On Thu, Oct 25, 2018 at 05:41:29PM +0530, Dolly Gyanchandani wrote: > A separate and more accurate time source is necessary. For measuring > the accuracy of SW timestamping you can use HW timestamping. > > What do you mean by this? Do we need to have PTP enabled NICs to be able to > test software timestamping? Not sure if we understand this point correctly. > Could you elaborate a bit?
A synchronized PTP clock is a good reference for measuring the accuracy of the system clock synchronized with SW timestamping. With SW timestamping a typical error is in microseconds or tens of microseconds and HW timestamping is usually good to hundreds of nanoseconds. > Thanks for your inputs. They were of great help. When we ran ptp4l with > hardware timestamping, without synchronization of the system clock with the > hardware clock on the master node, we are able to get around 25 nanoseconds > offset between the master and slave clock which seems to be a good and > desired value. > ptp4l[304.918]: master offset -39 s2 freq +80417 path delay > 1568 > ptp4l[305.919]: master offset 3 s2 freq +80447 path delay > 1568 > ptp4l[306.919]: master offset -14 s2 freq +80431 path delay > 1564 > ptp4l[307.919]: master offset 27 s2 freq +80468 path delay > 1563 The offset indicates the clock is stable to about 25 nanoseconds and the delay gives an upper bound on the error. If you don't know how much asymmetry is there in the timestamping and the network, you can tell only that this clock is accurate to 1.5 microseconds relative to the master, or grandmaster if synchronized directly to it. -- Miroslav Lichvar _______________________________________________ Linuxptp-users mailing list Linuxptp-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxptp-users
