Hello Florian, > Hallo Christian, > > On Fri, Feb 27, 2009 at 11:32:55AM +0100, Christian Gieseler wrote: > > The thing is that while looking for the source of the jitter i > > connected the master directly to the slaves and then i get all these > > messages about the timed out frames. Of course i still want to know > > what causes the jitter, but at the moment i would really like to know > > why i get loads of Messages about timed aut datagrams, even in when > > the master is in idle mode. Does someone have an advice for me? > > ok, I guess the jitter you're talking about and the lost frames are two > different problems.
Yes, you are right, they are probably caused by different sources. I just posted them together because i only discovered the problem with the timed out frames after looking for the reason of the jitter peaks. > > For the last one, please try to bisect the bus by plugging certain > slaves out to see, if the frames are 'swallowed' by a certain slave. I > experienced this some time ago. If this brings no difference, do you > have the possibility to try another Ethernet adapter (perhaps another > driver)? I tried to connect the slaves separated from each other, but i can´t point out one single slave that causes the problems. Even if i only connect a single EK1100 slave i still get the Timed out datagrams. And this also in the idle mode. No matter if it is a kernel, or a userspace application. This is quite annoying if are typing commands and the Warnings are scrolling up on the screen. > How are you measuring the jitter? I am measuring the jitter by writing an fpwr datagram to the register 0x0900 on a dSPACE slave, which has an FB1111-0140 module on board. This register is normaly used for distributed clocks, as you probably know. As the time of the next incomming frame is latched and i can read the nanosecond timestamp via the PDI Interface. Of course i have the latency from the master to the first slave, but this should always be the same. >What is your timing source? As timing source i use the same sources that are given in your user and mini examples. As my target is an pentium m embedded PC i have no chance to change the ethernet adapter, but i connected my hard disk to a pentium 4, also with an intel pro/100 adapter and i don´t get the timeouts. Do you have an idea if there is something to change in the kernel configuration expect from the architecture to solve the problem? I tried to keep the the kernel as small as possible. Normally one would expect that the switch would make things worse, but it doesn´t. What else could a switch do? Refresh the signal quality? Thank you for any help. Best Regards Christian -- Psssst! Schon vom neuen GMX MultiMessenger gehört? Der kann`s mit allen: http://www.gmx.net/de/go/multimessenger01 _______________________________________________ etherlab-users mailing list [email protected] http://lists.etherlab.org/mailman/listinfo/etherlab-users
