GB2551538A

GB2551538A - Method for the determination of cumulative error in clocks

Info

Publication number: GB2551538A
Application number: GB1610828.4A
Authority: GB
Inventors: George Hoptroff Richard
Original assignee: Hoptroff London Ltd
Current assignee: Hoptroff London Ltd
Priority date: 2016-06-21
Filing date: 2016-06-21
Publication date: 2017-12-27
Also published as: GB201610828D0

Abstract

Error 310 due to drift in a free-running clock, watch or other timepiece 302, and error 309 due to jitter, Allan deviation or random variation in a central primary clock 301 signal, are combined to provide the user of a timepiece with a display of the error 311 in the timepieces 302 local time. This may be in the form of error bars. When the local timepiece 302 is not receiving a signal from the primary clock 301, average drift error is summed over time until the signal is restored, providing a cumulative drift error value. When the primary clock 301 signal is received, it is used to determine an error value, which may be based on a moving average or root mean square (rms) recorded and calculated for the primary clock 301 signal error. The preferred embodiment is used with an atomic clock 302.

Description

Method for the Determination of Cumulative Error in Clocks

This invention relates to the absolute timekeeping error of clocks. It allows the cumulative timekeeping error to be estimated, recorded and displayed.

Clocks, from wristwatches to IEEE 1588 grandmaster clocks, are subject to two sources of error.

First the internal oscillator (e.g. pendulum, balance spring, quartz crystal or atomic resonator) of the 'local, clock will not beat precisely in time with the agreed SI standard of 9,192,631,770 transitions between the two hyperfine ground states of Caesium 133. Indeed, it will likely not be self-consistent ( stableJ over long periods of time due to aging, temperature changes, shock, etc.

Second, the local clocks time must be set with reference to an accurate 'primary, clock, which ultimately traces its timekeeping path back to the Bureau International des Poids et Measures, where time consensus is managed. Both the primary clock, and the path from it to the local clock, are subject to error. This error takes two forms: systematic error, for example transmission latency due the distance between the clocks, which can be corrected for by one-off calibration, and is not considered here, and random error, otherwise known as jitter or Allan deviation, being the variation between timing pulses over specified timescales. These errors may arise in the primary clock itself and also in the transmission media on the way to the local clock.

The complementary nature of the primary and local timekeeping systems is key to high accuracy timekeeping in a local clock. The primary clocks random deviations can be averaged out ('de-jittered J by the local clock, which is typically not subject to jitter over short timescales, producing a regular beat. The local clock can be calibrated ('trimmed, or 'disciplined J over longer timescales with reference to the primary source to compensate for its drift Such is the state of the art known as the Disciplined Oscillator, which de-jitters primary time when such primary time is available, trims an accurate local clock to high accuracy, and allows the local oscillator to free-run when the primary clock is not available ('holdover J.

For example, with reference to figure 1, a primary clock 101 and a local clock 102 both output one pulse per second (1PPS) signals. The rising edge of primary clock 101 s 1PPS will be, on average, at the start of a UTC second, but will be subject to random jitter. The leading edge of local clock 102 s 1PPS will be subject to very little jitter and will be very stable in terms of periodicity. However local clock 102 s 1PPS rising edge may differ in terms of actual periodicity (relative to 1 Hz), and phase (relative to the start of a UTC second). Local clock has a steering input 103 which allows the period P of the local clock to be varied slightly from its fundamental period without loss of stability.

Phase meter 104 measures the time difference between the rising edges of the primary clock 101 and a local clock 102 1 PPS signals and reports the value to phase controller 105 that uses a disciplining method, such as the example described below, to steer the frequency and phase of local clock 102 so that its 1 PPS, when averaged over longer timescales, coincides with the 1PPS of primary clock 101. This local clock 1PPS is used as the system output 106, being a de-jittered version of the primary clock.

Signal Loss Detector 107 detects if the primary clock 101 1PPS signal is lost and reports such loss to the phase controller 105. In the event of signal loss, disciplining is suspended and the local oscillator 102 is allowed to free-run at its current steering frequency 103 to provide holdover. If primary clock 101 1PPS signal resumes, then disciplining also resumes.

With reference to figure 2, an example of a phase steering algorithm is as follows.

1. Initialize pulse countert= Oand periodicity to a neutral value P = 1. (701J 2. Wait (702 J fora pulse from the primary clock 101. 3. Measure (703J the phase difference ^between the primary clock 101 pulse the closest local clock 102 pulse. 4. Once two or more phases differences j^are measured, calculate (705J a periodicity adjustment VP as follows:

where is a proportional term that adjusts the periodicity of local clock 102 so that it converges on the periodicity of primary clock 101 and § is a proportional term that adjusts the periodicity of local clock 102 so that it, on average, is in phase with the primary clock 101. Smaller values of and <f equate to longer averaging periods with slower convergence but greater long-term stability. 5. Apply (706 J periodicity adjustment VP to periodicity P.

6. Repeat from 2 for the next pulse t+1 (707J

Note that holdover is implicit in step 2, i.e. if primary clock 101 signal is lost, periodicity adjustment are suspended until the signal resumes. A limitation of the state of the art is that it does not estimate timekeeping error, which in part relates to the best estimate of how well the primary source is being de-jittered when it is available, and in part; the best estimate of how much the local clock has cumulatively drifted when the primary clock is not available.

The present invention describes a method for calculating and recording a best estimate for the accuracy of a local clock under these combined conditions, so the clock doesn't just offer a time value, but also 'error bars_ around it.

Accordingly, and with reference to figure 3, the invention is as follows: 1. Error estimating counter 312 maintains a count value such that: a. While signal loss detector 307 reports 310 that a primary clock 301 signal is available, error estimating counter 312 count value E is held in at the value Ep, an estimate of the dejittered local clock 302 error when a primary clock 301 is available. For example, Ep might be a root mean square moving average of recent error correction values VE. b. While signal loss detector 307 reports 310 that a primary clock signal is not available, error estimating counter 312 count value E monotonically increases by a value dependent on Eh each second, where Eh is an estimate of the local clock 302 error when allowed to free-run. For example, Eh might be based on a root mean square moving average of recent error correction values or the local clock 302 manufacturers data sheet specifications for Allen deviation, and where, for example, i. E is a cumulative summation of individual values Eh, reflecting local clock error that is expected to be consistent, or ii. E is the square root of a cumulative summation of the square of the individual values Eh, reflecting local clock error that is expected to exhibit random walk behaviour. c. If primary clock 301 resumes, error count returns to Ep, with the exception that where Ep is calculated as an accumulation of previous values, such previous values are discarded and the accumulation is reinitialized. 2. Error estimating counter 312 value E is output 311 in addition to the actual 1PPS output 306.

In a first embodiment the clock is an IEEE 1588 GNSS disciplined atomic grandmaster clock, where:

Primary clock 301 is a GPS satellite timing receiver;

Local clock 301 is an atomic clock;

Signal loss detection 307 and phase metering 304 are performed by a microcontroller;

Phase control 305, 303, and error estimation 312 are performed by a host Linux server application; C lock output 306 is a connection to an IE E E 1588 compliant network card with a 1PPS input such as an Intel^ i210 network card; E rror estimation 311 is output as a function call to the host Linux server application or an extension to the IE E E 1588 protocol.

In a second embodiment the clock is an IE E E 1588 timekeeping client, where:

Primary clock 301 is an IE EE 1588 grandmaster clock on a network;

Local clock 301 is an embedded high-accuracy clock, the timekeeping clients network card clock, or the Linux server system clock;

Phase control 305, 303, and error estimation 312 are performed by a host Linux server application;

Clock output 306 is steered by the host Linux server system clock;

Error estimation 311 is output as a function call from the timestamping application, the end result being not just an accurate time value, but error bars.around it

Claims

Claims CLAIMS

1. A ti mekeepi ng apparatus compri si ng: means for producing a timekeeping indicator in order to maintain a local time at the timekeeping apparatus; means for receiving a primary clock signal; and means for determining an estimate of an error in the local time maintained at the timekeeping apparatus, wherein the means for determining is arranged to determine the estimate of the error in the local time in dependence upon an estimate of free-run error associated with the means for producing a timekeeping indicator and/or, if available, upon a received primary clock signal.

2. A ti mekeepi ng apparatus as cl ai med i n cl ai m 1 wherei n the esti mate of the error i n the local time comprises a count value maintained by the means for determining.

3. A ti mekeepi ng apparatus as clai med i n clai m 1 or clai m 2 wherei n the esti mate of the error in the local time is a cumulative estimate.

4. A timekeeping apparatus as claimed in claim 1, claim 2 or claim 3 wherein the estimate of the error in the local time is determined in dependence upon at least one previous error correction value.

5. A timekeeping apparatus as claimed in any preceding claim wherein the estimate of the error in the local time is determined in dependence upon a summation, a product or a root mean square of previous error correction values or of previous error correction values raised to a power.

6. A timekeeping apparatus as claimed in any preceding claim wherein the estimate of the error in the local time comprises a moving average.

7. A timekeeping apparatus as claimed in any preceding claim wherein the means for determining is arranged to update a previous estimate of error in the local time in dependence upon the primary clock signal.

8. A timekeeping apparatus as claimed in any preceding claim wherein the timekeeping apparatus comprises a wristwatch, a desk clock, a wall clock or a free-standing clock.

9. A timekeeping apparatus as claimed in any preceding claim wherein the means for producing a timekeeping indicator comprises an oscillator.

10. A timekeeping apparatus as claimed in claim 9 wherein the oscillator is a pendulum, a balance spring, a quartz crystal or an atomic resonator.

11. A method for performance by a ti mekeepi ng apparatus, comprisi ng: produci ng a ti mekeepi ng i ndi cator i n order to mai ntai n a I ocal ti me at the ti mekeepi ng apparatus; and determining an estimate of an error in the local time maintained at the timekeeping apparatus, wherein the estimate of the error in the local time is determined in dependence upon an estimate of free-run error associated with the timekeeping indicator and/or, if available, upon a received primary clock signal.

12. A method as claimed in claim 11 additionally comprising receiving a primary clock signal.

13. A method as claimed in claim 11 or 12 wherein the estimate of the error in the local ti me comprises a count val ue mai ntai ned at the ti mekeepi ng apparatus.

14. A method as claimed in claim 11, 12 or 13 wherein the estimate of the error in the local time is a cumulative estimate.

15. A method as clai med i n any of clai ms 11 to 14 wherei n the esti mate of the error i n the local time is determined in dependence upon at least one previous error correction value.

16. A method as claimed in any of daims 11 to 15 wherein the estimate of the error in the local time is determined in dependence upon a summation, a product or a root mean square of previous error correction values or of previous error correction values raised to a power,

17. A method as cl ai med i n any of cl ai ms 11 to 16 wherei n the esti mate of the error i n the local time comprises a moving average.

18. A method as claimed in any of claims 11 to 17 comprising updating a previous estimate of error in the local time in dependence upon a primary clock signal.

19. A method as claimed in any of claims 11 to 18 wherein the timekeeping apparatus comprises a wristwatch, a desk clock, a wall clock or a free-standing clock.

20. A method as claimed in any of claims 11 to 19 wherein the timekeeping indicator is produced using an oscillator.

21. A method as claimed in claim 20 wherein the oscillator is a pendulum a balance spring, a quartz crystal or an atomic resonator.

22. A computer program comprising instructions which, when performed by a processor, cause performance of the following by a timekeeping apparatus: produci ng a ti mekeepi ng i ndi cator i n order to mai ntai n a I ocal ti me at the ti mekeepi ng apparatus; and determining an estimate of an error in the local time maintained at the timekeeping apparatus, wherein the estimate of the error in the local time is determined in dependence upon an estimate of free-run error associated with the timekeeping indicator and/or, if available, upon a received primary clock signal.

23. A computer program as claimed in claim 22 additionally causing receiving a primary clock signal.

24. A computer program as claimed in claim 22 or 23 wherein the estimate of the error in the I ocal ti me compri ses a count val ue mai ntai ned at the ti mekeepi ng apparatus.

25. A computer program as claimed in claim 22, 23 or 24 wherein the estimate of the error in the local time is a cumulative estimate.

26. A computer program as claimed in any of claims 22 to 25 wherein the estimate of the error in the local time is determined in dependence upon at least one previous error correction value.

27. A computer program as claimed in any of claims 22 to 26 wherein the estimate of the error in the local time is determined in dependence upon a summation, a product or a root mean square of previous error correction values or of previous error correction values raised to a power.

28. A computer program as claimed in any of claims 22 to 27 wherein the estimate of the error in the local time comprises a moving average.

29. A computer program as claimed in any of claims 22 to 28 comprising updating a previous estimate of error in the local time in dependence upon a primary clock signal.

30. A computer program as claimed in any of claims 22 to 29 wherein the timekeeping apparatus comprises a wristwatch, a desk clock, a wall clock or a free-standing clock.

31. A computer program as claimed in any of claims 22 to 30 wherein the timekeeping indicator is produced using an oscillator.

32. A computer program as claimed in claim 31 wherein the oscillator is a pendulum, a balance spring, a quartz crystal or an atomic resonator.