Welcome to the SeeMeCNC Community

After a week of test i'm not been able to obtain a right dimension of my printed object.
Usually the external dimensions are right or slightly smaller but every circle and hole are smaller just of 1 / 0.8mm
There is not correlation with layer width and hole error; i also tried to switch my nozzle from 0.5 to 0.3mm without benefit.
My extruder calibration looks good, on a solid cube all looks right.

My only solution is to scale up the object of 2% for obtain a good tolerance.

Any suggestion????

Do a google search for nophead poly hole.
He explains the phenomenon in his blog. Basically the issue is that holes will always shrink on a 3d printer, and round holes the most, his solution is to use polygon holes with a small number of faces, because they shrink less.

Personally I just upsize them slightly in the model, 3.2 will usually give a pretty good fit for an M3 bolt.

Polygonhell wrote:Do a google search for nophead poly hole.
He explains the phenomenon in his blog. Basically the issue is that holes will always shrink on a 3d printer, and round holes the most, his solution is to use polygon holes with a small number of faces, because they shrink less.

Personally I just upsize them slightly in the model, 3.2 will usually give a pretty good fit for an M3 bolt.

Interesting you mention that. I was always wondering why things like the Xnaron mag-arm effector had such basic screw holes (pentagon shape). I guess that might explain it?

Really interesting.

In practice the error is made by slicer that no do a compensation on the mesh.
IMO an little modification on slicer that scale up any circle or arch by a coefficient of they triangular face can correct the problem.

In fact if i measure the circle the first inner perimeter is just of it should end (inner face).

Reading the gcode also i see only G0 and G1, maybe a improvement with G2 and G3 could be a great idea

The problem with G2 and G3 is two fold, all the slicer has to deal with is the STL file which just contains a set of triangular faces, so in order to generate a G2 or G3 it has to identify what features are intended to be circular arcs.
Even if the slicer could do that reliably, the firmware just turns it back into a series of small line segments.

Yes, at second look the g2/3 commands not are very useful with this mcu (atmega2650).

I just open a issue on slic3r githup for ask they to add a "correction coefficient" for holes