IEC 61869-9 expects a time accuracy better than ±1 µs and prefers PTP with the IEC/IEEE 61850-9-3 profile, or 1PPS in legacy installations. If the signal is lost, the merging unit enters holdover: it keeps sending samples with the required timing, SmpSynch does not change and the minimum duration is 5 s with stable temperature. In free-running mode, SmpSynch is zero and the rate does not deviate by more than ±100×10⁻⁶.
Short answer
The samples of a merging unit are only useful for comparing phases if they are all referred to the same time. That is why IEC 61869-9, 2016 edition, devotes a whole clause to synchronization: where the time comes from, how it is encoded in each message and what the merging unit must do when the time is lost.
There are three states. Synchronized, in holdover and in free-running mode. Each one is recognized in the message by the SmpSynch attribute.
Where the time comes from
Clause 6.904.1 says the accuracy of merging units depends on time synchronization. They must be able to accept an external signal to synchronize their sampling with other merging units and with a time reference.
- Preferred method: the Precision Time Protocol, PTP (IEC 61588:2009, also known as 1588), with the profile subset of IEC/IEEE 61850-9-3.
- Legacy method: a one-pulse-per-second signal (1PPS), as an alternative to PTP or in place of it.
In both cases the accuracy of the time signal, measured as the mean error with respect to absolute time, is expected to be better than ±1 µs for accuracy characterization. The phase error in service gets worse according to how far the delivered time departs from that ideal. The standard refers to IEC 61850-5, table 3, for the time synchronization classes, which for synchronized sampling are usually T4 and T5.
The relationship between PTP, clocks and redundancy is covered in the guides on PTP IEEE 1588 power profile and on GNSS clock, holdover and redundancy.
With PTP, the merging unit distinguishes the origin of the signal by its clockClass. A signal with clockClass 6 or 7 comes from a global-area clock. Any other comes from a local-area clock. The 1PPS signal says nothing about its source, and so it is always interpreted as coming from a global clock.
The SmpCnt counter
The merging unit has an internal clock synchronized by the external signal. A sample counter, SmpCnt, identifies the samples within the current second and encodes their instant. It counts from zero up to the nominal number of samples per second minus one, and starts over. The instant of a sample is the fraction of a second of the internal clock that equals the counter value divided by the nominal rate.
With the merging unit synchronized, the counter is zero at the mark of the start of each second.
SmpCnt is a 16-bit unsigned integer, with values from 0 to 65,535. At 96 kHz, the rate for high-bandwidth DC control, the counter wraps to zero 65,536 samples after the second mark. Then it continues up to 30,463, which is 96,000 minus 65,536 minus 1. That way it represents the lower 16 bits of a 17-bit internal counter.
SmpSynch: what time the samples have
Clause 6.904.4 defines the SmpSynch attribute of the message:
| SmpSynch | Meaning |
|---|---|
| 2 | Samples synchronized to a global-area clock, with the accuracy the measuring class requires. Examples from the standard: GPS, NTP or time servers of national institutes |
| 1 or 5 to 254 | Synchronized to a local-area clock. The number from 5 to 254 identifies that clock if known |
| 0 | Not synchronized with the accuracy the class requires |
Samples synchronized to the same global-area clock are synchronized with each other. Those of the same local clock are too, but may not be with those of another local clock. The value 1, local unidentified, depends on the design of the time network: if all merging units receive from the same local clock, they are synchronized with each other; if they can receive from different clocks, it cannot be assumed.
SmpSynch is 0 when the signal was never received, when it was interrupted and the merging unit exceeded its holdover, when it failed to lock to the signal or because of another condition that leaves the samples without the required synchronization.
The three states of the merging unit
- Synchronized. SmpSynch at 2 (global clock) or 1 and 5 to 254 (local clock)
- The external signal is lost. It enters holdover
- Holdover. At least 5 s with stable temperature; SmpSynch does not change; SmpCnt continues as if there were a signal
- End of holdover without signal. SmpSynch goes to 0
- Free-running mode. Rate with maximum deviation of ±100×10⁻⁶; SmpCnt keeps counting
Holdover (clause 6.904.5). While it lasts, the merging unit keeps sending samples with the timing its accuracy class requires. SmpSynch does not change and SmpCnt increments and wraps as if the signal were present. The minimum duration is 5 s with stable temperature. If the signal returns before the end of holdover, messages continue as if the signal had been continuous.
Free-running mode (clause 6.904.6). When the merging unit is not synchronized, it sends samples at a rate whose maximum deviation from the nominal does not exceed ±100×10⁻⁶. SmpSynch is zero and SmpCnt keeps counting. Even if it is not synchronized with the outside, all the samples of one merging unit are synchronized with each other.
Time adjustments
Clause 6.904.7 covers three events: when the signal is restored after an interruption, when switching from one external source to another and when the external clock makes an adjustment. In all three there can be a difference between the time the merging unit had and the new one.
The adjustment is made like this:
- sampling jumps from the old time to the new one between two consecutive samples;
- the interval at the jump must be between one half and one and a half times the nominal interval;
- SmpCnt is discontinuous at the jump if the adjustment is larger than what an interval off the nominal can absorb, but it is always continuous in the samples immediately before and after the adjustment;
- the sample following the jump carries the adjusted values of SmpCnt and SmpSynch.
An adjustment larger than half the nominal interval causes a discontinuity in SmpCnt and a change in SmpSynch. Subscribers that use data during a change of synchronization state must tolerate that transition. The manufacturer declares in the PIXIT document the maximum time it takes to achieve synchronization at power-up and on resynchronization.
How it is verified
The loss of synchronization test is in clause 7.2.903. It verifies, in the worst case, that:
- on losing the signal, the merging unit keeps sending samples with the required timing during the published holdover;
- in that period SmpSynch does not change and SmpCnt keeps incrementing and wrapping;
- before the timing stops meeting the declared measuring class, SmpSynch goes to zero;
- when powered on without signal and after a suitable interval after the signal ceases, the rate does not deviate by more than ±100×10⁻⁶;
- with PTP, SmpSynch follows the transitions between a global clock and at least two local clocks.
For merging units with 1PPS, clause 7.2.904 verifies that SmpSynch is 2 with a valid signal and 0 without it. How a loss of synchronization is diagnosed in the substation network is explained in the guide on loss of PTP or IRIG-B time synchronization.
What not to do
- Do not assume 5 s of holdover is enough for a clock failure. It is the minimum; the manufacturer publishes its own.
- Do not compare samples from two merging units without looking at SmpSynch.
- Do not confuse SmpSynch at 1 with SmpSynch at 2. The 1 is a local clock with no guaranteed relation to other clocks.
- Do not ignore how your relay treats a resynchronization: the standard says subscribers must tolerate that transition.
Related topics
Synchronization is linked to delay, the sampling rate, sample quality and LPIT faults. The time reference is explained in GNSS clock and merging unit.
When to call us
At Tiza Group we design and test the synchronization of the digital substation: time sources, redundancy and the behavior of protection in each state of the merging unit. We offer IEC 61850 engineering and end-to-end testing, in Colombia or elsewhere on request. We quote the work according to scope.
Related guides
- Merging unit maximum processing delay
- Merging unit sampling rates: F4800S2I8U0
- Merging unit sample quality: bits and validity
- Merging unit: IEC 61869-9, 9-2LE and testing
- Substation GNSS clock: holdover, redundancy and risks
References
- IEC 61869-9, Instrument transformers - Part 9: Digital interface for instrument transformers, edition 1.0 of 2016, clauses 3.5.901, 3.5.904, 6.904.1 to 6.904.7, 7.2.903 and 7.2.904.
- IEC 61588:2009 and IEC/IEEE 61850-9-3, to which clause 6.904.1 refers.
Frequently asked questions
How long must the holdover of a merging unit last at a minimum?
Under IEC 61869-9, at least 5 s with stable temperature. During that time the merging unit keeps sending samples with the timing its accuracy class requires, SmpSynch does not change and SmpCnt keeps counting as if the synchronization signal were present. The manufacturer publishes its holdover duration.
What value does SmpSynch have depending on the time source?
It is 2 when samples are synchronized to a global-area clock, 1 or a number from 5 to 254 when synchronized to a local-area clock, and 0 when they are not synchronized with the accuracy the accuracy class requires. With PTP, a signal with clockClass 6 or 7 is considered to come from a global clock.
What happens to SmpCnt when the clock makes a time adjustment?
Sampling jumps from the old time to the new one between two consecutive samples. If the jump fits in an interval between 0.5 and 1.5 times the nominal one, SmpCnt stays continuous. If the adjustment is larger, SmpCnt is discontinuous at the jump and SmpSynch changes, but it is continuous just before and just after.