Hi A number of the chip guys will sell you micro's that have one or more 1588 ethernet ports on them. Some of them are in the sub $10 range. With a bit of Time Nut attention, they seem to be capable of well under 1 us performance. If you have a daisy chain network they will do what you need without any switches. As mentioned above, if your network has modern (gigabit rather than 10 megabit) switches on it, you can likely get to 1 us without any fancy time stamping switches. There are a few other minor details you would want to pay attention to, but they are manageable for a shop floor network. Bob On Jul 31, 2013, at 12:19 AM, Bill Hawkins <bill@iaxs.net> wrote: > Group, > > Has anyone used IEEE 1588 to synchronize clocks on an Ethernet network? > > I was involved in the design of time sync for Foundation Fieldbus circa > 2000. > We needed one millisecond accuracy, so we went with SNTP on local > networks. > I've just seen an ad for a switch that can do 1588, and looked up what > it does. > > Microsecond accuracy is impressive, but what does it cost? > > Industrial sensors are generally sampled at about 10 millisecond > intervals out > to several seconds. SNTP would appear to be very adequate for time > stamps as > there is uncertainty introduced by when the computer gets around to > sampling > the sensor in its sampling and control cycle. > > Any thoughts appreciated. > > Bill Hawkins > > _______________________________________________ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there.

This is true, but what´s needed to generate the sync packets? I think they call them "Grand Masters", and cost a lot of money... Is there any cheap approach for this end of the problem? Daniel Em 31/07/2013 08:03, Bob Camp escreveu: > Hi > > A number of the chip guys will sell you micro's that have one or more 1588 ethernet ports on them. Some of them are in the sub $10 range. With a bit of Time Nut attention, they seem to be capable of well under 1 us performance. If you have a daisy chain network they will do what you need without any switches. > > As mentioned above, if your network has modern (gigabit rather than 10 megabit) switches on it, you can likely get to 1 us without any fancy time stamping switches. There are a few other minor details you would want to pay attention to, but they are manageable for a shop floor network. > > Bob > > On Jul 31, 2013, at 12:19 AM, Bill Hawkins <bill@iaxs.net> wrote: > >> Group, >> >> Has anyone used IEEE 1588 to synchronize clocks on an Ethernet network? >> >> I was involved in the design of time sync for Foundation Fieldbus circa >> 2000. >> We needed one millisecond accuracy, so we went with SNTP on local >> networks. >> I've just seen an ad for a switch that can do 1588, and looked up what >> it does. >> >> Microsecond accuracy is impressive, but what does it cost? >> >> Industrial sensors are generally sampled at about 10 millisecond >> intervals out >> to several seconds. SNTP would appear to be very adequate for time >> stamps as >> there is uncertainty introduced by when the computer gets around to >> sampling >> the sensor in its sampling and control cycle. >> >> Any thoughts appreciated. >> >> Bill Hawkins >> >> _______________________________________________ >> time-nuts mailing list -- time-nuts@febo.com >> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts >> and follow the instructions there. > _______________________________________________ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there.

Cesium-based grandmasters can cost a lot of money but I think that GPSDO-based ones are a lot cheaper. On Thu, Aug 1, 2013 at 11:48 PM, Daniel Mendes <dmendesf@gmail.com> wrote: > > This is true, but what´s needed to generate the sync packets? I think they > call them "Grand Masters", and cost a lot of money... Is there any cheap > approach for this end of the problem? > > Daniel > > Em 31/07/2013 08:03, Bob Camp escreveu: >> >> Hi >> >> A number of the chip guys will sell you micro's that have one or more 1588 >> ethernet ports on them. Some of them are in the sub $10 range. With a bit of >> Time Nut attention, they seem to be capable of well under 1 us performance. >> If you have a daisy chain network they will do what you need without any >> switches. >> >> As mentioned above, if your network has modern (gigabit rather than 10 >> megabit) switches on it, you can likely get to 1 us without any fancy time >> stamping switches. There are a few other minor details you would want to pay >> attention to, but they are manageable for a shop floor network. >> >> Bob >> >> On Jul 31, 2013, at 12:19 AM, Bill Hawkins <bill@iaxs.net> wrote: >> >>> Group, >>> >>> Has anyone used IEEE 1588 to synchronize clocks on an Ethernet network? >>> >>> I was involved in the design of time sync for Foundation Fieldbus circa >>> 2000. >>> We needed one millisecond accuracy, so we went with SNTP on local >>> networks. >>> I've just seen an ad for a switch that can do 1588, and looked up what >>> it does. >>> >>> Microsecond accuracy is impressive, but what does it cost? >>> >>> Industrial sensors are generally sampled at about 10 millisecond >>> intervals out >>> to several seconds. SNTP would appear to be very adequate for time >>> stamps as >>> there is uncertainty introduced by when the computer gets around to >>> sampling >>> the sensor in its sampling and control cycle. >>> >>> Any thoughts appreciated. >>> >>> Bill Hawkins >>> >>> _______________________________________________ >>> time-nuts mailing list -- time-nuts@febo.com >>> To unsubscribe, go to >>> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts >>> and follow the instructions there. >> >> _______________________________________________ >> time-nuts mailing list -- time-nuts@febo.com >> To unsubscribe, go to >> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts >> and follow the instructions there. > > > _______________________________________________ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there.

Hi It all depends …. If you need to trace back to an ultimate standard, then yes you need some sort of uber master. If having everything within 1 us of everything else is fine, then no you don't need some sort of uber master. You do need to declare one of your gizmos to be the authority on what time it is. If nobody is in charge, then things get a bit nutty. Bob On Aug 1, 2013, at 5:48 PM, Daniel Mendes <dmendesf@gmail.com> wrote: > > This is true, but what´s needed to generate the sync packets? I think they call them "Grand Masters", and cost a lot of money... Is there any cheap approach for this end of the problem? > > Daniel > > Em 31/07/2013 08:03, Bob Camp escreveu: >> Hi >> >> A number of the chip guys will sell you micro's that have one or more 1588 ethernet ports on them. Some of them are in the sub $10 range. With a bit of Time Nut attention, they seem to be capable of well under 1 us performance. If you have a daisy chain network they will do what you need without any switches. >> >> As mentioned above, if your network has modern (gigabit rather than 10 megabit) switches on it, you can likely get to 1 us without any fancy time stamping switches. There are a few other minor details you would want to pay attention to, but they are manageable for a shop floor network. >> >> Bob >> >> On Jul 31, 2013, at 12:19 AM, Bill Hawkins <bill@iaxs.net> wrote: >> >>> Group, >>> >>> Has anyone used IEEE 1588 to synchronize clocks on an Ethernet network? >>> >>> I was involved in the design of time sync for Foundation Fieldbus circa >>> 2000. >>> We needed one millisecond accuracy, so we went with SNTP on local >>> networks. >>> I've just seen an ad for a switch that can do 1588, and looked up what >>> it does. >>> >>> Microsecond accuracy is impressive, but what does it cost? >>> >>> Industrial sensors are generally sampled at about 10 millisecond >>> intervals out >>> to several seconds. SNTP would appear to be very adequate for time >>> stamps as >>> there is uncertainty introduced by when the computer gets around to >>> sampling >>> the sensor in its sampling and control cycle. >>> >>> Any thoughts appreciated. >>> >>> Bill Hawkins >>> >>> _______________________________________________ >>> time-nuts mailing list -- time-nuts@febo.com >>> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts >>> and follow the instructions there. >> _______________________________________________ >> time-nuts mailing list -- time-nuts@febo.com >> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts >> and follow the instructions there. > > _______________________________________________ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there.