[ntp:questions] Trying to use Dimension 4 time keeper

W. eWatson wolftracks at invalid.com
Tue Oct 15 01:51:36 UTC 2013


On 9/9/2013 7:22 PM, Brian Inglis wrote:
> On 2013-09-08 23:22, unruh wrote:
>> On 2013-09-09, Brian Inglis <Brian.Inglis at SystematicSw.ab.ca> wrote:
>>> On 2013-09-08 11:51, W. eWatson wrote:
>>>> On 9/7/2013 9:16 PM, David Taylor wrote:
>>>>> On 07/09/2013 22:38, W. eWatson wrote:
>>>
>>> You may want to try the latest stable NTP release 4.2.6p5 from
>>> Meinberg, for
>>> ease of installation, on all three of your systems.
>>> With stable connections and remote NTP servers it should get your
>>> time offset
>>> within 100ms in a few hours, and within a few ms (or better depending
>>> on your
>>> network and sources) over a period of about three weeks, and the
>>> drift within
>>> PPB of your hardware clock over that period.
>>
>> ntpd is slow but not that bad. a few ms in a couple of hours. It is us
>> levels that tend to take a half day or so.
>
> We are talking about Windows here with remote network servers, not ref
> clocks, not Linux or BSD with kernel time discipline.
> It takes that long on Windows (7 64 bit) with current stable ntpd and a
> server selection which does not clock hop very much. I have a year of
> stats and graphs showing that. But that is much better than previous
> stable (which the OP is currently running) where the drift wandered
> +/-250ppB and the offset stayed around the 10ms range.
>
>>> If you are in the US, you may want to select some servers from
>>> http://www.ustiming.org/?page_id=458 or
>>> http://tf.nist.gov/tf-cgi/servers.cgi
>>> and keep the pool servers in ntp.conf as backup in case some of your
>>> selections
>>
>> Just use the pool for the accuracy he needs. (.1 sec) No need to
>> overload the bigger public servers.
I finally declared the slow clock PC useless, and transferred to the 
other PC in the same room, which keeps time very accurately.



More information about the questions mailing list