On Mon, Jul 01, 2019 at 08:10:43PM -0700, Richard Cochran wrote: > On Mon, Jul 01, 2019 at 03:53:20PM +0200, Armin HAMAR wrote: > > Like it's about 300ns for the PPS and -/+10ns for the output of linux PTP > > the delay is around 938 which looks > > for me pretty high for a point to point connection? > > No it isn't. You need to consider the delays from the MAC through the > PHY. These can be greater than one microsecond on the i210.
The igb driver compensates the delays for the I210. So, with two directly connected I210s (short cable) the measured delay should be close to zero. It's not perfect, but probably more accurate than anything else using MAC timestamping supported by the kernel. That also means the I210 can be used as a reference to measure the delays of another card (on <= 1Gb speeds). Those ~938 nanoseconds is the TX+RX timestamping error of the other end. If the clocks can be synchronized independently from their HW timestamping (e.g. PPS input/output), the measured offset should correspond to the (TX-RX)/2 error. When both TX+RX and (TX-RX)/2 are known, TX and RX can be calculated. -- Miroslav Lichvar _______________________________________________ Linuxptp-users mailing list Linuxptp-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxptp-users
