On Fri, 20 Aug 1999, Alex P. Rudnev wrote:

Through I did not see anything to worry about - your GPC may be show you the wrong date, but why it can affect the accuracy at all (except some short time around the very moment itself).

This is not true. There is things called the catelog and ephemerus, which are time based and give detailed corrections of the orbital position of the satellites that GPS derives locaton from. If the receiver does not handle rollover correctly, it will not correctly return time or position. Most receivers built since the mid 90s have handles this, and even more of the precision time sources have handled this, but nothing is perfect. It's really easy for the manufacturer to test this, but almost impossible for a user to (you need a GPS simulator.) The good thing about modern NTP systems is that they don't accept times that are way off (there was a bad incident of a wacko clock many years ago) so if the GPS reports a 1980 date, the software would not believe it. That would mean losing synch with the GPS, but that should not be the end of time for a reasonalbly configured system. I'll be watching all my clocks, but a lot of people won't be. jerry

Most telcos do use use GPS for timing, however they also use Cesium standards for backup also. Don't forget CDMA systems use GPS/Cesium for clocking also, and CDMA won't work at all without accurate clocking. There are only about 2 companies in the world making precision time/frequencey references for high speed telecommunications networks. The NTP term "stratum" is derived from the old primary refence clocks used by the Bell network to provide timing. The little birdies tell me that some of this equipment failed its week rollover test the first time, but patches were quickly made. Hopefully all the telco's using the equipment read the engineering field notices.... In message <19990820125326.F24455@puck.nether.net>, Jared Mauch writes:

I'm interested in knowing if there are any telcos that are using a GPS for their ckt timing, and this will cause that timing to break, and those of us that take "clock source line" from M13's, etc.. will have problems with our channelized ckts (dial, ct3, etc..?)

Anyone here privy to that type of information, and can you comment?

- jared

On Fri, Aug 20, 1999 at 09:42:30AM -0700, Jerry Scharf wrote:

...

On Fri, 20 Aug 1999, Alex P. Rudnev wrote:

...

Through I did not see anything to worry about - your GPC may be show you the wrong date, but why it can affect the accuracy at all (except some short time around the very moment itself).

This is not true. There is things called the catelog and ephemerus, which are time based and give detailed corrections of the orbital position of the satellites that GPS derives locaton from. If the receiver does not handle rollover correctly, it will not correctly return time or position. Most receivers built since the mid 90s have handles this, and even more of the precision time sources have handled this, but nothing is perfect. It's really easy for the manufacturer to test this, but almost impossible for a user to (you need a GPS simulator.)

The good thing about modern NTP systems is that they don't accept times that are way off (there was a bad incident of a wacko clock many years ago) so if the GPS reports a 1980 date, the software would not believe it. That would mean losing synch with the GPS, but that should not be the end of time for a reasonalbly configured system.

I'll be watching all my clocks, but a lot of people won't be.

-- Jared Mauch | pgp key available via finger from jared@puck.nether.net clue++; | http://puck.nether.net/~jared/ My statements are only mine. | "Waste Management Consultant" VOYN

--- jerry@fc.net Freeside/ Insync Internet, Inc.| 512-458-9810 | http://www.fc.net #include <sys/machine/wit/fortune.h>

At my last job we had a Telecom Solutions setup. It consisted of a redundant receiver that spoke to the satelites and thus the cesium clocks and a redundant set of rhubidium clocks. If I recall correctly the rhubidium clocks were supplying the T1 interfaces with a clock source and they would adjust themselves off of the reference cesium clocks. According to the manufacturer, the whole thing could remain autonomous for a month if the cesium reference was unavailable. Very nice equipment. At 11:11 AM 8/20/99 -0700, Jerry Scharf wrote:

On Fri, 20 Aug 1999, Jeremy Porter wrote:

...

Most telcos do use use GPS for timing, however they also use Cesium standards for backup also. Don't forget CDMA systems use GPS/Cesium for clocking also, and CDMA won't work at all without accurate clocking. There are only about 2 companies in the world making precision time/frequencey references for high speed

telecommunications

...

networks. The NTP term "stratum" is derived from the old primary refence clocks used by the Bell network to provide timing.

The little birdies tell me that some of this equipment failed its week rollover test the first time, but patches were quickly made. Hopefully all the telco's using the equipment read the engineering field notices....

In message <19990820125326.F24455@puck.nether.net>, Jared Mauch writes:

...

I'm interested in knowing if there are any telcos that are using a GPS for their ckt timing, and this will cause that timing to break, and those of us that take "clock source line" from M13's, etc.. will have problems with our channelized ckts (dial, ct3, etc..?)

Anyone here privy to that type of information, and can you comment?

- jared

There are two types of timing that are used by telco networks, relative timing for signal recovery and absolute timing for timestamps, billing and the like. The one most people care about is the signal recovery timing. There has been a major move from the classic Bell time distribution system to independent GPS based timing sources, it's a whole load cheaper and as/more reliable. The idea behind these systems are that you have a stable oscillator (rhubidium or cesium (usually two of them in an oven) and something that figures out how far off of the desired frequency the oscillator is. The GPS is used as the "discipline" source since it's accuracy doesn't drop over time like the oscillator. The same basics are true for both land based and mobile base station clocks (as well as TV networks ...). Depending on the specifics of the failure mode for the GPS side, the clock could either fail completely, or loose the discipline and slowly drift off exact time.

The reality is that no one selling these systems (there are about a half a dozen now) hasn't done full testing against a simulator and gotten patches out to all their customers in case of problems. Also, since the problem has been well known in the GPS community for many years, these kinds of problems shouldn't be a surprise this week. Probably Y2K has helped the customers take this more seriously. I'm not too worried about my circuits still working after the rollover.

I think absolute time for billing is a whole different can of worms, and not subject to good generalizations. (If someone couldn't bill me for a while, I'd just have to live with it.)

jerry

Michael Heller Sr. Systems Engineer Earthweb, Inc. 212.448.4175 mikeh@earthweb.com