On Mon, Sep 11, 2017 at 02:38:07PM +0100, Alan Young wrote: > On 11/09/17 13:45, Miroslav Lichvar wrote: > > I think there is a requirement that both SW and HW transmit timestamps > > are before the actual transmission. A timestamp captured after the > > transmission could make the measured network delay shorter than it > > actually is. This might be more important for NTP than PTP. > > That is a pity, because I think that the callback to tx_complete will have > relatively low jitter with respect to the actual transmission. My guess was > consistency was more important than correctness; does it matter if the > measured delay is less than it actually is so long as it is consistently off > by the same amount?
It does in some applications. If the measured round-trip time is X, an upper bound on error due to asymmetric network delay in offset calculated from the same timestamps is X/2 (assuming frequency of the clock is accurate). For example, if the actual network delay from A to B was 10 microseconds, from B to A it was 50 microseconds, and a measurement had an accurate RX timestamp, but the TX timestamp was late by 20 microseconds, the measured delay would be only 40 microseconds and the assumed maximum error (20 microseconds) would be smaller than the actual error (30 microseconds). -- Miroslav Lichvar ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Linuxptp-devel mailing list Linuxptp-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxptp-devel
