Missing first 5 pulses on direction change

This has been frustrating me for some time and i could really use some help here... it has got to be something stupid that ive overlooked

anyway, i will try and describe my problem as best as i can.

i recently switched over to a 7i96 and since i have, i consistently am off 0.5mm on my x axis and nothing i do seems to fix this. i have recalibrated and recalibrated my recalibrations and i am 0.5mm off every time. climb, conventional, fast, slow, aluminum, brass, hardwood, softwood, hdpe - it makes absolutely no difference.

it is a belt driven gantry so i understand that there are accuracy limits and backlash, but it is a very robust machine and before the switch i could reasonably cut within +/- 0.05-7mm consistently. now i am an order of magnitude off and its driving me crazy

so, today i was trying to again recalibrate and i discovered something very interesting... i put a dial indicator against my spindle to check movement, and thats when i noticed that whenever i change direction, it will lose the first 5 commands and then resume like normal. it does this both directions. for example, i would move it and set the indicator at 0 and then in 0.001 steps (i was testing in imperial) i would move it 10x until the dial was at 0.010. then i would go left 10x and the dial would be at 0.005. then once again i would go right 10x and i would be at 0.010. if i continued another 10x then i would be at 0.020 no problem. but if i switched and go back 20x, i still end up at 0.005 and 10 more puts me at -0.005

so on a whim, i put my hand on the x axis to see if i was possibly missing steps if it was backlash and i could hold the spindle back and thats when i noticed that whenever i switch direction, the first 5 move commands are dropped. the dro shows that i am moving, but there are 0 pulses going to the stepper until after the 6th keyboard press. this happens every time any direction, no matter where at along the x axis.

so my question is, what could be causing this? is it hardware, mechanical or software? could this be a mesa issue or lcnc issue or configuration issue? how can i test further to eliminate variables? has anybody else had this problem?

oh before i forget, my specs are:
- xyyz gantry, 20mm htd3 belts, rack and pinion style
- linuxcnc 2.8 master
- intel core i3 dell optiplex
- stepperonline dm542-y digital stepper drivers @48v
- mesa 7i96

thanks for any help anybody has!

Can you clarify if you are judging the steps reaching the stepper just based on the DTI, or also by some other means? It does sound like you are just describing mechanical lash.

i know that its unscientific, but basically i had my hand on the stepper and would feel the vibrations of the individual steps. for the first 5 movements, there is nothing and then you can feel each individual step after that. is there a way i could possibly check with an ohmmeter?

I’d test that motor disconnected from the machine.
Then if possible swap that motor for one from another axis.
If nothing changes I’d swap the drivers over.
I assume you didn’t change the stepper timings ?

Are you getting this behaviour at all points of the axis ? If you test at close to each end are you seeing the same ? What about at the middle ?
Does it change if you change your jog speed ?
How about using MIDI ?
Do G0 moves and G1 moves do the same ?
What about different speeds for G1 ?

Isn’t there a direction hold timing that is configurable ?

Yes, there are direction setup and hold timings that can be set.
Typically there is never any reason to set these to less than say
20000 ns (20 usec) which will work with pretty much any drive.

so i got home tonight and tried some of your suggestions...

BeagleBrainz wrote:

I’d test that motor disconnected from the machine.
Then if possible swap that motor for one from another axis.
If nothing changes I’d swap the drivers over.
I assume you didn’t change the stepper timings ?

Are you getting this behaviour at all points of the axis ? If you test at close to each end are you seeing the same ? What about at the middle ?
Does it change if you change your jog speed ?
How about using MIDI ?
Do G0 moves and G1 moves do the same ?
What about different speeds for G1 ?

so i
- swapped drivers - same result
- drove x as the z axis and works fine
- tested all along axis and same result at every point
- changing jog speed had no effect, fast or slow same behavior
- g0 g1 were identical no matter the speed

so mdi was interesting. i was moving in 0.5" increments because my dial indicator goes slightly over an inch. whenever i would switch direction, i would always come up 0.005 short, but as soon as i went back the other way, i would hit my mark exactly. the reason that i find this interesting, is that the scale on my machine is pretty close to 0.001 per step and when i am making 0.001 moves, i am losing the first 5 after a direction change

tommylight wrote:

Has this same issue been in another topic?
Does this happen only on 1 axis ?
Did you try changing the drive?
Is it direct driven ?
Changing the drive would prove the issue is with the drive or the motor or settings in hal.
If it is direct drive, to much microstepping can cause that, to much load also.

i searched and searched, but honestly, i really had a hard time describing this problem, so it may very well have been addressed before, but i really wouldnt even know where to begin.

yes its direct drive, belt as a rack and pinion, and my microstepping is set at 1/8 and they are 425 oz/in nema23 steppers driven @4.2a, so they are pretty beefy for their size

PCW wrote:

Typically there is never any reason to set these to less than say
20000 ns (20 usec) which will work with pretty much any drive.

the manual calls for 2.5µ sec timing and i am running 5000/5000/2500/2500 in my ini.

at first i thought that it was just backlash, so i started checking the usual suspects and when i figured out that the motor wasnt being driven, then it baffled me. now if for whatever reason the driver wasnt getting the direction change, then it would still continue to go in the original direction until directed otherwise, right?

also could the board or firmware be the issue? i have noticed that the leds on the endstops flicker lightly whenever its powered on, and none of my other ones do that...

the manual calls for 2.5µ sec timing and i am running 5000/5000/2500/2500 in my ini.

In general its not a good idea to use drive datasheet values as these are usually minimums.
Its better to use the largest values of steplen and stepspace you can that does not
interfere with your maximum speed. Also its good to use very generous values of dirseup and dirhold
since these have no significant influence on performance (as I said 20000 is a good number
for these)

Have you tried connecting the Z axis step gen to the X driver and X step gen to the Z axis ?
That should eliminate any issue with the 7i96.
If you still get the same issues I start chasing a mechanical issue.

As to it being a hal\ini issue, your hal\ini files would help.