Re: [Linuxptp-devel] [PATCH RFC 0/6] Improve accuracy with software timestamping

[Richard C. <ric...@gm...>]

On Mon, Mar 09, 2015 at 02:20:19PM +0100, Miroslav Lichvar wrote:
> I think the difference is clearly visible. With exponential
> distribution you can also see how the asymmetry in the delay
> distribution is included in the filtered offset.

The generated offsets from the script are the *inputs* to the PI
controller, and they don't look to far apart to my eyeball analysis.
(Yes, the dispersion looks a bit better in the "raw delay" plots.)

The really interesting comparision would be the actual offset after
PI servo control.  I am looking to compare the result of your patch
set on a small but real network...

Thanks,
Richard