[ntp:questions] Re: CDMA and Leap Seconds

David L. Mills mills at udel.edu
Sun Jan 1 11:46:15 UTC 2006


Bruce,

Your candid explanation is much appreciated. On the other hand, 
everybody else got it right and your unit didn't. There is an obvious 
remedy here. If your unit implements Autokey, and it does implement just 
about everything else, it could run in TAI and deliver the UTC offsets 
in an extension field. I would be happy to collaborate on an RFC to that 
effect.

Dave

Bruce Penrod wrote:
> Apparently a little explanation of how the CDMA mobile phone system 
> keeps time would be welcome.  As most of you are aware, each basestation 
> in the system must maintain absolute synchronization to GPS time to the 
> 10 us level in order to keep from interfering with the other 
> basestations.  To do this, each has a GPS timing receiver with high 
> stability oscillator, either quartz or rubidium.  This so as to be able 
> to maintain that level of sync even during hours of holdover after the 
> antenna has been shot off the tower.
> 
> UTC time is not used by the CDMA system for obvious reasons, e.g. 
> leapseconds.  The GPS timescale is strictly monotonic.  However, in 
> order to make the phones show the correct local time, embedded in the 
> sync channel message which all basestations broadcast continuously, is 
> the value of the UTC offset to GPS time and the local offset to UTC, 
> i.e. timezone.  However, since these offset values do not affect the 
> operation of the mobile phone system, great care was not taken to 
> implement the changeovers rigorously.
> 
> Though until now there have been no leap second insertions since we 
> began shipping our CDMA based products, we have had over the years a 
> handful of customers report basestations transmitting leap second values 
> that were off by one second.  In response, three years ago we created a 
> user enterable leap second mode for our CDMA products to allow 
> overriding the system transmitted value if necessary.  This mode also 
> included the ability to set up the future leap second value prior to the 
> next insertion so as to be independent of the behavior of the 
> basestation.  This has all been well documented in the product manuals 
> and a webpage dedicated to leap seconds has been maintained continuously 
> on our site to show the current and future values of the leap second so 
> that customers may easily determine how to configure their boxes.
> 
> When the IERS Bulletin C arrived in July, we updated our website and 
> notified via e-mail all CDMA product customers of the impending leap 
> second, and recommended that all users operate their NTP servers in the 
> user leap second mode.  All NTP servers we shipped after July 1 were 
> preconfigured to user leap second mode with the current and future leap 
> seconds set appropriately.
> 
> We are pleased to announce that our GPS NTP servers and our CDMA NTP 
> servers configured in user leap mode performed the leap properly this 
> afternoon.  We are aware that some of the cellular basestations set the 
> leap second a day early, and some PCS basestations have still not set 
> it.  We regret that some customers apparently had not configured their 
> NTP servers to operate in user leap second mode, and apologize for any 
> problems this might have caused.
> 
> Unfortunately, judging from the other recent posts, it appears to me 
> that the long dearth of leap seconds may have resulted in a step 
> backwards in the proper handling of leap seconds.  Many products that 
> had been debugged back when we were having regular leap second events 
> have been replaced by new ones that have no experience with them.  I'm 
> not optimistic that we'll ever have a ho-hum leap second event with all 
> posting to say how smoothly everything went.
> 
> If we're lucky, we'll get rid of the pesky things before there's another 
> one...
> 
> Bruce Penrod
> EndRun Technologies
> 




More information about the questions mailing list