Accuracy of small circles on perimeter  

  RSS
KjsiderPrinter
(@kjsiderprinter)
Eminent Member

Hi all, I'm hoping someone has a lead on what settings I can apply to get better dimensional accuracy of small features on the perimeter of my object. This is what I'm trying to print: https://www.4dbrix.com/products/train/2-04-tracks/005/index.php.  

The parts where the tracks connect to each other are not working that well when I use Prusa Slicer. The extruding part is supposed to be 3.90 mm wide and the accepting part is supposed to be 3.90mm at the entry. However, using the default 0.15mm Original Prusa profile, the parts are coming in at 3.70 mm for the extruding connector, and 3.80 mm for the accepting connector. This means the tracks don't grip each other correctly.

Scaling the entire piece up doesn't seem like it's the right thing to do since it'll just exacerbate the issue. I did try setting the XY Compensation to different values and found that changing it to something in the range of 0.025 allowed the parts to get gripping power. However, they're still undersized to what they're supposed to be. 3.7mm actual vs 3.9mm expected. What other settings are there to play with on Prusa Slicer to get better accuracy in this part of the model? 

What's interesting is that Cura gets better values here using its stock settings. The pieces will grip together fine. Both Cura and Prusa Slicer are producing similar results in terms of the overall width of the test models so it seems like just an issue in these particular places. 

p.s. dimensions of 20mm cube are coming pretty close at 19.94 mm for X and Y. The XY calibration cross yielded similar numbers at the 20mm widths. I could make adjustments here to get "bigger" overall, but I don't think this will affect the connector parts of the model.

Posted : 22/02/2021 6:55 am
Dan Rogers
(@dan-rogers)
Estimable Member

For dimensional accuracy - be prepared to do micro-scaling of certain parts that are meant to fit together.   You'll want to do some test prints to find the right scale #'s for each piece, until you get a fit that you want.

Posted : 22/02/2021 4:30 pm
KjsiderPrinter
(@kjsiderprinter)
Eminent Member

@dan-rogers

What is micro scaling? Is that scaling the object in very small scale or scaling only portions of the model? The problem I’m having is that only some parts of the model are wrong by a significant amount. Unless you think scaling the whole object will affect certain areas more than the other areas. 

Posted : 22/02/2021 5:51 pm
Diem
 Diem
(@diem)
Reputable Member

Try Print Settings > Layers and perimeters and tick External perimeters first - you need to be in advanced mode.  I've not tried it in quite your situation but it's often useful in close tolerance print-in-place parts.

Cheerio,

Posted : 22/02/2021 8:10 pm
KjsiderPrinter
(@kjsiderprinter)
Eminent Member
Posted by: @diem

Try Print Settings > Layers and perimeters and tick External perimeters first - you need to be in advanced mode.  I've not tried it in quite your situation but it's often useful in close tolerance print-in-place parts.

Cheerio,

From what PrusaSlicer puts into the description of that option (on their website too), it seems like it is beneficial when the printed size ends up too big? The description says it'll move extra filament away from the outer wall so it goes towards the inside of the object? 

I'll give it a try anyways, but seems like it would make the problem a bit worse because the outerwall is not slicing to specs. A different slicer(Cura) seems to be providing better dimensional accuracy in these specific areas of the model, but I get nicer looking overall print from PrusaSlicer. 

Posted : 22/02/2021 8:22 pm
Diem
 Diem
(@diem)
Reputable Member

By putting a second instance of the part on the plate and offering them up it's possible to see the degree of fit in the .stl

The yellow disc on the right is 3.9mm.

Now let's slice it:

and we find those locator rings are so small there is only room for two perimeters and no fill between them.

The horns of the socket cannot print narrow enough at the tip to enter the narrow space behind the ring and retain properly.

This is never going to fit well with the default 0.4mm nozzle.  A smaller nozzle is the only hope for greater precision.  A slightly different extrusion multiplier might close the gap but only at the expense of accuracy elsewhere.

The solid cylinder has more perimeters and will probably retain the fit better.

It would be better to redesign the connectors, preferably without the central hole; a conical profile would locate more accurately. A longer, crossword piece style stalk to the connecting disc would leave room for the horns to enclose properly.

This part seems to have been provided by a third party who doesn't understand 3D print design. If you really need to make these you might be better to take the rail parameters and design a substitute part that will print well.

Cheerio,

Posted : 23/02/2021 2:30 am
KjsiderPrinter
(@kjsiderprinter)
Eminent Member

@diem

That’s great info. Thanks for the tips.

I do see a difference in how Cura handles it and PrusaSlicer. Even Cura doesn’t reach 3.9 but it seems to maintain the size a bit better than here. I’m wondering if it’s Cura behaving well or PrusaSlicer behaving badly. 

Posted : 23/02/2021 3:30 am
KjsiderPrinter
(@kjsiderprinter)
Eminent Member

Or maybe there’s settings I need to tweak in the slicer? 

Posted : 23/02/2021 3:39 am
Diem
 Diem
(@diem)
Reputable Member

@kjsiderprinter

I'd say neither.

The two-perimeter-and-no-fill sections are severely constrained.  The slicer has to lay down the strands close enough to stick together - so they musn't go far apart - but the design doesn't allow them to go far enough apart to insert any fill, or another perimeter in between.  There has to be a compromise.

The default extrusion width for a 0.4mm nozzle is 0.45mm.  For a two or three strand section you must use multiples.   The slicer doesn't know whether the internal or external perimeter is the most important.  If you have a master to copy you might be able to modify the extrusion width to suit.

You are discovering that not all published designs are suitable for printing.  My guess is this has been created by an enthusiast for one particular printer/slicer/filament combination without accounting for variation - this might even be a tweaked scan.

If all else fails, preassemble your track and glue it down to prevent movement.

You are not alone building railway models here, take a look at some of @jsw's work:

https://forum.prusaprinters.org/forum/english-forum-awesome-prints-hall-of-fame/model-power-plant-building/

Cheerio,

Posted : 23/02/2021 10:30 am
KjsiderPrinter
(@kjsiderprinter)
Eminent Member

@diem

I have been able to get a "gap fill" in between the two perimeters by setting a XY compensation of around 0.025mm. This does push the size out of that circle to 3.8mm. It can go bigger with bigger XY compensation, but at that point the rest of the model is unusable. I believe the printer is capable of printing that circle with this nozzle size and setup, so I guess there must be a way to setup the model to tell PrusaSlicer to do this. 

Anyways, this was an interesting case for me on how off this section was. There's probably many ways around this, including modifying the model specifically for specific slicers. I've eliminated filament and printer from the possibilities for now since a different slicer, Cura, gets much closer to the correct size, but it does suffer the same issue of being smaller than intended and being smaller than what my printer's dimensional accuracy would indicate. 

 

Posted : 23/02/2021 9:50 pm
Diem
 Diem
(@diem)
Reputable Member

I have been able to get a "gap fill" in between the two perimeters by setting a XY compensation of around 0.025mm. This does push the size out of that circle to 3.8mm. It can go bigger with bigger XY compensation,

And you can limit the effect to just that subsection by applying it inside a modifier.

There's probably many ways around this, including modifying the model specifically for specific slicers.

Yes, this is the essential point. The basic model is designed for a different technology and can only be printed sucessfully if the printing parameters are specifically tweaked.

We haven't even addressed the trickier issue, the horns of the socket cannot be printed narrowly enough at the tip to enclose the fitting and retain enough strength to be of use.

Again a modifier, a different one, would allow you to print a finer tip but only a fragile one.

Already the effort that has gone into attempting to print this object has been much greater than would be required to redesign it to be printable on a wide range of machine/software combinations.

It should be relatively simple to design a functional replacement:

The principal constraint must be the dimensions and placement of the rails, I assume the profile is critical but the length can be parametric.  Simple sleepers are just cuboids, programmatic placement is a version of the old fencepost problem.  The surface knobs are, I assume, some sort of Lego compatibility feature and can probably be lost.

As to the connectors, the reason for this thread; if you are making parametric rails then you can use your own, sturdier version between them. If they must mate with pre-existing manufactured track then simplify them; fill the hole, remove the enclosing horns from the sockets and devise a bridging clamp piece to clip them together.  I guess the holes are for screws to hold down the track - place screw holes in the centre of selected sleepers to replace them.

If the track parameters are not published you will have to measure a manufactured example.  Do not start from this part as any compromises made by the designer will be compounded by any you must make.

Fire up your CAD program of choice: This would be straighforward in Fusion 360 or FreeCAD but my choice would be OpenSCAD. If you're new to it this is a decent project to start on. Once you're up to speed an afternoon's work should be enough to create a template from which any length of rail may be generated at will.

Cheerio,

Posted : 24/02/2021 9:13 am
Share:

Please Login or Register