Hi Janos, In fact I have developed another time synchronization protocol and I have tested it using the same testbed. I can get 2-3 miroseconds accuracy. Thus, I feel that the timestamping mechanism works well (but just a feeling :)). I just suspect about if some modification to TimeSyncP is needed in order to get accurate results.
I have a final detail that the results I presented in the previous email are taken by setting the "skew" variable to zero in TimeSyncP. Otherwise, the result is much more worse... Sinan 2010/10/21 Janos Sallai <sal...@isis.vanderbilt.edu> > I can make some wild guesses, but can't give you a solution. > > - The 7.3mhz crystal has a precision of 50ppm. The beaconing rate of > FTSP is 10 seconds by default (can be set with > PFLAGS+=-DTIMESYNC_RATE=<value> in the Makefile), which can result in > an 500 us error per hop. However, FTSP's linear regression should take > care of this. Also, 500 us is really the worst case scenario. You can > try decreasing TIMESYNC_RATE and see how the results change. > > - I am not sure if the packet-level time synchronization > implementation of the IRIS is calibrated. There's a chance that it's > not, which can result in a constant one-hop synchronization error. > Miklos could clarify this. > > Janos > > On Thu, Oct 21, 2010 at 11:17 AM, Sinan Yildirim <sinanyi...@gmail.com> > wrote: > > Hi Janos, > > > > I have done the modifications as you have said and also used the NoSleepC > > component of your implementation in tinyos-2.x-contrib repository. But I > > don't get tight skew values on a line of 5 iris sensor nodes. > > > > Here are the results after a 15 minute run: > > > > 1287678021203: [nodeid=0x1] [clock=0x1f3fc724] > > 1287678021250: [nodeid=0x2] [clock=0x1f3fc76a] > > 1287678021294: [nodeid=0x3] [clock=0x1f3fc75d] > > 1287678021350: [nodeid=0x4] [clock=0x1f3fc6ae] > > 1287678021398: [nodeid=0x5] [clock=0x1f3fc47d] > > > > Why am I measuring such big clock skew in this small network?? Any > > suggestions? > > > > 2010/10/18 Janos Sallai <sal...@isis.vanderbilt.edu> > >> > >> Sinan, > >> > >> It's not there, but it's not terribly complicated to accomplish. Just > >> take a look at TimeSyncC and TimeSync32kC, and based on those, come up > >> with your TimeSyncMicroC. As far as I can tell, this is the only file > >> you'll need to cook for yourself, you don't need to touch anything > >> else. > >> > >> A couple of notes: > >> - On the iris, the interface provided by TimeSyncMessageC that gives > >> you microsecond precision packet-level timesync is called > >> TimeSyncAMSendRadio. > >> - You can get microsecond-precision local time from LocalTimeMicroC, > >> however, time will stop when the mote goes to sleep. You will need to > >> keep the MCU awake all the time to get FTSP to work properly. There > >> are a couple of different ways to do this, for example by having a > >> component that provides McuPowerOverride and the lowestState function > >> always returns ATM128_POWER_IDLE. You'll need to wire this component > >> to McuSleepC. > >> > >> Janos > >> > >> On Sat, Oct 16, 2010 at 4:33 AM, Sinan Yildirim <sinanyi...@gmail.com> > >> wrote: > >> > Hi all, > >> > > >> > I have a question on getting microsecond precision using FTSP on Iris > >> > platform. Current version of tinyos 2.1.1 does not include components > >> > and > >> > modules in tos/lib/ftsp directory in order to achieve microsecond > >> > precision. As far as I know, previous version of tinyos (1.x) had a > >> > support > >> > for getting microsecond precision using FTSP. > >> > > >> > Has anybody a modified version of the FTSP code for tinyos 2.1.1 in > >> > order to > >> > get microsecond precision global time value? Is it possible to share > it > >> > here? I want to test this code on a line topology of 20 Iris motes. > >> > > >> > Thanks for your helps, > >> > > >> > Sinan. > >> > > >> > > >> > > >> > _______________________________________________ > >> > Tinyos-help mailing list > >> > Tinyos-help@millennium.berkeley.edu > >> > > https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help > >> > > > > > >
_______________________________________________ Tinyos-help mailing list Tinyos-help@millennium.berkeley.edu https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
