[Tinyos-devel] FTSP testing result

Andreas Köpke koepke at tkn.tu-berlin.de
Fri Feb 6 06:02:21 PST 2009


For a quick check of the timesync implementation I:

1. predict the new global time given the local time when the packet was 
   generated (timestamp - offset between the event time and the SFD interrupt 
   time)

2. compute the difference between the predicted global time and the global time 
   announced in the beacon message.

This approach captures the effects due to wrong slopes and offsets, although it 
does miss certain errors in the computation that cancel each other. For a 
secondary check a reference broadcast approach is needed. (wiring pins is more 
controllable).

Best, Andreas

Branislav Kusy wrote:
> so it seems the problem fixed?
>
> it takes about 8 minutes for all nodes to synchronize, but then they
> stay synced.
>
> would be interesting to see
> - after more hours
> - after more hours with trickle beacon
> - timesync error (to confirm that it really works)
>
> any ideas of how we can estimate timesync error without a reference
> broadcast?
>
> brano
>
> Omprakash Gnawali wrote:
> > Top graph shows the result from about 1 hr run:
> > http://enl.usc.edu/~om_p/net2/ftsp/
> >
> > - om_p
> >
> > On Thu, Feb 5, 2009 at 10:40 PM, Branislav Kusy <kusy at stanford.edu> wrote:
> >> fixes checked in (both CC2420 stack and the timesync patch from thomas)!
> >>
> >> brano
> >>
> >> Omprakash Gnawali wrote:
> >>> I can certainly try Motelab but waiting for the fixes to be patched
> >>> in...
> >>>
> >>> - om_p
> >>>
> >>> On Tue, Feb 3, 2009 at 5:23 PM, Matt Welsh <mdw at eecs.harvard.edu> wrote:
> >>>> Can you try on Motelab?
> >>>>
> >>>> On Feb 3, 2009, at 2:18 AM, Omprakash Gnawali wrote:
> >>>>> I ran FTSP for a few hours on Tutornet and the result is not good.
> >>>>> The nodes go out-of-sync frequently enough to make the system
> >>>>> unusable.
> >>>>>
> >>>>> Setup: I updated my tos directory from CVS head earlier today. I
> >>>>> changed tests/TestFtsp to log FTSP time every 10s. 56 nodes total.
> >>>>> "PFLAGS += -DTIMESYNC_RATE=30" added to the Makefile.
> >>>>>
> >>>>> Here are some results:
> >>>>> http://enl.usc.edu/~om_p/net2/ftsp/
> >>>>>
> >>>>> Every 10s, for each node, I plot a point on the first graph if the
> >>>>> getGlobalTime() returns a FAIL indicating unavailability of global
> >>>>> time. If there are no dots on the graph, it would mean the nodes have
> >>>>> a valid global time at all times.
> >>>>>
> >>>>> With Trickle timer, the results are worse, but maybe we should fix
> >>>>> the periodic case before we get to Trickle. Turns out the
> >>>>> consistency/PULL-bit is more complex than I had thought but it should
> >>>>> be doable.
> >>>>>
> >>>>> - om_p
> >>>>> _______________________________________________
> >>>>> Tinyos-devel mailing list
> >>>>> Tinyos-devel at millennium.berkeley.edu
> >>>>>
> >>>>> https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-d
> >>>>>evel
> >>>
> >>> _______________________________________________
> >>> Tinyos-devel mailing list
> >>> Tinyos-devel at millennium.berkeley.edu
> >>> https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-dev
> >>>el
>
> _______________________________________________
> Tinyos-devel mailing list
> Tinyos-devel at millennium.berkeley.edu
> https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-devel



More information about the Tinyos-devel mailing list