re-calibrating X/Y size of print in mm

[jbeale 0]

I tried printing out the "Nickel Calibration Disc" from http://www.thingiverse.com/thing:84795

which is described in more detail here: http://blog.mcmelectronics.com/post/3D-printer-calibration

I find my UM2 is printing too small by a factor of 0.994 in both X and Y axes (per my calibrated Mitutoyo calipers). It is at least consistent, I did two separate print runs some hours apart of this disc and the OD measurements were consistent between the two parts in both X and Y directions within 0.02 mm which I think is pretty good. (Measured OD was 37.87 mm, CAD file was 38.10 mm). On the Z axis, the parts were too thick by 0.27 to 0.30 mm but I assume that is a fixed offset, as I have set Initial Layer Thickness: 0.3 mm.

So now I would like to enter this calibration coefficient into the machine so it can enlarge the X and Y outputs to the steppers by a factor of 1/0.994 and print closer to nominal dimensions, but is this done in Cura, or the machine firmware? I can't find any mention of it in the UM2 manual, or the LCD maintenance menu. The Cura user manual at https://www.ultimaker.com/spree/uploads/38/original/Cura_User-Manual_v1.0.pdf

does not describe this kind of calibration constant.

 

[anon4321 16]

It is not a calibration problem. PLA and other thermoplastics shrink while they cool so you will always get parts that are slightly smaller than the model.

There are three solutions to this listed below in most recommended to least recommended order -

Most - Use Cura to scale the model before printing by a small percent.

Medium - Use a CAD program to design or alter a model to account for shrinkage

Not recommended - Change the steps per unit in the firmware.

UM has calculated the correct steps per unit based on the mechanisms (pulleys) used. When the firmware needs to move the head a distance of say 10mm, the steps per unit are calculated such that the head accurately moves 10mm. It's not recommended you change these values as the error comes from another issue namely the shrinkage.

The next one, changing the model, means that the model becomes specific to the type of material. If a model is altered to account for the shrinkage of PLA then it won't be accurate when it is printed in ABS which has a different shrinkage factor.

The Z axis error is probably do to the bed leveling/distance to the nozzle which is very hard to get accurate.

 

[jbeale 0]

It is not a calibration problem. PLA and other thermoplastics shrink while they cool so you will always get parts that are slightly smaller than the model.

 

Thanks for your reply.

It does seem to me, given that the firmware knows I am printing in PLA since I selected that in the menu, I should also be able to tell the printer that "this particular brand of PLA shrinks by 0.6% after printing". Or if all brands of PLA shrink about that amount, then the printer firmware should already compensate for it. Is there a reason why the printer should pretend that PLA doesn't shrink at all? Is there a problem with having a firmware calibration constant for this? If there was such a factor, I could print the same .gcode file in either PLA or ABS (after selecting the material in the printer menu of course) and have them both dimensionally correct.

 

[illuminarti 18]

The shrinkage amount not only varies with the color, brand and even production run of the plastic (since all can introduce different additives), but more importantly with the geometry of the object being printed and the rate of cooling (which depends on fan settings, ambient temperature, humidity, etc). It's not in any way a simple linear factor.

 

[jbeale 0]

I see. If that is the case, it seems the calibration method described here http://blog.mcmelectronics.com/post/3D-printer-calibration#.U7QY30BCmdw

is actually pointless...?

 

[drayson 75]

Tried the same at the beginning of my "3D print passion" and can confirm: pointless... :-)

[Dim3nsioneer 557]

Any calibration based on printing an object is pointless IMHO. However, I think it makes sense to check if a travel of e.g. 100mm is 100mm or 99.3mm or 101.1mm...and adjust the steps per unit...

 

[illuminarti 18]

It's one thing to calibrate the total distances moved - especially for the extruder as that's not an exact number as it depends on the pressure applied to the filament, how far the teeth penetrate into the material (and hence the effective diameter of the drive gear) and also the effects of http://www.extrudable.me/2013/04/18/exploring-extrusion-variability-and-limits/, if you really want to go there.

However, measuring the size of finished parts is futile looking for a generally applicable magic number is futile.

 

[jbeale 0]

Saying it is futile seems to me like we can say nothing about the final size, but are there not some approximate values, and some limits? Does PLA always shrink on cooling, or does it ever expand? I have read some people say ABS shrinks about half a percent, up to as much as 1% (due to thermal expansion coefficient of ~ 70 ppm/K), and PLA shrinks a variable amount but it is less than that. Is PLA part shrinkage ever more than a few percent? Let us say for designs around 10 cm in size (eg not something tiny like a single strand wisp of fiber).

Correct me if I'm wrong and it is possible to get a solid part that is, for example 8 cm long, or 12 cm long, due to variable plastic effects, when it was designed to be 10 cm. If that were true, I don't think 3D printing would be possible; layers would not align well enough to adhere to each other.

This link https://thingiverse-production.s3.amazonaws.com/assets/a7/3c/8f/7e/89/Zero_to_Hero_Rep2_Guide_DigiFabLab_2013-04-06.pdf

says "For precision work, it is useful to keep in mind that PLA shrinks slightly. Shrinkage is about 0.2% (or one-fifth of one percent)."

for what that's worth.

 

[jbeale 0]

Note: I tried printing two versions of an object nominally 100 x 10 x 2 mm (printed on edge, with the 100x2 mm face down) and the results (in PLA) were 99.74 x 10.19 x 1.86 and 99.72 x 10.16 x 1.89 mm so that is consistent with 0.2% shrinkage and 0.2mm slop, with some extra Z thickness for the inexact levelling on the first layer.

It was this one: http://www.thingiverse.com/thing:382463

also available as https://www.youmagine.com/designs/10-cm-ruler

I printed "5mm-ruler.stl" at 40 mm/s and "cm-ruler.stl" at 20 mm/sec. No surprise that the slower print had smoother surface texture, with less sign of mechanical resonances from head stop/start.

larger images at http://umforum.ultimaker.com/index.php?/gallery/image/6977-ruler-compare/

 

[drayson 75]

Maybe my findings help: I run test cubes with 10, 20 and 100 side length and 3 different colours to check my calibration. Comparing the results, the factor was not linear nor constant...

This was only at cubes. With more complex structures, it became more inconsistent as illuminarti mentioned.

Nevertheless, let us please know your findings!