On 20/01/11 12:24 AM, "wzhty86" <[email protected]> wrote: > Dear Professor Brano Kusy: I am a student in Institute of Computing > Technology, Chinese Academy of Sciences. I contacted you some days ago and you > did a great favour for my paper. In your FTSP paper,I still have some > questions with respect to MAC layer timestamp? Q1:As you described in paper,a > timestamp for packet was made by getting the value of timer counter when the > packet was transmitted or reveived.Because the data type of globalTime and > localTime is uint32_t,I want to know what we should do when the timer > overflow?
note that the timestamping code is taking care of time overflows. in all calculations (e.g., when calculating a new skew, or a global time) we use differences of uint32_t numbers which are overflow-safe. that is not to say that the global time will not overflow, your application needs to take care of that. however, this happens every ~1.5 day for a 32khz crystal, so it should not be hard. > Q2:Because of different initial startup time between nodes, the > offset(globalTime-localTime) may larger than one second.In this case,what we > should do in order to get a better regression line? I am eagerly looking the initial offset of times does not affect accuracy of the regression line. the accuracy would only get compromised if the skew of local/global time is much bigger than 1, as we assume (1-skew) is close to 0. brano > forward to your reply.Thanks! Best wishes! Zehua Wang At 2010-12-03 _______________________________________________ Tinyos-help mailing list [email protected] https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
