A timekeeping device and method for automatic switching of multiple master clocks

Abstract

The present invention provides a timekeeping device and method for automatic switching of multiple master clocks. Relying on seamless switching equipment, the reference of the physical signal output by the timekeeping system can be selected between multiple atomic clocks, which expands the number of candidate master clocks while saving the use of phase trimmers. At the same time, a master clock switching strategy is provided to reduce the impact of random noise fluctuations on the system time signal of a single atomic clock. The performance of the master clock of the present invention is no longer determined by a single atomic clock, and it can be flexibly selected from multiple atomic clock frequency signals; the master clock switching process is different from the previous switching method. The switching is performed before the master clock signal undergoes frequency control. It is a switch on the frequency source of the timekeeping system, which can reduce the number of phase trimmers used; at the same time, the master clock selection strategy provided by the method also ensures that the system time will not have a large phase deviation due to the noise fluctuation of a single clock.

Description

Technical Field

[0001] The present invention relates to the field of timekeeping, in particular to a timekeeping method. Background Art

[0002] Time is the basis of modern society and is indispensable to human daily life. With the development of science and technology, the accuracy of timekeeping is also increasing. In order to generate and maintain a stable time scale, most timekeeping laboratories in the world currently use multiple atomic clocks to build atomic clock groups, and use internal measurement and analysis of the clock group to weaken the impact of single clock noise on the time scale.

[0003] In simple terms, timekeeping is to use atomic clocks to reproduce the time reference and output real-time physical signals. The atomic clock that provides a frequency reference for the physical signal is called the master clock. In the timekeeping system, the atomic clock with the most stable performance and the best predictability in the atomic clock group is usually selected as the master clock, and the frequency and phase deviation of the master clock are adjusted using a phase fine-tuner. However, in any case, signal interruption or performance degradation will occur due to problems with the atomic clock itself, and the continuous stability of the output physical signal cannot be guaranteed. To solve this problem, most timekeeping laboratories currently usually select two master clocks and two phase fine-tuners to form the output of the main and standby signals, keep the phase and frequency of the two signals consistent, and when the main signal fails or recovers, the system time signal is manually controlled by switching the switch to quickly switch between the main and standby signals to ensure the reliability of the system time. However, for some miniaturized timekeeping systems, the number of atomic clocks involved in timekeeping is small, and in order to save costs, it may not be possible to build two relatively independent physical signal output means. On the other hand, if the performance of each atomic clock in the clock group participating in timekeeping is not much different, the number of master clocks can be further expanded to improve the reliability of local physical signals. However, the current switching of the main and standby master clocks is to switch the signals of the atomic clocks after being controlled by the phase fine tuner, which is switched at the output end of the timekeeping system signal. Therefore, each master clock requires a phase fine tuner, which increases the demand for phase fine tuners. Therefore, it is necessary to improve the existing master clock frequency control method. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the present invention provides a timekeeping device and method for automatic switching of multiple master clocks. According to the frequency control principle of atomic clocks, the reference of the physical signal output by the timekeeping system can be selected among multiple atomic clocks by relying on seamless switching equipment, which expands the number of alternative master clocks while saving the use of phase fine-tuners. At the same time, a master clock switching strategy is provided to reduce the impact of random noise fluctuations of a single atomic clock on the system time signal.

[0005] The technical solution adopted by the present invention to solve its technical problem is:

[0006] A timekeeping device with automatic switching of multiple master clocks, comprising an atomic clock group unit, a measurement and comparison unit, a master clock signal generating unit and a frequency control unit;

[0007] The atomic clock group unit provides multiple master clock frequency signals;

[0008] The measurement and comparison unit completes the measurement of clock difference and phase difference data between each atomic clock signal, remote traceability signal and timekeeping system time signal through a counter and a phase comparator;

[0009] The master clock signal generating unit adopts a seamless switching device, taking the input multiple master clock frequency signals as references, to achieve seamless switching of the output signal between multiple frequency source references, and at the same time, based on the digital frequency synthesis technology, to achieve phase and frequency adjustment of the output signal;

[0010] The frequency control unit processes and analyzes the comparison data obtained by the measurement and comparison unit. First, a local time reference is obtained based on the atomic time algorithm using the local comparison data; then the local time reference is corrected using the remote traceability data; finally, the frequency control amount of each master clock relative to the corrected local time reference is calculated by forecasting, and the frequency of the signal output by the current seamless switching device is controlled by a phase fine tuner to generate a stable system time signal, and the frequency control amount is updated after the master clock is switched to ensure that the system time does not experience frequency jumps after the switch.

[0011] In the atomic clock group unit, the number of atomic clocks is greater than three, and the atomic clocks are compared in pairs to determine the noise fluctuation of the master clock.

[0012] The lossless switching device includes a data sampling unit, a frequency fusion unit and a DDS frequency module; each master clock frequency signal is input into a data sampling unit respectively, and phase detection is performed using down-conversion technology to measure the frequency difference and phase difference of the atomic clock relative to the local clock; the output signal of the data sampling unit is input into the frequency fusion unit, and the selected atomic clock 10MHz frequency is output through the frequency fusion unit and the DDS frequency module based on the local clock; the frequency control method of its output signal is the same as that of the phase fine tuner, and the frequency and phase of the output signal are directly adjusted when the master clock is switched.

[0013] The atomic time algorithm adopts Kalman algorithm, ALOGS algorithm or ATI1 algorithm.

[0014] The present invention also provides a timekeeping method using a timekeeping device with automatic switching of multiple master clocks, comprising the following steps:

[0015] 1) The local 10MHz signal output by the phase trimmer is used as the frequency reference of the multi-channel time interval counter, and the output 1PPS signal is used as the door opening signal. The 1PPS signal of the atomic clock and the 1PPS signal of the external traceability are used as the door closing signals respectively. The host computer software in the industrial computer realizes the collection of comparison data, which is recorded as MC-clock i and MC-UTC k , MC refers to the 1PPS second signal output by the phase trimmer, clock i and UTC k They represent the atomic clock i and the remotely traced 1PPS second signal respectively; the externally traced 1PPS signal is directly given by UTC(k) as a physical signal. If there is no direct signal, it is necessary to obtain the comparison data with the external traceability through satellite common view, two-way satellite time transmission or other methods, and store it in the industrial computer;

[0016] After obtaining the original comparison data, in order to reduce the impact of accidental errors in measurement, the MC-clock i The gross errors in the data are eliminated, and the eliminated parts are interpolated using linear interpolation;

[0017] 2) Calculation of integrated atomic time;

[0018] The calculation of the comprehensive atomic time refers to the ALGOS algorithm used by BIPM in calculating UTC. The basic core idea is to weight the atomic clocks involved in the calculation according to their predictability, and then perform weighted average calculation on the comparison data of the atomic clocks; the calculation flow chart is shown in the figure below: Figure 3 As shown, the steps are as follows:

[0019] 2.1) First, read the comparison data from the local file according to the set comprehensive atomic time calculation frequency and the length of historical data used;

[0020] The calculation of comprehensive atomic time is to ensure the local time retention capability after the traceability is interrupted. The setting of calculation frequency is related to the traceability link used. If the traceability adopts direct connection, common view or two-way time comparison, the traceability update frequency is faster, and the calculation frequency of atomic time can be calculated at a faster frequency. If the UTC or UTC(r) of the BIPM bulletin is used for traceability, the calculation frequency of atomic time should not be too fast; direct connection physical signal is used for traceability, so the calculation frequency of comprehensive atomic time is set to 1 hour; using too long historical data will cause the calculation of atomic time to fail to accurately reflect the recent fluctuations of the atomic clock, and too short will cause the stability of atomic time to decrease. The present invention selects 36 historical data;

[0021] 2.2) Atomic time calculation;

[0022] The weight of the atomic clock depends on the predictability of the atomic clock, that is, the degree to which the atomic clock conforms to a model. The present invention uses a first-order polynomial as the calculation model:

[0023] y=a+b*t

[0024] Get 36 MC-clock i After comparing the data in hours, the residual is obtained by deducting the model value of the atomic clock at that moment from each data. In the initial calculation, a and b are 0, and the atomic clock weight obtained in the last calculation is used to weight the residual to get the atomic time MC-TA. The last MC-TA is saved in a local file as the atomic time for this calculation. In the initial calculation, the weight of each clock is 1 / N, where N is the number of atomic clocks involved in the calculation, and N is 3.

[0025] By MC-clock i Get 36 TA-clocks for each clock with MC-TA i , use the least squares method to update the a and b values ​​of each clock, and use the TA-clock i Subtract the new model value and calculate the weight of each clock according to the sum of squares of the residuals. The weight calculation formula is as follows:

[0026]

[0027] The weight calculation formula needs to be iterated to ensure the stability of the calculation. The present invention iterates 3 times. At the same time, an upper limit is set for the weight of the atomic clock to prevent a single atomic clock from being completely dependent on the operation of each atomic clock, which affects the reliability of the atomic clock. The upper limit of the weight of the present invention is 2 / N. For atomic clocks that exceed the upper limit, the excess weight part needs to be redistributed to other atomic clocks in proportion. At the same time, when the weight of a certain atomic clock is too low, it means that the predictability of the atomic clock is extremely poor, and the weight needs to be set to 0 in this calculation. The lower limit of the weight of the present invention is 1%;

[0028] 2.3) Atomic time frequency calibration;

[0029] The MC-TA calculated in step 2.2) is weighted by the local atomic clock comparison data only. The atomic time needs to be relatively consistent with the external traceability source, and the atomic time needs to be calibrated in phase and frequency.

[0030] By MC-UTC k Get TA-UTC with MC-TA k , the number of data is selected as 24, and the least square method is used to calculate the 24 TA-UTC kPerform a polynomial modeling y=a+b*t, and deduct the fitting value from the MC-TA to obtain the more frequent MC-TA. Each newly generated MC-TA needs to deduct the fitting value y. The update frequency of the a and b values ​​in the model is related to the update frequency of the external traceability and the atomic time calculation frequency. In the present invention, it is updated once every hour.

[0031] After each atomic time phase and frequency calibration is completed, MC-TA and MC-clock i Get TA-clock i , stored in a local file;

[0032] 3) Calculation of master clock frequency control quantity;

[0033] The frequency control amount of the master clock is related to the latest MC-TA phase difference, the frequency deviation of the current master clock relative to TA and the control frequency. The control frequency of the master clock should be greater than the update frequency of the external traceability and atomic time. The control frequency of the master clock of the present invention is 1 hour, which is consistent with the calculation frequency of atomic time. After determining the current master clock, the calculation formula of the control amount offset is:

[0034]

[0035] TA-clock m is the phase difference between the most recently obtained atomic time and the master clock, k is the adjustment coefficient, and is an empirical constant. In the present invention, 2 is used. m is the frequency deviation of the master clock relative to atomic time, for the 24 most recent TA-clock m The least square method is used to perform a polynomial modeling, and the number of data items can be adjusted according to actual needs;

[0036] 4) Switching control of the master clock;

[0037] The switching judgment of the master clock needs to be made according to the performance status of the current master clock. When the master clock has abnormal behaviors such as phase hopping or frequency hopping, the master clock signal needs to be switched in time to ensure stable local physical signal output. The present invention adopts the dynamic Allan variance method to complete the abnormal judgment of the atomic clock, and the implementation steps are as follows:

[0038] 4.1) For the MC-clock obtained every second i , get the clock by pairwise comparison i -clock j Calculate, compared to the clock 100s ago i -clock j The difference between the last 50 values ​​is counted to calculate the clock i -clock jWhen the 100s stability calculated at the current moment exceeds twice the nominal 100s stability of the atomic clock in use, it is considered that one of the atomic clocks i and j is abnormal. After the three sets of comparison values ​​are judged, the abnormal atomic clock can be determined; the average time of stability and the magnification of judgment are determined according to the actual performance of the atomic clock;

[0039] 4.2) When an abnormality occurs in the master clock, the master clock is switched to an atomic clock that has not experienced an abnormality. In the present invention, there are only three atomic clocks. If two atomic clocks have abnormalities at the same time, it is impossible to determine the specific atomic clock, and the master clock is not switched at this time;

[0040] 4.3) When the master clock is switched, the frequency control value of the phase trimmer must be corrected in time according to the control value calculation formula in step 3).

[0041] The beneficial effect of the present invention is that the performance of the master clock is no longer determined by a single atomic clock, and can be flexibly selected from multiple atomic clock frequency signals; the master clock switching process is different from the previous switching method, and the switching is performed before the master clock signal undergoes frequency control, which is a switch on the frequency source of the timekeeping system, and can reduce the number of phase fine-tuning instruments used; at the same time, the master clock selection strategy provided by this method also ensures that the system time will not have a large phase deviation due to the noise fluctuation of a single clock. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is the principle block diagram of lossless switching device;

[0043] Figure 2 It is a schematic diagram of an embodiment of a multi-master clock timekeeping method and equipment connection;

[0044] Figure 3 It is a comprehensive atomic time calculation flow chart. DETAILED DESCRIPTION

[0045] The present invention is further described below in conjunction with the accompanying drawings and embodiments. The present invention includes but is not limited to the following embodiments.

[0046] The first problem to be solved by the present invention is the switching problem of frequency source references between different atomic clocks. The solution to this problem is as follows: Figure 1As shown, the function of switching the master clock signal between multiple atomic clock signals is completed by relying on the lossless switching device. The lossless switching device uses down-conversion technology to perform high-precision phase detection on the input of multiple external atomic clock frequency signals, and can quickly measure the frequency difference and phase difference of multiple atomic clocks relative to the local clock in real time. Further, the local clock is used as the clock reference, and the selected atomic clock 10MHz frequency is output through frequency fusion and DDS control. The frequency control method of its output signal is the same as that of the phase trimmer. When using this method to switch the master clock, the frequency and phase of the output signal can also be directly adjusted. The adjustment principle is consistent with the common phase trimmer, thereby playing the role of replacing the phase trimmer and saving hardware costs.

[0047] The second problem that the present invention needs to solve is the calculation of the driving amount of the current master clock. Since the master clock is no longer determined by a single atomic clock, in order to maintain the consistency of the system time and the time reference during the master clock driving process, it is necessary to predict the rate of each atomic clock participating in the lossless switching of the master clock, and update the corresponding frequency control amount in real time according to the currently selected master clock to ensure that the output signal is always consistent with the reference frequency.

[0048] The third problem that the present invention needs to solve is when to switch the master clock. In a general timekeeping system, the frequency control of the master clock is divided into two parts, one is to correct the relative frequency deviation of the master clock relative to the reference time scale, and the other is to correct the phase deviation caused by the noise fluctuation of the master clock to ensure the absolute synchronization of the system time and the reference signal. Therefore, the selection of the master clock should select an atomic clock with small noise fluctuation as the master clock to avoid a large phase deviation between the system time and the reference signal within the control frequency of the master clock, so it is necessary to monitor the noise fluctuation of the current master clock in real time. Usually, the noise of the atomic clock is characterized by frequency stability, that is, statistical estimation is performed through Allan variance. When the atomic clock leaves the factory, different models of atomic clocks will provide corresponding frequency stability nominal values. When performing noise monitoring of the atomic clock, it can be judged whether the atomic clock has generated a large noise fluctuation according to the nominal value or the frequency stability index requirements of actual application, so as to judge whether to switch the current master clock.

[0049] The present invention firstly ensures that the timekeeping system can output a stable system time, and secondly ensures the switching of multiple master clocks under the premise of signal stability. It is necessary to be able to accurately calculate the frequency deviation of each atomic clock relative to the reference signal during switching, and correct the master clock control amount in real time. At the same time, it is necessary to monitor the phase fluctuation of the current master clock. When a large phase fluctuation occurs, the current master clock is switched. In order to achieve these three purposes, the present invention needs to include an atomic clock group unit, a measurement and comparison unit, a master clock seamless switching unit, and a frequency control unit.

[0050] The atomic clock group unit provides multiple stable master clock signals for the timekeeping system to ensure the reliability of the timekeeping system. The noise fluctuation of the master clock needs to be judged by comparing the data of the atomic clocks one by one, so for the timekeeping system, the number of atomic clocks should be greater than three.

[0051] The measurement and comparison unit uses counters and phase comparators to complete the clock difference and phase difference data measurements between the atomic clock signals, remote traceability signals and timekeeping system time signals, providing data support for subsequent atomic clock master clock switching judgments, atomic time calculations and master clock frequency deviation control quantity calculations.

[0052] The master clock signal generating unit relies on the existing frequency source seamless switching equipment, and uses the input multiple master clock frequency signals as reference to perform phase-locked control on the voltage-controlled oscillator to achieve seamless switching of its output signal between multiple frequency source references. At the same time, based on digital frequency synthesis technology, the phase and frequency adjustment of the output signal is achieved.

[0053] The frequency control unit processes and analyzes the comparison data obtained by the data comparison unit. First, a local time reference is obtained based on the atomic time algorithm (Kalman algorithm, ALOGS algorithm, ATI1 algorithm, etc.) using the local comparison data. Then, the local time reference is corrected using the remote traceability data. The purpose of calculating the local time reference is to provide autonomous timekeeping capability after the external traceability link is interrupted. Finally, the frequency control amount of each master clock relative to the corrected local time reference is calculated by forecasting, and the frequency of the signal output by the current seamless switching device is controlled by the phase fine-tuning instrument to generate a stable system time signal. The frequency control amount is updated in time after the master clock is switched to ensure that the system time does not have a frequency jump after the switch.

[0054] The hardware of this embodiment includes three cesium atomic clocks, seamless switching equipment, phase trimmer, multi-channel time interval counter, multi-channel phase comparator and industrial computer. During the design process, the type and number of atomic clocks and the number of acquisition devices need to be adjusted according to actual application requirements. The specific device connection method is as follows: Figure 2 shown.

[0055] A timekeeping device with automatic switching of multiple master clocks, comprising an atomic clock group unit, a measurement and comparison unit, a master clock signal generating unit and a frequency control unit;

[0056] The atomic clock group unit provides multiple master clock frequency signals;

[0057] The measurement and comparison unit completes the measurement of clock difference and phase difference data between each atomic clock signal, remote traceability signal and timekeeping system time signal through a counter and a phase comparator;

[0058] The master clock signal generating unit adopts a seamless switching device, taking the input multiple master clock frequency signals as references, to achieve seamless switching of the output signal between multiple frequency source references, and at the same time, based on the digital frequency synthesis technology, to achieve phase and frequency adjustment of the output signal;

[0059] The frequency control unit processes and analyzes the comparison data obtained by the measurement and comparison unit. First, a local time reference is obtained based on the atomic time algorithm using the local comparison data; then the local time reference is corrected using the remote traceability data; finally, the frequency control amount of each master clock relative to the corrected local time reference is calculated by forecasting, and the frequency of the signal output by the current seamless switching device is controlled by a phase fine tuner to generate a stable system time signal, and the frequency control amount is updated after the master clock is switched to ensure that the system time does not experience frequency jumps after the switch.

[0060] In the atomic clock group unit, the number of atomic clocks is greater than three, and the atomic clocks are compared in pairs to determine the noise fluctuation of the master clock.

[0061] The lossless switching device includes a data sampling unit, a frequency fusion unit and a DDS frequency module; each master clock frequency signal is input into a data sampling unit respectively, and phase detection is performed using down-conversion technology to measure the frequency difference and phase difference of the atomic clock relative to the local clock; the output signal of the data sampling unit is input into the frequency fusion unit, and the selected atomic clock 10MHz frequency is output through the frequency fusion unit and the DDS frequency module based on the local clock; the frequency control method of its output signal is the same as that of the phase fine tuner, and the frequency and phase of the output signal are directly adjusted when the master clock is switched.

[0062] The frequency control part of this embodiment needs to complete four processes: sampling and preprocessing of time difference data, calculation of integrated atomic time, calculation of frequency control amount, and master clock switching control.

[0063] 1) Sampling and preprocessing of time difference data

[0064] The time difference data is collected through an industrial computer and a multi-channel time interval counter. The local 10MHz signal output by the phase trimmer is used as the frequency reference of the multi-channel time interval counter, and the output 1PPS signal is used as the door opening signal. The 1PPS signal of the atomic clock and the 1PPS signal of the external traceability are used as the door closing signals respectively. The host computer software in the industrial computer realizes the collection of comparison data. It is denoted as MC-clock i and MC-UTC k , MC refers to the 1PPS second signal output by the phase trimmer, clock i and UTC kThe 1PPS second signal for each atomic clock and remote traceability. The traceability 1PPS signal of this embodiment is directly given by UTC(k) as a physical signal. If there is no direct signal, it is necessary to obtain comparison data with external traceability through satellite common view, two-way satellite time transfer or other methods, and store it in the industrial computer.

[0065] After obtaining the original comparison data, in order to reduce the impact of accidental errors in measurement, it is necessary to use the 3-Sigma principle to calibrate the MC-clock i The gross errors in the data are eliminated, and linear interpolation is used to interpolate the eliminated parts.

[0066] 2) Calculation of integrated atomic time

[0067] The calculation of the comprehensive atomic time in this embodiment refers to the ALGOS algorithm used by BIPM when calculating UTC. The basic core idea is to weight the atomic clocks involved in the calculation according to their predictability, and then perform weighted average calculation on the comparison data of the atomic clocks. The calculation flow chart is as follows: Figure 3 As shown, the steps are as follows:

[0068] 2.1) First, read the comparison data from the local file according to the set comprehensive atomic time calculation frequency and the length of the historical data used.

[0069] The calculation of comprehensive atomic time is to ensure the ability to maintain local time after the traceability is interrupted. The setting of the calculation frequency is related to the traceability link used. If the traceability adopts direct connection, common view or two-way time comparison, the traceability update frequency is faster, and the calculation frequency of atomic time can be calculated at a faster frequency. If the UTC or UTC (r) of the BIPM bulletin is used for traceability, the calculation frequency of atomic time should not be too fast. This embodiment uses a direct physical signal for traceability, so the calculation frequency of comprehensive atomic time is set to 1 hour. Using historical data that is too long will cause the calculation of atomic time to fail to accurately reflect the recent fluctuations of the atomic clock. If it is too short, it will cause the stability of atomic time to decrease. This embodiment selects 36 historical data.

[0070] 2.2) Atomic time calculation

[0071] The weight of an atomic clock depends on its predictability, that is, how well it conforms to a model. In this example, a first-order polynomial is used as the calculation model:

[0072] y=a+b*t

[0073] Get 36 MC-clock iAfter comparing the data in hours, the model value of the atomic clock at that moment is deducted from each data to obtain the residual. In the initial calculation, a and b are 0, and the atomic clock weight obtained in the last calculation is used to weighted average the residual to obtain the atomic time MC-TA, and the last MC-TA is saved in a local file as the atomic time for this calculation. In the initial calculation, the weight of each clock is 1 / N, where N is the number of atomic clocks involved in the calculation, and N is 3 in this embodiment.

[0074] By MC-clock i Get 36 TA-clocks for each clock with MC-TA i The a and b values ​​of each clock are updated using the least squares method and are obtained from the TA-clock i Subtract the new model value and calculate the weight of each clock according to the sum of squares of the residuals. The weight calculation formula is as follows:

[0075]

[0076] The above calculations usually need to be iterated to ensure the stability of the calculations. This embodiment iterates 3 times. At the same time, an upper limit needs to be set for the weight of the atomic clock to prevent a single atomic clock from being completely dependent on the operation of each atomic clock, which affects the reliability of the atomic clock. This embodiment uses a weight upper limit of 2 / N. For atomic clocks that exceed the upper limit, the excess weight needs to be redistributed to other atomic clocks in proportion. At the same time, when the weight of a certain atomic clock is too low, it means that the predictability of the atomic clock is extremely poor, and the weight needs to be set to 0 in this calculation. The lower limit of the weight in this embodiment is 1%.

[0077] 2.3) Atomic time frequency calibration;

[0078] The MC-TA calculated in step 2.2) is weighted only by the local atomic clock comparison data. The atomic time needs to be relatively consistent with the external traceability source, and the atomic time needs to be calibrated in phase and frequency.

[0079] By MC-UTC k Get TA-UTC with MC-TA k In this embodiment, 24 data items are selected, and the least square method is used to calculate the 24 TA-UTC k Perform a polynomial modeling y=a+b*t, and deduct the fitting value from MC-TA to obtain the more frequent MC-TA. Each newly generated MC-TA needs to deduct the fitting value y. The update frequency of the a and b values ​​in the model is related to the update frequency of the external traceability and the atomic time calculation frequency. In this embodiment, it is updated once every hour.

[0080] After each atomic time phase and frequency calibration is completed, MC-TA and MC-clock iGet TA-clock i , stored in a local file.

[0081] 3) Calculation of master clock frequency control quantity;

[0082] The frequency control value of the master clock is related to the latest MC-TA phase difference, the frequency deviation of the current master clock relative to TA, and the control frequency. The control frequency of the master clock should be greater than the update frequency of the external traceability and atomic time. The control frequency of the master clock in this embodiment is 1 hour, which is consistent with the calculation frequency of atomic time. After determining the current master clock, the calculation formula of the control value offset is:

[0083]

[0084] TA-clock m is the phase difference between the most recently obtained atomic time and the master clock, k is the adjustment coefficient, which is an empirical constant. In this embodiment, it is 2, and b m is the frequency deviation of the master clock relative to atomic time, for the 24 most recent TA-clock m The least square method is used to perform a polynomial modeling, and the number of data items can be adjusted according to actual needs.

[0085] 4) Switching control of the master clock;

[0086] The switching judgment of the master clock needs to be made according to the performance status of the current master clock. When the master clock has abnormal behaviors such as phase hopping or frequency hopping, the master clock signal needs to be switched in time to ensure stable local physical signal output. This embodiment uses the dynamic Allan variance method to complete the abnormal judgment of the atomic clock, and the implementation steps are as follows:

[0087] 4.1) For the MC-clock obtained every second i , get the clock by pairwise comparison i -clock j Calculate the difference between the clock value 100 seconds ago and i -clock j The difference between the last 50 values ​​is counted to calculate the clock i -clock j When the 100s stability calculated at the current moment exceeds twice the nominal 100s stability of the atomic clock used, it is considered that one of the atomic clocks i and j is abnormal. After the three sets of comparison values ​​are judged, the abnormal atomic clock can be determined. The average stability time and the magnification of the judgment can be determined according to the actual performance of the atomic clock.

[0088] 4.2) When an abnormality occurs in the master clock, the master clock can be switched to an atomic clock that has not experienced an abnormality. In this embodiment, there are only three atomic clocks. If two atomic clocks have abnormalities at the same time, it is impossible to determine the specific atomic clock, and the master clock will not be switched at this time.

[0089] 4.3) When the master clock is switched, the frequency control value of the phase trimmer must be corrected in time according to the control value calculation formula in 3).

Claims

1. A timekeeping device with automatic switching of multiple master clocks, comprising an atomic clock group unit, a measurement and comparison unit, a master clock signal generating unit and a frequency control unit; characterized in that: The atomic clock group unit provides multiple master clock frequency signals; The measurement and comparison unit completes the measurement of clock difference and phase difference data between each atomic clock signal, remote traceability signal and timekeeping system time signal through a counter and a phase comparator; The master clock signal generating unit adopts a seamless switching device, taking the input multiple master clock frequency signals as references, to achieve seamless switching of the output signal between multiple frequency source references, and at the same time, based on the digital frequency synthesis technology, to achieve phase and frequency adjustment of the output signal; The frequency control unit processes and analyzes the comparison data obtained by the measurement and comparison unit. First, a local time reference is obtained based on the atomic time algorithm using the local comparison data; then the local time reference is corrected using the remote traceability data; finally, the frequency control amount of each master clock relative to the corrected local time reference is calculated by forecasting, and the frequency of the signal output by the current seamless switching device is controlled by a phase fine tuner to generate a stable system time signal, and the frequency control amount is updated after the master clock is switched to ensure that the system time does not experience frequency jumps after the switch.

2. The timekeeping device with automatic switching of multiple master clocks according to claim 1, characterized in that: In the atomic clock group unit, the number of atomic clocks is greater than three, and the atomic clocks are compared in pairs to determine the noise fluctuation of the master clock.

3. The timekeeping device with automatic switching of multiple master clocks according to claim 1, characterized in that: The seamless switching device includes a data sampling unit, a frequency fusion unit and a DDS frequency module; each master clock frequency signal is input into a data sampling unit respectively, and a down-conversion technology is used for phase detection to measure the frequency difference and phase difference of the atomic clock relative to the local clock; The output signal of the data sampling unit is input into the frequency fusion unit, which uses the local clock as a reference and outputs the selected atomic clock 10MHz frequency through the frequency fusion unit and the DDS frequency module. The frequency control method of the output signal is the same as that of the phase fine-tuner, and the frequency and phase of the output signal are directly adjusted when the master clock is switched.

4. The timekeeping device with automatic switching of multiple master clocks according to claim 1, characterized in that: The atomic time algorithm adopts the Kalman algorithm.

5. A timekeeping method using the timekeeping device with automatic switching of multiple master clocks as claimed in claim 1, characterized in that The steps include: 1) The local 10MHz signal output by the phase trimmer is used as the frequency reference of the multi-channel time interval counter, and the output 1PPS signal is used as the door opening signal. The 1PPS signal of the atomic clock and the 1PPS signal of the external traceability are used as the door closing signals respectively. The host computer software in the industrial computer realizes the collection of comparison data, which are recorded as MC-clocki and MC-UTCk. MC refers to the 1PPS second signal output by the phase trimmer, clocki and UTCk represent the atomic clock i and the remote traceability 1PPS second signal respectively; the 1PPS signal of the external traceability is directly given by UTC (k) The physical signal. If there is no direct connection signal, it is necessary to obtain the comparison data with the external traceability through satellite common view, two-way satellite time transmission or other methods, and store it in the industrial computer; After obtaining the original comparison data, in order to reduce the influence of accidental errors in the measurement, the gross errors in the MC-clocki data were eliminated according to the 3-Sigma principle, and linear interpolation was used to interpolate the eliminated parts; 2) Calculation of integrated atomic time; The calculation of the comprehensive atomic time refers to the ALGOS algorithm used by BIPM when calculating UTC. The basic core idea is to weight the atomic clocks involved in the calculation according to their predictability, and then perform a weighted average calculation on the comparison data of the atomic clocks. 3) Calculation of master clock frequency control quantity; The frequency control value of the master clock is related to the latest MC-TA phase difference, the frequency deviation of the current master clock relative to TA and the control frequency. The control frequency of the master clock should be greater than the update frequency of the external traceability and atomic time. The control frequency of the master clock is 1 hour, which is consistent with the calculation frequency of atomic time. After determining the current master clock, the calculation formula of the control value offset is: TA-clock m is the phase difference between the most recently obtained atomic time and the master clock, k is the adjustment coefficient, b m is the frequency deviation of the master clock relative to atomic time, for the 24 most recent TA-clock m The least square method is used to perform a polynomial modeling, and the number of data items can be adjusted according to actual needs; 4) Switching control of the master clock; The switching judgment of the master clock needs to be made according to the performance status of the current master clock. When the master clock has abnormal phase or frequency hopping behavior, the master clock signal needs to be switched in time to ensure stable local physical signal output.

6. The timekeeping method of the timekeeping device with automatic switching of multiple master clocks according to claim 5, characterized in that: The specific steps of step 2) are as follows: 2.1) First, read the comparison data from the local file according to the set comprehensive atomic time calculation frequency and the length of historical data used; Direct physical signal is used for tracing, so the calculation frequency of comprehensive atoms is set to 1 hour; 36 historical data are selected; 2.2) Atomic time calculation; The weight of an atomic clock depends on its predictability, that is, how well it conforms to a model. A first-order polynomial is used as the calculation model: y=a+b*t Get 36 MC-clock i After comparing the data in hours, the residual is obtained by deducting the model value of the atomic clock at that moment from each data. In the initial calculation, a and b are 0, and the atomic clock weight obtained in the last calculation is used to weight the residual to get the atomic time MC-TA. The last MC-TA is saved in a local file as the atomic time for this calculation. In the initial calculation, the weight of each clock is 1 / N, where N is the number of atomic clocks involved in the calculation, and N is 3. By MC-clock i Get 36 TA-clocks for each clock with MC-TA i , use the least squares method to update the a and b values ​​of each clock, and use the TA-clock i Subtract the new model value and calculate the weight of each clock according to the sum of squares of the residuals. The weight calculation formula is as follows: The weight calculation formula needs to be iterated to ensure the stability of the calculation. It is iterated three times. At the same time, an upper limit needs to be set for the weight of the atomic clock to prevent a single atomic clock from being completely dependent on the operation of each atomic clock, which affects the reliability of the atomic clock. For atomic clocks that exceed the upper limit, the excess weight needs to be redistributed to other atomic clocks in proportion. At the same time, if the weight of an atomic clock is too low, it means that the predictability of the atomic clock is extremely poor, and the weight needs to be set to 0 in this calculation. 2.3) Atomic time frequency calibration The MC-TA calculated in step 2.2) is weighted by the local atomic clock comparison data only. The atomic time needs to be relatively consistent with the external traceability source, and the atomic time needs to be calibrated in phase and frequency. By MC-UTC k Get TA-UTC with MC-TA k , the number of data is selected as 24, and the least square method is used to calculate the 24 TA-UTC k Perform a polynomial modeling y=a+b*t, and deduct the fitting value from the MC-TA to obtain the calibrated MC-TA. Each newly generated MC-TA needs to deduct the fitting value y. The update frequency of the a and b values ​​in the model is related to the update frequency of the external traceability and the atomic time calculation frequency, and is updated once every hour. After each atomic time phase and frequency calibration is completed, MC-TA and MC-clock i Get TA-clock i , stored in a local file.

7. The timekeeping method of the timekeeping device with automatic switching of multiple master clocks according to claim 5, characterized in that: In step 4), the dynamic Allan variance method is used to complete the abnormal judgment of the atomic clock, and the implementation steps are as follows: 4.1) For the MC-clock obtained every second i , get the clock by pairwise comparison i -clock j , calculate its difference with the clock 100 seconds ago i -clock j The difference between the last 50 values ​​is counted to calculate the clock i -clock j When the 100s stability calculated at the current moment exceeds twice the nominal 100s stability of the atomic clock in use, it is considered that one of the atomic clocks i and j is abnormal. After the three sets of comparison values ​​are judged, the abnormal atomic clock can be determined; the average time of stability and the magnification of judgment are determined according to the actual performance of the atomic clock; 4.2) When the master clock is abnormal, the master clock is switched to an atomic clock that has not been abnormal. There are only three atomic clocks. If two atomic clocks are abnormal at the same time, it is impossible to determine the specific atomic clock. In this case, the master clock will not be switched; 4.3) When the master clock is switched, the frequency control value of the phase trimmer must be corrected in time according to the control value calculation formula in step 3).

8. The timekeeping method of the timekeeping device with automatic switching of multiple master clocks according to claim 5, characterized in that: The adjustment coefficient k is 2.

9. The timekeeping method of the timekeeping device with automatic switching of multiple master clocks according to claim 6, characterized in that: The upper limit of the weight is 2 / N, and the lower limit of the weight is 1%.

Images