[time-nuts] Lost GPS lock or 1PPS recently?

jimlux jimlux at earthlink.net
Fri Sep 7 19:51:17 EDT 2018

On 9/7/18 4:01 PM, Scott McGrath wrote:
> You are SO convinced that GPS will ALWAYS be there,   I’m NOT (Think Carrington Event) and i’ve been part of a few disaster exercises where both Internet and GPS were considered ‘down’ for the exercise and these exercises are done in conjunction with the military so PPS was also ‘off the table’.
> It was quite an eye opener to see how many networks could not keep time synchronized within 5 minutes much less 5 seconds because of the cheap XO’s used in servers and workstations(NTP will ALWAYS BE AVAILABLE).  Just like the old Sun workstations.

I assume you're using network in the sense of "group of computers that 
has to keep common time, but are not physically connected to the same 
ethernet", because NTP *will* keep everyone together - *if properly 

I suspect the latter is not always true - people assuming that 
time.microsoft.com is always available, for instance.

In which case the exercise where "external internet" goes away is a 
valuable one, because it exposes the vulnerability.

But in general, if you've got physical infrastructure to connect the 
boxes via network, keeping within "seconds" should be straightforward. 
Appropriate failover time sources, etc.

Yes, GPSDOs provided an "easy" way to get good time accuracy across a 
geographical area without having to think about disaster recovery. It 
solves the "simultaneity of database updates" problem for a few thousand 
bucks. While hiding the bigger synchronization issue under the rug.

> It was also fun watching the multimillion dollar Harris radios drift once they no longer had a 10 Mhz input from a GPSDO.   One would think the local timebase would be a bit better than it was.

But what sort of requirement is there?  Obviously, it's not the NIST HF 
SSB of 20Hz (roughly 1 ppm), because I'd find it hard to believe a GPSDO 
couldn't hold 1ppm for a year, without a GPS signal.

> The older ‘Pacer Bounce’ and Falcon series radios did much better because they had good local timebases and made no assumptions of the availability of a external timebase.  Whereas the new radios depend upon it.

But nothing says that you can't have a reliable external timebase - an 
OCXO and a battery backup, if nothing else.

> These exercises are intended to practice restoring government communications after a large scale natural disaster.    And without readily available precision time it aint easy.

Yes - it's non trivial - it's well out of most people's "I know how to 
do that" bucket, and needs a significant amount of network and system 
knowledge and "block to block" interaction knowledge that typically 
doesn't exist.

> Its also fun watching executives realizing that their phone during the exercise is a paperweight useful only in weighting down stacks of Form 213’s
> Its not for nothing that Symmetricom is building more 5071’s than HP/Agilent ever did.

I wonder if CSAC devices will have significant traction - The CSAC 
provides ppb sort of accuracy out of the box without having to think 
about it, and doesn't cost anywhere near what a 5071 does.


More information about the time-nuts mailing list