This follows on from a thread of this title which has been cold for months so I have started a new one.
Hi cncnovice, rcaffin and Dan13,
I know this thread has been untouched for several months but I have a bit to add which may help others. I too had the problem of lost steps: apparently a single step missing on direction change. How did I know? I was cutting a 288 tooth gear by profiling with a 2mm endmill, four passes for each tooth with the indexing by a rotary table on A axis. So the g-code was a set of instructions to cut one tooth. As a result there were around 1200 changes of the direction state while the axes were moving. But when the table had finished its 288 steps I found I had a cut not a circle but a helix with the end point inside the radius of the start point.
I tried all the tests you did, except for the scope which I do not have, and came to similar conclusions. The error was very repeatable which seemed to eliminate noise, it appeared when different computers were running Mach and seemed impervious to any of the changes in motor tuning, ports and pins etc. One difference from yours, however, was that it appeared on all three axes even when only a single axis was in use.
Eventually I decided that as it seemed to be most likely a parallel port problem (that's how little I know
) I would try a Smooth Stepper which I had been considering since they appeared. The result: 48hours delivery from Minnesota to Devon UK, an hour installation and setup and then.......... no lost steps. As always, you think "why did I not do this before and save the hours of fruitless testing?" I guess the answer is that we are always trying to understand a bit more.
So the point of this post is firstly a big "thank you" for pointing me in the right direction and teaching me a lot I did not know about Mach. Secondly, I wonder how many systems have this error which only appears with multiple repetitions and does not otherwise intrude. I had been using my setup under mach for three years before I stumbled over it.
Cheers, Peter.