Not getting precise alignment, even with Probing phase

[DavidYon47]

So I'm back to doing a Pi Hat, using the template. I'm running into the alignment issue I had before where the unit clips the side of a hole while probing. The symptom last time was I killed a nozzle because the probing worked as the alignment was sufficiently off that the nozzle clipped the edge of the hole.

But now, no matter how precisely I align, it dies in the probing phase:

You can see that when it intended to measure just to the left of the hole, it went a smidgen too far right, dove into the hole, then zoomed sideways digging a scar across the board. It then dies on an error (the Z probe is triggered when it shouldn't be).

I also notice that as I probe around, the alignment can be off by as much as a 1/4 diameter of a through-hole. I've tried being ultra-precise, using different alignment pads (even aligning to the "problem" hole), making sure there's no warp in the clamping, etc.

For reference, here's the circuit:

It's carving a scar where that trace is on the lower-right. On one alignment-check pass I've noticed that the bottom of the vertical header can be noticeably more misaligned than the top, for what that's worth.

Any thoughts?

My X-axis backlash is zeroed out, as the test print doesn't indicate any. For reference:

[mewertow]

Hi David,

My first guess would be that the hole is just slightly oversized or shifted relative to what the V-One is expecting. If you have a look at the design, the trace going into the pad around the problem hole terminates right at the edge of the hole:

If the hole is at all oversized or slightly shifted, the V-One will try to probe the edge of that trace and fall in. To get around this, I'd suggest the following:

1) Terminate the trace at the top or bottom of the pad, rather than right at the left edge.
2) Oversize the hole / pad radii by about 0.2mm.

That should resolve the issue - if not, let me know.

• Matt

[DavidYon47]

Matt,

That makes sense to me. I'll look at making that trace come in at a diagonal, and identify others that might have similar issues.

David

[DavidYon47]

That seems to have cured this particular board. Would be nice for the probing algorithm to give through-holes more breathing room to account for this type of error. After curing the problem with the through hole I had to re-route a couple of traces that led to vias due to the same problem.

I'm still a bit worried about inaccuracies across the board after doing an extremely precise alignment (like, through-a-loupe precise). I get skew of 0.1mm or maybe more when I do alignment checks. The errors appear to have a center of gravity, where one area of the board is ok but the farther away you get the less accurate the probe, almost always on the X axis.

At this point I'm just not even checking, as either the detailed Z-probing will work or it will snag a hole and fail, doesn't seem to matter how precise I am with the initial alignment. If Z-probing succeeds, the prints will sometimes noticeably-but-not-fatally drift from dead-center precision.

[DavidYon47]

Managed to get a print of this board. Only issue I had was with the FPC connector (0.5mm spacing, so I know I'm pushing the limit). One was that the traces in the Gerber didn't quite touch the pads on the connector (there was a tiny gap). It took a few tries for me to trust that the ink was going to make a connection. This was partially due to the fact that I had to push the hairy edge of low-flow ink calibration, otherwise 0.5mm pads will simply glob into each other and short out.

I also ended up having to manually reflow the connector. Same as the last test print I did, the connector rotated a small amount, causing misalignment. I'm guessing that the unevenness of the pads overcomes the surface tension of the melting solder. Or if the solder melt happens earlier on one side vs the other, then you get some torque.

At any rate, manually dispensed paste and careful work with the air station was enough to get the connector solidly on the footprint. Which I count as a victory given the limitations.

I have some through-hole components and some hand-soldering to do, and since the circuit is working so far (at least what I've populated) I'd rather not wreck it with the touch-and-go luck I have with hand-soldering. I'm printing another practice board (with fresh ink this time) to see if I can get the Zen of the new wire.

[jazozaya]

Hi @DavidYon47,

Quick question unrelated to latest posts - in the picture you posted above about your X-Axis backlash. Did you print that with a broken nozzle?

The lines look much more bubbly than what they should be. It looks like way to much ink was being extruded or your print height is off. Do you have ink residue build up in the switches?

[DavidYon47]

Nozzle wasn't broken. I did repeat a couple of traces but otherwise went normally.

I do find that I'm usually feeling around in the dark in terms of dialing in flow, especially height vs rate. I'm getting ready to print a much more involved version of the Pi hat, any way I can get more guidance on that this time around?

I'm going to again be fighting 0.5mm pad spacing, as well as trying as best I can to keep the height of the traces even and consistent. If there's a magic balance between nozzle height and flow rate that accomplishes this, I'd be thrilled.

[michael]

Follow up

Because I'm having similar alignment precision problems, I decided to adjust the x axis backlash as well, and see on my late 2016 machine that it's similar to your print.

Namely, it seems that the first line in both of our prints with perfect alignment is '0' and thus the backlash gcode value should be zero. So:

Alt+C
8 > M507 X0
< Offsets Stored
< ok
8 > M400
< ok

Questions

Can someone please confirm that the command M507 X0 is correct for the machine(s) in question?

Also: How can we read the value stored in the Voltera One's gcode configuration? Is something like 8 > print M507 possible?

Nozzle

By the way, I've read advice about broken nozzles. The picture above (with black PCB and taped over holes) was done with a brand new nozzle and all default configuration (reset using restore defaults) except for a -4 print conductive adjustment.

The taped over holes are my idiot attempt at avoiding a probe sinking into them.

By the way, the backlash compensation guide is misleading or at least confusing. It shows a print clearly illustrating a backlash value of 1 and then explains how to consider the pair labeled 4 as the first aligned lines. Why not keep it simple by using consistent illustration?

[mewertow]

@DavidYon47 In general I'd suggest keeping the print height at the recommended 0.1mm, and adjusting only the flow rate - if you're having issues with height, that's generally a sign that something may have gone wrong with height mapping. Also, not sure if I've mentioned this here, but make sure not to remove the dispenser after it's been calibrated - if you do, just hit the Home button to reset the positional reference.

For your case, since you're trying to print 0.5mm pin-to-pin pitch, you can try reducing the print height to ~ 0.08 - 0.07mm, and trying out the 150um nozzles. This will help with getting a smaller minimum resolution.

The only other case when you may need to adjust height is if you're printing on a template with 2oz copper pads - those have a pretty significant height, so it's possible that print height may need to be bumped up a bit.

Hope that helps!

• Matt

[michael]

Thanks @mewertow, yes your advice helps. Can you please look at my black PCB (with tape) picture and tell me your opinion of which backlash value you believe I should use when setting the M507 gcode configuration? Thanks.

[mewertow]

Hey @michael, I agree with your assessment of backlash in the image - the 0 mark is dead-on, so your unit was well-calibrated when you took the picture, meaning the backlash value shouldn't need to be adjusted. If you'd like to see your X axis backlash value, use M520 (I don't have a V-One around at the moment to grab a screencap, but you should see the value beside 'Backlash Compensation:').

However, I would recommend running the test again to ensure it's still well-calibrated. For V-Ones that we shipped after Sept 2016, we run a calibration in-house to ensure good x-axis backlash, and using M507 X0 may have overwritten this value... seems like we might have to be smarter with how we handle this on our end, or provide clearer instructions in the guide. For now, I'll update the guide on the website.

• Matt

[jazozaya]

By the way, the backlash compensation guide is misleading or at least confusing. It shows a print clearly illustrating a backlash value of 1 and then explains how to consider the pair labeled 4 as the first aligned lines. Why not keep it simple by using consistent illustration?

Hey Michael,

The 'print' is actually an image of the Gerber. It illustrates what no backlash would look like.

But you are correct, we should include an example with proper pictures.