[time-nuts] IEEE1588

[Magnus Danielson]

On 01/02/12 18:52, Chris Albertson wrote:
> On Wed, Feb 1, 2012 at 9:20 AM, Bob Camp<lists at rtty.us>  wrote:
>> Hi
>>
>> The "gotcha" with a public grandmaster is routing to it. Without 1588
>> routers / hubs / switches / what ever, the result is compromised. You fall
>> back into the same routing delay mess as NTP. Since public pretty much means
>> internet accessible, you would need to upgrade a lot of stuff. Since that
>> includes the traditional "last mile" gear, it's not a simple proposition at
>> all. At least in my case, I get more delay from the ends of most connections
>> than I get from all the routing in-between.
>
> I think except for testing you need the grand master to live on your
> own Ethernet.   And I mean "Ethernet" in the most precise way.  Two
> Ethernets with a router between them is not an Ethernet it is two
> Ethernets with a router.     sub-uSec synchronization is very hard
> over over any network if packets get stored and forwarded and that is
> what routers do.
>
> NTP is better if the network path is complex and you don't have
> end-to-end control over all the equipment.
>
> I think it's best to start from requirements rather then starting from
> solutions.   But if your requiremt is to get time to within a few uSec
> and all the machines are in the same lab PTP can work.
>
> Next you have to figure out some way to TEST is the machines are all
> in fact in sync at the uSec level and if yu are using PTP the test
> should not also use PTP.   Testing, I think is the hardest part.

Do you want a plain Ethernet switch or a IEEE 1588 capable Ethernet 
switch? Which PTP capable switch did you choose?

Cheers,
Magnus