I have also heard the 1us number for substations. Substations are relatively small, engineered networks. Probably the 1us can be met by PTP, with the default profile, and some guidelines about network design, including the recommended use of transparent clocks or boundary clocks for switches. This is very important, but I don't see a role for the IETF. I have heard another requirement, which is more relevant to the IETF. That is a requirement for 1 ms timing accuracy, CONSISTENTLY, across the communications infrastructure of regional power grids. Such a grid might contain several SONET rings, as well as Ethernet and numerous routers in the path between master and slave. 1 ms sounds easy, except the power utility would want to count on it. Besides a lot of queuing delay variation, such networks likely have large asymmetries. Neither NTP, nor PTP can meet this spec consistently without the the "on path support" (whatever that turns out to be) which TICTOC promises. //Doug Arnold
________________________________ From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of [EMAIL PROTECTED] Sent: Monday, September 22, 2008 8:21 AM To: STUART VENTERS Cc: [email protected] Subject: Re: [TICTOC] Requirements draft - call for assistance/power systems Hi Stuart, The goal of time correlation accuracy in sub-stations for power distribution systems is 1uS. Across power distribution networks per power generating station is 10uS. These goals are the targets defined by EPRI, Electric Power Research Institute. In reality the needs for correlation come down to as Stuart says looking at the slope of the change. The goal is to identify the voltage event. The frequency is derived from the rotation of the armature in the power plant. Quick and sudden changes in voltage drop across distribution environments are the real tough ones to spot before the become breaker trips. The networks used are almost always private. They are mostly SNA types many over frame relay with spoofing routers at the end points. As you can imagine the post event processing is almost useless in regards to intelligent power grid "self healing" grids. The use of the public internet to carry this mission critical data is not going to happen given todays operating paradigms. Regards, Pat "STUART VENTERS" <[EMAIL PROTECTED]> 09/22/2008 07:41 AM To <[EMAIL PROTECTED]> cc <[email protected]> Subject [TICTOC] Requirements draft - call for assistance/power systems Pat, (or any power expert listening) For power system timing requirements, any idea what sort of accuracy is desired or in common practice? Without clues, as a starting point, my guess would be some combination of the following: 1) For correlating meter readings across the grid for billing purposes. (Perhaps 1 second would be good enough for this?) 2) For correlating significant events (like breaker trips) over the grid to aid in determining the root cause of a failure. (Perhaps 1 to 10 mS would be good enough for this?) 3) For correlating voltage and current phase angles over the grid. (At 60Hz, 4.6 uS would be 1/10 of a degree?) But it seems like we should have some additional clues as to if these numbers are reasonable and if these or other applications are what is actually needed before adding specific requirements Also, what sort of packet network would it need to maintain the accuracy over. (On a LAN inside the substation, outside over a dedicated link, other?) Regards, Stuart -----Original Message----- ---------------------------------------------------------------------- Message: 1 Date: Thu, 18 Sep 2008 00:44:01 -0700 From: [EMAIL PROTECTED] Subject: Re: [TICTOC] Requirements draft - call for assistance To: Yaakov Stein <[EMAIL PROTECTED]> Cc: "[email protected]" <[email protected]> Message-ID: <[EMAIL PROTECTED]> Content-Type: text/plain; charset=US-ASCII Hi Yaakov, I can comment on the electric power and instrumentation bit. Each power grid sub-station has a large number of monitoring devices. These devices historically have been served time via IRIG-B distributed time systems. The substations will typically have a gps receiver for time and 1pps to drive the IRIG generators. the backhaul networks are usually private circuit based networks. In past years many of these were on SNA or some other similar network. As these substations are upgraded they are being implemented with local LAN's and will have time distribution systems based on 1588 or ntp. This allows the removal of the IRIG systems. The backhaul networks can then be migrated to packet based with whatever access makes sense for each location. The time distribution systems are local to the substations. In test and measurement the LXI forum, LAN extensions for Instrumentation, has mandated the use of 1588 for compliance. Again this is a local time distribution paradigm with a local time source. Many of the "time" based event correlation systems use a local source of time, typically gps, for time stamps. This is then sent in non-real time to the monitoring systems for overall system correlation processing. I agree with your thought that most of these are not applicable to the TICTOC work. Hopefully this info is helpful. Pat
_______________________________________________ TICTOC mailing list [email protected] https://www.ietf.org/mailman/listinfo/tictoc
