Parallelism, other issues

Hi everyone,

K8800 is my first delta, although not my first 3D printer (about 6th in total, including SLA, and various cartesian FDM printers - just so happens i never trusted or liked deltas).

Assembly went quickly and easily (about an hour), first prints were nice for a 0-configuration, ready-to-assemble kit, but then i started noticing issues…

First and foremost, parallelism, or lack thereof. My test cubes, and various prototype prints end up with sides that are not straight, but curved. Not all of them, just one, and it’s consistently the same one. Any clues on what to look for?

All three belts are tensioned just enough (slight twang, but not overly tight), everything’s lubed and oiled up, there is no binding in any of the arms, and all carriages move freely.

Second, i had one of the belts always slip near the bottom of the range, until i pulled it out, bent it the other way a bit, and put it back. Now it’s fine (so far). Whoever thought of using idlers that are unsupported on either side must’ve never printed for longer than the demo keychain. That’s a serious design flaw, just like the bed that’s not fixed, but can be nocked around by the nozzle.

I’ve noticed someone posted a fix in the shape of a clip - has anyone tried simply adding a small washer and then a large washer to the front and back of the bearing stack? (Small to stay in contact with the stator of the bearing, then a big one to prevent the belt from sliding off - the issue with this design is the belt would rub against it, instead of it rotating freely with the washer.)

And finally, third thing - i occasionally get X-Y drifts - i can’t call them skips, because they happen over very gradually, over a multitude of layers. To me this smells awfully like the pulleys on the steppers are slipping. Before i start looking into that, does anyone have a different suggestion / debugging starting point?

Oh, great, i’ve jinxed it.
Now i’m getting layer shifts too.

Apparently, concave/convex lines are usually due to mismatched rod lengths and/or wrong rod length input in the firmware. As much as 0.1mm delta can cause noticable curvedness on long enough lines.

Going to take all the rods down after the current print and check them out for lengths - i did tighten them down quite a bit, but maybe one of them is overtightened and slightly compressed, or maybe bent.

Some photos:



Took all the rods down, measured them, they’re all same length, although some had the ends cut and/or had leftover swarf from cutting, making the endcaps sit at a very very slight angle. Doubted that affected anything, but in order to exclude it, sandpapered them very lightly until perfectly true.

Remeasured everything, put them back, and - still getting the exact same bend in the test print, in exact same place.

I’m now half-convinced the delta constants (rod length, distance to center) are set slightly wrong in the firmware.

Welcome Orcinus,

Although I didn’t had this problem so far (I own this printer a year now and printed 3km of filament) I think your problem could be connected with the optical end stop positions. The end stops should be on the same height as good as possible.
Connect your K8800 to a computer and send a G33 to calibrate your printer. This will report you some X,Y,Z heights try to adjust the end stops so that difference is minimal.

Here is a example of my k8800 with an installed heated bed:

09:39:40.289 : G33 Auto Calibrate
09:39:42.612 : Checking... AC
09:39:42.614 : .Height:293.17    Ex:+0.00  Ey:-0.20  Ez:-0.40    Radius:107.20
09:39:42.615 : .Tower angle :    Tx:-0.29  Ty:-0.06  Tz:+0.00
09:39:59.357 : Iteration : 01                                    std dev:0.030
09:39:59.359 : .Height:293.14    Ex:+0.00  Ey:-0.20  Ez:-0.44    Radius:106.96
09:39:59.361 : .Tower angle :    Tx:-0.28  Ty:-0.05  Tz:+0.00
09:40:21.820 : Calibration OK                                    rolling back.
09:40:21.822 : .Height:293.20    Ex:+0.00  Ey:-0.20  Ez:-0.40    Radius:107.20
09:40:21.823 : .Tower angle :    Tx:-0.29  Ty:-0.06  Tz:+0.00
09:40:21.825 : Save with M500 and/or copy to Configuration.h

Thanks, unfortunately, they’re already -0.02, -0.08, +0.00.

It seems like the source of the problems is the Z tower - the concaveness/convexness is along the axis going through that one. Printing a bigger test print to establish if it might be an angle issue, after that’s done, i’ll unscrew that tower completely, and retighten slowly again - perhaps it’s twisted.

Z tower is also the one that had the belt slipping off, which is a bit too much of a coincidence to be a coincidence :slight_smile:

Just by the way, the endstop offset is not really that important (as long as it’s roughly in the same ballpark). Marlin’s auto-calibration adjusts for it, as well as tower angle (offset from 120 deg).

What it doesn’t adjust for is differences in rod lengths - that’s why that was my first suspicion.

Do you see this effect also on different models or just on this test cube? What kind of infill did you use and does it change anything if alter it? Which Slicer did you use?

About the pulleys slipping I did have that issue, I re did them with red thread locker and they are fine. And my belts never slipped off, they are perfectly in the center and I think that when they slip off it means something is wrong.

Can you please check wether the prints are square to the bed in Z axis?

Also I’ve remixed the test cube to be a prism. You might find it useful :wink:

1 Like

They’re square to the bed in the Z axis, and i’ve checked the pulleys and they’re rock solid.
I’ve also unscrewed the top plate, bottom plate, and the case from the bottom plate, to let everything realign, then retightened everything in a star pattern.

No luck, same exact thing.

I’m using Cura, with grid infill. Slicing configuration doesn’t really affect the straightness, from what i’ve tested. I’ve also tried printing calibration hexagons, asterisks and various other circular / 3-axis-symmetrical objects. On those one of the axes has a 0.5-1% size error.

Do note that on a delta, any error in rod lengths / any other non-orthographic error will typically be exhibited as circles turning into tri-lobes, and lines turning into arcs - in some cases the dimensions of the extremes might even end up being exact, but the shape is deformed.

Don’t wanna jinx it, but i think i might’ve found the culprit!

The Z axis carriage is slightly tilted!
With the printer completely level, a spirit level indicates X and Y carriages are level, but Z is slightly angled towards the Y side of the printer, tilting the effector sideways as well. If i’m not mistaken, that should produce concave lines bending in the Y direction, pretty much as i’m experiencing.

Now i need to figure out what’s causing the tilt…

I forced the bushings in the zipties up on one side and down on another, using some toothpick tips jammed under the zipties, as a temporary kludge, to test the hypothesis out - guess what…

It reduced the curve / made the lines straighter.
It’s still not perfect, but there’s a noticeable difference when i put the new test object next to the old.

I’ll need to strip that carriage down, and figure out a jig and a way to ensure everything’s perfectly square to the rods. I think i’ve got some LM10UUs somewhere too, might give that a try too.

1 Like

A semi-relevant observation.
The top plate is mismachined…
Consistently so - so it looks like a design issue, not production issue.

It forces the rods apart to a distance wider than the centers of the bushing receptacles on the carriages. So when you mount the carriage, it’s deformed by the bushings (displaced by rods), causing unnecessary deformation and wear on the walls of the bushings.

I’m glad you figured it out.

@riodoro1 are you using the original sleds or Dr_Vegetable’s, SusisStrolch’s, Psychos LM10UU/Drylin modifications? What are your jerk and acceleration setting to get acceptable ghosting effects on your prism?

It’s not the top plate. It’s the carriages!
It was always the carriages!

Whoever did the injection moulding on them did a piss-poor job.

I took all three down, and checked them out.
They’re all twisted/bent to a slight degree, and force the rods to bow inwards / carriage to bend outwards, causing all manner of stuff, including that twist/tilt i was getting. Z was the worst, X was almost fine, Y was somewhere in between.

I ended up heating them up with a hairdryer, and flattening them against a kitchen sink until they fit the bushings without forcing (tip: put a towel between your hand and the carriage so you don’t burn yourself). I’ve then replaced all the wimpy zip ties with thicker ones, that match the grooves on the bushings better. These don’t fit the slot in the carriage, though, so they get twisted a bit as they go through - this is not a problem, as long as one side is twisted in the mirrored direction from the other side. In fact, it has the consequence of forcing bushings and the carriage to align better, as they offer some upward / downward force when tightening (depends on the twist chosen).

After doing this for all three towers, suddenly, all the lines are straight.
Now i just need to calibrate the dimensional accuracy, and it’s gonna start looking like something usable…

I’ve some videos of the carriage issue + the post-fix result.
Going to upload that later tonight, after work.


@shelxle I’m using stock carriages/bushings my acceleration is 3000 and jerk is 20. I’m talking about values set in marlin. I usually print around 35mm/s and there is some ripple or ghosting but I don’t mind it at all.

@orcinus Great news that you found your issue. I remember there was a fix somewhere here from Velleman for an issue where the carriages were too narrow and they put too much force on the bushings so maybe your issue is related to that. They cited that early manufactured carriages shrunk too much but maybe they also warped.
While you have your hair dryer out you can also bend the bed holding hooks a little inward to avoid the bed moving around, that can sometimes ruin the finish on long prints and a little tension does not impede bed leveling at all. I’ve also seen my heatbreak loosening sometimes and this too resulted in poorer quality. Despite all that I think it’s a good machine and the quality can be pretty great.

Glad to have another problem solver on board :wink: