Strange motion offsets in one direction on one axis

Those OP pics are not cool. Re-check all your terminations when you'll have a spare 143hrs. Crimps, joints, everything linked to motion/control, look for something loose. I know, it's a nightmare, takes quite a bit of time and efforts, but I'll have to throw in the highly suspect "caused by vibration".

Checked a few things- still need to check filddling with error limits on the servo (requires me to dredge up and set up a Windows PC by the mill- servo software doesn't play nice with WINE):

• "Differential" noise isn't quite as bad as it initially looks- yes, still awful, but mostly the noise between the A+/A- pins at the servo end is in phase. The high-frequency noise that is out of phase is well below the minimum step time (750 ns). So... awful, but I think "OK". At least seems less likely to be the problem. See picture with math in it, measured at the end of the cable right where it would connect to the motor.
• Tying the A- pin to GND (single-ended signals instead of pseudo-differential) mostly just gives different noise, similarly bad. See picture without math in it, measured at the end of the cable right where it would connect to the motor.
• The noise gets far worse when I disable the e-stop on the machine. The e-stop removes power from the spindle drive and servos, so could be either one- though disconnecting the servo power plugs individually had no impact on noise. Spindle is a DMM DYN4 with all the recommended filtering on the power inputs, but still could be a problem.
• Drive strength seems OK- the 0-5V transitions are really sharp.
• Connectors do tend to be the bane of happiness (or at least reliable signals), but they seem to check out OK. Crimps all seem good (can pull on them with no movement), soldered joints all look fine, connectors are clean, secure, and haven't been exposed to any contaminants. All signals are reading ~0.3 ohms from the end of the cable to the screw terminal on the 7i76E, which is maybe 0.1~0.2 ohms higher than I get just touching the DMM leads together. I retorqued the screw terminals, which made no difference.
• I disconnected the servo from the leadscrew and turned it by hand. It appears to be slightly bent- not enough that it's hard to turn with my fingertips poking on the coupling, but definitely goes through phases of being easier/harder to turn. Might replace that on principle since it's easy to do and "not right", but it's not obviously the issue.

Will test removing spindle power to see if that's the signal noise issue, and also see if powering the 7i76E from an isolated 5V supply impacts anything. Should've tested those before posting this, but they just occurred to me now. (also changing the motor error threshold)

Thanks all for the help, even though it's not AFAICT LinuxCNC related! Really helps to have outside eyes on problems when your head is in the weeds.

1. i really do not think this is software/electronics but hardware
2. are the last scope signals actually at 8.33 and 8.6MHz ???
3. check spindle for proper clamping of the end mill, although chances are slim
4. use a different end mill and run the same gcode and another gcode, check results
5. while the motors are locked/enabled push on the end mill from all sides, check for deviation on one side, use something as a reference behind the end mill to see easier if there is play
6. try exporting the gcode for milling in the opposite direction, is result the same.
-
as for interference, from what you explain it seems you have no good grounding there, or grounding is not connected
if the above is OK, check wiring for ground loops, check if grounding is done to a single point in the cabinet, check if motor outputs from VFD and servos have ferrite beads on, if not add them.
if all that noise is still present when VFD and servos are disabled, check power supplies in use, had terrible issues with some new cheaper ones lately.

1. Entirely plausible

2. Yes. The point of those was to highlight the noise seen is very high frequency, and seemingly unlikely to cause steps

3. Spindle is a BT30 taper with a manual drawbar- no ATC/belleville washers/pull studs/etc. End mill is held in a Nikken SK10 collet holder torqued to 35~40 ft-lbs, which is the recommended value. Tools are changed with an impact driver on the drawbar.

4. I should definitely re-run the code and see if I can reproduce. Haven't wanted to spend the time/material doing so, but I've wasted enough other time on this to dwarf both costs. I'll rerun both with the same tools originally used and also with a different one. May or may not have time to do that today.

5. I put an indicator on the tool and pushed on it. I can get it to deflect ~0.0005 (5 tenths) before I don't really want to push on it any harder. Seems to apply going any direction.

6. So climb vs conventional? I've thought about this, but the pocket was roughed with an adaptive/HSM toolpath, which started in the center of the pocket- so the cutter is roughing -Y to +Y on the +X side of the pocket, and +Y to -Y on the -X side. Correct me if my logic is wrong here, but if cut direction were the problem, wouldn't you expect to see issues in (eg) the +X+Y and -X-Y quadrants instead of the -Y half?


Grounding: One ground loop snuck in (shielded cable with shield connected at both ends), though disconnecting that has no impact on noise I can see. Will fix it anyway.

Beyond that, I wouldn't rule ground issues out, but am not seeing an obvious issue. The entire cabinet is aluminum and grounded. It's not a single point due to eg the filters, PSUs, and spindle drive all grounding through their chassis. The axis motors have a completely isolated supply (transformer-based)- as far as they're concerned, the world is the isolated 65V and a couple opto-couplers. The 24V and 5V negatives are all tied to a single bus bar- which is NOT ground. A multimeter shows an open circuit between 24V negative and the enclosure. The machine body has no dedicated ground, but is grounded indirectly via the spindle motor. The axis motor bodies are electrically connected to the machine, but that's isolated from their power supply.

Spindle drive has all the recommended filtering. I deviated slightly from the recommendations by omitting fuses on the logic power, replacing the contactor with an SSR, and placing said SSR before all the filtering, though I don't see an issue with that. Pictures show datasheet drawing and the schematic of how my machine is wired.

Noise is much better when motors are disabled- got it on my to-do list to see what's the root causes there. Guess is the spindle drive, but need to verify. No ferrites on anything at the moment- and no mention of them in the DYN4 (spindle) nor ClearPath (axis) manuals. Not to say they wouldn't help, just that the manuals don't call them out as important. The PSUs are MeanWell units, so at least not no-name. SOP is to not buy dubious PSUs- they can cause too many problems and take out too much valuable hardware to be worth the risk. Not to mention noise, as you say.

2. from what ? Aliens?
3. damn, that is a big machine, i thought it was some small hobby thing.
Does it use ballscrews?
If yes, try machining the same part with the same gcode but in another place on the table. Might be a broken ball.
Or might be a broken ball in linear rails, so decoupling the motor and spinning the screw by hand slowly might find a bump.

Oh, it is a little baby machine by many measures- 1 kW spindle, 12x6x10" XYZ travel or so, etc. Weighs, I dunno, maybe 500 lbs? I built it with a welded steel frame filled with epoxy granite, with proper linear bearings and ballscrews (to be fair, I got nice THK linear bearings, cheaped out on the ballscrews), so hobby bench mill sized, but with way more rigidity. The BT30 spindle is probably overkill, but it was available, and there's a lot of tooling options. TTS didn't seem like a great option, and ISO20 seemed too small.

Aside from this issue popping up, it's been a great experience coming from the gantry-style machine I'd been tweaking for years. The rigidity, Z height, and toolholders are the huge upgrades, but the MPG, touch probe, covered ways, coolant control, spindle/axis stall-detection, and servos are very nice as well. Having a spindle with enough torque at reasonable RPMs to actually handle (small) drills is still blowing my mind.

Sorry for the crappy picture- literally just walked over to snap it, so you can see it's a bit disassembled for debugging (and with random scrap taped over the t-slots so I didn't have to clean them out). That's a Kurt 4" DX4 vise for scale.