Delta printer dimensional accuracy not square

Hello, before I go into the problem, allow me to give some background so that we avoid the noob talk. I am a mechanical engineer and I have been 3d printing since 2013 I believe. I have experience with Rostock MAX V2 and now also the HE3D K280 (yea it was a very cheap option and hard to pass up). I am very familiar with Marlin 1.1.8 and 1.1.9.

For the purposes of printing trinkets, the K280 with its auto bed leveling feature is great. However, when I try to print something that requires dimensional accuracy, and let’s say I print a rectangle that is supposed to be 5.310" by 8.070", then I would get 5.334" on one end by 5.394" on the other. I’m not concerned about the scaling, it’s more about the .060" difference between both ends of the part. I think the other axis had a .020" difference. This is a huge problem for me and so I updated to Marlin 1.1.9 with unified bed leveling in the hopes that I can adjust this difference (steps per mm of each arm) and then the mesh table could just lower or raise the print head regardless of the unoffset tool head path. Well that did not solve anything. UBL does a great job at allowing you to print the first layer but it did not make the parts more square.

I decided to re-frame the whole printer with stronger and stiffer 2040 aluminum towers instead of the 2020 and use metal corners and doubled up on the corner joiners. I test printed and now my difference is about .030". I made sure to build this thing very carefully and using a metal 90degree aluminum angle piece to check if everything is being assembled together squarely.

.030" is still too much. I have made sure that the print bed is at 90 degrees versus all the towers and it is.

Here are the 2 things I have been fighting:

  1. Keeping the print bed at 90 degrees to all the towers does not cause the mesh in UBL to be remotely uniform. The print head is compensating 2mm more near the Z tower area versus the rest and there looks to be a lean to the print bed when there physically isn’t.
  2. Knowing this, I decided to compensate using the steps/mm for each tower until the bed looked level in the UBL mesh. I thought this was success but then I printed a calibration piece where I have 150mmx4mm rectangles printed from the tower towards the center of the print bed. I took measurements and they were all off substantially.

I tried fixing the mesh using endstop corrections but they don’t seem to react at all when using UBL. The mesh remains the same even though the comps are saved.

So the biggest problem I am having is not the absolute size of the part but the fact that it doesn’t print squarely. It either skews to make smaller or to make bigger towards the outer edges of the print bed.

Has anyone dealt with this or tried to fix it? I believe that the printer itself has been built properly since I am able to confirm perpendicularity of the bed to the towers. Regardless, is there any way to address this through firmware compensations?

I’ve only printed my calibration piece when I made the mesh more or less equal across the whole bed. Perhaps I’ll try my rectangular object and see if it will print more consistently then and if the scaling is off, then I will adjust that in the slicer.

Also, if I just went with a Cartesian, would all these issues disappear??

Delta printers are a pain like that. Cartesians are simpler to debug. At least that is my understanding. I have had 3 delta printers now and I have grown to detest delta math and complexity. I hear it is all just slants and scales that you have to worry about on cartesian printers for dimensional accuracy.

If your gearing and stepper motors are identical, all 3 towers must have identical stepping.

If you print the rectangle 5" or so tall, is there a difference in dimensions at the top compared to at the bottom? If so, you have a leaning tower.

Are you 100% sure your diagonal rods are the same length? Are you 100% sure your effector is uniform for all 3 sides? Are you 100% sure the parts that slide up and down the towers are identical? If either of those are off, it will cause a problem.

I hope you are measuring the squareness from each tower to each other tower and not just straight across since sideways slants can screw things up too.

Here is a website that may help, but it is a little bit unclear on what the pictures really mean.

Yikes! You had to wait 29 days for a reply. That sucks. I hope you had the problem fixed before my reply. 29 days with printing issues is too long.

Hi Nathan,

I figured it out on my own. Apparently no one in the world has this issue besides me or everyone who owns a delta only prints trinkets.

Anyways, I wrote up a tutorial on the reprap forum:,861588,862433#msg-862433

Now my printer is accurate within a few thousandths. This would not be possible without UBL mesh tables.

1 Like

Something seems odd about your instructions. You mention setting the steps/mm for each tower and it makes it sound like the different towers can have different steps/mm. Unless the pulleys are different on a delta, the 3 towers should have the same steps/mm.

Nope. Unless you have a belt deflection gauge that is incredibly accurate, there is slight stretch(to varying degrees) to the belts that you need to compensate for. We are talking small fractions but they do affect accuracy in all directions. All my towers have a slight difference in steps/mm and I print very accurately now. It could also be that like you are stating, the pulleys are slightly different. Makes no difference to me what it is if I can compensate with the steps/mm and kill 2 birds with one stone.