S5 How do I calibrate? Uneven X Y and Z scales

[cthermo 0]

I have been using our S5 to print prototype housings for a new product and just found that the X, Y and Z axes are not printing equally.  I discovered this after parts were not fitting together properly.  I have a 3.0 x 3.0 x 1.5 (XYZ) inch housing and I am getting 2.994 in X axis, 3.006 in Y axis and 1.485 in Z axis.  How do I correct for this?

 

Is there a setting in Cura? An offset on the printer?

 

I have used other printers where you print a calibration cube in your desired material and then enter offsets to adjust.  Then you repeat the print to verify/readjust. 

 

I have searched and looked through the manual and website and can't find anything to help.

 

Suggestions of compensating in the CAD model are simply ridiculous.  Surely Ultimaker has a better method.  I just can't find it.

 

Andy

 

 

 

[dxp 23]

8 hours ago, cthermo said:

Suggestions of compensating in the CAD model are simply ridiculous.

Ok, but what about scaling the STL in Cura? As a start...

The longer way would be fiddling around with the steps-per-mm in the configuration files on your S5 in Dev-Mode. But i would not recommend that cause you can brick your S5 easily unless you know what you do.

[cthermo 0]

Thanks DXP,

 

I had seen this discussed and it does seem like a good workaround.

 

But is there not a calibration procedure for the S5 to correct these errors?  My error on X, Y and Z may not seem much to many, but what if it were 0.5mm (0.020") and part fitment was a joke.  Am I to keep typing in a scale correction in Cura forever?

 

 

[Carla_Birch 116]

Keep in mind that with FDM printing you are melting a plastic, this will change in size due to heat and cooling and different brands of filament will change more or less then others going by how good the filament is, then swap the type of filament and you got the same problem but even bigger.

 

The size of the parts can also play a part due to more or less heat, so i really don't think you will get a FDM printed part down to perfect or even within a 0.01 range. Even resin MSLA printing as the same problem of parts that shrink due to heat. 

Edited June 5, 2020 by Carla_Birch

[dxp 23]

On 6/5/2020 at 7:48 PM, cthermo said:

But is there not a calibration procedure for the S5 to correct these errors?

As far as i know there is no official procedure, you are on your own.

 

On 6/5/2020 at 7:48 PM, cthermo said:

My error on X, Y and Z may not seem much to many, but what if it were 0.5mm (0.020") and part fitment was a joke.

You know the general tolerance tables?

Half a millimeter is quite common. Precision CNC-Shops have temperature controlled areas for their machines...

[cthermo 0]

DXP, Carla,

 

I am completely aware of the  tolerances one can expect from FDM and many other manufacturing processes.  Many of us in manufacturing strive to get the best accuracy we can from our machines and processes.  We have CNC machines (in temp controlled spaces) here that have tight tolerances, but a good user and a well built machine will allow you to exceed those tolerances when its needed. 

 

My main point is the LACK of a correction/calibration process for this 3D printer.  Printers costing USD$400 to $800 have it, but not a USD$6000/$10000 USD printer?  My errors in X and Y are within tolerance.  My Z is not.  I can compensate either through the CAD model or by scaling in CURA prior to printing, but this is, at best, a cumbersome workaround. 

 

I am also aware that this "correction process" would be dependent on the material and print parameters used.  So, if I changed from ABS to Nylon (or adjusted temperatures or slice thickness etc.), I should repeat the correction/calibration process. I would expect users to not use this every time, but only when they wanted the best dimensional accuracy.

 

I like our S5.  It prints nicely and consistently in a wide range of materials and we have used it extensively.  But I expected a more polished and refined machine for a USD$6000/$10000 price point.  Old issues such as the camera dropout and the near useless WiFi connection should have been fixed early on but are still an annoying problem almost 2 years later.  I fear some of these issues will never get corrected if a new "flagship" printer is released.  

 

Andy

[Carla_Birch 116]

For your Z with the S5 it really pushes down the first layer to make sure the print sticks. If you are out basically the same no matter how tall the print is, then maybe try the "z offset settngs" plugin from the market place so you can have it not push down as much into the glass to see if it effects the overall z..

[dxp 23]

A cold guess: adding e-step-corrections to the start a/o end-scripts can be an easy accessible workaround. But it is not on a material based profile.

After all: Ultimaker are generally well enough calibrated, that is why there is no easy routine.

[robinmdh 100]

We are planning a bunch of changes to improve accuracy for everyone regardless of calibration but nothing like what you suggest, I'll pass it along to those working on this .
that said here are some tips:

• The Z reliability would be hugely improved by running manual leveling to make sure your buildplate is in the same plane as the printhead. We do squash layers in the Z direction to slowly remove skew and buildplate non flatness compensation. Similar to what is shown here (though we now probe multiple points, etc):

• Further you could make sure your X and Y belts are tight!
• Did you print with the engineering profile in Cura? Because the feeds and speeds (changes in flow rate)  and layer order also influence the dimensions of the print.
• There is also a slice error in the Z axis dependent on the layer height of the machine, is your object an exact multiple of this? Even in that case you will likely miss up to a whole layer height, sue to the slicing in layers process the print is prone to shrink a bit.

We do have calibration of the X and Y offset between the 2 nozzles but indeed nothing to calibrate the precision.

 

I would instinctively question if you print a cube that small would the error in that part be representative for the whole travel range of the machine? if you were to print a cube the size of the printer's volume it might be better but it would still miss things like backlash (and take ages).

[jwsocool 0]

In Cura and the Marketplace Setting Plugin I found the X, Y and Z MM per step settings (50 steps per mm).  As my print is off by .004" (in many material types) in Y why couldn't I just change the step setting to 45 or 55  (.0007874 inches per step) to make the correct ..  It looks like I could make this change and save in my profile.  What am I missing??

 

I realize I could scale my part by ~1.2% in the Y directions also.

 

Again why could I use the step settings to make the required correction now and when the machine starts showing signs of wear.

 

I have a brand new Ultimaker S5 Bundle.

 

Thanks for any response...

[ris-pert 0]

Just realized that our S5 is inaccurate. Y-axis is ok, but X-axis if off by 0,4-0,5%. It would be really nice to have adjustment to compensate that. Scaling the model in cura is not that good solution.

 

For instance thin walls designed to work exactly according to nozzle width causes slight problems. In worst case might even need one more line. But at least more or less flow according to wall alignment.

[gr5 2,172]

.5% across 200mm is 1mm and is huge.  I doubt you have that much error.  0.5% across 20mm I believe.  How did you measure the error?  Did you print something and then measure it with a caliper?  Better to measure such that you avoid the corners as those tend to be off due to slowing down and other reasons.

 

There are so many things that cause errors, e.g. if the Z axis doesn't move enough, then that layer will be overextruding and it will stick out in both X and Y, or one of the pulleys could be out-of-round and you get +.3mm error and then as the pulley rotates you get -.3mm error but over the course of the print area it should average out.  These are only 2 of about 50 things.

 

Before trying to compensate e.g. Y scale, I'd first check if the movement of the print head is also off by 0.5% and is it off by that much from one end to the other or does the error vary (e.g. the pulley thing).

[RaidenUltimaker 3]

My S5 has  a 1.4mm error in a distance of 140mm on the Y axis.  

I noticed this after printing two Frame v2.stl from

Disney World Magic Kingdom 3D Miniature on printables, printed with 2 printers and they weren't the same dimensions, but only for the Y axis!  

Should be 134.75 but is 137mm and this is in the V (Very coarse) category from above. 

 

Does anyone have a dev-mode fix for this?  Since this is marlin firmware it should be easy to adjust the e-steps/mm, correct? 

Edited April 8 by RaidenUltimaker

[gr5 2,172]

There's 2 computers - the unix computer and the "arduino" computer that has Marlin.  They talk a lot.  I'm not sure but I think the python code on the unix computer sends this kind of thing to Marlin on boot up.  You could grep through the python code for the steps/mm command: "M92".  The nice thing about all that python code is that it isn't compiled so all the source code is visible.  Note that any changes to python will get replaced if/when you update the firmware.

 

Even just 1mm seems like a LOT for an S5.  I'd characterize it a bit with a ruler measuring actual nozzle movement and not measure the plastic which can vary by quite a bit due to cura settings, over/under extrusion, warping.  For example if one of the gantry pulleys isn't drilled out in the center and is off center by 1mm that could cause that kind of error but UM knows this dearly and has higher quality pulleys than the old days.

 

Maybe binder clip some paper to the bed, move the head around to specific positions and mark the positions with a pencil then remove the paper and measure with a good quality ruler.

[gr5 2,172]

A more common error mode is when the X and Y axis are not perpendicular.  This happens usually in shipping where the shipping company apparently has an elephant tilt the printer on one corner and step on it.  Is there any chance the error is more on a diagonal?

 

Often this error in orthogonality can be seen visually or with a square against the outer rods in the gantry.