Time synchronization method suitable for satellite communication system

Abstract

The invention discloses a time synchronization method suitable for a satellite communication system, and the method comprises the steps: analyzing a forward broadcast signal, and obtaining the reference time of a network clock; constructing a geometric prediction handle based on satellite orbit data, and predicting one-way propagation time delay between a terminal and a satellite by using the geometric prediction handle; estimating the processing delay of the terminal according to the symbol rate and the equipment working condition of the terminal and the delay calibration data; fusing the network clock reference time with the processing delay estimate to synthesize a terminal local time; and the sending time of the return link is calculated based on the local time and the propagation delay prediction. According to the method, the synchronization precision can be improved to a microsecond level, the convergence speed in a dynamic scene is accelerated, adaptive compression of the guard interval is supported, and the stability and spectrum efficiency of a satellite communication system are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of satellite communication, in particular to a time synchronization method suitable for a satellite communication system. BACKGROUND

[0002] With the surge of global demand for broadband communication, high-throughput satellite communication systems have become the core direction of industry development, with a global market size expected to reach 18-22 billion US dollars in 2030, with a compound annual growth rate of more than 15%. High-throughput satellite systems have realized a leap from tens of Gbps to more than 1 Tbps in single-satellite capacity by adopting technologies such as multi-point beam and frequency reuse. To match this trend, user terminals are also developing towards miniaturization and phased array. In a complex network composed of hundreds of fine management and fast switching point beams, in order to ensure the accurate time slot alignment of return link signals at the satellite receiving end, avoid interference between users and maximize the spectral efficiency of the system, it is of great significance and value to establish a high-precision terminal time synchronization mechanism to improve the synchronization accuracy from the traditional millisecond level to the microsecond level.

[0003] Currently, the time synchronization method of satellite communication systems is usually based on the framework of timing-ranging-compensation. Specifically, the master station obtains a unified network time reference (such as Network Clock Reference, NCR) from high-precision clock sources such as GPS, and broadcasts it to all terminals under its jurisdiction through the forward link. After receiving the time reference, the terminal estimates the one-way propagation delay of the signal based on its geographical location and the satellite's orbital ephemeris data for geometric calculation. On the other hand, the terminal will compensate for its internal processing delay according to the symbol rate used by the current service and refer to the pre-stored calibration data. The terminal integrates the network time, propagation delay and processing delay to calculate the accurate sending time of the return link burst. To cope with the changes in propagation delay, the system usually reserves a fixed guard interval (GP) in the time slot design, the size of which is set according to the maximum delay difference within the beam coverage, for example, a point beam covering 200 kilometers has a maximum delay difference of about 667 microseconds.

[0004] However, the prior art has some problems in dealing with modern high-throughput satellite systems with high dynamics and high efficiency, resulting in slow system synchronization convergence, serious waste of guard interval resources, and limitation of overall performance improvement. In dynamic scenarios such as beam rapid switching, the group delay of the terminal processing link will have a stepwise nonlinear jump due to the instantaneous change of the symbol rate, and the traditional static calibration table cannot accurately model this transient effect, resulting in a relatively obvious transient error of the synchronization loop after switching, which needs a long time to re-converge. For low-orbit satellite scenarios with high-speed relative motion, the existing method fails to fully consider the real-time path changes caused by the Doppler effect within the frame, forcing the system to use an excessively conservative fixed guard interval to absorb the worst-case delay drift, resulting in persistent waste of spectrum resources. Further, even if the above delay estimation problem is improved, the overall synchronization accuracy will be limited by the sampling clock resolution of the terminal digital system, and the conventional time synthesis method cannot perform fine tuning of sub-sampling periods, setting a lower limit for synchronization error that is difficult to overcome, hindering the breakthrough of synchronization accuracy to a higher level. SUMMARY

[0005] The purpose of the application is to provide a time synchronization method suitable for a satellite communication system to solve the above problems existing in the prior art.

[0006] Technical scheme, according to one aspect of the application, a time synchronization method suitable for a satellite communication system, comprising:

[0007] Analyzing the forward broadcast signal to obtain a network clock reference time;

[0008] Based on satellite orbit data, a geometric prediction handle is constructed;

[0009] Using the geometric prediction handle, the one-way propagation delay between the terminal and the satellite is predicted;

[0010] According to the symbol rate and equipment working condition of the terminal, and referring to the delay calibration data, the processing delay of the terminal is estimated;

[0011] Fusing the network clock reference time and the processing delay estimation, the terminal local time is synthesized;

[0012] Based on the terminal local time and the one-way propagation delay prediction, and according to the non-reciprocal bias, the sending time of the return link is calculated.

[0013] According to one aspect of the application, a possible implementation mode, fusing the network clock reference time and the processing delay estimation, synthesizing the terminal local time, comprising:

[0014] Integrating the network clock reference time and the processing delay estimation, a baseline local time is generated;

[0015] determining a fractional delay control quantity for the baseline local time according to the one-way propagation time delay prediction;

[0016] generating a fractional delay adjustment quantity by processing the fractional delay control quantity with a variable fractional delay filter;

[0017] correcting the baseline local time with the fractional delay adjustment quantity to synthesize the terminal local time with sub-sample level precision.

[0018] According to a possible implementation manner of one aspect of the present application, synthesizing the terminal local time further includes:

[0019] determining whether a frequency-dependent error introduced by the variable fractional delay filter is greater than a threshold value;

[0020] if yes, retrieving a group delay compensation quantity from the group delay calibration data by referring to the fractional delay control quantity and spectral characteristics of the signal;

[0021] applying the fractional delay adjustment quantity and the group delay compensation quantity to the baseline local time to synthesize the terminal local time.

[0022] According to a possible implementation manner of one aspect of the present application, the variable fractional delay filter processing and the retrieval process of the group delay compensation quantity are configured in a processing module that is bound to each other;

[0023] when real-time performance of a state of the processing module or an internal parameter of the processing module does not satisfy a preset condition, the processing module is disabled as a whole, and the synthesized terminal local time is rolled back to the baseline local time.

[0024] According to a possible implementation manner of one aspect of the present application, estimating a processing delay of the terminal according to a symbol rate of the terminal and a device working condition and referring to delay calibration data includes:

[0025] constructing a feedforward prediction value of the processing delay for a switching event of the symbol rate of the terminal;

[0026] obtaining a delay deviation measured by a probe signal;

[0027] performing feedback correction on the feedforward prediction value according to the measured delay deviation to determine the processing delay estimate.

[0028] According to a possible implementation manner of one aspect of the present application, obtaining the delay deviation measured by the probe signal includes:

[0029] identifying a beam switching protection window according to a beam switching time table, and instructing the terminal to transmit a dedicated probe burst during the window;

[0030] performing relevant estimation on the received probe burst by the network side to obtain an initial measured delay deviation and a corresponding measurement confidence.

[0031] The initial measured delay deviation is time-stamped aligned by using the frame sequence number, cross-frame measurement error is eliminated, and an aligned measured delay deviation for feedback correction is generated.

[0032] According to one aspect of the present application, a possible implementation manner is that, according to the measured delay deviation, feedback correction is performed on the feedforward prediction value, and the processing delay estimation is determined, including:

[0033] The update weight is dynamically determined based on whether the current state is in a beam switching protection window and a measurement confidence level;

[0034] The first weight value is configured to the update weight when and only when the current state is in the window and the measurement confidence level is not lower than a preset threshold, and the second weight value is configured to the update weight in other conditions;

[0035] The determined update weight is used to perform weighted fusion on the feedforward prediction value and the aligned measured delay deviation, and the processing delay estimation is determined.

[0036] According to one aspect of the present application, a possible implementation manner is that, the one-way propagation delay between the terminal and the satellite is predicted by using a geometric prediction handle, including:

[0037] The geometric propagation delay based on a geometric distance is calculated by using the geometric prediction handle;

[0038] The relative velocity between the terminal and the satellite is obtained by using the geometric prediction handle, a Doppler frequency shift is calculated according to the relative velocity, and the Doppler frequency shift is mapped to a time delay correction term in a time domain range;

[0039] The time delay correction term is superimposed on the geometric propagation delay, and a one-way propagation delay prediction reflecting real-time dynamics is generated.

[0040] According to one aspect of the present application, a possible implementation manner is that, the method further includes:

[0041] The Doppler frequency shift or a slant range change rate derived from the relative velocity is continuously monitored;

[0042] The dynamic statistical characteristics of the Doppler frequency shift or the slant range change rate in a preset time window are evaluated;

[0043] The protection interval of the return link is adaptively configured based on the dynamic statistical characteristics, wherein when the dynamic statistical characteristics do not exceed a threshold, the protection interval is contracted, and otherwise, the protection interval is expanded to a conservative value.

[0044] According to one aspect of the present application, a possible implementation of the method for predicting the one-way propagation delay between the terminal and the satellite comprises: superimposing the time-domain correction term mapped by the Doppler frequency shift on the geometric propagation delay to perform feedforward correction; estimating the processing delay of the terminal comprises: constructing the feedforward prediction value of the symbol rate switching, and performing feedback correction on the feedforward prediction value according to the delay deviation measured by the probe signal; synthesizing the terminal local time comprises: applying variable fractional delay filtering to realize fine adjustment of the sub-sampling stage, and synchronously compensating for the group delay introduced by the filtering.

[0045] Technical effects, through the above technical solutions, the present application solves the problems of insufficient synchronization accuracy, slow convergence and low resource utilization rate in high-speed dynamic scenarios, can improve the synchronization accuracy to the microsecond level, accelerate the convergence speed in dynamic scenarios, and support adaptive compression of guard intervals, effectively improve the stability and spectral efficiency of the satellite communication system. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 It is a whole flowchart of a time synchronization method suitable for a satellite communication system.

[0047] Figure 2 It is a flowchart of synthesizing the terminal local time by fusing the network clock reference time and the processing delay estimation.

[0048] Figure 3 It is a flowchart of estimating the processing delay of the terminal according to the symbol rate and the equipment working condition of the terminal, and referring to the delay calibration data.

[0049] Figure 4 It is a flowchart of obtaining the delay deviation measured by the probe signal. DETAILED DESCRIPTION

[0050] In order to enable personnel in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0051] The terms "first", "second", and the like in the description and the drawings of the present application are used to distinguish different objects, and are not used to describe a particular order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, or product that includes a series of steps or units is not limited to the listed steps or units, but can optionally include other steps or units not listed, or can optionally include other steps or units inherent to the process, method, product or end.

[0052] In this paper, the term "embodiment" means that the specific features, structures or characteristics described in conjunction with the embodiment can be included in at least one embodiment of the present application. The phrase appears at various places in the specification does not necessarily refer to the same embodiment, nor is it independent or alternative to other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0053] The present specification uses uniform symbols and data item conventions, which are defined as follows for the main characters to be used in the embodiments of the present application:

[0054] t _NCR_rx Indicates the network clock reference time parsed by the terminal from the forward broadcast signal, in seconds (s).

[0055] geo _handle Indicates the geometric prediction handle, which functions to input any time, and can output the geometric relationship parameters (such as distance, relative velocity, etc.) between the terminal and the satellite at that time.

[0056] τ _path_pred Indicates the predicted one-way propagation delay between the terminal and the satellite, in seconds (s).

[0057] device _conditions Indicates the device working condition of the terminal, which is a vector containing multiple parameters such as temperature, power supply voltage, circuit working mode, etc.

[0058] delay _stat Indicates the estimated terminal processing delay, in seconds (s).

[0059] t _term_local Indicates the local time finally synthesized at the terminal, in seconds (s).

[0060] bias _nonrecip Indicates the forward / backward link non-reciprocal bias provided by the network side or estimated by the terminal, in seconds (s).

[0061] t _txThis indicates the return link transmission time without non-reciprocal bias correction, in seconds (s).

[0062] t _tx_corrected This indicates the final return link transmission time after non-reciprocal bias correction, in seconds (s).

[0063] t _base The baseline local time is represented by t, which is directly synthesized from the network clock reference time and the processing delay estimate. _base =t _NCR_rx +delay _stat .

[0064] u _k This represents the fractional delay control value, which is a dimensionless normalized value, typically in the range of [-0.5, 0.5), and is used to control a variable fractional delay filter.

[0065] vfd(u _k This indicates that the variable fractional delay filter is based on the control quantity u. _k The generated fractional delay adjustment is expressed in seconds (s).

[0066] Δ _group_delay This represents the amount of compensation used to compensate for the group delay error introduced by the variable fractional delay filter, expressed in seconds (s).

[0067] delay _pred This represents the predicted feedforward processing delay, including the step term, in seconds (s).

[0068] e _meas_aligned It represents the end-to-end delay deviation measured and aligned by the network side using probe signals, in seconds (s).

[0069] α _k This represents the dynamically updated weights used for feedback correction. It is a dimensionless value and is typically in the range [0,1].

[0070] Example 1: This example describes the overall framework and basic process of a time synchronization method suitable for satellite communication systems, such as... Figure 1 As shown. Specifically, it includes the following steps:

[0071] Step S101: Analyze the forward broadcast signal and obtain the network clock reference time.

[0072] For example, the terminal continuously receives downlink forward broadcast channels from satellites. Specific physical channels or higher-layer signaling within this channel carry the network clock reference count (ncr). _count_rx With the corresponding frame number _index The terminal configures NCR according to the preset network clock reference. _config(where the bit width of the counter, the update frequency, the duration of each count, etc. are defined), and the ncr _count_rx is converted into absolute time t _NCR_rx . This conversion process preferably takes into account the rule that the counter rolls over from the maximum value to zero, in order to ensure continuity of time. For example, a 32-bit counter, updated at a frequency of 10 MHz, has a roll-over period of approximately 429.5 seconds, and algorithms within the terminal need to correctly handle time calculations that cross the roll-over point.

[0073] Step S102, based on satellite orbit data, construct a geometric prediction handle.

[0074] Specifically, in the terminal initialization or network entry phase, the terminal obtains the satellite orbit data orbit _data (e.g. two-line orbital elements or more accurate ephemeris data) from the network or reads it from local storage. Based on the above data, the terminal internally constructs a geometric prediction handle geo _handle . This handle can be a function, an object or a service in terms of function, which receives an absolute time stamp as input and outputs the position and velocity vector of the satellite at that time in a predetermined coordinate system (such as the Earth-Centered Inertial System or the Earth-Centered Earth-Fixed System). A preferred implementation is that the module pre-computes the satellite positions at a series of discrete time points, and uses methods such as cubic spline interpolation or piecewise linear interpolation to provide high-precision position information for any input time. The module can also record a geometric accuracy indicator geo _accuracy , such as the maximum interpolation error, for subsequent quality assessment.

[0075] Step S103, use the geometric prediction handle to predict the one-way propagation delay between the terminal and the satellite.

[0076] At each period when the transmission time needs to be calculated, the terminal obtains its own accurate geographical position (e.g. through GNSS global navigation satellite system positioning), and then inputs the current time into geo _handle to obtain the real-time position of the satellite. Based on the two position coordinates of the terminal and the satellite, the straight-line geometric distance between them is calculated. Dividing this distance by the speed of light c, the propagation delay based on geometric distance τ _geometry is obtained. In the basic embodiment, this τ _geometry can be taken as the one-way propagation delay prediction τ _path_pred .

[0077] Step S104, estimate the processing delay of the terminal according to its symbol rate and device operating conditions, and refer to the delay calibration data.

[0078] The processing delay of a terminal refers to the time experienced by a signal from the interface of a baseband processing module to the output of an antenna port (or vice versa), which varies significantly with symbol rate and device conditions (such as chip temperature and operating voltage). To accurately estimate this delay, the method uses a calibration data-based strategy. Specifically, the terminal pre-stores a processing delay calibration table delay _calibration_table , which records the measured or simulated processing delay values at different symbol rates and typical conditions. When the delay needs to be estimated, the terminal obtains the symbol rate symbol _rate of the current service configuration and the current device condition device _conditions , and calculates the processing delay estimate delay _stat under the current conditions by looking up delay _calibration_table or using interpolation / extrapolation algorithms.

[0079] An exemplary implementation is that if the calibration data is sufficient, a quadratic polynomial function of the delay with respect to the reciprocal of the symbol rate delay _stat =a0+a1*(1 / symbol _rate )+a2*(1 / symbol _rate ) 2 can be fitted, where a0, a1, and a2 are fitting coefficients.

[0080] Step S105: fuse the network clock reference time and the processing delay estimate to synthesize the terminal local time.

[0081] After obtaining the unified time reference t _NCR_rx from the network and the processing delay delay _stat characterizing its internal characteristics, the terminal fuses the two. Specifically, by a simple addition operation t _term_local =t _NCR_rx +delay _stat , the terminal local time t _term_local can be synthesized. The physical meaning of t _term_local is that when the network side time is t _NCR_rx , the local time aligned with the network after the processing delay in the terminal. When performing this addition operation, attention should be paid to numerical precision, for example, time-related calculations can be uniformly represented by 64-bit floating-point numbers and follow a pre-set rounding _rule rule (such as rounding to the nearest nanosecond), to avoid the accumulation of precision loss.

[0082] Step S106: based on the terminal local time and the one-way propagation delay prediction, and optionally corrected according to the non-reciprocal bias, the transmission time of the return link is calculated.

[0083] In order to make the return link signal sent by the terminal arrive at the satellite at the predetermined time accurately, the terminal needs to send in advance. The time in advance is the propagation time of the signal in space. Therefore, the basic sending time t _tx is calculated as t _tx =t _term_local -τ _path_pred . In some systems, due to the difference in frequency bands of uplink and downlink, the difference in filtering characteristics of the on-board transponder, and other reasons, the propagation path is non-reciprocal, that is, the uplink and downlink propagation delays are not completely equal. In order to compensate for the slowly varying systematic deviation, a non-reciprocal bias bias _nonrecip may be introduced for correction. At this time, the sending time t _tx_corrected is calculated as t _tx_corrected =t _tx -bias _nonrecip . The terminal sending scheduling module will accurately control the sending of the return burst according to t _tx_corrected .

[0084] In an exemplary embodiment, the step of running the preparation and network clock reference time acquisition can also include two parts:

[0085] The first part is data loading and network clock reference time acquisition, which is as follows:

[0086] Reading satellite orbit data orbit _data , beam coverage and switching schedule beam _plan , network clock reference configuration ncr _config , processing delay calibration table delay _calibration_table , physical and system constants constants, completing file / message integrity check, and obtaining loading state load _ok .

[0087] When the loading state load _ok is true, the network clock reference count ncr _count_rx and the frame number frame _index are extracted from the forward broadcast frame; according to the bit width and step length of ncr _config , ncr _count_rx is converted into network clock reference time t _NCR_rx (considering the count roll-back rule).

[0088] Based on the satellite orbit data orbit _data , a geometric prediction handle geo _handle (three times spline or linear segment interpolation; the step length comes from the time resolution in constants) is constructed, and the geometric accuracy geo _accuracy (maximum position interpolation error, for example) is recorded.

[0089] The second part is a running environment self-check and calibration coverage check. Specifically, read delay _calibration_table The symbol rate set symbol _rate_set_calib , which can be issued by the network. _rate_set_net Compare and form a calibration coverage label calib _ready_flag (including the missing rate list).

[0090] Embodiment two, the embodiment provides a local time synthesis method with sub-sampling level precision. The method breaks through the limitation of terminal digital system sampling rate on time synchronization precision, and realizes sub-sampling level fine tuning.

[0091] In the embodiment, as Figure 2 shown, the step of synthesizing the terminal local time by fusing the network clock reference time and the processing delay estimation further includes:

[0092] Step S201, integrate the network clock reference time and the processing delay estimation to generate a baseline local time; determine a fractional delay control quantity for the baseline local time by comparing with the one-way propagation delay prediction; generate a fractional delay adjustment quantity by filtering the fractional delay control quantity through a variable fractional delay filter; correct the baseline local time by using the fractional delay adjustment quantity to synthesize a terminal local time with sub-sampling level precision.

[0093] Specifically, according to the manner of embodiment one, t _NCR_rx is added to delay _stat to obtain the baseline local time t _base . The precision of t _base is limited by the sampling clock period Ts of the system (Ts=1 / Fs, where Fs is the sampling rate). In order to realize finer adjustment, the decimal part of the difference between the target adjustment time and the current achievable time needs to be calculated. An ideal transmission time should be t _base -τ _path_pred , but since t _base is discrete, the decimal part of the ideal time cannot be directly realized. The decimal part that needs to be corrected is called fractional delay, and the fractional delay control quantity u _k is calculated accordingly. For example, u _k may be calculated as the remainder of the ideal adjustment quantity divided by the sampling period, and normalized to obtain the target fine adjustment amount Δ _t_frac =(t _base -τ _path_pred )modTs, then u _k =Δ _t_frac / Ts. u _k is a dimensionless value usually in the range of (-0.5, 0.5).

[0094] u _k is sent into a variable fractional delay filter (VFD). For example, a Farrow structure can be employed, which is characterized by filter coefficients being polynomials of u _k , and its introduced group delay can be flexibly adjusted by changing u _k . The VFD filter processes the timing signal stream internally and generates an accurate time adjustment, i.e. fractional delay adjustment vfd(u _k ), according to u _k . This adjustment is applied to the baseline local time, and thus the precision correction at sub-sampling level is achieved.

[0095] Further, in order to eliminate the frequency-dependent error introduced by the variable fractional delay filter itself, the step of synthesizing the terminal local time further comprises: judging whether the frequency-dependent error introduced by the variable fractional delay filter is greater than a threshold, if yes, retrieving a group delay compensation from the group delay calibration data, with reference to the fractional delay control and the spectral characteristics of the signal; and applying the fractional delay adjustment and the group delay compensation to the baseline local time to synthesize the terminal local time.

[0096] The group delay response of the VFD filter is not flat at all frequencies while implementing fractional delay, which will cause signal waveform distortion. In order to solve this problem, the embodiment introduces a group delay secondary compensation mechanism. The system pre-stores a group delay calibration table gd _calibration_table , which records the equivalent group delay introduced by the VFD filter under different u _k values and different signal spectral characteristics. In runtime, the group delay compensation Δ _group_delay is obtained by table lookup or interpolation according to the current u _k and the signal spectrum spectrum. The terminal local time with sub-sampling level precision and no frequency-dependent error is synthesized as: t _term_local = t _base + vfd(u _k ) + Δ _group_delay .

[0097] In order to ensure the stability and robustness of the system, preferably, the variable fractional delay filter processing and the retrieval process of the group delay compensation are configured in a mutually bound processing module; when the real-time performance of the processing module execution state or the internal parameters do not meet the preset conditions, the processing module is disabled as a whole, and the synthesized terminal local time is rolled back to the baseline local time.

[0098] In other words, the VFD filtering and the group delay compensation are regarded as inseparable functional units. The system will monitor the running state of the module, for example, if vfd(u _k ) or Δ _group_delaythe processing delay exceeds a preset real-time threshold, or u _k the value of the processing delay is out of the stable working range, the functional unit is disabled as a whole. At this time, the time synthesis process automatically falls back to the basic mode, i.e. _term_local =t _base This design guarantees the stability and reliability of the entire synchronization loop at the cost of temporary precision degradation.

[0099] As a balance strategy between performance and resource consumption, the step of processing the fractional delay control quantity through the variable fractional delay filter can further include an order self-adaptation mechanism. Specifically, the system continuously monitors the alignment error performance of the return link and the calculation resource occupation of the terminal processor. When the alignment error does not meet the target threshold (for example, less than 100 nanoseconds) for continuous multiple periods, and the calculation resource occupation rate is lower than the preset budget (for example, lower than 80%), the system can temporarily increase the filter order of the variable fractional delay filter (for example, increase the order of the Lagrange interpolator of the Farrow structure from 3 to 4) to enhance the interpolation precision and converge the error faster. When the alignment error performance recovers to the standard, the filter order falls back to the baseline level (for example, 3) to save the calculation resource and power consumption.

[0100] Embodiment three, the embodiment provides an adaptive and accurate processing delay estimation method, which solves the problems that the static calibration method is difficult to cope with the processing delay step caused by symbol rate switching, and the tracking problems caused by slow time-varying factors such as temperature drift and aging, and realizes dynamic, adaptive and high-precision estimation of the processing delay.

[0101] In the embodiment, as shown in Figure 3 , the step of estimating the processing delay of the terminal according to the symbol rate of the terminal and the equipment working condition and referring to the delay calibration data further includes:

[0102] Step S301, for the switching event of the symbol rate of the terminal, a feedforward prediction value of the processing delay is constructed. The switching of the symbol rate will cause the approximate step change of the group delay of the internal digital signal processing link (such as the FIR filter, the decimation / interpolation module, etc.) of the terminal, which is a deterministic change that can be predicted in advance. Specifically, the system not only acquires the current symbol rate symbol _rate , but also records the symbol rate symbol _rate_prev before the switching. Based on the two and the current equipment working condition device _conditions , the system will construct a feedforward prediction value delay _pred . The calculation method of the prediction value is:

[0103] delay _pred =delay _stat (symbol_rate )+J _step (symbol _rate_prev ,symbol _rate ,device _conditions ). Wherein, delay _stat (symbol _rate ) is the static delay estimation from the current symbol rate according to the method of embodiment one; and J _step is the rate switching step item obtained in advance by offline calibration or theoretical modeling. The step item J _step accurately quantifies the additional delay variation introduced by changes in filter order, pipeline depth, etc. when the processing link is switched from symbol _rate_prev to symbol _rate . By introducing J _step for feedforward compensation, delay _pred can approximate the true delay value to the maximum extent before the feedback correction loop intervenes, shortening the convergence time of the system.

[0104] Step S302, obtaining the delay deviation measured by the probe signal. In order to correct the residual error of feedforward prediction and track the unpredictable slow-varying factors such as temperature drift, the embodiment introduces a feedback correction mechanism based on measurement. As shown in Figure 4 , the specific process of obtaining the measured deviation includes:

[0105] According to the beam switching schedule beam _plan , the beam switching protection window is identified, and the terminal is instructed to transmit a dedicated probe burst during the window. The probe burst is a physical layer signal, and its waveform structure is designed, for example, a sequence with good autocorrelation characteristics (such as ZC sequence) is used, so as to facilitate the network side to estimate the arrival time with high precision. The probe is selected to be performed in the beam switching protection window, because this window is a pause gap for service data transmission, and inserting the probe signal will not affect the normal communication, realizing zero overhead of the measurement process.

[0106] Further, the network side performs correlation estimation on the received probe burst to obtain the initial measured delay deviation and the corresponding measurement confidence. After the network side receiver captures the probe burst, it will perform correlation operation with the locally stored ideal probe sequence copy. The position of the correlation peak corresponds to the actual arrival time of the signal, which is compared with the expected ideal arrival time of the network side, and the initial end-to-end delay deviation e _meas is obtained. According to the peak value, signal-to-noise ratio, sidelobe suppression ratio, etc. of the correlation peak, the measurement confidence meas _conf can be calculated, which is used to represent the reliability of this measurement.

[0107] Preferably, in order to improve the measurement accuracy, before performing the correlation estimation at the network side, a pre-calibrated impulse response kernel can be applied to the received probe burst signal to perform a transient deconvolution pre-processing, so as to purify the signal and reduce the distortion caused by the transient effect of the transmitting link or channel, and then the correlation estimation is performed on the purified signal obtained after the deconvolution pre-processing. Specifically, the devices such as the power amplifier of the terminal can have a memory effect or a filtering transient at the initial stage of the burst signal, which can contaminate the waveform of the probe signal and cause the correlation peak to spread or shift. By pre-calibrating the equivalent impulse response kernel h _txrx of the transmitting link, the fast deconvolution operation on the signal at the receiving end can eliminate the distortion to a greater extent and obtain a sharper and more accurate correlation peak, thereby improving the purity of the delay deviation e _meas .

[0108] On this basis, the initial measured delay deviation e _meas_aligned is time-stamped and aligned, the cross-frame measurement error is eliminated, and the aligned measured delay deviation e _meas for feedback correction is generated. Since there can be a slight time deviation between the measurement time at the network side and the frame structure of the terminal, directly using e _meas may introduce a cross-frame error. Therefore, the time stamp of e _index needs to be strictly aligned with the frame sequence number frame _meas_aligned at the terminal side, so as to compensate for the inter-frame deviation and obtain e _k that can be used for local recursive updating.

[0109] In step S303, the feedback correction is performed on the feedforward prediction value according to the measured delay deviation, and the processing delay estimate is determined. The specific implementation manner is as follows: based on whether the current state is in the beam switching protection window and the measurement confidence, the update weight a _k is dynamically determined. When and only when the current state is in the window and the measurement confidence is not lower than a preset threshold conf _th , a higher first weight value a _k_hi is configured for the update weight, and a lower second weight value a _k_lo is configured for the update weight under other conditions. For example, a _k_hi may be set to 0.5, and a _k_lo may be set to 0.01. In the window period of the rate switching, the system error is mainly caused by the step, at this time, the measurement value e _meas_aligned is very critical, and a high weight should be given to achieve fast tracking and convergence; and in the stable period outside the window, the system error is mainly the slow change of the temperature drift, at this time, a low weight should be given to smooth the noise and avoid excessive tracking of the measurement jitter.

[0110] The determined update weight is used to weight and fuse the feedforward prediction value and the aligned measured delay deviation, so as to determine the processing delay estimate delay_stat The recursive formula is:

[0111] delay _stat_k =(1-α _k )*delay _pred_k +α _k *e _meas_aligned_k ; where the subscript k represents the current period.

[0112] As a more refined weight determination scheme, the step of dynamically determining and updating weights can be specifically implemented as a piecewise recursive update mechanism with multiple time constants. This mechanism includes: when within the beam switching protection window and the measurement confidence level meets the requirements, configuring the update weights to a fast update constant α. _fast (equivalent to α) _k_hi This allows for rapid correction of transient step errors; when outside the window, if the detected rate of change of device temperature dT / dt is greater than a preset threshold, the update weight is configured as the intermediate update constant α. _mid (e.g., 0.1) to accommodate faster temperature drift; when outside the window and the device temperature changes steadily, the update weight is configured to the slow update constant α. _slow (equivalent to α) _k_lo This allows for stable tracking of slow time-varying drift and aging.

[0113] To ensure the stability and reliability of the entire estimation process, after determining the processing delay estimate, the method also includes a step of performing boundary and saturation control. Specifically, this involves verifying the determined processing delay estimate. _stat Does the calibration exceed the preset allowable range? _bounds or its updated residual r _k =e _meas_aligned_k -delay _pred_k The system checks whether the absolute value of the delay estimate is greater than a preset threshold. If any test result is true, a clamping operation is performed on the processing delay estimate to constrain it within the calibrated allowable range, and a delay status flag indicating an abnormal current estimation state is generated. _bound_flag =1; otherwise, maintain the original processing delay estimate, and delay _bound_flag =0. The delay _bound_flag It is an important indicator of the system's health status and can be used by other control modules (such as adaptive gating of protection intervals).

[0114] This embodiment also includes a fallback strategy in the event of a probe burst loss. Specifically, when a dedicated probe burst loss is detected (e.g., via a probe loss identifier), _loss_flag=1 indicates that the feedback correction based on the measured delay deviation is paused; instead, a device condition compensation term g (device) is generated based on the terminal's device conditions, such as temperature. _conditions The equipment condition compensation term is then added to the feedforward prediction value to generate the processing delay estimate for the current cycle, i.e., delay. _stat =delay _pred +g(device _conditions Among them, g(device) _conditions The expression is a linear or nonlinear compensation function based on temperature changes. A backoff mechanism ensures that even when measurement information is temporarily unavailable, the system can still maintain basic tracking capability against slow drift using other auxiliary information (such as temperature sensor data), achieving progressive degradation.

[0115] For example, the steps for processing delay estimation can also be:

[0116] Read the symbol rate symbol from the business configuration. _rate With equipment operating conditions _conditions (Temperature, power supply, circuit mode), and with frame _index Alignment. Based on delay. _calibration_table symbol _rate device _conditions Calculate the static estimation of processing delay. _stat When the calibration covers the calib mark _ready_flag When insufficient coverage is detected, interpolate adjacent rates or use a fitted delay function. _stat (symbol _rate ) = a0 + a1 * symbol _rate -1 +a2*symbol _rate -2 Perform extrapolation and generate the upper limit of the residual. _bound .

[0117] Alternatively, the steps for processing delay estimation can also be:

[0118] In handling static estimation of delay _stat Based on this, the processing link step term caused by symbol rate switching is explicitly modeled to obtain the feedforward prediction delay. _pred :

[0119] delay _pred =delay _stat (symbol _rate )+J _step (symbol _rate_prevsymbol _rate ,device _conditions );

[0120] When the beam switching protection window comes, the network inserts an ultra-short probing burst in a very short gap without affecting the service, and the network side returns the end-to-end delay deviation e _meas ; a larger update weight is used within the window to recursively update delay _pred to delay _stat : delay _stat =(1-α)*delay _pred +α*e _meas ;

[0121] Outside the window, a smaller value of α is taken to slowly track temperature drift and aging. The embodiment outputs delay _stat .

[0122] Optionally, the step of processing the delay estimation can also be a step of containing rate hopping feedforward and window short measurement recursion, specifically:

[0123] The rate hopping step item modeling is performed, symbol _rate_prev , symbol _rate , device _conditions are read and aligned with frame _index to form the rate hopping description rate _step . Based on delay _calibration_table , a static delay estimation delay _stat_base is obtained; on this basis, a step item J _step (rate _step , device _conditions ) is introduced to obtain a feedforward processing delay prediction delay _pred . The feedforward confidence pred _conf (comes from the coverage degree and device working condition weight) is recorded.

[0124] Further, the needle-shaped short measurement of the beam switching window is performed, the beam _plan switching protection window is determined according to beam _plan , an ultra-short probing burst (1 beam symbol) is inserted, and a short measurement plan short _probe_plan is generated. The network side estimates the end-to-end delay deviation according to short _probe_plan , obtains the measured deviation e _meas and the measurement confidence meas _conf (considering the transmitter memory effect and filtering transient state). The e _meas timestamp is aligned with frame _index , the cross-frame deviation is eliminated, and the aligned e _meas_aligned is output.

[0125] Further, perform recursive update processing of delay, read input e _meas_aligned , delay _pred , meas _conf , window _flag (flag of beam switch window), update weight a based on measurement confidence and window state _k : if window _flag =1 and meas _conf is high, take larger weight; take smaller weight in stable period or low confidence. Calculate residual r _k : read input e _meas_aligned , delay _pred , take r _k =e _meas_aligned -delay _pred , output r _k . Recursive to get static estimation of processing delay delay _stat : read input delay _pred , a _k , e _meas_aligned , take delay _stat =(1-a _k )*delay _pred + a _k *e _meas_aligned , output delay _stat . Perform boundary and saturation control, read input r _k , delay _stat , calibration _bounds (allowable range of calibration), if delay _stat goes out of range or |r _k | exceeds threshold, perform clipping, output estimated terminal processing delay delay _stat_clamped after performing clipping and state flag delay _bound_flag indicating that the current estimation state is abnormal. Update rate switching state, assign symbol _rate to symbol _rate_prev .

[0126] Further, handle exception and fallback (only when exception is triggered), exception criterion: read input meas _conf , r _k , probe _loss_flag , trigger exception when meas _conf is low, |r _k | is too large or probe _loss_flag =1, output delay _est_status =degraded (fallback). Fallback strategy: read input delay_pred , delay _est_status , if degraded, set estimated terminal processing delay backoff delay _stat_fallback = delay _pred , and mark update _α_reduction (decrease alpha _k for subsequent time). Output delay _stat_fallback , update _α_reduction .

[0127] Exemplarily, a sub-sampling level precision local time synthesizing method (fractionally delayed execution) step can also be:

[0128] Read input t _NCR_rx , delay _stat (or delay _stat_clamped / delay _stat_fallback ), get baseline local time t _base =t _NCR_rx +delay _stat .

[0129] Calculate the fractionally delayed control quantity, specifically, read input t _base , tau _path_pred , form the target fine adjustment quantity delta _t_frac =round _to_fraction (t _base -tau _path_pred , 1 / Fs) remaining fraction, calculate the vfd control quantity u _k (unit sample).

[0130] Perform fractionally delayed filter execution, specifically, use Farrow structure or equivalent implementation, apply u _k to the local timing link to get fractionally delayed vfd(u _k ), output vfd(u _k ) and execution state vfd _status .

[0131] Further, perform group delay quadratic compensation, specifically: read input u _k , spectrum, vfd _status , gd _calibration_table (group delay calibration table), find group delay compensation delta _group_delay (u _k , spectrum) according to the spectrum and control quantity. Synthesize terminal local time, read input t _base , vfd(u _k ), delta _group_delay , calculate t _term_local =t _base+ vfd(u _k ) + Δ _group_delay .

[0132] Further, the sending time and non-reciprocal correction are performed, specifically: reading input t _term_local , τ _path_pred , calculating t _tx =t _term_local -τ _path_pred ; non-reciprocal correction, reading input t _tx , bias _nonrecip (if any), calculating t _tx_corrected =t _tx -bias _nonrecip .

[0133] Further, alignment checking and adaptive protection interval are performed, specifically: scheduling sending and measuring: scheduling return burst with input t _tx_corrected , the network side obtains alignment error ε _align and measures the label align _conf , adaptive protection interval: reading input GP _static , ε _align , align _conf , delay _bound_flag , update _α_reduction , τ _path_pred_conf (one-way propagation delay prediction confidence) (if any), calculating protection interval configuration GP _config : when ε _align meets the threshold and the label is good and has no degradation flag, GP _config converges to the upper limit given by the adaptive protection interval; otherwise, it falls back to GP _static . Output GP _config .

[0134] Synchronization state and statistics are performed: aggregate ε _align , align _conf , delay _est_status to form synchronization state sync _state , output sync _state as the end point for operation and performance statistics.

[0135] Embodiment four, the embodiment provides a propagation delay prediction and adaptive protection interval method based on Doppler correction, specifically, in high-speed relative motion scenarios such as low earth orbit (LEO) satellites, only relying on static geometric distance to calculate propagation delay will introduce obvious errors, because in the duration of a frame, the distance between the satellite and the ground will change. This embodiment introduces first-order feedforward correction of Doppler to improve the dynamic accuracy of delay prediction, and on this basis, realizes adaptive configuration of return link protection interval.

[0136] In this embodiment, the step of predicting the one-way propagation delay between the terminal and the satellite using the geometric prediction handle specifically includes:

[0137] In step S401, the geometric propagation delay based on the geometric distance is calculated using the geometric prediction handle. This step is the same as the basic part of S103 in Embodiment 1, that is, the geometric distance d between the terminal and the satellite is calculated based on the instantaneous position coordinates of the terminal and the satellite _geometry , and the geometric propagation delay τ _geometry = d _geometry / c is obtained. Wherein, c is the speed of light. This τ _geometry constitutes the main part of the propagation delay.

[0138] In step S402, the relative velocity between the terminal and the satellite is obtained using the geometric prediction handle, and the Doppler shift is calculated accordingly. The geometric prediction handle geo _handle not only provides position information, but also provides velocity vector information. By obtaining the velocity vector v _term of the terminal (for a stationary terminal, this velocity is the linear velocity of the earth rotation at this position) and the velocity vector v _sat of the satellite, the relative velocity vector v _rel =v _sat -v _term between them can be calculated. The relative velocity vector is projected in the direction of the satellite-terrestrial line of sight to obtain the radial relative velocity, that is, the range rate range _rate . The relationship between the Doppler shift doppler _hz and the range rate range _rate is doppler _hz =-range _rate *f _c / c, wherein f _c is the carrier center frequency of the signal. The doppler _hz is positive, indicating that the satellite-terrestrial distance is shortening.

[0139] In step S403, the Doppler shift is mapped to a time delay correction term in the time domain. The physical nature of the Doppler shift is the continuous change of the wave path. A non-zero doppler _hz means that in a basic time unit (such as the period T _frame of a wireless frame), the length of the propagation path will change slightly Δ _d =range _rate *T _frame . The change in path length directly leads to the change in propagation delay. Therefore, the Doppler shift can be mapped to a time delay correction term Δ _τ_dopp, which aims to compensate the delay drift due to range change from current time to next frame. _τ_dopp which can be approximated as _τ_dopp = -doppler _hz * T _frame / f _c . This correction term is a first-order linear prediction of the delay change in the near future, which belongs to feed-forward correction.

[0140] Step S404, superimpose the delay correction term to the geometric propagation delay to generate a one-way propagation delay prediction which reflects the real-time dynamics more accurately. The one-way propagation delay prediction τ _path_pred after Doppler first-order correction is calculated as: _path_pred = τ _geometry + Δ_τ _dopp . Thus, τ _path_pred is no longer a static snapshot of geometry, but a predicted value containing first-order dynamic trend, which can more accurately reflect the actual propagation time experienced by the signal after being emitted, especially in the satellite over-the-top or low elevation phase with large range rate of change.

[0141] On this basis, the method further comprises: step S405, continuously monitoring the Doppler frequency shift or the range rate of change derived from the relative velocity; evaluating the dynamic statistical characteristics of the Doppler frequency shift or the range rate of change within a preset time window; and adaptively configuring the guard interval of the return link based on the characteristics. The traditional guard interval (GP) is usually fixedly configured according to the worst case (i.e. the maximum possible delay change range), which will cause waste of spectrum resources in most of the time.

[0142] This embodiment uses the real-time acquired doppler _hz or range _rate information to realize adaptive adjustment of the guard interval GP _config . Specifically, the terminal will maintain a sliding window (for example, containing the range _rate measurements of the past 100 frames), and continuously calculate the statistical characteristics within the window, such as the maximum, minimum, mean and standard deviation of range _rate . Based on these statistical characteristics, the system executes the following adaptive rules: when the dynamic statistical characteristics represent a smooth environment (for example, the standard deviation of range _rate is less than a certain smooth threshold th _stable , which usually occurs in the satellite high-elevation overpass phase), the guard interval is shrunk to improve the link efficiency, and GP _config can be set to a value based on the measured statistical upper bound and additional small margin. When it represents a dynamic environment (for example, the standard deviation of range _rate is greater than a certain dynamic threshold th_dynamic , which is usually in the stage of low elevation and high speed of satellite), the guard interval is extended to a conservative value GP _static to guarantee the robustness of alignment. The size of guard interval is matched with the real risk of channel dynamics, which maximizes the spectral efficiency and system throughput under the premise of ensuring communication reliability.

[0143] Exemplarily, the steps of geometry and propagation delay prediction and guard interval configuration can be specifically:

[0144] According to geo _handle , beam _plan , constants, frame _index , the geometric distance from the current service area to the satellite is calculated to obtain the one-way propagation delay prediction τ _path_pred and the accompanying timestamp ts _pred . Output τ _path_pred , ts _pred for sending time calculation. The coverage radius of beam _plan and the speed of light in constants are used to calculate the conservative guard interval GP _static (e.g. GP _static = the upper limit of one-way coverage radius / speed of light, or take multiple according to system rules). Output GP _static for guard interval configuration. For τ _path_pred and GP _static , quality _tag is generated (e.g. too low elevation, interpolation extrapolation out of bounds).

[0145] Optionally, geometry and propagation delay prediction and guard interval configuration can also be: the relative speed output by the geometry prediction handle geo _handle introduces a first-order Doppler correction on the single-frame scale, and the advance correction term obtained by mapping doppler _hz is added when calculating the one-way propagation delay prediction τ _path_pred to reduce the alignment drift caused by beam switching or relative motion; on the guard interval, GP _static is adaptively upper limited to converge according to the statistical upper limit of Doppler and range rate of change, and falls back to a conservative value when the environment is stable.

[0146] Optionally, the steps of geometry and Doppler first-order correction can be:

[0147] The delay prediction of the first-order Doppler correction is obtained by geo _handle , frame _index , the relative speed of satellite position and velocity and the center of the service area, the Doppler frequency shift doppler _hz and the range rate range _rate. Doppler _hz mapping to propagation delay advance Δ _τ_dopp (Linear approximation based on frame period and carrier wavelength), superimposed with a geometric term to form one-way propagation delay prediction τ _path_pred : τ _path_pred = geometric delay + Δ _τ_dopp . Record prediction timestamp ts _pred and Doppler confidence doppler _conf (given by elevation and range _rate estimation), for risk weighting.

[0148] Adaptive protection interval upper bound calculation, estimate Doppler _hz , range _rate upper bound over a period of time, calculate adaptive protection interval upper bound GP _static (revert to conservative value in stationary environment), and output upper bound confidence.

[0149] Embodiment five, the embodiment provides a chain-coupled system-level time synchronization method.

[0150] Specifically, in the embodiment: the step of predicting the one-way propagation delay between the terminal and the satellite includes superimposing the time domain correction term mapped by the Doppler frequency shift on the geometric propagation delay to perform feedforward correction; the step of estimating the processing delay of the terminal includes constructing a feedforward prediction value of symbol rate switching, and performing feedback correction on the feedforward prediction value according to the delay deviation actually measured by the probe signal; and the step of synthesizing the local time of the terminal includes applying a variable fractional delay filter to realize fine adjustment at a sub-sampling level, and synchronously compensating for the group delay introduced by the filter.

[0151] The design points in the scheme are as follows: the gain of geometric feedforward correction on processing delay estimation, specifically, the Doppler-corrected τ _path_pred provided by embodiment four provides more accurate data for the entire synchronization system. When the network side measures the arrival time of the probe burst under the framework of embodiment three, and calculates the delay deviation e _meas , since the ideal arrival time it relies on is itself closer to the true value due to Doppler correction, the calculated e _meas can reflect the processing delay error of the terminal, and is less polluted by geometric prediction error. In other words, the feedforward correction on the geometric side reduces the observation noise of the processing delay estimation loop, so that the feedback correction loop in embodiment three can work in a condition with higher signal-to-noise ratio, converge faster, and the estimation result is also more stable.

[0152] Support of sub-sampling synthesis by fast loop tracking of processing delay, specifically, embodiment three uses the mechanism of step feedforward + window probe fast loop to make the processing delay estimation delay_stat The real delay variation caused by symbol rate switching and temperature drift can be tracked in real time and quickly. Therefore, the baseline local time t _base = t _NCR_rx + delay _stat provided to Embodiment Two is already very close to the ideal value, which is the optimal estimate on the level of integer sampling periods. The burden of the fractional delay executor in Embodiment Two is alleviated. Instead of compensating for large errors on the level of integer sampling periods, it is only necessary to focus on processing the last remaining small residual errors on the level of sub-sampling periods. This enables the VFD filter to work in the fine-tuning interval with the highest linearity and accuracy, ensuring that the final synthesized t _term_local can achieve high accuracy.

[0153] The sub-sampling accuracy enables the closed-loop of the adaptive guard interval, and specifically, the whole chain of improvements is embodied in the extremely low alignment error ε _align provided by Embodiment Two. Only when the system can continuously and reliably stabilize ε _align within a very small threshold, does the strategy of adaptive contraction of the guard interval in Embodiment Four have the premise of implementation. If the system alignment error itself is large and unstable, then the behavior of trying to compress the guard interval will bring the risk of disconnection. Therefore, the sub-sampling level execution accuracy provided by Embodiment Two provides a closed-loop decision basis and feasibility guarantee for the spectrum efficiency improvement strategy of Embodiment Four.

[0154] Embodiment Six, this embodiment describes the step of optionally correcting according to the non-reciprocal bias in Embodiment One. In actual satellite communication systems, since the uplink and downlink usually use different frequency bands, the delays experienced by signals when passing through the ionosphere and the troposphere are not exactly the same, and this difference is the non-reciprocal bias. This bias is not a fixed constant, but slowly changes with factors such as satellite elevation angle and atmospheric conditions. This embodiment provides a light-weight, slow-loop tracking compensation method based on proxy data, which is particularly suitable for terminals without dual-frequency measurement hardware.

[0155] In this embodiment, the step of optionally correcting according to the non-reciprocal bias specifically includes:

[0156] Step S601, obtaining the current elevation angle of the terminal and the meteorological proxy data representing the propagation path conditions. The terminal calculates the apparent elevation angle e _handle of the current satellite relative to the local through its positioning module and geo _est At the same time, the terminal also obtains a set of meteorological proxy data meteo _proxy . There are multiple implementation levels for meteorological proxy data:

[0157] Basic implementation: Can be the parameters output by the standard atmospheric model based on geographic location, season and time query, such as ITU-RP.835-6 model, for example, total water vapor content.

[0158] Enhanced implementation: If the terminal is equipped with basic sensors (such as temperature, humidity, air pressure sensors), meteo _proxy These measured ground meteorological parameters can be included.

[0159] Preferred implementation: If the terminal has network connection capability, it can obtain more accurate and real-time local weather data such as troposphere zenith delay (ZTD) through Internet API. These data collectively constitute the basis for evaluating atmospheric effects.

[0160] Step S602, based on the elevation and meteorological proxy data, the current non-reciprocal bias update is estimated and calculated. Based on the above input, the system estimates the current non-reciprocal bias through the preset non-reciprocal bias model f(elevation _est , meteo _proxy ). The ionospheric delay is inversely proportional to the square of the signal frequency, and the tropospheric delay is related to the mapping function of the elevation angle. Therefore, a simplified mathematical model or an empirical lookup table of these physical effects can be established in advance. For example, the model f can be specified as: f = f _iono (elevation _est ) + f _tropo (elevation _est , meteo _proxy ). Wherein, f _iono is an ionospheric non-reciprocal delay model related only to the elevation angle (because the uplink and downlink frequency bands are known, the delay difference can be modeled), and f _tropo is a tropospheric non-reciprocal delay model related to the elevation angle and the ground water vapor pressure (from meteo _proxy ). The output is the instantaneous estimate of the bias under the current environment.

[0161] Step S603, using a recursive update algorithm, the estimated bias update is fused into the existing non-reciprocal bias estimate to continuously track the slow changes of the bias. Because the atmospheric parameters change slowly, and the model itself has errors, directly using the instantaneous estimate value may introduce noise. Therefore, the embodiment adopts a low-pass filtering method for recursive update to achieve smooth and stable tracking. The update formula is:

[0162] bias _nonrecip_k = (1-β) * bias _nonrecip_(k-1) + β * f(elevation _est_k , meteo _proxy_k ). Wherein, bias_nonrecip_k is the updated non-reciprocal bias; bias _nonrecip_(k-1) is the bias value of the last period; β is a filter coefficient with a value much smaller than 1 (for example, 0.001), which determines the bandwidth of the tracking loop. The smaller the value of β, the slower the tracking speed, but the better the smoothing effect on the model noise.

[0163] Step S604, the updated non-reciprocal bias is used to correct the transmission time. This step is the same as S106 in Embodiment One, that is, when calculating the final transmission time, subtract the bias updated by the slow loop tracking _nonrecip_k value: t _tx_corrected = t _tx -bias _nonrecip_k Through the above steps, this embodiment can effectively model and track compensation for key slow-varying system error sources without additional hardware costs, further improving the accuracy of time synchronization.

[0164] Embodiment Seven: The present embodiment provides a gating method for adaptive contraction of the protection interval, which is a preferred implementation of step S405 in Embodiment Four for adaptively configuring the protection interval of the return link. Embodiment Four proposes to dynamically contract or expand the protection interval according to the channel, and this embodiment sets up a set of strict multi-index consistency gating rules for the contraction operation. Only when all internal state indicators of the system show health and stability, can the operation of compressing the protection interval, which may bring risks, be performed, avoiding the decline of communication quality due to wrong decisions.

[0165] In this embodiment, the step of adaptively configuring the protection interval of the return link specifically includes:

[0166] Step S701, multi-index consistency gating judgment is performed. Before attempting to contract the protection interval GP _config in each decision period, the system must pass the gating judgment. The passing conditions of the judgment are that the following multiple sub-conditions must be met simultaneously:

[0167] (i) The alignment error of the return link satisfies the preset threshold: that is, ε _align <= ε _th . ε _align is the performance output of the system, and its value must have converged in a sufficiently small range (for example, ε _th = 50 nanoseconds), indicating that the main synchronization loop is working properly.

[0168] (ii) The measurement confidence associated with the alignment error satisfies the preset threshold: that is, align _conf >= conf _th . The network side must have ε _alignThe measurement itself also has a confidence level _conf , which must be high enough (e.g. _th =95%) to ensure that the ε _align data is reliable, not noise or outliers.

[0169] (iii) The internal status flag of the processing delay estimation is normal: i.e. delay _bound_flag =0 from embodiment three. This flag indicates that the processing delay estimation loop is working stably, without any abnormal situation such as out-of-bound or saturation clipping. If the loop itself is in degraded state, its output should not be trusted, let alone aggressive shrinking operation.

[0170] (iv) The confidence level of the single-pass propagation delay prediction meets the preset condition: i.e. quality _tag does not contain serious alarms, and / or τ _path_pred_conf >conf _geo_th . It is ensured that the geometric prediction used for calculating the transmission time is of high quality. For example, if the current satellite elevation angle is too low or the orbit data is extrapolated, even if the alignment error temporarily meets the requirements, it is not allowed to shrink the protection interval due to potential risks.

[0171] Step S702, the shrinking operation of the protection interval is performed only when the gating decision passes. If all the above (i) to (iv) conditions are met, the gating decision passes, and the system performs the shrinking operation of the protection interval. One preferred shrinking method is stepwise shrinking: GP _config_k =max(GP _min ,GP _config_(k-1) -Δ _GP ). Wherein, GP _min is the minimum protection interval allowed by the system, and Δ _GP is the preset shrinking step. Gradual shrinking can avoid oscillation caused by too large single adjustment.

[0172] Step S703, if the gating decision does not pass, the protection interval is rolled back to the conservative value. In any decision cycle, as long as any of the above gating decision sub-conditions is not met, the gating is closed. At this time, the system will immediately take the conservative strategy and reset the protection interval GP _config to the conservative value GP _static . GP _static is the maximum protection interval calculated according to the beam coverage, which can ensure that no conflict occurs. Through the strict gating mechanism, the embodiment pursues the spectral efficiency while ensuring the robustness and reliability of the entire adaptive process.

[0173] Embodiment eight describes an alternative or parallel propagation delay prediction refinement scheme. While embodiment four predicts intra-frame delay variation through first order (linear) Doppler compensation, this embodiment adopts higher order interpolation methods to more accurately fit the non-linear motion trajectory of the satellite within a single frame period, further reducing quantization and model errors of geometric prediction.

[0174] In this embodiment, the step of predicting the one-way propagation delay between the terminal and the satellite using the geometric prediction handle specifically includes:

[0175] Step S801, within the period corresponding to the current network frame, a series of geometric distances at sub-time points are obtained by using the geometric prediction handle for multi-point sampling according to a preset intra-frame sub-time table. Assuming that the starting time of the current frame is T _start , and the frame period is T _frame , the system will call geo _handle multiple times within the time interval [T _start , T _start + T _frame ] of the frame according to the preset sub-time table. For example, a simple three-point sampling table can be to sample at T _start , T _start + 0.5*T _frame , and T _start + T _frame , to obtain three corresponding satellite-geometric distances d _0 , d _1 , and d _2 .

[0176] Step S802, the density of multi-point sampling is increased within the sub-time interval near the beam switching event. As a preferred adaptive implementation, the sampling density is not fixed. When the system predicts that beam switching will occur in the next frame or several frames according to beam _plan , the accuracy of the geometric position becomes crucial. At this time, the system will automatically increase the sampling points. For example, only 3 points may be sampled in a regular frame, but in the key frame before switching, the number of sampling points may be increased to 5 or 7 to capture the trajectory details of the switching edge with higher time resolution.

[0177] Step S803, a spline interpolation algorithm is applied to the obtained series of geometric distance sampling points to generate a refined one-way propagation delay prediction with higher time resolution. After obtaining the sampling point pairs of (time, distance), the system applies a spline interpolation algorithm (such as cubic spline interpolation) to fit a smooth curve function d(t) that continuously describes the change of distance with time within the frame. The effective domain of the function d(t) is t in [T _start , T _start + T_frame ] interval. The one-way propagation delay prediction τ _path_pred is no longer a single value, but a time function τ(t) = d(t) / c obtained by dividing the function d(t) by the speed of light c. When calculating t _tx , it is necessary to solve the equation t _tx = t _term_local - τ(t _tx + τ(t _tx )) or to approximate the solution using numerical methods such as iteration. The present embodiment can improve the accuracy of the propagation delay prediction from the frame level to the intra-frame continuous level at a relatively small computational cost, effectively suppress the prediction error introduced by the nonlinear motion of the satellite, and provide more accurate input for the entire synchronization system.

[0178] In some embodiments, the steps of local time synthesis, transmission time calculation and alignment verification can also be:

[0179] Synthesizing the terminal local time t _term_local , specifically, using t _NCR_rx and delay _stat to perform t _term_local = t _NCR_rx + delay _stat , and adding frame _index to ensure timing consistency.

[0180] Calculating the transmission time t _tx , specifically, using t _term_local and τ _path_pred to perform t _tx = t _term_local - τ _path_pred , and recording the numerical precision rounding _rule (e.g., nanosecond-level rounding).

[0181] Non-reciprocal correction, specifically, calculating the transmission time correction t _nonrecip = t _tx_corrected - bias _tx if bias _nonrecip exists; otherwise t _tx_corrected = t _tx .

[0182] Alignment verification and guard interval configuration, specifically, scheduling the return burst according to t _tx_corrected , and measuring the alignment error ε _align at the network side; if ε _align is less than the target threshold, using the conservative guard interval GP _static as the guard interval configuration GP _config ; otherwise, recording the deviation and quality label quality _tagUpper limit of residual _bound Association, GP _config Maintain a conservative approach. Output ε _align GP _config This is the end point of the process; proceed to statistics.

[0183] Statistics and reports, specifically, summarizing ε _align With quality _tag geo _accuracy ,residual _bound Generate a synchronization state sync _state Used for operations and maintenance visualization. Output sync. _state Provided for external system calls.

[0184] Optionally, the steps of local time synthesis, transmission time calculation, and alignment verification can also be as follows:

[0185] At the terminal's local time t _term_local During synthesis, a variable fractional delay filter and its group delay quadratic compensation term are introduced to overcome the time resolution limitation of 1 / sampling rate:

[0186] t _term_local =t _NCR_rx +delay _stat +vfd(u _k )+Δ _group_delay (u _k (spectrum);

[0187] Where vfd(u _k ) is a fractional delay actuator, Δ _group_delay (u _k ,spectrum) is its frequency-dependent group delay compensation (provided by calibration tables or small-scale online updates).

[0188] Sending time t _tx Still press t _tx =t _term_local -τ _path_pred Calculate; if a non-reciprocal bias exists. _nonrecip t is still calculated _tx_corrected =t _tx -bias _nonrecip With t _tx_corrected The scheduler returns a burst, and ε is obtained. _align When ε _align If the threshold is met, the protection interval can be configured with GP. _config The adaptive approximation is based on the upper limit given in the protection interval configuration, and automatically reverts to a conservative value when environmental disturbances increase.

[0189] Output: t _term_local t_tx_corrected , ε _align , GP _config .

[0190] Example Nine, describes the calculation process through an example in a practical scenario.

[0191] Suppose a LEO satellite communication terminal working in Ka band (carrier frequency f _c = 30 GHz), the sampling rate of its digital baseband system Fs= 40 Msps (sampling period Ts= 25 ns). At a certain time, the terminal needs to perform a service switching according to the beam switching plan beam _plan , the symbol rate symbol _rate will be switched from 5.12 Msps (symbol _rate_prev ) to 9.1 Msps.

[0192] The terminal parses the network clock reference time t _NCR_rx = 12345.678910000 s from the forward broadcast signal at the beginning of the period.

[0193] Through geo _handle calculation, the current time pure geometric propagation delay τ _geometry = 0.012500100 s (12.5001 ms). At the same time, the slant range change rate range _rate = -1500 m / s (negative value means the distance is shortening).

[0194] 5.12 Msps corresponds to the processing delay of 23.966620 μs; 9.1 Msps corresponds to the processing delay of 22.252590 μs. Assuming that through offline calibration, the delay step J _step = -1.500000 μs is known from 5.12 Msps to 9.1 Msps.

[0195] Within the beam switching protection window, the terminal transmits a probe burst, and the network side measures and returns the aligned delay deviation e _meas_aligned = -0.000000050 s (-50 ns).

[0196] Set the radio frame period T _frame = 1 ms. The fast update weight of feedback correction α _fast = 0.5. The non-reciprocal bias bias _nonrecip is tracked by the method of example six, and the current value is +0.000000020 s (+20 ns).

[0197] Calculate the Doppler shift:

[0198] doppler _hz =-range _rate *f _c / c=-(-1500)*(30e9) / (3e8)=150,000Hz=150kHz.

[0199] Calculate the delay correction term Δ _τ_dopp =-doppler _hz *T _frame / f _c =-(150e3)*(1e-3) / (30e9)=-5e-9s=-5ns.

[0200] Calculate the final propagation delay prediction:

[0201] τ _path_pred =τ _geometry +Δ _τ_dopp =0.012500100s-0.000000005s=0.012500095s.

[0202] It can be seen that Doppler correction brings nanosecond-level correction to the prediction of propagation delay.

[0203] Obtain the static delay estimate at the new symbol rate (9.1 Msps). _stat (9.1) = 0.000022253s (22.253μs, an approximate value is taken for ease of calculation).

[0204] Constructing feedforward predictions:

[0205] delay _pred =delay _stat (9.1)+J _step =22.253μs+(-1.500μs)=20.753μs=0.000020753s.

[0206] Perform feedback correction, and use fast update weight α because it is in a switching window and the measurement is reliable. _fast =0.5.

[0207] Updated processing delay _stat_k =(1-α _fast )*delay _pred +α _fast *(delay _stat (5.12)+e _meas_aligned Note: e here _meas_aligned It is the deviation relative to the delay of the previous state, therefore the correction target should be the delay of the previous state. _stat(5.12). A more accurate recursion should be delay _stat_k =delay _pred_k +α _fast *(e _meas_aligned_k ). For simplicity, we take the form delay _stat_k =(1-α _k )*delay _pred_k +α _k *(delay _pred_k +e _meas_aligned_k ),

[0208] i.e. delay _stat_k =delay _pred_k +α _k *e _meas_aligned_k .

[0209] delay _stat_k =0.000020753s+0.5*(-0.000000050s)=0.000020753s-0.000000025s=0.000020728s.

[0210] After the two steps of feedforward + feedback, the processing delay is accurately estimated as 20.728 μs.

[0211] Calculate the baseline local time:

[0212] t _base =t _NCR_rx +delay _stat_k =12345.678910000s+0.000020728s=12345.678930728s.

[0213] Calculate the fractional delay control u _k . Calculate the relationship between the ideal adjustment amount and the sampling period:

[0214] The fractional part of the ideal sending time reference = (t _base -τ _path_pred ) mod Ts

[0215] = (12345.678930728 - 0.012500095) mod 25e-9 = 12345.666430633 mod 25e-9.

[0216] 12345.666430633 / 25e-9 = 493826657225.32, with a fractional part of 0.32.

[0217] Thus, the ideal time instant is ahead of the last sampling point clock period by 0.32 times. A delay of +0.32Ts is needed. The fractional delay control u _k = 0.32.

[0218] According to u _k = 0.32, the VFD filter and group delay compensation module are queried, and it is assumed that the following results are obtained:

[0219] The fractional delay adjustment vfd(u _k ) = u _k *Ts = 0.32*25ns = +8ns = +0.000000008s.

[0220] The group delay compensation Δ _group_delay = -1.5ns = -0.0000000015s.

[0221] The final local time is synthesized as:

[0222] t _term_local = t _base + vfd(u _k ) + Δ _group_delay = 12345.678930728s + 0.000000008s - 0.0000000015s = 12345.6789307345s.

[0223] The transmission time instant is calculated as:

[0224] t _tx = t _term_local - τ _path_pred = 12345.6789307345s - 0.012500095s = 12345.6664306395s.

[0225] The non-reciprocal bias is applied for correction,

[0226] t _tx_corrected = t _tx - bias _nonrecip = 12345.6664306395s - 0.000000020s = 12345.6664306195s.

[0227] The terminal will initiate transmission of the return link burst at t _tx_corrected = 12345.6664306195s. The calculation of this time instant takes into account the millisecond-level propagation delay, the microsecond-level processing delay step and tracking, and the nanosecond-level Doppler effect, sub-sampling fine tuning and non-reciprocal bias.

[0228] The application is an adaptive estimation method combining feedforward prediction and feedback correction, i.e., a feedforward prediction value delay _step containing a step term J _pred is constructed in advance, and a high-weight fast feedback correction is performed on the measured deviation e _meas_aligned obtained by using a probe burst in a beam switching window period. Most of the step errors can be actively compensated, and the measured data is used to quickly converge the residual error, so that the synchronization loss time after switching is shortened from the hundreds of milliseconds in the traditional scheme to within a few frame periods, and the stability and service continuity of the system in the dynamic switching scene are improved. The transient delay step problem caused by symbol rate switching is solved.

[0229] The application refines the propagation delay prediction by introducing feedforward correction based on Doppler shift. The accurate quantization of the real-time change of the intra-frame path provides higher time resolution and accuracy for delay prediction. On this basis, combined with the multi-index consistency gating mechanism, the system can safely and adaptively shrink the guard period GP _config to closely match the real channel dynamics, convert the wasted spectrum resources of the traditional fixed interval into effective user data throughput, and improve the link efficiency and guard period resource utilization.

[0230] By applying the variable fractional delay (VFD) filtering and synchronously compensating the group delay scheme, the system is given the ability to fine-tune the time between integer sampling clock points. This scheme breaks through the quantization accuracy bottleneck determined by the sampling period Ts, so that the alignment error ε _align can be stably controlled in the order of microseconds or even sub-microseconds. Not only is it a direct manifestation of synchronization performance, but it is also the basis for reliable implementation of the aforementioned adaptive guard period and other advanced functions, improving the time synchronization accuracy of the entire system.

[0231] In the propagation delay prediction, the time domain correction term mapped by the Doppler shift is superimposed for feedforward correction; in the delay estimation, the feedforward prediction value of symbol rate switching is constructed, and feedback correction is performed according to the delay deviation measured by the probe signal; in the local time synthesis, variable fractional delay filtering is applied and the group delay introduced by it is synchronously compensated, achieving fine adjustment at the sub-sampling level.

[0232] The preferred embodiments of the application are described in detail above, but the application is not limited to the specific details in the above embodiments. Within the technical concept of the application, various equivalent transformations can be made to the technical solutions of the application, and these equivalent transformations all belong to the protection scope of the application.

Claims

1. A time synchronization method suitable for satellite communication systems, characterized in that, include: Analyze the forward broadcast signal to obtain the network clock reference time; A geometric prediction handle is constructed based on satellite orbit data; Using a geometric prediction handle, predict the one-way propagation delay between the terminal and the satellite; Based on the terminal's symbol rate and equipment operating conditions, and referring to delay calibration data, the terminal's processing delay is estimated; The network clock reference time and processing latency estimate are combined to synthesize the terminal's local time; Based on the terminal's local time and one-way propagation delay prediction, and corrected according to the non-reciprocal bias, the transmission time of the return link is calculated.

2. The method according to claim 1, characterized in that, By fusing network clock reference time and processing latency estimation, a composite terminal local time is obtained, including: Integrate network clock reference time with processing latency estimation to generate a baseline local time; Based on the one-way propagation delay prediction, a fractional delay control amount is determined for the baseline local time; The fractional delay control quantity is processed by a variable fractional delay filter to generate a fractional delay adjustment quantity; The baseline local time is corrected using a fractional delay adjustment, and a terminal local time with subsampling accuracy is synthesized.

3. The method according to claim 2, characterized in that, Synthesized terminal local time also includes: Determine whether the frequency correlation error introduced by the variable fractional delay filter is greater than the threshold. If so, refer to the fractional delay control quantity and the spectral characteristics of the signal, and retrieve the group delay compensation quantity from the group delay calibration data; The fractional delay adjustment and the group delay compensation are applied together to the baseline local time to synthesize the terminal local time.

4. The method according to claim 3, characterized in that, The variable fractional delay filtering process and the retrieval process of group delay compensation are configured in mutually bound processing modules; When the real-time performance or internal parameters of the processing module do not meet the preset conditions, the processing module is disabled as a whole, and the synthesized terminal local time is reverted to the baseline local time.

5. The method according to claim 1, characterized in that, Based on the terminal's symbol rate and equipment operating conditions, and referring to delay calibration data, the terminal's processing delay is estimated, including: For the switching event of terminal symbol rate, construct a feedforward prediction value for processing delay; Obtain the delay deviation measured from the probe signal; Based on the measured delay deviation, feedback correction is applied to the feedforward prediction value to determine the processing delay estimate.

6. The method according to claim 5, characterized in that, Obtain the delay deviation measured from the probe signal, including: The beam switching protection window is identified based on the beam switching schedule, and the terminal is instructed to transmit a dedicated detection burst during this window. The network side performs correlation estimation on the received probe bursts to obtain the initial measured delay deviation and the corresponding measurement confidence level; The initial measured delay deviation is timestamped using the frame sequence number to eliminate cross-frame measurement errors and generate an aligned measured delay deviation for feedback correction.

7. The method according to claim 5, characterized in that, Based on the measured delay deviation, feedback correction is applied to the feedforward prediction value to determine the processing delay estimate, including: The updated weights are dynamically determined based on whether the beam switching protection window is currently open and the measurement confidence level. Among them, a higher first weight value is configured for updating the weight only when the measurement confidence is within the window and not lower than the preset threshold, and a lower second weight value is configured under other conditions. Using determined update weights, the feedforward predicted values ​​and the aligned measured delay deviations are weighted and fused to determine the processing delay estimate.

8. The method according to claim 1, characterized in that, Using a geometric prediction handle, predict the one-way propagation delay between the terminal and the satellite, including: Using the geometric prediction handle, the geometric propagation delay based on geometric distance is calculated; Using the geometric prediction handle, the relative velocity between the terminal and the satellite is obtained, and the Doppler frequency shift is calculated accordingly, which is then mapped into a time delay correction term in the time domain. The delay correction term is superimposed on the geometric propagation delay to generate a one-way propagation delay prediction that reflects real-time dynamics.

9. The method according to claim 8, characterized in that, The method also includes: Continuously monitor the Doppler frequency shift or the slant range change rate derived from the relative velocity; Evaluate the dynamic statistical characteristics of Doppler frequency shift or slant range change rate within a preset time window; Based on dynamic statistical characteristics, the protection interval of the return link is adaptively configured. When the dynamic statistical characteristics do not exceed the threshold, the protection interval is reduced; otherwise, the protection interval is expanded to a conservative value.

10. The method according to claim 1, characterized in that: Predicting the one-way propagation delay between the terminal and the satellite includes: superimposing the time-domain correction term mapped by the Doppler frequency shift onto the geometric propagation delay for feedforward correction; Estimating the processing delay of the terminal includes: constructing a feedforward prediction value for symbol rate switching and performing feedback correction on the feedforward prediction value based on the delay deviation measured by the probe signal; Synthesize the local time of the terminal, including: applying variable fractional delay filtering to achieve fine adjustment at the subsampling level, and synchronously compensating for the group delay introduced by the filtering.

Images