Hi, I've been experimenting with two Beaglebone Blacks and Linuxptp to measure the accuracy of the clock synchronization. I use a GPS receiver with a 1PPS Signal as a reference clock. On my "master" Beaglebone, the system clock is synchronized to the GPS/PPS with NTP and the kernel PPS driver. I then use phc2sys to synchronize the master's PHC to the system clock. On the slave I use phc2sys to synchronize the system clock to the PHC and finally NTP to measure the offset of the slave's system clock to the GPS/PPS reference.
The master and slave machines are connected directly to each other via a single Ethernet cable and the slave synchronizes its PHC to the master's using ptp4l. I'm using UDPv4, E2E delay measurement and hardware timestamping. Using this setup and plotting a histogram of the offsets measured on the slave (between system clock and GPS reference) over two days, I noticed that while the offsets roughly follow a normal distribution, they are centered at around +500ns. I would've expected them to center around 0. (See attached figure.) Can anyone explain what might introduce this behavior? I can't determine if this is caused by phc2sys, by NTP, by the kernel, by PTP or by the network. Is there a way for me to check if it happens on the network or on one (or both) of the boundary clocks? Unfortunately, I have not found a way to have a PHC timestamp the PPS signal directly, so right now I can only rely on what the various synchronization processes tell me. Is there a way to have a Beaglebone's PHC timestamp an external PPS signal? I know I can have it timestamp the PWM output of one of the DM timers... Thanks, Mino
_______________________________________________ Linuxptp-users mailing list Linuxptp-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxptp-users
