QuestionHow Does a Leap Second Affect My GPS Time Server
AnswerIn the nominal case of a Spectracom master clock / network time server synchronizing its system time to GPS to serve NTP requests to clients, there is no user interaction required to ensure correct handling and operation during a leap second event. They automatically detect the leap second indication in the GPS signal and pass the indicator according to the NTP convention for proper handling by the NTP client's operating system.
Testing downstream devices such as NTP clients
We recommend that you evaluate the ability of the NTP client to successfully manage the leap second event and any possible impact on your application. Spectracom's SecureSync and NetClock 9400 models can be used for leap second testing in a test network/laboratory environment. A GPS simulator, such as Spectracom's GSG series, can also be used to test the impact of a leap second event on any GPS master clock or time server.
Additional information about Spectracom master clocks and time servers
Contact us for more information about the capabilities of some Spectracom devices:
- Checking for the presence of a pending leap second once it is included in the GPS signal.
- The affect of the leap second on other, non-GPS, external timing references and/or user-configurable timescales that may require intervention during a leap second.
At the time of a positive leap second event, UTC midnight of the designated day, the leap second is inserted in the UTC timescale as an extra second:
23:59:60 <--- Leap Second
Any device connected to the master clock that uses the UTC timescale explicitly will show 23:59:60 such as the front panel display or other digital display clock and most time codes. However there is no allowance for an extra second in the design of NTP (1). Essentially a positive leap second makes NTP time stop for the duration of the second. If you query a NTP server multiple times within that second, you will see the same time. Note that this will never happen in normal operation since a client never queries a server that fast. Then it is solely the role of the operating system to either ignore a leap second and deal with a one second time difference immediately after the leap second event like it would any other time jump, or make use of advanced leap second notification to perform some specific action (2) such as step back time by a second (with obvious problems), hold the time for the second (most accurate), or skew the time before and after a leap second to minimize the discontinuity.
references to NTP.org:
(1) Detail description about the NTP timescale and leap seconds (get all the way to the end, it is worth the read for anyone managing a precision timing system.
(2) What happens during a leap second summarizes the issues with the OS' managing of a leap second and provides additional references for kernel clocks and kernel disciplining. in an OSfor more about A leap second insertion NTP.org documents 2 scenarios proper handling of an NTP client's operating system.