hi *, Am 26.02.19 um 18:25 schrieb Chris Radek:
> Have you considered the quadrature scheme, mode 2? It has the same > benefit of one edge per step, but does not have the drawbacks > requiring consideration of setup and hold times (the dirsetup and > dirhold parameters) this was exactly the hint I was searching for! now it sends a packet with the quadratur-signal every RT-tick. I hope so... ;-) next day, I will check out, how stable the signal is. what a can say now: with the linux-tcpdump-timestamp it is not possible, to read the correct packet-time, if the packets are running in faster (or near) the RT-ticks. with a stepgen.maxvel of 30 I get values, which seems to be correct: setp stepgen.0.maxvel 30 (distance between two 60byte packets) Time | last packet timediff ----------+---------------------------- 55.713276 - 0.00005000 55.713350 - 0.00007400 55.713426 - 0.00007600 55.713475 - 0.00004900 55.713550 - 0.00007500 55.713625 - 0.00007500 55.713676 - 0.00005100 55.713750 - 0.00007400 55.713825 - 0.00007500 but a maxvel of 40 (which is the highest value without an warning - see below) results in impossible timestamps: 0.00000200sec == 2usec - thats impossible - setp stepgen.0.maxvel 40 (distance between two 60byte packets) Time | last packet timediff ----------+---------------------------- 54.504630 - 0.00000200 54.504631 - 0.00000100 54.504632 - 0.00000100 54.504633 - 0.00000100 54.504635 - 0.00000200 54.504636 - 0.00000100 54.504637 - 0.00000100 54.504639 - 0.00000200 54.504640 - 0.00000100 54.504641 - 0.00000100 54.504643 - 0.00000200 54.504644 - 0.00000100 54.504645 - 0.00000100 54.504646 - 0.00000100 54.504647 - 0.00000100 STEPGEN: Channel 0: The requested maximum velocity of 45000 steps/sec is too high. if I count the packets, I get 170848 packets in 10 seconds. 17000 pkt./sec ==> 0,000058824 sec./packet. that sounds a little bit more realistic. but it is less than the 40k packets, which should be possible with a 25usec-thread. why ? regards wicki _______________________________________________ Emc-developers mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-developers
