Rough movement in one direction, but only when running a job in Axis

I'm setting up a new machine and encountered a problem where the stepper is smooth and quiet in one direction, but moves erratically in the other direction BUT.... this is only when running a job in Axis. Manual jogging is fine, and the stepconf 'test this axis' is also fine.


In the problem direction, about 90% of the time the stepper either does not move at all (when it should be moving very slowly) or it moves very roughly (when desired speed is faster). The rough movements captured in the video above are actually pretty tame - doing rapids causes nasty vibrations that shake the whole table.

I've double checked the phase wiring and dir/step connections, can't see any problems there. The manual jogging and stepconf controlled movement being very nice leads me to believe wiring is not the cause. Swapping two wires of a phase only causes the problem to change to the opposite direction.

Hardware is parport with cheapie Mach3 BoB going into Leadshine EM806 and Leadshine 86HS85 at 60V. The driver is pre-tuned for this specific motor. Currently it's only moving a light MDF mockup gantry which is a very low load for this motor. Computer is the one listed here wiki.linuxcnc.org/cgi-bin/wiki.pl?Latency-Test (search for "As above but with RTAI wheezy")

Control signal requirements for the driver: www.manualslib.com/manual/1987319/Leadsh....html?page=10#manual
Step time and step space are set at a conservative 7.5us which is three times as long as required. Direction hold and setup are 20us which is four times the requirement.

So I'm wondering, is there some reason that running a job would cause rough movement when the other methods don't?

At the beginning of the video I was using the left and right arrow keys on the keyboard to jog smoothly in both directions. If the step pin of some other axis was wired to dir pin of X I can't imagine how jogging X in both directions like that would work at all. Since the other axis is not being stepped, wouldn't the direction of X movement be stuck going one way?

So far I cannot find any MDI command that causes the rough movement, either X alone or with other axes involved. Here's a test of using an MDI command in comparison to what the job execution does, with a better example of the nastier type of vibration:

Okay, brace yourself..... stupid mistake discovered..... in all my testing yesterday USB power was not connected to the BoB. I had only connected the 24V power and somehow completely forgot about the 5V power side. All my 'meticulous checking' of connections became tunnel-vision on stepper and signal wiring.

Seems like the parport signal from the PC was sufficient to run steppers on their own, but I'm guessing that when spindle PWM became non zero  (for running a job) there just wasn't enough power to go around. As for why it was still able to move correctly in one direction, presumably that was the direction for which the dir pin was low.

Thanks for the help Tommy, sorry to be such a dummy!