Hi Martyn, I work with other people in the timing system for CERN's particle accelerator complex in Geneva (Switzerland), which includes the Large Hadron Collider (LHC), hopefully to start operation in spring 2008. Our timing system consists basically of a lot of complicated high level software orchestrating the transfer of particles from one accelerator to another, and the subsequent acceleration cycles. Hardware-wise, we have home-made VME modules which receive the 10 MHz and PPS signals from a commercial GPSDO (Symmetricom XLi) and use these two signals to encode messages (as decided in advance by the high level software) on a serial link. Using the 10 MHz, the second is broken in millisecond slots and then further in 125 microsecond slots. Each 125 us slot can carry a message signaling some event to occur in the accelerators. On the receiving side, we have home-made PCI, PMC and VME boards which can react to any of these messages by starting a counter, generating an output pulse on their front panel or an interrupt on the bus. This allows hardware and software synchronization all around the complex (the LHC is 27 km in circumference). Now, to come to your question: we rely heavily in the PPS containing exactly 10 million ticks of the 10 MHz. All our UTC time-tagging logic is based on that, so we definitely need a stable phase between the 10 MHz and the PPS. But as Tom says, I thought you got that by design in every GPSDO. Maybe we have misunderstood the question? Cheers, Javier

At 06:24 PM 5/25/2007 +0200, you wrote: >Hi Martyn, > >I work with other people in the timing system for CERN's particle >accelerator complex in Geneva (Switzerland), which includes the Large >Hadron Collider (LHC), hopefully to start operation in spring 2008. Our >timing system consists basically of a lot of complicated high level >software orchestrating the transfer of particles from one accelerator to >another, and the subsequent acceleration cycles. Hardware-wise, we have >home-made VME modules which receive the 10 MHz and PPS signals from a >commercial GPSDO (Symmetricom XLi) and use these two signals to encode >messages (as decided in advance by the high level software) on a serial >link. Using the 10 MHz, the second is broken in millisecond slots and >then further in 125 microsecond slots. Each 125 us slot can carry a >message signaling some event to occur in the accelerators. On the >receiving side, we have home-made PCI, PMC and VME boards which can >react to any of these messages by starting a counter, generating an >output pulse on their front panel or an interrupt on the bus. This >allows hardware and software synchronization all around the complex (the >LHC is 27 km in circumference). > >Now, to come to your question: we rely heavily in the PPS containing >exactly 10 million ticks of the 10 MHz. All our UTC time-tagging logic >is based on that, so we definitely need a stable phase between the 10 >MHz and the PPS. But as Tom says, I thought you got that by design in >every GPSDO. Maybe we have misunderstood the question? I read his original message to say his customers were asking for coicidence of pulse edge, rather than a finite (not zero) phase relationship. >Cheers, > >Javier > >_______________________________________________ >time-nuts mailing list >time-nuts@febo.com >https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts