On Mon, May 20, 2019 at 07:18:22AM -0400, Sanjay Bhandari wrote: > Thanks for the response. Between what you wrote and some more spelunking, I > think I finally understand. I'll summarize it here in case it's useful to > others.
That is a good summary. > 1. The PPS from the GPS that you feed into the PHC has nothing to do with > the kernel PPS system. It is simply an input on one of the pins of the PHC. > The PHC has the capability of timestamping such an input event (once > programmed correctly). It timestamps the event with the PHC time. This > timestamp then effectively gives you the offset between the GPS time and > PHC time. Which can be used to train the PHC clock to match the GPS time. > > 2. The PPS that the PHC itself generates is fed into the kernel PPS system, > and that works as usual whereby the kernel timestamps the PPS event with > the timestamp from the sysclock. This gives you the offset between the PHC > and sysclock, which can be used to train the sysclock to match the PHC > (which is what phc2sys does). Note that phc2sys can (and usually does) work without that kernel PPS event. In fact, depending on ISR latency, the synchronization probably works better without the PPS. New Intel CPUs and MACs can use the ART to synchronize the two in hardware, and phc2sys does support this mode. Thanks, Richard _______________________________________________ Linuxptp-users mailing list Linuxptp-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxptp-users
