Andrea Forni wrote: > Hi Jan, thank's for your suggestions.... > > I have two different situations: > > 1) Logical control: i have five or more nodes in my network which communicate > in random mode (depend from machine's operations), in this situation I don't > need a critical period for the cycle (cycle: 5000us, it's acceptable). > > 2) Robot control: i have a control station (node A) and a station which > calculate the robot's trajectory (node B), in this situation i would a > comunication, between nodeA and nodeB, of 100/150us so that I don't miss the > trajectory's points. >
And all stations will be in the same network segment? Full TDMA network may not be required if you only transfer data at known maximum rate with known fixed order (e.g. polled) between two nodes. But if you attach several nodes with more complex communication behaviour and maybe also tunnelled non-RT traffic (for diagnosis etc.), then TDMA management is recommended. > > In second case I would understand if I can poll a node more than one time in > a cycle of TDMA. > > E.g. : > > # Primary master > master: > ip 10.0.0.3 > cycle 900 > slot 0 150 > slot 2 450 > slot 3 750 > > > # Slave A > slave: > ip 10.0.0.1 > slot 0 0 > slot 2 300 > slot 3 600 > > In this mode, slave send a message to master every 300us. I would understan > if this situation is feasible... I tried in the mode above (code in the old > mail), but it don't work... I read the documentation of RTmac and I see a > similar example, for this motif I ask you that.... You mean it doesn't worked at the rate of 300 us, or it doesn't work at all? I.e. is this effect load-depended or could it rather be a generic communication problem (I had no closer look at your code yet)? What architecture/CPUs do you use for your nodes? Jan
signature.asc
Description: OpenPGP digital signature
