The Use of Atomic Clock Holdover for Precision Time Applications in the Field

Although the use of primary reference time clocks (PRTC) in Holdover mode is not often recommended for field-oriented Precision Timing test applications, such as Time Error, Wander, or One-Way Delay measurements, which still need to be performed in many places that have no access to any other traceable or accurate reference clock. So, the Clock Holdover mode is often proposed as an alternative workaround, often based on overhyped characteristics of atomic clocks. However, one must be fully aware of the compromises being made, the uncertainties, the process and precautions required, and their effects on the accuracy, uncertainties, and repeatability of the measurement results.

Atomic Clock holdover is not inherently unsuitable for field testing; however, expectations and workflow constraints must reflect the practical limitations of precision timekeeping. Common requests, such as “be ready to test in one hour or less,” “guarantee nanosecond-level accuracy all day,” or “make it easy,” can underestimate the preparation required and overestimate the capabilities of atomic clocks and GNSS timing references. Meeting the accuracy requirements of a field measurement requires an understanding of the underlying physics and limitations of precision timekeeping, even when operational demands favor a faster, simpler process. Overlooking these factors can lead to frustration and unreliable measurement results. If you choose to use timing holdover for field testing, proceed with care, understand the associated uncertainties, train your people, write detailed and realistic procedures, and allow sufficient time for preparation and testing.

An Atomic Clock can only be considered in Holdover mode if it has been properly disciplined and locked to an accurate and traceable timing reference source, such as the 1PPS signal from a precision timing GNSS receiver, before losing its input reference. Once the external reference is disconnected, the Atomic Clock 1PPS output starts to drift slowly, decreasing its accuracy as time goes by.

Holdover is meant to bridge GNSS signal outages in long-term always on precision timing equipment, in which atomic clocks have been continuously disciplined by a good 1PPS reference, at stable temperatures, and for a long time. They do an excellent job for the intended application.

Note that the main concern is that clock holdover performance specs are often written based on theoretical calculations or highly controlled ideal test conditions and long disciplining times, which are considerations that don't usually apply to portable (field) test equipment applications.

Should you decide to use the one pulse-per-second (1PPS), from an Atomic Clock, in holdover for timing and delay measurements in the field, you must:

Adjust your expectations accordingly. Keep in mind that field instruments have to deal with time constraints, poor satellite reception (1PPS reference), temperature variation, airflows vibration/shock, among other internal and external factors.
Have a clear goal of what are the minimum required accuracy for the type of applications and measurements being performed, as well as the expected resolution (ns, µs, ms).
Develop a test procedure defined to minimize the effects of external environmental conditions in its Time Error or Timing drift (Time Error accumulation over time).
Experiment with multiple GNSS Survey and Atomic Clock disciplining thresholds (profiles), to identify a good compromise between the required disciplining (preparation) time and the resulting 1PPS holdover accuracy and stability, in order to be practical and meet the application's requirements.

For example, if you plan to measure one-way Ethernet packet delay with ±1 μs resolution, then a well-trained clock holdover may provide that level of uncertainty for up to a few hours. If the expected resolution is ±0.01 ms or lower, then holdover certainly starts making a much more sense, since the required accuracies and expected uncertainties are more attainable, and in such scenario the reference time clock could even last a full working day in holdover mode. You must know the application to make the necessary informed decisions.

Although VeEX does not officially recommend using Atomic Clocks in holdover mode for precision timing measurements, the following suggestions are provided in an effort to improve the chances of good measurement results for users who choose to do so, despite the uncertainties introduced by the changing environmental conditions portable instruments are exposed to.

1. General Application Workflow

Single-Tech OWD Test Procedure Example

The following scenario considers one of the worst-case scenarios for a single field technician's workflow, to measure the one-way delay (OWD) between two sites, with initial referencing (disciplining) at a central (known, trusted) site. It allows plenty of time for the test sets' internal timing references to drift, before the measurements even start.

Connect the test set's GNSS receiver to the roof antenna at the central location, turn the test set ON and let it's GNSS receiver perform the Location Survey (find its most accurate position) and Discipline the internal Atomic Clock (calibrate its frequency and align its timing to the standard second) for 24 hours or longer (or at the very least, overnight).
Launch the desired test application to let the test set warm up while the clock disciplining process is taking place. Some examples of time-based test application are: 1PPS Clock Phase (Time) Error, One-Way Delay (Ethernet), Recovered Clock Wander Measurements, SyncE, etc.)
Keeping the test set ON, disconnect the antenna to enter holdover mode, and transport the test set to the test location. Avoid sudden changes in temperature during the trip.
At the location, connect the test ports to the device or link under test, select Atomic_1PPS as the Reference Clock for time error oriented measurements or the Atomic_10MHz for clock frequency and wander applications. Note: The use of Clock Holdover is not recommended for high-precision timing applications, like PRTC or PTP Time Error verification, requiring nanoseconds-level (ns) accuracy.
Since the holdover time accuracy slowly degrades over time, for most test applications, it is recommended to keep the total test time under four (4) hours, including the transit time. If the measurement doesn't require high accuracy, like measuring One-Way Delay with 1 μs resolution, the total time could be extended to six (6) hours.
When finished, it is recommended to keep the test set ON during the return to the central location. Then go back to step 1.

Whenever possible, keep the test set's AC/DC adapter plugged in, to maintain a fully charged battery.

2. Recommended Configuration for VeEX Test Sets

Use a flexible coaxial (adapter) cable to connect the GNSS antenna feed to the test set's GNSS receiver SMA connector . Turn the test set ON.

2.1 GNSS Receiver Configuration

Go to > >Utilities >Settings >More >High Precision Clock Source and select the GNSS Receiver tab.

The test set may be configured with one of the two Precision Timing GNSS Receivers offered by VeEX and their matching portable antennas:

Precision-timing Single-band Dual-constellation GNSS Receiver (P/N: Z88-00-009P), supporting up to two constellations at a time, including GPS (L1C/A), GLONASS (L1OF), Galileo (E1B/C) and BeiDou (B1I). It can be identified by the GNSS Receiver label on the configuration and status page.
High-precision Multi-band Quad-constellation GNSS Receiver (P/N: Z88-00-010P), supporting up to four simultaneous constellations with two frequency bands each, including GPS (L1C/A, L2C), GLONASS (L1OF, L2OF), Galileo (E1B/C, E5b) and BeiDou (B1I, B2I). It can be identified by the Multi-band GNSS Receiver label on the configuration and status page.

Select all the GNSS Systems and Frequency Bands supported by the GNSS receiver and antenna being used, for best performance. You may exclude any GNSS system or bands that are not considered trustworthy or have problems in your region (e.g., jamming, spoofing, etc.). The use of multi-constellation and multi-band receivers is recommended for best performance.
If the active GNSS antenna is not already powered, turn the Antenna 5V ON.
Location Survey - The Site Survey is a fine-tuning process in which the GNSS receiver reduces the position uncertainty to a desired level of accuracy, in order to lock the coordinates (no longer updated) to produce a more accurate and stable 1PPS timing signal (one pulse per second). Enable the Location Survey for precision timing applications and Atomic Clock disciplining.
Window (s) - Is the amount of time the position must remain below the target accuracy, before the GNSS receiver locks the coordinates and enters the precision time mode. The recommended value is 600 s.
Accuracy (m) - Is the target position accuracy for the survey to lock the coordinates and enter precision time mode. Use 1 m for roof antennas with excellent satellite signal quality (multiple satellites with >42 dB-Hz C/No) or 3 m for portable antennas, with unobstructed view of the sky. Note: If the GNSS signal reception is poor, the position may never reach the target accuracy (3D Deviation).
Antenna Cable Delay Compensation (ns) - Enter the coaxial cable delay measured from the antenna to the GNSS Receiver's SMA connector, including any additional patch cord cables. If the exact cable delay is not known, users may approximate it by using the total antenna cable Length (L) and multiply it by the Velocity of Propagation (VP) stated on the cable's datasheet or use the generic 5 ns/m. Refer to the following article for more details on How to Measure Antenna Cable Delay. If using the 5 m portable antenna, supplied by VeEX, the typical delay is 26 ns.
Time Base - Select the satellite constellation to be used as the time reference. GPS or Galileo are recommended.
UTC Offset - Select the proper time zone offset for the region. This doesn't affect the measurements; however, it assures that the correct time stamps are shown in reports.
Connect the antenna cable to the test set's GNSS receiver SMA port, using a flexible coaxial cable. Turn the GNSS Receiver ON, to start the location survey (site's coordinates) and time alignment process.

The 3D Deviation value should start to decrease until it reaches the desired Accuracy (m) target. The Location Survey process will take some time to complete. The total time will depend on the targeted Window (s) and Accuracy (m) thresholds, as well as the quality of the satellite view and reception. Poor RF signal quality will make the process take longer or not complete at all, if the signals' quality (C/No) are poor.

A. GNSS Status shows Locked when at least four satellites have been acquired and the GNSS receiver is cable of calculating the local coordinates and can perform an initial rough alignment of the internal GNSS 1PPS timing reference pulse.

B. Satellites displays the total number of GNSS satellite vehicles that the antenna can currently "see".

C. Antenna Status - If the test set is the one powering the antenna (Antenna 5V = ON), this field will give an indication of the antenna and cable status. OK indicates that an antenna is connected at the far end of the cable. Short indicates that a short circuit or high current draw has been detected. Open indicates that there is no antenna connected.

D. Reset GNSS - Use this button to make the GNSS receiver forget all the parameters that has learned from the satellites and start the satellite acquisition process again (cold start).

E. Site Survey - Status of the Location Survey process. Active indicates that the survey is still in progress. Locked indicates that the survey has finished and that the coordinated have been locked (no longer updated). You MUST allow time for the survey to Lock the 3D coordinates, in order to provide a more stable 1PPS reference to the Atomic Clock to discipline to. If the survey does not lock, you may need to consider increasing the Accuracy threshold.

F. 3D Dev. - Tridimensional position deviation, in meters (m). If the deviation stops decreasing and doesn't reach the target Accuracy (m), check the GNSS antenna installation and the Signals (C/No), to make sure that there are at least four satellites with >38 dB-Hz at all times.

G. Antenna's 3D Coordinates displayed in Degrees, Minutes and Seconds (DMS) and in Decimal Degree (DD) formats, as well as the antenna's altitude. Once the survey has locked, these coordinate values must remain constant.

H. Save - Once the Location Survey has finished (Site Survey = Locked), users can save the coordinates with a descriptive name. The saved location can be used in future tests to save time, by avoiding the survey process.

I. ToD - Displays the current universal (UTC) Time of Day (date and time) in YYYY/MM/DD hh:mm:ss format.

J. Sync ToD - Applies the current universal date and time (ToD) obtained from GNSS to the test set, with the appropriate local time zone offset correction.

Once the Site Survey = Lock, the 3D coordinates stop updating and the GNSS clock goes into Precision Timing mode. Users can tap on the Save button to store the coordinates for commonly used locations, with a name, then save time next time by selecting Location Survey = Manual and selecting the saved location. In Manual mode, the test set doesn't have to go through the Site Survey process, saving up to an hour of preparation time.

2.2 Atomic Clock Configuration

If the factory-installed Atomic Clock hardware option is installed on the test set, go to >Utilities >Settings >More >High Precision Clock Source and select the Atomic Clock tab.

Disciplining Profile - Users can select between multiple pre-loaded settings or Manual mode. Each profile has different a Time Constant and a hint of the minimum time required to achieve Disciple Lock, with a good reference signal.
Time Constant (s) - This is the time the Atomic Clock's disciplining circuitry (servo) will require the relative Time Error to be below the configured Discipline Threshold (ns), before declaring 1PPS Discipline Status = Locked. For a multi-band quad-constellation GNSS receiver, with good roof antenna installation, the recommended Time Constants are 900 s or 1800 s. Choose depending on the application, time available, required accuracy and stability, workflow, etc. The larger the time constant is, the more stable the 1PPS reference is required to be and the longer it will take the 1PPS Disciplining Status to lock.
Discipline Threshold (ns) - This is the maximum relative Time Error variation allowed by the Atomic Clock's disciplining circuitry (servo), which will be required to enter the 1PPS Discipline Status = Locked status. For a multi-band quad-constellation GNSS receiver, with good roof antenna installation, the recommended Discipline Threshold is 20 ns. If the best attainable GNSS satellite reception is not that good, users may have to consider loosening the threshold to 50 ns and the larger uncertainty that comes with it.
Discipline Source - Selecting a 1PPS clock reference starts the Atomic Clock synchronization or disciplining process. For this application, select GNSS 1PPS.
1PPS Signal Health - Verify that the Atomic Clock is receiving a valid 1PPS clock reference signal. If it shows invalid, check the GNSS Receiver configuration and that the GNSS Status = Locked.
10MHz Oscillator Status - Indicates the status of the internal Atomic Clock oscillator frequency. It should display Locked after it has finished its quick warm up and internal calibration sequence, which should happen within 180 seconds after the test set is turned ON.
1PPS Disciplining Status - Once the Discipline Source is selected, the status turns into Acquiring, to indicate that the Atomic Clock is being trained by the 1PPS reference clock. Once the Atomic 1PPS output meets the relative Time Error threshold (Discipline Threshold) for at least the amount of time defined by the Time Constant, the discipline process is declared Locked and ready to use. Before disconnecting the 1PPS reference, to bring the Atomic clock into Holdover mode, it is recommended to:
1. Keep disciplining the Atomic Clock for several minutes, so it has time to learn from the environmental changes.
2. Launch the required test application to bring up the internal temperature of the test set to normal operation and let the Atomic Clock adjust to it.
Lock Time - Time that has passed since the internal Atomic Oscillator achieved Locked status. It can also serve as an indicator of how long the test set has been ON.
Total Holdover - This is the indication of how long the test set has been in Atomic Clock Holdover mode (without a 1PPS Reference).
Phase Graph - Displays the relative Time Error between the 1PPS Reference (from GNSS) and the Atomic Clock 1PPS output (Atomic 1PPS). The Atomic Clock time correction (disciplining) is a slow process that can take several hours (with the most stringent settings). As the Atomic 1PPS timing aligns with the standard second the relative time error should converge towards zero. The phase monitor graph shows the last 10 minutes (600 s). Zero means that the Atomic Clock time is aligned with the 1PPS Reference from the GNSS receiver.
Atomic Clock 1PPS Timing status Indicator (always visible).
GNSS Receiver Satellite Lock status indicator (always visible). Green indicates that satellites are being tracked and a valid GNSS 1PPS timing is available.

Once the Disciplining Status = Locked the test set's Atomic Clock 1PPS and 10 MHz signals have achieved their most accurate state, according to all the settings. The Atomic Clock will continue to be disciplined, and correcting any remaining Time Error, for as long as the 1PPS reference is present.

2.3 Entering 1PPS Holdover

Once the test set's Atomic Clock is discipled and locked , it will automatically enter Holdover mode any time it loses the reference 1PPS clock being used for disciplining (GNSS 1PPS).

To force the test set's Atomic Clock into Holdover, users can quicky disconnect the GNSS antenna cable from the SMA port of the GNSS receiver. The icons at the bottom of the screen will show Holdover and loss of satellite signal .

3. Field Testing

Time and timing-based measurements are performed in the same way, whether the reference time clock is being actively synchronized by GNSS or in holdover mode. Follow the procedure for the required test and use Atomic 1PPS or Atomic 10 MHz as the measurement reference clock.

It is highly recommended to load the test application right after the disciplining process has started at the central location, so the disciplining process takes the internally generated heat into account, while disciplining the Atomic Clock.

Do not turn the test set OFF or it will lose its frequency and time synchronization.

Avoid sudden or significant ambient temperature changes during holdover (ΔT < 10 ºC, 18 ºF).

Tests that Use 1PPS as a Reference for Measurements

One-Way Delay (OWD): Uses two test sets, at two different locations, both synchronized to the standard second (1PPS), so they can measure the payload (data) delay (or latency) in each direction independently. This type of test is available for different types of links, such Ethernet, G.703 Codirectional, IEEE C37.94, etc. Refer to this Ethernet One-Way Delay Measurement article for more details.
Clock Wander & Phase Measurements: Uses 1PPS reference to measure network elements' physical clock output Time Error (accuracy) and Wander (stability), or 10 MHz to measure frequency wander, stability and offset.
PTP Slave Emulation with Recovered Clock Time Error Measurements: Uses the 1PPS clock reference to measure Time Error (2WayTE). Reference clocks in holdover mode may only be advisable for quick clock checks at the edge of the link, where the maximum allowance is large enough (e.g., 1,100 ns).

Note that in all those cases, any cumulative Time Error (drift) in the holdover reference will be reflected in the measurement results. The longer the holdover period, the greater the error. Also referred as uncertainty, because (in the field) there is no way to know or measure the current timing accuracy of the Atomic 1PPS reference at any given time.

4. Important Considerations

Can the 1PPS Pulse be Quickly Re-Aligned, at the Location, Before Testing?

Reconnecting the portable antenna in the field, to re-sync the 1PPS (e.g., outside of the customer's building), is NOT recommended. There are several reasons why this could increase the Reference Time Error:

The antenna cable delay compensation for the 5 m portable antenna is typically 26 ns, while the compensation used for the longer (roof) antenna cable, at the central site, will be much larger. The difference would be reflected in the measurements as added Time Error.
After the Location Survey at the central office, the GNSS receiver goes into Precision Time mode, fixing the coordinates used for time calculations. If the portable antenna is connected at a different location (different coordinates), the time calculations will be wrong (more added Time Error). In certain cases, if the distance exceeds certain GNSS receiver thresholds, it may trigger to restart the process again. One option to reduce the effects of different coordinates is to set Location Survey = Disabled, to allow the GNSS receiver to get the new coordinates, but this would also reduce the timing accuracy and stability.
These issues could not only introduce the difference as an extra error to the TE or OWD measurements, but if the difference (in ns) is large enough, it could trigger the atomic clock to restart the disciplining process.
The RF signal quality (C/No) and multi-path conditions at street level are mainly unpredictable. Bad GNSS signal could lead to bad performance and increased Time Error.

What About Temperature Changes?

Temperature changes, especially temperature shocks, must be avoided! An example of temperature shock would be if in a cold winter or hot summer, the user takes the time to discipline the test set outside and then quickly brings it inside to an air-conditioned building at 22 ºC (72 ºF). That could mean an instant change in ambient temperature >±10 ºC (18 ºF). That will definitely affect the characteristic of the Holdover's time drift behavior.

During the disciplining process the atomic clock's tuning system learns about the effects of slow temperature changes and it will use that information during the Holdover state to try compensating for it. In general, try to keep the temperature changes slow and no more than 10 ºC (18 ºF) within one hour.

One should also keep into account the heat generated by the instrument itself, which is many cases helps maintain a somewhat constant temperature. So, it is important to keep the test set ON all the time and with the required test application loaded and at their operational temperature. (Loading test applications just before the start of the test may create extra heat, that was not accounted for.)

DO NOT operate the test set below 10 ºC (50 ºF) or above 40 ºC (104 ºF), while in Holdover. Operate it inside an air-conditioned van/truck if necessary and keep it between 18~24 ºC (64~76 ºF), for best results.

DO NOT place a powered-ON test set inside the carrying case, as it can overheat. Keep the fans airflow from being obstructed.

Keep the Test Set Running at All Times

In general, Atomic Clocks are designed to be running continuously, to stay in top condition, but that can be a challenge for portable test equipment. For that reason, it is recommended to keep test sets used for precision timing applications always ON and disciplining against a good GNSS antenna, even when not in use.

Why the Uncertainty? Shouldn't Atomic Clocks be Highly Predictable?

Actively disciplined Atomic Clocks are highly predictable, accurate and stable, in their steady-state (long-term) operation. However, Holdover is not considered normal operation and their performance in this condition will depend on the clock quality, among other parameters.

While in normal operation (locked mode) the disciplining circuitry continuously changes the Atomic Clock's frequency, to adjust the 1PPS rising edge and keep it aligned with the start of the standard second. Although those variations (frequency offsets) can be just a few parts-per-trillion up and down, the atomic oscillator will try to retain the last correction, in the event of a reference loss. That means, the atomic oscillator's frequency is never 100% accurate, because the priority is on keeping the 1PPS timing aligned. However, in the field, it is not easy to quantify, predict or estimate what the last frequency offset was. Any frequency error will cause the Time Error to accumulate (drift) over time.

Note that atomic clocks' Holdover performance measurements or calculations, if published, would be based on highly ideal and controlled lab conditions (e.g., high precision 1PPS reference, constant temperature, no airflow, no vibration, plenty of time to discipline, etc.). None of those would apply to portable field instruments.

The graph below shows a theoretical worst-case scenario and the possibilities for a Time Clock drift in holdover, in absence of any other environmental factors. The datasheet of the chip-scale Atomic Clock, used in VeEX products, specifies a free running frequency accuracy of ±5E-11 (0.05 ppb) at shipment. Of course, disciplining helps correct any intrinsic error from the atomic oscillator. The dark-red lines would be the expected Time Error drift if not disciplined (free running). In those extreme cases, it would only take around 33 minutes to drift to 100 ns. The dark-green lines show an example of a disciplined (corrected) clock, drifting to 100 ns in about three hours, because its last frequency offset for the last correction was 10 parts-per-trillion (1E-11). But there are thousands of possible last correction possibilities in between, considering that a disciplined atomic clock can achieve accuracies better than 1E-12.

That is to show that, even if the frequency error (slope) doesn't change over time, the Time Error is cumulative and adds up very quickly, no matter how "good" a portable atomic clock is.

For example, if we wanted to maintain a reference time clock error (in holdover) <±100 ns at the end of the day (let's say 10 hours), to measure values around 1,100 ns, then the frequency error of the Atomic Clock must be better than 2.8E-12 at all times. Although this could be possible, it is certainly not repeatable. Every time you start a new holdover, there are multiple paths that cumulative Time Error function could take.

That is what makes Time Holdover unpredictable. Then we can add temperature changes, GNSS short-term stability issues and vibration to the mix.

For supporting VeEX products, the Atomic Clock's Phase Graph can be used in the field to get an idea of the current state of the time alignment corrections, to try reducing the frequency error and its effect on Holdover Time Error drift. First, wait until the graph is "flat" and completely horizontal (theoretically, the lowest frequency offset possible). Then zoom in to 10 ns/Div and you will notice that the yellow line is not actually flat, but it is changing all the time with small amounts of short-term phase noise (mainly coming from the GNSS). You may wait until the average trend is zero (flat) to disconnect the antenna, in order to gett the lowest frequency offset possible. However, there is a long-term wander component to the GNSS that is difficult to characterize and take into account in the field.

For more detailed information on disciplining and holdover, refer to: The Use of Atomic Clock Disciplining with GNSS/GPS & Holdover for Synchronization and Time Error Testing on the Field.

The Use of Atomic Clock Holdover for Precision Time Applications in the Field - Quick Guide

Although using PRTC in Holdover mode isn't recommended for Precision Timing test applications; Time Error, Wander, One-Way Delay measurements are needed in places with no access to traceable reference clock. So, it's often seen as a fair alternative.