Karl Reichert wrote: > Karl Reichert wrote: >> Am Samstag, 28. Juli 2007 15:02 schrieb Jan Kiszka: >>> Karl Reichert schrieb: >>>> Jan Kizka wrote: >>>>> Karl Reichert schrieb: >>>>>> Sorry, some mistakes, please see correct settings ... >>>>>> >>>>>>> Hi everybody, >>>>>>> >>>>>>> I have a question concerning TDMA: >>>>>>> >>>>>>> When I set up RTnet with TDMA and a cycle length of 5000 ms, I get >> an >>>>>>> average master jitter (difference between scheduled and real >>>>> transmission time >>>>> >>>>>>> of synchronisation frames) of less then 100 us, max 160 us. >>>>>> My setup was cycle length = 5000 us = 5 ms >>>>>> >>>>>>> When I do the same with cycle length 1000000 ms (1 second), I get >> an >>>>>>> average jitter of 8500 us (more then 8 ms). >>>>>> My setup was 1000000 us = 1000 ms = 1 s >>>>>> >>>>>>> With cycle length 2 seconds, I get an average jitter of 4500 us. >>>>>> My setup was 10000000 us = 10000 ms = 10 s >>>>> What jitter do you get with the "latency" test running at the same >>>>> periods? >>>> I assume we're talking about /usr/xenomai/bin/latency: >>>> With 1 ms cycle length and latency tool running, jitter is much lower, >>>> between 5-25 us. >>>> >>>> The other two cycle lengths & latency-tool I didn't gave a try, >> because >>>> this seems already strang to me. I'm very busy ATM but if testing with >>>> those other cycles would help you with the problem, I could give you >> the >>>> values at end of august (when I'm finished with diploma). >>>> >>>>> What are the distributions of the jitters? All only around 4.5 >>>>> or 8.5 ms, or are they spread over a far wider range than with 5 ms >>>>> cycles? >>>> I general they spread little bit, but not very much, so I don't get >>>> values higher then 5 ms / 9 ms. So this seems to be normal. >>>> >>>>> I can't look into this soon, but it has high priority for time I will >>>>> have test equipment in reach again! >>>> Thanks Jan! Keep informed the list then, please ;) >>>> Would really like to know why this happens, although it's not so >> critical >>>> to my project. >>> So far I didn't find any systematic error with large cycle times. >>> Capturing the outgoing master traffic, I got typical deltas of a few ten >>> microseconds (unloaded system) according to Wireshark. That quick test >>> took place over Xenomai trunk, but I would rather expect RTnet to be the >>> problem here - if at all. >>> >>> One further important question to you: How did you measure / read out >>> the master timing? Are you sure you didn't confuse micro- and >>> nanoseconds here as well? Note that e.g. Wireshark reports all TDMA >>> frame timing fields in _nanoseconds_ (all TDMA fields are in nanoseconds >>> - to answer your older question). >>> >>> Jan >> I read it via wireshark (with respect to nanoseconds) and also on my own >> slave >> implementation on TriCore 1130 via debugger. The values are really like >> that. >> I will send you a wireshark capture file on monday, so you can see on your >> own >> eyes! >> >> Karl >> > Please see attached wireshark capture file!
Yep, I can see it. Something must be different with our setups. Once there are a few minutes, I will try to recreate your environment. I can only imagine some overflow, maybe only occurring with your CPU frequency or your kernel config. Thanks, Jan
signature.asc
Description: OpenPGP digital signature
------------------------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________ RTnet-users mailing list RTnet-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rtnet-users
