Thinus Viljoen wrote: > Hi > > I was wondering if it is possible to synchronize the RTNet TDMA cycles with > an external sync source? In our application we will have two seperate > network segments (ArcNet and RTNet), and it would be preferable if the RTNet > cycle could be sync'ed with the ArcNet cycle. The RTNet master wil probably > be a PC. The h/w design is not finalized, so a discrete "LanSync" signal > could be added, for example. Maybe you can connect such a signal to the > serial/parallel port and use that for synchronization?
Definitely feasible. If you just "hack" a solution, you may simply replace the sleep-until call in the TDMA master implementation with some sleep-on-event. This event can then be triggered by an external source, i.e. via an interrupt handler. But any proposal or patch to create a clean framework for such a scenario would also be welcome. Well, basically, the TDMA discipline would just have to be asked for the event object to be triggered. Then some other driver module could implement the interrupt handler that signals it. > > On a related note, where can I find measured examples of RTNet timing > results? I understand that it will vary from system to system, but an > example or two would be helpful. I was thinking about TDMA cycle time > accuracy and jitter, accuracy and jitter of the delay between SOF and a > external node's message, etc. > I only have a few German thesis at hand, which unfortunately only deal with pre-0.8 RTnet (less precisely clock synchronisation) or just measured payload round-trip times. But as a rule of thumb you can estimate the frame timing precision as being the order of the IRQ jitter of your box (from <10 us on high-end to some 10 us on low-end systems). Jan
signature.asc
Description: OpenPGP digital signature
