A method and system for frequency-free time synchronization based on dynamic characteristics of atomic clocks

Abstract

The application discloses a frequency-time synchronization method and system based on atomic clock dynamics characteristics, and relates to high-precision time synchronization technology, which comprises the following steps: calculating macro time deviation according to a time stamp message; obtaining local environment parameters as input characteristics; measuring distortion time deviation through DDMTD; predicting micro frequency deviation based on PINN; constructing data loss with macro time deviation; constructing physical loss with atomic clock dynamics characteristics; weighting and summing as overall loss to update network weight online; fixing the amplification coefficient of DDMTD to dynamic amplification coefficient according to the predicted value to correct the real time deviation; and generating calibration instructions to compensate the local digital time stamp accumulator in a closed loop. The application can realize high-precision time synchronization without physical frequency synchronization, and is immune to frequency link injection fraud threats.

Description

Technical Field

[0001] This application relates to the field of high-precision time synchronization technology, and in particular to a frequency-free time synchronization method and system based on the dynamic characteristics of atomic clocks. Background Technology

[0002] Time synchronization aims to ensure that heterogeneous devices within a distributed system or network have a unified time reference through specific technical means. With the rapid development of fields such as power, communications, and aviation, the demand for time synchronization accuracy has increased to the nanosecond or even picosecond level.

[0003] In traditional high-precision time synchronization network architectures, the time synchronization process requires deep coupling of frequency synchronization and phase synchronization: frequency synchronization ensures synchronization by locking the operating rates of the master and slave node clock sources, while phase synchronization aligns the times of the master and slave node clock sources. However, this tightly coupled architecture faces certain physical limitations and security challenges during application deployment: on the one hand, the system mandates that all intermediate nodes in the link have dedicated hardware supporting frequency pass-through, significantly increasing network deployment costs; on the other hand, the physical layer frequency link is significantly vulnerable, highly susceptible to external electromagnetic interference or malicious frequency injection attacks. Once the frequency link loses its lock due to a momentary interruption or attack, the time synchronization accuracy of the entire system will experience a non-linear collapse.

[0004] Digital Dual Mixer Time Difference (DDMTD) measurement technology exhibits extremely high phase difference measurement resolution in high-precision time synchronization systems due to its equivalent time sampling principle. However, when the atomic clocks of the master and slave nodes are in an asynchronous free-running state, the real-time fluctuating frequency deviation, influenced by random factors such as the drift of atomic clock physical characteristics and environmental temperature drift, causes severe nonlinear dynamic changes in the period of the beat frequency signal generated by DDMTD mixing. This nonlinear distortion directly triggers the failure of conventional phase measurement logic, becoming the core technical bottleneck for achieving high-precision time synchronization in environments without frequency synchronization. Summary of the Invention

[0005] This application provides a frequency-free time synchronization method and system based on atomic clock dynamics characteristics. By introducing a Physics-Informed Neural Network (PINN), the physical prior constraints of atomic clock dynamics characteristics are deeply integrated with the data constraints provided by the application layer's time message calculation. This enables high-precision dynamic prediction of micro-frequency deviations, thereby supporting DDMTD to complete dynamic time difference measurement without establishing a physical frequency synchronization link. This also immunizes against injection spoofing threats targeting physical frequency links from the bottom layer of the time synchronization network architecture.

[0006] This application provides a frequency-free time synchronization method based on the dynamic characteristics of atomic clocks, applied to a time synchronization system consisting of a master node and slave nodes. The master node is equipped with a master atomic clock as a time reference, and the slave nodes are equipped with independently operating slave atomic clocks. The master node and the slave nodes exchange information only through timestamp messages and do not establish a physical frequency synchronization link. The method includes: The master node and the slave node exchange timestamp messages, and the slave node calculates the macroscopic time deviation within a synchronization period T based on the exchanged timestamp messages. ; Obtain the local environment parameters of the slave node, and then calculate the macroscopic time deviation. The local environment parameters are used as input features; The distortion time deviation between the local clock signal of the slave node and the recovered master node clock signal is measured using the digital double mixer time difference (DDMTD) method. ; The micro-frequency deviation of the atomic clock is analyzed based on the Physical Information Neural Network (PINN). To make a prediction, wherein the macroscopic time deviation is used. The data loss is constructed using data labels, and the physical loss is constructed using atomic clock dynamics characteristics. The weighted sum of the data loss and the physical loss is used as the overall loss. The network weights are updated online through backpropagation of the overall loss, and the predicted value of the micro-frequency deviation is output. ; Based on the predicted value of the micro-frequency deviation Reconstruct the fixed amplification factor N of DDMTD into a dynamic amplification factor. And using the dynamic amplification factor Regarding the distortion time deviation Correction is performed to obtain the true time deviation. ; Based on the actual time deviation Generate calibration instructions to perform closed-loop compensation on the local digital timestamp accumulator, achieving high-precision time synchronization without physical frequency synchronization.

[0007] This application provides a frequency-free time synchronization system based on the dynamic characteristics of atomic clocks, including a master node and a slave node; The master node includes a master atomic clock and a timestamp message sending and receiving module. The master atomic clock serves as the time reference for the entire system, and the timestamp message sending and receiving module of the master node is used to interact with the slave nodes using timestamp messages. The slave node includes a slave atomic clock, a timestamp message transmission and reception module, a time deviation measurement module, a frequency deviation prediction module, a dynamic coefficient reconstruction and compensation module, and a clock update module, wherein: The atomic clock operates independently and freely, and no physical frequency synchronization link is established between the slave node and the master node; The time deviation measurement module extracts the distortion time deviation between the master node and the slave node based on the DDMTD principle. ; The frequency deviation prediction module, based on the PINN principle, uses macroscopic time deviation and local environmental parameters as input features to predict the microscopic frequency deviation of the atomic clock. ; The dynamic coefficient reconstruction and compensation module predicts the value based on the micro-frequency deviation. Reconstruct the fixed amplification factor N of DDMTD into a dynamic amplification factor. And using the dynamic amplification factor Regarding the distortion time deviation Correction is performed to obtain the true time deviation. ; The clock update module updates according to the actual time deviation. Generate calibration instructions to perform closed-loop compensation on the local digital timestamp accumulator, achieving high-precision time synchronization without physical frequency synchronization.

[0008] The synchronization method described in this application eliminates the need to establish a physical frequency synchronization link, thus immunizing the time synchronization network architecture from the ground up against injection spoofing threats targeting the physical frequency link. In traditional solutions, the physical layer frequency link is vulnerable to external electromagnetic interference or malicious frequency injection attacks, and loss of lock-on can lead to a nonlinear collapse in time synchronization accuracy. This invention uses PINN to predict micro-frequency deviations and reconstruct the dynamic amplification factor, achieving high-precision time synchronization even in environments without frequency synchronization, significantly improving the security of high-precision time synchronization networks in complex adversarial environments.

[0009] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0010] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a schematic diagram of a time synchronization system corresponding to the frequency-free time synchronization method in this application embodiment; Figure 2 This is a schematic diagram of the time synchronization process of the frequency-free time synchronization method in this application embodiment. Detailed Implementation

[0011] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0012] This application provides a frequency-free time synchronization method based on the dynamic characteristics of atomic clocks, applicable to a time synchronization system consisting of a master node and slave nodes. For example... Figure 1 As shown, the time synchronization system consists of two parts: a master node and a slave node. The two nodes only exchange timestamp message information and do not transmit frequency information at the physical layer.

[0013] The master node is equipped with a high-precision atomic clock A (master atomic clock) as the time reference for the entire system, and is responsible for exchanging timestamp messages with the slave nodes through the timestamp message transmission and reception module. The slave node includes an independently operating high-precision atomic clock B (slave atomic clock), a timestamp message transmission and reception module, a time deviation measurement module, a frequency deviation prediction module, a dynamic coefficient reconstruction and compensation module, and a clock update module.

[0014] The master node and the slave node exchange information only through timestamp messages and do not establish a physical frequency synchronization link. The timestamp message transmission and reception module is responsible for encoding, transmitting, decoding, and recovering time information and signals. It provides the time deviation measurement module with a clock representing the current true oscillation state of the master node. Simultaneously, it provides the frequency deviation prediction module with macroscopic time deviation calculated using timestamp information. .

[0015] The time deviation measurement module extracts the original time deviation between master and slave nodes based on the DDMTD principle.

[0016] The frequency deviation prediction module is based on the PINN principle to achieve high-precision dynamic prediction of micro frequency deviations.

[0017] The dynamic coefficient reconstruction and compensation module is responsible for solving the DDMTD measurement failure problem in the time deviation measurement module. It reconstructs the traditional fixed amplification coefficient into a dynamic amplification coefficient, eliminating the nonlinear signal distortion caused by frequency drift.

[0018] The clock update module is responsible for converting the compensated time deviation sequence into absolute calibration instructions for the system timestamp, directly performing closed-loop compensation on the local digital timestamp accumulator driven by atomic clock B (the slave atomic clock). Based on this... Figure 2 As shown, the method in this application embodiment includes: In step S1, the master node and the slave node exchange timestamp messages, and the slave node calculates the macroscopic time deviation within a synchronization period T based on the exchanged timestamp messages. In some embodiments, an auxiliary sampling clock with a small frequency difference from the slave atomic clock is generated via an internal phase-locked loop. The auxiliary sampling clock The frequency satisfies: in, Let N be the frequency of the atomic clock, and N be the theoretical amplification factor of DDMTD.

[0019] In step S2, the local environment parameters of the slave node are obtained, and the macroscopic time deviation is... The local environment parameters are used as input features.

[0020] In step S3, the distortion time deviation between the local clock signal of the slave node and the recovered master node clock signal is measured using the digital double-mixer time difference (DDMTD) method. Specifically, when using a D flip-flop... For sampling pulses, the clock from the master node When performing equivalent sampling, due to the lack of frequency synchronization between the master and slave atomic clocks, there is a relative frequency deviation in the master node clock. This causes the original beat frequency signal at the mixer output to exhibit nonlinear dynamic changes. An internal high-speed counter captures the edges of the two beat frequency signals to output a time deviation sequence containing nonlinear measurement distortion characteristics to subsequent modules. Because there is no frequency synchronization between the master atomic clock and the slave atomic clock, the master node clock signal has a relative frequency deviation, causing the period of the mixing beat signal to change nonlinearly and dynamically.

[0021] In step S4, the micro-frequency deviation of the atomic clock is analyzed based on the Physical Information Neural Network (PINN). To make a prediction, wherein the macroscopic time deviation is used. The data loss is constructed using data labels, and the physical loss is constructed using atomic clock dynamics characteristics. The weighted sum of the data loss and the physical loss is used as the overall loss. The network weights are updated online through backpropagation of the overall loss, and the predicted value of the micro-frequency deviation is output. .

[0022] In step S4, based on the predicted value of the micro frequency deviation... Reconstruct the fixed amplification factor N of DDMTD into a dynamic amplification factor. And using the dynamic amplification factor Regarding the distortion time deviation Correction is performed to obtain the true time deviation. .

[0023] In a specific example, based on the actual dynamic amplification factor in the aforementioned embodiments, for: However, the backend measurement logic still assumes the signal has been amplified by N times, thus generating a measurement ratio error. : Where M is the actual dynamic amplification factor and N is the theoretical amplification factor. This error means that the time deviation of each measurement by the system will be smaller than the true value. Without dynamic coefficient compensation, this measurement error will accumulate and diverge over time.

[0024] In step S5, based on the actual time deviation Generate calibration instructions to perform closed-loop compensation on the local digital timestamp accumulator, achieving high-precision time synchronization without physical frequency synchronization.

[0025] The method described in this application eliminates the need to establish a physical frequency synchronization link, thus immunizing the time synchronization network architecture from the ground up against injection spoofing threats targeting the physical frequency link. In traditional solutions, the physical layer frequency link is vulnerable to external electromagnetic interference or malicious frequency injection attacks, and loss of lock-on can lead to a nonlinear collapse in time synchronization accuracy. This invention uses PINN to predict micro-frequency deviations and reconstruct the dynamic amplification factor, achieving high-precision time synchronization even in environments without frequency synchronization, significantly improving the security of high-precision time synchronization networks in complex adversarial environments.

[0026] The method described in this application reduces network deployment costs. Traditional solutions require all intermediate nodes in the link to have dedicated hardware that supports frequency pass-through. This invention only requires the master and slave nodes to exchange timestamp message information, without requiring intermediate nodes to support frequency pass-through, which greatly reduces network deployment costs and improves deployment flexibility.

[0027] The frequency deviation prediction module uses the PINN principle to predict micro-frequency deviations. High-precision dynamic prediction. The input features of this module include local temperature and humidity sensor data, as well as macroscopic time deviations calculated by the timestamp message transceiver module within a synchronization period T. Since macroscopic time deviation is equivalent to the integral of microscopic frequency deviation over time, it is used as the data label for the PINN model to construct the data loss. In some embodiments, the local environmental parameters include temperature sensor data and humidity sensor data local to the slave node; The macroscopic time deviation Equivalent to the integral of the micro-frequency deviation over the time axis, the data loss Construct it as follows: in, The frequency of the atomic clock, For the PINN at time The predicted value of the micro-frequency deviation is output.

[0028] Due to the lack of physical layer frequency synchronization, time synchronization messages are susceptible to network latency jitter. If macroscopic time deviations are used to infer microscopic frequency drifts, the neural network model is prone to making sudden jump predictions that do not conform to objective physical laws, directly leading to the failure of DDMTD dynamic coefficient compensation. Therefore, the dynamic characteristics of atomic clocks are forcibly embedded into the training logic of the neural network. In some embodiments, the physical loss... It includes an aging trend constraint and a random walk constraint. The aging trend constraint is used to constrain the net increment of the frequency deviation within the synchronization period T to conform to the physical aging trend of the atomic clock. The random walk constraint term is used to constrain the average value of the square of the short-term increment of the frequency deviation within the synchronization period T to approximate the frequency modulation random walk variance derived from the Allen variance of the atomic clock. The constraint that the net increment of frequency deviation within the time window T conforms to the physical aging trend of atomic clocks prevents unreasonable overall tilting or jumps in the predicted frequency curve. On the other hand, the constraint that the average value of the square of the short-term increment of frequency deviation within the time window T approximates the variance of the random walk limits unreasonable spikes in the predicted curve within a very short period of time.

[0029] In some embodiments, the physical loss for: Wherein, D is the intrinsic aging rate calibrated when the atomic clock leaves the factory. The sampling step size of PINN is... The frequency of the atomic clock.

[0030] In some embodiments, the overall loss L of the PINN model is constructed as follows: in, and To balance the data loss With the physical loss The weight hyperparameters are determined by a custom method or an adaptive adjustment method; the PINN updates the network weights online through backpropagation based on the overall loss L.

[0031] The frequency deviation prediction module updates the network weights online through backpropagation of the loss function. After training, it can output high-precision micro-frequency deviation prediction values ​​in real time. .

[0032] The dynamic coefficient reconstruction and compensation module is responsible for resolving the DDMTD measurement failure problem in the time deviation measurement module. It reconstructs the traditional fixed amplification factor into a dynamic amplification factor, eliminating the nonlinear signal distortion caused by frequency drift. First, based on the input from the frequency deviation prediction module 24... The current dynamic amplification factor is calculated in real time. In some embodiments, the dynamic amplification factor Refactor as follows: in, The frequency of the master atomic clock, The frequency of the auxiliary sampling clock for DDMTD, This is the predicted value of the micro-frequency deviation.

[0033] Then, the distortion time interval input into the time deviation measurement module is... Divide by dynamic coefficient This accurately reconstructs the time deviation sequence of the actual changes. In some embodiments, the actual time deviation... Calculate as follows: in, This refers to the distortion time deviation. is the dynamic amplification factor.

[0034] In some embodiments, the closed-loop compensation is: adjusting the actual time deviation. The absolute calibration command, converted into a system timestamp, directly compensates the local digital timestamp accumulator without changing the physical voltage control voltage of the slave atomic clock, thereby achieving high-precision time synchronization without physical frequency synchronization.

[0035] This application also proposes implementation examples of a frequency-free time synchronization method based on the dynamic characteristics of atomic clocks: Taking the exchange of timestamp messages between master and slave nodes based on the PTP protocol as an example, without Sync_E enabled for frequency synchronization, the master and slave atomic clocks are in an asynchronous, free-running state. Assume that due to differences in the deployment environment, the relative frequency drift between the atomic clocks is +0.8Hz.

[0036] Assuming the time deviation measurement module has two input signals and Theoretically, both are 10MHz, with a theoretical amplification factor N of 16384. The auxiliary sampling clock for the slave node is generated by the FPGA. : The beat frequency signal output from the DDMTD mixer at node DDMTD is: Because the master node clock frequency has drifted by +0.8Hz, i.e., the actual... This causes a non-linear stretching of the beat frequency signal output from the DDMTD mixer: At this point, the actual dynamic amplification factor M is: However, the backend measurement logic will still consider the signal to have been amplified by N times, which will result in a measurement ratio error. : That is, the time deviation result of each measurement by the system will be reduced by 0.13% compared with the true value.

[0037] The timestamp message transmission module 22 calculates the macroscopic time deviation every 1 second using the PTP timestamp and uses it as input samples for training the PINN model. The frequency deviation prediction module 24 uses the macroscopic time deviation and temperature parameters as model inputs and adjusts the data loss weights accordingly. Physical loss weights And parameters such as backpropagation are used to complete model performance training. Assume the PINN model outputs frequency bias prediction values. .

[0038] Upon receiving the predicted frequency deviation value, the dynamic coefficient reconstruction and compensation module immediately reconstructs the originally fixed amplification factor into a dynamic amplification factor in real time, based on the reconstruction. formula: Based on this coefficient, the nonlinear signal distortion caused by frequency drift is removed, restoring the true time deviation. Then, the clock update module performs closed-loop compensation on the local digital timestamp accumulator to achieve high-precision time synchronization without physical frequency synchronization.

[0039] The high-precision time synchronization process without frequency synchronization based on the dynamic characteristics of atomic clocks is as follows: First, the master and slave nodes exchange timestamp messages. Then, the macroscopic time deviation and local environmental parameters are calculated to construct the input features for the frequency deviation prediction model. Simultaneously, the distorted time deviation between the slave node's local clock signal and the recovered master node clock signal is measured. After the frequency deviation prediction model has been trained, the dynamic amplification factor of the time deviation measurement module is reconstructed based on the predicted frequency deviation values, and the true time interval is restored. Finally, local system timestamp compensation and updates are performed.

[0040] It should be noted that the PTP protocol in this embodiment is only an example. Those skilled in the art will understand that the present invention is also applicable to NTP protocol and dual / unidirectional time synchronization systems.

[0041] This application deeply integrates the physical prior constraints of atomic clock dynamics characteristics with the data constraints provided by the application layer's time message calculation through a physical information neural network. On the one hand, it constructs data loss using macroscopic time deviation as data label, and on the other hand, it forces predictions to conform to the physical laws of atomic clocks through aging trend constraints and random walk constraints. This effectively suppresses sudden jump predictions that do not conform to physical objective laws caused by network latency jitter, and achieves high-precision dynamic prediction of microscopic frequency deviations, thereby supporting DDMTD to complete dynamic time difference measurement in a frequency-synchronous environment.

[0042] This application also proposes a frequency-free time synchronization system based on the dynamic characteristics of atomic clocks, including a master node and a slave node; The master node includes a master atomic clock and a timestamp message sending and receiving module. The master atomic clock serves as the time reference for the entire system, and the timestamp message sending and receiving module of the master node is used to interact with the slave nodes using timestamp messages. The slave node includes a slave atomic clock, a timestamp message transmission and reception module, a time deviation measurement module, a frequency deviation prediction module, a dynamic coefficient reconstruction and compensation module, and a clock update module, wherein: The atomic clock operates independently and freely, and no physical frequency synchronization link is established between the slave node and the master node; The time deviation measurement module extracts the distortion time deviation between the master node and the slave node based on the DDMTD principle. ; The frequency deviation prediction module, based on the PINN principle, uses macroscopic time deviation and local environmental parameters as input features to predict the microscopic frequency deviation of the atomic clock. ; The dynamic coefficient reconstruction and compensation module predicts the value based on the micro-frequency deviation. Reconstruct the fixed amplification factor N of DDMTD into a dynamic amplification factor. And using the dynamic amplification factor Regarding the distortion time deviation Correction is performed to obtain the true time deviation. ; The clock update module updates according to the actual time deviation. Generate calibration instructions to perform closed-loop compensation on the local digital timestamp accumulator, achieving high-precision time synchronization without physical frequency synchronization.

[0043] It should be noted that, in the embodiments of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0044] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0045] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0046] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims. All of these forms are within the protection scope of this application.

Claims

1. A frequency-free time synchronization method based on the dynamic characteristics of atomic clocks, applied to a time synchronization system consisting of a master node and slave nodes, wherein the master node is equipped with a master atomic clock as a time reference, and the slave nodes are equipped with independently operating slave atomic clocks, and the master node and the slave nodes exchange information only through timestamp messages without establishing a physical frequency synchronization link, characterized in that, The method includes: The master node and the slave node exchange timestamp messages, and the slave node calculates the macroscopic time deviation within a synchronization period T based on the exchanged timestamp messages. ; Obtain the local environment parameters of the slave node, and then calculate the macroscopic time deviation. The local environment parameters are used as input features; The distortion time deviation between the local clock signal of the slave node and the recovered master node clock signal is measured using the digital double mixer time difference (DDMTD) method. ; The micro-frequency deviation of the atomic clock is analyzed based on the Physical Information Neural Network (PINN). To make a prediction, wherein the macroscopic time deviation is used. The data loss is constructed using data labels, and the physical loss is constructed using atomic clock dynamics characteristics. The weighted sum of the data loss and the physical loss is used as the overall loss. The network weights are updated online through backpropagation of the overall loss, and the predicted value of the micro-frequency deviation is output. ; Based on the predicted value of the micro-frequency deviation Reconstruct the fixed amplification factor N of DDMTD into a dynamic amplification factor. And using the dynamic amplification factor Regarding the distortion time deviation Correction is performed to obtain the true time deviation. ; Based on the actual time deviation Generate calibration instructions to perform closed-loop compensation on the local digital timestamp accumulator, achieving high-precision time synchronization without physical frequency synchronization.

2. The frequency-free time synchronization method based on atomic clock dynamics as described in claim 1, characterized in that, An auxiliary sampling clock with a slight frequency difference from the slave atomic clock is generated through an internal phase-locked loop. The auxiliary sampling clock The frequency satisfies: in, Let N be the frequency of the atomic clock, and N be the theoretical amplification factor of DDMTD.

3. The frequency-free time synchronization method based on atomic clock dynamics as described in claim 2, characterized in that, The distortion time deviation between the local clock signal of the slave node and the recovered master node clock signal, measured using the digital double-mixer time difference (DDMTD) method, includes: With the auxiliary sampling clock The sampling pulse is obtained by equivalent sampling of the recovered master node clock signal through a D flip-flop to generate a mixed beat frequency signal; The edges of the two mixed beat frequency signals are captured by a high-speed counter, and a time difference sequence containing nonlinear measurement distortion characteristics is output as the distortion time deviation. ; Since there is no frequency synchronization between the master atomic clock and the slave atomic clock, the master node clock signal has a relative frequency deviation, causing the period of the mixing beat signal to change nonlinearly and dynamically.

4. The frequency-free time synchronization method based on atomic clock dynamics as described in claim 1, characterized in that, The local environmental parameters include temperature sensor data and humidity sensor data from the slave node. The macroscopic time deviation Equivalent to the integral of the micro-frequency deviation over the time axis, the data loss Construct it as follows: in, The frequency of the atomic clock, For the PINN at time The predicted value of the micro-frequency deviation is output.

5. The frequency-free time synchronization method based on atomic clock dynamics as described in claim 1, characterized in that, The physical loss It includes an aging trend constraint and a random walk constraint. The aging trend constraint is used to constrain the net increment of the frequency deviation within the synchronization period T to conform to the physical aging trend of the atomic clock. The random walk constraint term is used to constrain the average value of the square of the short-term increment of the frequency deviation within the synchronization period T to approximate the frequency modulation random walk variance derived from the Allen variance of the atomic clock. The physical loss for: Wherein, D is the intrinsic aging rate calibrated when the atomic clock leaves the factory. The sampling step size of PINN is... The frequency of the atomic clock.

6. The frequency-free time synchronization method based on atomic clock dynamics as described in claim 1, characterized in that, The total loss L is constructed as follows: in, and To balance the data loss With the physical loss The weight hyperparameters are determined by a custom method or an adaptive adjustment method; the PINN updates the network weights online through backpropagation based on the overall loss L.

7. The frequency-free time synchronization method based on atomic clock dynamics as described in claim 1, characterized in that, The dynamic amplification factor Refactor as follows: in, The frequency of the master atomic clock, The frequency of the auxiliary sampling clock for DDMTD, This is the predicted value of the micro-frequency deviation.

8. The frequency-free time synchronization method based on atomic clock dynamics as described in claim 1, characterized in that, The actual time deviation Calculate as follows: in, This refers to the distortion time deviation. is the dynamic amplification factor.

9. The frequency-free time synchronization method based on atomic clock dynamics as described in claim 1, characterized in that, The closed-loop compensation is: to adjust the actual time deviation... The absolute calibration command, converted into a system timestamp, directly compensates the local digital timestamp accumulator without changing the physical voltage control voltage of the atomic clock.

10. A frequency-free time synchronization system based on the dynamic characteristics of atomic clocks, characterized in that, Includes master nodes and slave nodes; The master node includes a master atomic clock and a timestamp message sending and receiving module. The master atomic clock serves as the time reference for the entire system, and the timestamp message sending and receiving module of the master node is used to interact with the slave nodes using timestamp messages. The slave node includes a slave atomic clock, a timestamp message transmission and reception module, a time deviation measurement module, a frequency deviation prediction module, a dynamic coefficient reconstruction and compensation module, and a clock update module, wherein: The atomic clock operates independently and freely, and no physical frequency synchronization link is established between the slave node and the master node; The time deviation measurement module extracts the distortion time deviation between the master node and the slave node based on the DDMTD principle. ; The frequency deviation prediction module, based on the PINN principle, uses macroscopic time deviation and local environmental parameters as input features to predict the microscopic frequency deviation of the atomic clock. ; The dynamic coefficient reconstruction and compensation module predicts the value based on the micro-frequency deviation. Reconstruct the fixed amplification factor N of DDMTD into a dynamic amplification factor. And using the dynamic amplification factor Regarding the distortion time deviation Correction is performed to obtain the true time deviation. ; The clock update module updates according to the actual time deviation. Generate calibration instructions to perform closed-loop compensation on the local digital timestamp accumulator, achieving high-precision time synchronization without physical frequency synchronization.

Images