> I think the question was intended to be more theoretical and asks about > "exactly" synchronizing commands. The LinuxCNC/SPI solution is not > that. SPI works only because it is so fast that the error in > synchronization is tiny and goes unnoticed.
Using a Micro controller it is possible to connect pwm output to CS signal and use it to trigger interrupt to send a message, receive may clock input so if needed it is possible to get really good synchronization with SPI and standard Micro controller if needed. In doubt there is any case this accuracy is needed or make a difference in practice but syncronize clock drift may be needed, especially if a FIFO is used. > Here is a harder problem. Let's say I am in North America and by buddy > lives in Europe and we want to each run clocks and we want them to stay in > phase at a high level of accuracy. To make matters worse assume this is > the mid-1800s and the radio is not yet invented. They actually solved this > problem. The solution was "mutually observed events" and we use this same > solution today to keep widely dispersed machines in sync. In the old > days, they would observe one of Jupiter's moons from both America and > Europe and assume they both say the moon transit the planet at the same > time. Orchestras use a conductor waving a stick who is "mutually > observed" by all musicians. Same with a CAN bus, you could, if needed > use a high priority "clock tick" message that all nodes see at the same > time. Know about it and it could be sursingly tricky to agree about event ordering. > But in real-life. We accept "close enough" and just us a SPI signal that > is fast enough that no one notices the error. Agree, used this approach many times and quite often there is plenty of other random variations. _______________________________________________ Emc-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-users
