Re: [Linuxptp-users] Issue with initial high jitter(Path Delay Variation)

[Miroslav L. <mli...@re...>]

On Mon, Feb 15, 2016 at 02:16:13PM +0530, Sujatha Guguloth wrote:
> In Failire case (Logs provided below)
> - slave clock is taking 340 - 278 = 62 seconds for stabilization
> -  jitter is maintained below 100ns after stabilization but it is very very
> high compared to previous case during the stabilization stage

It looks like the initial step is off by one second for some reason.
It could be a driver issue. What HW and kernel version do you use?

As a workaround you could try setting step_threshold to 0.1 for
example and allow ptp4l to correct that one second error by stepping
instead of slewing.

> *Failure Logs:*
> ptp4l[270.981]: selected /dev/ptp0 as PTP clock
> ptp4l[271.014]: driver changed our HWTSTAMP options
> ptp4l[271.020]: tx_type   1 not 1
> ptp4l[271.024]: rx_filter 1 not 12
> ptp4l[271.028]: port 1: INITIALIZING to LISTENING on INITIALIZE
> ptp4l[271.040]: port 0: INITIALIZING to LISTENING on INITIALIZE
> ptp4l[272.323]: port 1: new foreign master 001ec0.fffe.8614b5-1
> ptp4l[276.323]: selected best master clock 001ec0.fffe.8614b5
> ptp4l[276.328]: port 1: LISTENING to UNCALIBRATED on RS_SLAVE
> ptp4l[278.496]: master offset -1475113437970290 s0 freq      +0 path delay
>      7176
> ptp4l[279.496]: master offset -1475113437987402 s1 freq  -17112 path delay
>      7176
> ptp4l[280.496]: master offset -1000000284 s2 freq -32767999 path delay
>  7176
> ptp4l[280.501]: port 1: UNCALIBRATED to SLAVE on MASTER_CLOCK_SELECTED
> ptp4l[281.501]: master offset -967516976 s2 freq -32767999 path delay

-- 
Miroslav Lichvar