtering cleverly combines the advantages of matched filtering and least squares, avoiding the shortcomings of both, and has both high propagation delay resolution and good noise resistance.

[0152] In the embodiments of this application, the structurally underconstrained least-squares pulse compression filter is generated by an ideal pulse compression matrix U, and the matrix is ​​obtained by setting the elements of several adjacent rows before and after the l-th row corresponding to the main lobe of U to zero. by As filter coefficients.

[0153] In the embodiments of this application, frequency domain weighting employs any one of the following: Hamming window, 3:1 cone ratio window, cosine square window, cosine cubic window, or cosine fourth power window. Many other windows can also be used for frequency domain weighting, such as Hanning window, Blackman window, Taylor window, Kaiser window, Chebyshev window, etc. All window functions with symmetrical forms and smooth transitions at the boundaries (i.e., continuous differentiability) can be used for frequency domain weighting.

[0154] In the embodiments of this application, in step S6, the path with the largest amplitude or the smallest propagation delay is used as the direct wave signal.

[0155] In the embodiments of this application, the period T0 satisfies T0≥τ max , τ max To maximize the identifiable propagation delay, in order to avoid overlap of compression results between adjacent periodic pulses.

[0156] In the embodiments of this application, the cascading order of spectrum correction and frequency domain weighting is to perform spectrum correction first and then frequency domain weighting.

[0157] Within a synchronization period T0, the hardware must complete operations such as multipath identification and crystal oscillator frequency adjustment. Shortening the synchronization period T0 can better suppress temperature and time drift of the crystal oscillator and improve frequency synchronization performance. Therefore, a low-latency hardware logic design is needed to implement the joint multipath identification algorithm based on least squares and matched filtering proposed in this application.

[0158] Based on the algorithm described above, the overall framework for low-latency hardware logic design can be obtained as follows: Figure 3 As shown. Among them, h1[n] and h2[n] are the least squares pulse compression filter and the matched filter pulse compression filter that are designed and stored in the hardware in advance, respectively. V is the coefficient matrix of the linear equation system, q is the constant term vector of the linear equation system, and g is the unknown vector of the linear equation system (i.e., the amplitude and additional phase of each LFM signal).

[0159] In the aforementioned hardware framework, the two convolution operations have the highest computational complexity, with a time complexity of O(k). 2 (k is y) N Even using FFT and IFFT to replace time-domain convolution with frequency-domain multiplication, the time complexity is still O(klogk), and it also incurs additional hardware resource overhead. Considering that after the pulse compression result is suppressed by sidelobes, multipath signals with a large time delay difference from the direct wave signal will hardly affect the estimation of the propagation delay and additional phase of the direct wave signal, the convolution result of only a few points near the direct wave signal can be calculated, reducing the time complexity to O(k).

[0160] In the embodiments of this application, the position of the first path signal is detected by setting an amplitude threshold, and convolution operations are performed only on Q sampling points on both the left and right sides of that position, where Q is a preset integer. Specifically, the threshold can be set using y N [n] Detect the approximate location of the first path signal (Q sampling points on each side), and then calculate the convolution result only for a few points near this approximate location (the direct wave signal is generally the first path signal or adjacent to the first path signal). The optimized hardware framework is as follows: Figure 4 As shown.

[0161] The proposed real-time dynamic multipath identification method for wireless frequency synchronization addresses the impact of dynamic multipath effects caused by environmental changes on the accuracy and stability of custom-protocol-based wireless frequency synchronization methods. When F s When the frequency is 600MHz, T=10ns, and B=100MHz, under the condition of a signal-to-noise ratio of 25dB, in the face of dynamic multipath effects, the technical solution proposed in this paper achieves a propagation delay resolution of 5ns and an additional phase estimation error of less than 0.02° for the direct wave signal.

[0162] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the above embodiments do not limit the present invention in any way, and all technical solutions obtained by equivalent substitution or equivalent transformation fall within the protection scope of the present invention.

Claims

1. A dynamic multipath real-time identification method for wireless frequency synchronization, characterized in that, Includes the following steps: S1: The master node sends a wireless frequency synchronization signal in period T0; S2: Receive the signal r from the node after it has passed through the superposition of N propagation paths. N (t), and for r N (t) performs down-conversion to obtain the baseband signal y N (t); S3: Using a structurally underconstrained least-squares pulse compression filter to apply y N (t) Perform pulse compression to obtain the first compression result, and detect the N maxima in the first compression result to determine the propagation delay τ1…τ of the N propagation paths. N ; S4: First, use a spectrum-corrected matched filter to apply the filter to y. N (t) Perform pulse compression, and then perform frequency domain weighting on the result to suppress sidelobes, thereby obtaining the second compression result; S5: Based on the determined propagation delay τ1…τ N Extract the complex values ​​at the corresponding positions from the second compression result as constant vectors, construct a system of linear equations and solve for the amplitude and additional phase of N propagation paths; S6: Using the path with the largest amplitude or the smallest delay as the direct wave signal, extract its propagation delay and additional phase, and adjust the slave node crystal oscillator frequency accordingly to achieve high-precision frequency synchronization with the master node.

2. The dynamic multipath real-time identification method for wireless frequency synchronization according to claim 1, characterized in that, The structurally underconstrained least-squares pulse compression filter is generated from an ideal pulse compression matrix U, and the matrix is ​​obtained by setting several adjacent rows of elements corresponding to the l-th row of the main lobe in U to zero. ,by As filter coefficients.

3. The dynamic multipath real-time identification method for wireless frequency synchronization according to claim 1, characterized in that, The frequency domain weighting adopts any one of the following: Hamming window, 3:1 cone ratio window, cosine square window, cosine cubic window, or cosine fourth power window.

4. The dynamic multipath real-time identification method for wireless frequency synchronization according to claim 1, characterized in that, In step S6, the path with the largest amplitude or the smallest propagation delay is taken as the direct wave signal.

5. The dynamic multipath real-time identification method for wireless frequency synchronization according to claim 1, characterized in that, The position of the first path signal is detected by setting an amplitude threshold, and convolution operation is performed only on Q sampling points on the left and right sides of that position, where Q is a preset integer.

6. The dynamic multipath real-time identification method for wireless frequency synchronization according to claim 1, characterized in that, The period T0 satisfies T0≥τ max , τ max To maximize the identifiable propagation delay, in order to avoid overlap of compression results between adjacent periodic pulses.

7. The dynamic multipath real-time identification method for wireless frequency synchronization according to claim 1, characterized in that, The cascading order of the spectrum correction and frequency domain weighting is to perform spectrum correction first and then frequency domain weighting.

8. The dynamic multipath real-time identification method for wireless frequency synchronization according to claim 1, characterized in that, The latency resolution is 5ns.

9. The dynamic multipath real-time identification method for wireless frequency synchronization according to claim 1, characterized in that, The additional phase estimation error of the direct wave signal is less than 0.02°.

10. The dynamic multipath real-time identification method for wireless frequency synchronization according to claim 1, characterized in that, The wireless frequency synchronization signal is a linear frequency modulation (LFM) reference signal.

Images