Re: [Linuxptp-devel] Optimal P, I constants

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

On Thu, May 23, 2013 at 04:11:36PM +0200, Miroslav Lichvar wrote:
> On Wed, May 22, 2013 at 07:41:14PM +0200, Richard Cochran wrote:
> > Right, so after seeing your simulation results, I am going to try and
> > validate them using an actual PPS.
> 
> Great, I'm looking forward to see your results.

First results are in:

  http://linuxptp.sourceforge.net/igb-master_bcc-slave_pps/

  - plot1.png  Initial convergence at Sync 2^0
  - plot2.png  Initial convergence at Sync 2^-4
  - plot3.png  Steady state at Sync 2^0
  - plot4.png  Steady state at Sync 2^-4

For this test, I used the m5234bcc (phyter) as a slave to the Intel
i210 card. The 1 PPS from the i210 was time stamped by the phyter.
You can see a static offset of 400 nanoseconds, due to path
asymmetry. Overall, method B is looking better, but that is just my
impression, and I haven't really quantified that impression.

I hope to repeat the test today, with the roles reversed.

Thanks,
Richard