z height shift

[rockethacker]

my machine setup

• rostock max v2
• injection molded carriage upgrade
• ball cup arm upgrade
• belt tensioner screw jack upgrade (my own design)
• e3d hotend
• stock firmware (repetier 0.91)

i'm having a very odd issue with z height shifting. let me see if i can clearly explain my observations.

• this issue gets worse when printing close to a tower
• issue it unnoticeable at the center of the platform
• z height will rapidly shift up to 0.2mm when the travel direction of the tower it is closest to changes
• z height is lower when traveling toward the tower, and higher when traveling away from the tower
• this issue existed before i installed the carriage/arm upgrades

i've tried the following

• tightened belts until they play a note. installed belt tensioner screw jacks and maxed out their travel
• checked all motors for loose screws or pulley set screws
• squared up the towers
• loosened belt tension jacks four screw turns and it didn't get better or worse
• grabbed the extruder by the nozzle, gave it a wiggle and verified that no hardware on the effector is loose

questions

• movement tends to indicate backlash in the tower, as the carriage appears to lag where it is supposed to be?
• if it is backlash, should i enable backlash correction in the firmware?

[skappes]

Did you change the horizontal radius and recalibrate when changing to the new arms?
The manual says to start at 140. I ended up at something like 144.

[rockethacker]

Did you change the horizontal radius and recalibrate when changing to the new arms?
The manual says to start at 140. I ended up at something like 144.

yes. that was checked out.

horizontal radius will not cause this behavior. it only influences the concavity of the plane. here the z height is shifting significantly in just a few millimeters. if you do zig-zags, it will print alternately tall and short extrusions.

[rockethacker]

ping. any ideas anyone?

[bot]

It is my experience that this is the best a Rostock can be calibrated. What you need to do is average out the difference as best you can, so that the extrusion at the center of the bed is the same thickness as the average of those spots you point out. That is, if the thinnest point is .12 and the thickest point is .32 (of those near the tower points) You'll want to have the center of the bed's extrusion to be about .22mm thick. You'll then want to verify that the extrusion is as close to .22mm at points around the bed as possible. It will likely vary by .1mm minumum. In order to use the entire surface area of the bed, this is how my rostock is calibrated. I must use a .25mm layer minimum for prints that use the entire area.

[IMBoring25]

With the thin spots being consistently on the clockwise side of each tower, it would seem to me that would be consistent with a whole build that's twisted as it goes up, so there's a consistent circumferential lean on each tower.

[bot]

With the thin spots being consistently on the clockwise side of each tower, it would seem to me that would be consistent with a whole build that's twisted as it goes up, so there's a consistent circumferential lean on each tower.

I'm not sure if the effect of a frame twist would be so dramatic. I have a theory that this is caused by backlash in the belt system. The farther out to the edge near a tower the effector is, the more the positioning depends a lot on that one tower. As the effector passes a tower, the tower has to move rapidly up, then rapidly down. As it is moving up, it's taking up some of the lash, so the effector is lower than it should be (because some of the upward movement was wasted on backlash), conversely, once the effector starts moving past the tower, the carriage is dropping, and thus the backlash comes into play again. the carriage needs to be forcing the effector down, but the lash prevents it from pulling down as far as needed.

So, in essence, the belt tension might be a factor... do you have astrosyn dampers on?

[rockethacker]

Yeah, this is definately not an artifact from twisted towers. the height shifts dramatically when the carriages change direction of travel. if you're not convinced, i can do another demonstration that further discounts it.

It appears to be backlash of some sort, but i'm not sure where it is coming from. I've tried tight belts and loose belts with relatively the same results (i really though belt tension would play a bigger role...?). I'm using the stock setup with injection moulded carriages. I don't have the astrosyn dampers either.

I think it could be caused by excess friction in the arm ball joints. I'll try adding some oil to them next time i work on it. Does this seem like a reasonable fix?

[blackrabt]

Any word on the effectiveness of the oil?

Have you tried playing with print speed to see if it changes the harshness of the backlash?

[teoman]

Have you looked at the motors, i did discorver that the motors themselves were a bit loose. The melamine had compressed i think.

[rockethacker]

OK, i've done another 20 hours of work and did the following

1. addressed my non-flat build surface. The heater was bulging in the center. i went down to only using three clips and shimmed the clips with kapton tape until it was mostly flat. i used a metal ruler and feeler gauges to work this out.
2. i did some hard core observation when the effector is at the perimeter and noticed that the z shift is NOT coming from belt backlash. it is actually due to the effector experiencing rotational deflection as the two far arms shift direction and go from static to dynamic friction. i backed this hypothesis up by tightening and then loosening the belts to the extremes and noticed that it really didn't impact the z shift i'm trying to fix. note that this was a really tedious test because adjusting belt tension requires a machine re-calibration. i settled on a medium tension based on the seemecnc recommendation (it turns out they know what they're talking about).
3. disconnected the ball bar (for lack of a better term) from the effector and moved it around by hand. i noticed that the ball joints have less resistance when close to the edge of the platform if the bar balls are installed on the effector with the ejection registration marks facing upward and the carriage bar balls are installed with the ejection registration marks facing inward
4. This made me realize that the balls are exactly spherical, so i detached them and used a piece of ribbon wrapped around them to spin them in their sockets. figured that this may help break them in
5. I had an idea that this behavior may improve if the balls are broken in a bit, so i did the absolute best calibration i could. got it where i could print on the entire platform and get reasonable adhesion if i printed with a first layer height of 0.225 on the machine. then i put about 200 hours of printing on the machine.

After all of this, i re-printed the test array and this is what it looks like:
[img]160105-IMG_20160105_204300.jpg[/img]

it looks much better. i'm seeing a 0.10 to 0.15 mm variation on the outer squares and less than 0.10 variation on the next square in. The effect isn't gone, but it looks like it has improved 20% or so. Because this is a ball joint friction effect, i don't think you'll ever get this to go away entirely, but i'm not happy enough for a machine at this price point.

Takeaways:

1. tower squareness doesn't have a significant impact
2. belt tension doesn't have a significant impact. medium tension is adequate
3. changing belt tension requires a re-calibration. don't muck with it
4. build platform flatness has waaaay more impact on first layer adhesion than backlash. focus on that first. it is easy because it only requires a ruler and some tape. don't forget to do it when it is hot! imho, this should be added to the assembly guide.
5. i can't say for sure if oiling the balls helped. i wouldn't recommend it. i suspect that over time it will add more resistance as the oil attracts dust and "dries up"
6. beyond the first layer issues, the z-height causes issues with support material. when printing an array of ten objects on the platform that all require support to print, some support would remove easily and some would be glued on.
7. Because I think that these issues are inherent to deltas, i'll probably avoid delta printers for my next machine purchase.

[thedoble]

Interesting reading, are you saying that the glass bulges in the centre? Or is it the heated bed that is bulging, and you're 'levelling' the glass by adding the tape as a spacer?

Thanks for posting your findings

[rockethacker]

correct. the onyx rev 5 bulges upward in the center. i though that maybe the mounting screws were binding it, but after loosening them, the bulge remained.

Yes, my glass is perfectly flat. i started by measuring it by itself to verify.

I shimmed with kapton tape under the clips. you can see it in the picture above. i think i used 3 pieces of tape on two corners and 5 on the other. i should probably come back in there at some point and even those up...

The bulge is due to the fact that the center gets heated and expands, but the edges do not receive as much heat so expand less. this seems kind of hard to avoid with this design. imho, i think you could add some stress relief slots in the board, but that will force you to use some complex trace (heater coil) routing and will make even heating hard to achieve.

note that there are newer onyx versions out there, maybe they don't have this issue?