| Re: [chrony-users] asymmetric phy delay? |
[ Thread Index |
Date Index
| More chrony.tuxfamily.org/chrony-users Archives
]
- To: chrony-users@xxxxxxxxxxxxxxxxxxxx
- Subject: Re: [chrony-users] asymmetric phy delay?
- From: Nuno Gonçalves <nunojpg@xxxxxxxxx>
- Date: Mon, 10 Aug 2020 14:09:29 +0100
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :content-transfer-encoding; bh=YKe7kYnY99Ro3jMq/8DE8d52EhS055uBcy5038nrGs0=; b=dMNELkhoqMryEeLNDBQ2lFdOzfQtjVmyrkX9PGaj8T+uOB9wuysCMTomU70I93JwCC fdKNcR3NSW2lnoSPAkZ3SOY07SUwwSJseY9lXhcj8t/7ubfgtJpa7kOnQx7ULqNk/9mR yont2Rbq+Kq+QmfT4TMNAXfXGXEjgGQXTlx29+4/+KEQRJcgkF4bQ2ikjaF0eFay2/Wx ykLio9VKPZxRDRNpbiIoBnGwh/ykrPanQeIkrQo85GYSoGMULQsvNADSTgTEoFYL+ZLd +XQqIkaf4KRpXLHA51yWNkZryN1dcRCYdXwAxUhdnHiyDkOlbJD0k6KIMGBraqeqFfR3 1ipQ==
Thanks for the great support Miroslav.
I had no idea about xleave mode. I confirm it changes the behaviour to
what I expected.
Wouldn't it be possible to provide a fixed bias/calibration to be
always added to the server timestamp to calibrate a known device when
hw timestamps are unavailable?
Thanks,
Nuno
On Mon, Aug 10, 2020 at 1:46 PM Miroslav Lichvar <mlichvar@xxxxxxxxxx> wrote:
>
> On Mon, Aug 10, 2020 at 12:55:59PM +0100, Nuno Gonçalves wrote:
> > SOC A to A shows a offset of about 50us+-25us
> > SOC A to B shows a offset of about 0us+25us
> > SOC B to B shows a offset of about 75us+-25us
> > SOC B to A shows a offset of about 130us+-25us
>
> > I understand that a eventual asymmetric tx/rx delay on the phy could
> > cause this unbalance, but my understanding of the NTP protocol is that
> > this would cancel each other on a query/response cycle of the same
> > board type, which is not happening.
>
> The asymmetry is likely in the timestamping. If you don't enable the
> interleaved mode, only three of the four timestamps from the NTP
> exchange will be kernel or hardware timestamps. Only when the
> interleaved mode is enabled, the fourth timestamp (server TX) will be
> symmetric to the client TX timestamp.
>
> The client could be easily modified to ignore the more accurate
> kernel/HW timestamp when not using interleaved mode, but the idea was
> to prefer stability over symmetry between two identical machines.
> There are many sources of asymmetry and this is just one of them.
>
> --
> Miroslav Lichvar
>
>
> --
> To unsubscribe email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
> with "unsubscribe" in the subject.
> For help email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
> with "help" in the subject.
> Trouble? Email listmaster@xxxxxxxxxxxxxxxxxxxx.
>
--
To unsubscribe email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
with "unsubscribe" in the subject.
For help email chrony-users-request@xxxxxxxxxxxxxxxxxxxx
with "help" in the subject.
Trouble? Email listmaster@xxxxxxxxxxxxxxxxxxxx.