Dimensional accuracy: part design or slicer/printer setup?

[Harblar]

Hey all!

So what I'm trying to figure out is how to correct for ABS shrinkage. My second print was a self designed part (an oven knob) and when done the overall size was .5-1mm smaller than designed, but the central shaft hole was off by 1.5mm or so. Too smal to fit. To correct I had to scale up by a factor of 1.062. This got the center hole correct but really magnified the overal size. Fine for the knobs, but this won't work for other parts I'm planning on printing. For instance, I'm planning on building spacers and PCb mounts designed to fit inside of metal tubes. For example, the inner diameter of the tube might be, say 50mm. I would need the spacer to have an outer diameter within .5mm of that to provide a nice snug fit. Some of the spacer will be more donut like with a hole in the center that another pipe would need to fit into, if that makes sense.

Anyway, this got me thinking that my printer needed a lot more calibrating to maximize the accuracy. My first print however was the peek fan shroud and it fit perfectly on the first go. Fan fit snug with out binding and it snapped perfectly onto the hotend. So... Does that mean I need to be tweaking the slicer more for my own parts or were the fan shrouds designed with abs shrinkage in mind?

Any links to good guides on printing/ designing for tight tolerances out there anybody could point me towards?

Aside from the basic calibration for the end stops, bed leveling, temp, and extruder are there any other calibration factors that one should be doing to apply to the slicer and EEPROM software?

Thanks!

[geneb]

The problem isn't in the machine, it's in the hole.

http://hydraraptor.blogspot.com/2011/02/polyholes.html" onclick="window.open(this.href);return false;

g.

[Harblar]

Very interesting read! The more I thought about how accurate the fan shrouds were the more I kept thinking it had to be in the model or software. That link really clears up exactly what is happening... at least in the case of small holes. When I designed my knob, I should have made the 'D' shaped center post hole as a 5-6 sided polygon instead of an Arc and flat line. Will remember that in the future!

When it comes to screw and bolt holes I'm not as concerned. It isn't hard to drill those out to the correct size or use self tapping screws. What I would be worried about is hole spacing. If a PCB, for example, needs two or three hold down screws in specific spots, any shrink or inaccuracies could be problematic. Any slicer specific settings to combat this or would applying a slight scale factor into the mix when designing the part be the best course? Obviously, like mentioned in the link, that's not necessarily the best plan if sharing the part for use on other printers, but maybe it'd be better for one off designs. I'm going to have to make up a few size tests to see how pronounced it is in the applications I'm looking at I guess. Seems to be much less of an issue once you get above the 10mm diameter range or so. For the most part , I can live with that. Just want to get every last bit of precision I can squeeze out of this beast!

[geneb]

The hole center points will be accurate. The only potential error is with the actual hole itself. I've made a number of little enclosures for myself and the hole centers are never off.

g.

[RegB]

Given that I am not and never have been a mechanical engineer - I have to ask WHY is this a problem ?
Aren't holes through which parts pass ALWAYS spec'd for CLEARANCE ?

Simplified understanding; a hole for a 1/2 inch bolt to pass through is spec'd for max tolerance bolt in combination with min tolerance drill bit.
If not the line stops every time a few fat bolts come in, or the tool crib issues skinny drill bits.

I understood the article that Gene pointed to, more or less - but I think clearance tolerances in "common practice" should be adequate.
========================================
OTOH a 1/2 inch diameter circle approximated as a polygon SHOULD be INCLUDED in the polygon, not the polygon included in the circle.
Should "Fix Normals" (or something similar) resolve this ?

[Jimustanguitar]

Given that I am not and never have been a mechanical engineer - I have to ask WHY is this a problem ?
Aren't holes through which parts pass ALWAYS spec'd for CLEARANCE ?

Simplified understanding; a hole for a 1/2 inch bolt to pass through is spec'd for max tolerance bolt in combination with min tolerance drill bit.
If not the line stops every time a few fat bolts come in, or the tool crib issues skinny drill bits.

I understood the article that Gene pointed to, more or less - but I think clearance tolerances in "common practice" should be adequate.
========================================
OTOH a 1/2 inch diameter circle approximated as a polygon SHOULD be INCLUDED in the polygon, not the polygon included in the circle.
Should "Fix Normals" (or something similar) resolve this ?

The issue is that the polygons in an STL file and the linear segments in GCode... Neither are curved, everything is a bunch of little straight lines. This is an overly exaggerated version of what's happening in case anybody is still scratching their head:

[img]http://etc.usf.edu/clipart/43400/43451/8c_43451_lg.gif[/img]

Maybe you can specify in your CAD software whether to land the critical dimension on the inside or outside of the polygon? Sounds like a plugin or feature that wouldn't be a bad idea. Maybe a library of holes listed by their resultant dimension and not their digital representation. I think it needs to be part of the part design phase (CAD) instead of something that has some algorithm search and correct for later in the slicing process or gcode manipulation.

[RegB]

I got that, what I don't get is why anyone would spec a 1/2 inch hole for a 1/2 inch part to fit into (or through).
In just about any assembly operation the goes_in_ta parts will sometimes be "fat" and where they go will sometimes be TIGHT, hence there are tolerances on clearances - unless an "interference fit" is the spec, in which case Ya just hammer it home.

SOMEWHERE there are tables for this, I suspect that the "polygonizing" error is less than regular manufacturing clearance tolerances - though I really SHOULD look it up before espousing (-:

I think I'll just spec octagons when I need approximately round holes, the filament drag will likely round off the "points" anyway.
A good enuf compromise.

[DocHogan]

I got that, what I don't get is why anyone would spec a 1/2 inch hole for a 1/2 inch part to fit into (or through).
In just about any assembly operation the goes_in_ta parts will sometimes be "fat" and where they go will sometimes be TIGHT, hence there are tolerances on clearances - unless an "interference fit" is the spec, in which case Ya just hammer it home.

SOMEWHERE there are tables for this, I suspect that the "polygonizing" error is less than regular manufacturing clearance tolerances - though I really SHOULD look it up before espousing (-:

I think I'll just spec octagons when I need approximately round holes, the filament drag will likely round off the "points" anyway.
A good enuf compromise.

Get a physical copy of (*highly recommended*), or google references to, The Machinist's Handbook; it has fantastic references for clearances and tolerances for press fit, interference fit, etc, as well as *almost* everything you would ever need to know for any mechanical makering you'd ever want to do.

[RegB]

http://www.engineershandbook.com/Tables/taphole.htm" onclick="window.open(this.href);return false;
Hole sizes for "close fit" and "free fit" - ignores tolerance on drill size.

[RegB]

Repetier.h

I have been reading this,
Around the line with some typos;
"startet to correct the circle interpolation"
It seems like the scene is set here for the poly approximations ?

[Jimustanguitar]

I got that, what I don't get is why anyone would spec a 1/2 inch hole for a 1/2 inch part to fit into (or through).

Software can't fix stupid

BTW, I knew you understood it, it just took me a second so I thought others might appreciate the visual.

[RegB]

I am having some trouble understanding some of the OTHER things affecting accuracy of these delta printers.
I probably try to oversimplify, but at this stage of my understanding a few things in the EEPROM don't seems to jive.

X max length and Y max length are both set to 250 mm - but printable radius is 135 mm.
In Slic3r we set the bed size to be X:280 mm and Y:280 mm
If these are consistent, or if their inconsistency doesn't affect accuracy I'll move on, though my curiosity will continue to itch.

I am still having trouble understanding why my PEEK fan shroud came out with such a narrow opening for the fan.
I have also printed out the nickel test gauge and a 5c piece will not fit into the circle or through either the X or the Y slot. Although the slots are PLENTY wide enough, they are way too short.

[Harblar]

Great info guys! I'm definitely going to be referencing a lot of this in the next few months.

The hole center points will be accurate. The only potential error is with the actual hole itself. I've made a number of little enclosures for myself and the hole centers are never off.

g.

I went back and measured the first knob I printed before scaling it up. The overall diameter was 44.3mm give or take a hundreth. The designed diameter was 44.44mm, so about a .15 mm error overall. Not bad at all in the grand scheme of things and when compared to mass produced parts. Is this about the max tolerance for the V2? I'm thinking I could probably get it into the .05 -.1mm range with a smaller nozzle and a few other setup tweaks (slower speeds, heated chamber, different slicer settings.) Is that feasible or am I really just grasping for a level of perfection that isn't needed? Afterall, I'm already way more precise with the rostock than I could do any of this by hand.

[Polygonhell]

X max length and Y max length are both set to 250 mm - but printable radius is 135 mm.
In Slic3r we set the bed size to be X:280 mm and Y:280 mm
If these are consistent, or if their inconsistency doesn't affect accuracy I'll move on, though my curiosity will continue to itch.

I am still having trouble understanding why my PEEK fan shroud came out with such a narrow opening for the fan.
I have also printed out the nickel test gauge and a 5c piece will not fit into the circle or through either the X or the Y slot. Although the slots are PLENTY wide enough, they are way too short.

The only things in the EEPROM that really have any effect on position, are those setting directly attributed to the steppers that result in steps/mm being computed and the delta geometry settings, delta radius, arm length and the tower positions.

How short?
I would I not make the assumption that someone on thingiverse understood that a parts need clearances, the vast majority of printer users have never manufactured or had manufactured anything before. You are much better off calibrating against simple geometric primitives, squares, cylinders etc.

The Max EEPROM settings should be really close minus the XYZ heights, because of the nature of a delta platform, if you have all the steppers set the same on the towers, any asymmetry you see in error by dimension has to be the result of the mechanical assembly.

I have 1/2 a dozen printers here, both Cartesian and delta they all print slightly undersized a 20mm cube is usually in the 19.9 range give or take. There are several reasons holes shrink one was covered earlier in this thread, the second is predominantly because as you print the filament it's stretched and as a result has a tendency to cut interior corners.

[RegB]

This is the test piece;
http://www.thingiverse.com/thing:11261" onclick="window.open(this.href);return false;

The slots come out on my machine as;
X:19.99 Y:20.00 vs a nickel that has edges in fair shape that is 21.23 across.
The circular (polygon) hole is pretty much on 20.0 in any direction, it is hard to know when you are EXACTLY on the diameter.

Point is X & Y are about equally short and it is a little over 5%.

Both slots are 3.16 or 3.15 wide, not easy to get with these verniers.
The physical nickel is right around 1.8, so the slot width is WAY off.
I'll put this down the the model just being "wrong".

[Harblar]

Nice find on the nickel test, RegB. I'll have to give that a try when I get a sec.

Did a couple other prints over the weekend. One was for replacing a piece on a model kit (that I had overdrilled and cracked a LONG time ago). I designed it in the CAD program with an oversized hole in the center. The hole needed to be fairly tight to fit a round a shaft with no gaps (mainly for the aesthetics of the model where a straight edged hole would look wierd). Even with the hole scaled up by 1.062 it was a really tight fit. Tighter than the oven knobs and that got me thinking... The model piece I printed at .1mm and the oven knob at .2mm. With the standard .5mm nozzle, I'm probably overextruding and introducing more error into these holes than I would be at a .2mm height. I'm going to try it again at .2mm just to verify, but my guess is, if I want to do .1mm prints I need to invest in a smaller nozzle (.35mm) too further minimize this error. Maybe do a new hotend while I'm at it!

I was already planning on upgrading to an E3D V6 hot end and then I started thinking, maybe I should get two so I could print with dissolvable supports (like PVA). And that started me thinking, How would that effect hole size? If I were to fill the holes with the PVA material while printing, would this reduce the hole shrink and other factors when the part was cooled and cleaned? (is it possible to program the slicer to fill all round holes with support material in the first place?)

Just something I started thinking about yesterday while I was at work...

[RegB]

Nice find on the nickel test, RegB. I'll have to give that a try when I get a sec.

Did a couple other prints over the weekend. One was for replacing a piece on a model kit (that I had overdrilled and cracked a LONG time ago). I designed it in the CAD program with an oversized hole in the center. The hole needed to be fairly tight to fit a round a shaft with no gaps (mainly for the aesthetics of the model where a straight edged hole would look wierd). Even with the hole scaled up by 1.062 it was a really tight fit. Tighter than the oven knobs and that got me thinking... The model piece I printed at .1mm and the oven knob at .2mm. With the standard .5mm nozzle, I'm probably overextruding and introducing more error into these holes than I would be at a .2mm height. I'm going to try it again at .2mm just to verify, but my guess is, if I want to do .1mm prints I need to invest in a smaller nozzle (.35mm) too further minimize this error. Maybe do a new hotend while I'm at it!

I was already planning on upgrading to an E3D V6 hot end and then I started thinking, maybe I should get two so I could print with dissolvable supports (like PVA). And that started me thinking, How would that effect hole size? If I were to fill the holes with the PVA material while printing, would this reduce the hole shrink and other factors when the part was cooled and cleaned? (is it possible to program the slicer to fill all round holes with support material in the first place?)

Just something I started thinking about yesterday while I was at work...

From my newbie understanding I would guess that;
"YES, filling holes with support material, would PROBABLY solve a lot of issues".

But, or "Ahh, Butt";
Unless you have managed to build a dual extruder machine (I haven't) that can push out (soluble ?) support material into those holes first and THEN make the rest of the part - then the "How Ya gonna DO that ?" question is begged (-:

As to whether or not a slicer such as Cura could be fooled into believing that there is a ghost extruder to do that... ?? I don't know, but it may be worth trying to tell it that the holes are already "filled"... err, SOMEHOW.

[RegB]

I changed my mind;
I think support material is only AUTOMATICALLY built into holes that create overhang as the material around them is built. It is to support the tendency of unset plastic to droop while still hot/warm.

So;
"maybe" for holes in a vertical plane
"probably not" for holes in a horizontal plane
"it may depend" for holes on an inclined axis.

Ya can't COUNT on it.

[Captain Starfish]

I fix this with a combination of two things.

First, wherever possible, I treat the printed hole (?) as I'd treat a hole in a casting - good start, but it needs drilling out. The centre's fine so as long as the wall thickness is sufficient and I can reach the hole, I can ream it out to spec with either a ream or even a twist drill.

Second, if I can't get to the hole with a drill bit, if the walls are too thin, if I don't have a bit of the right size, I'm lazy, whatever, I can't do the first approach: print a drill guide which shows the printed diameter of a bunch of smaller sizes and allows you to get the vernier in there. I then copy the part I'm working on in the CAD app, throw the new copy in a "print fudge" layer, and expand the bores in the model appropriately so they print to spec.

[RegB]

I think I'll just make them a bit bigger, for clearance.
If it becomes a problem (unlikely) I'll try polygons, octagons look like a fair compromise for the likely size range I'll use.

[Generic Default]

The best way to get around small hole problems is to use a tiny nozzle and go slower. The viscosity and molecular length of each plastic determines how it will print; ABS and PLA cling to the hot extruded stream of plastic less than nylon, and fiber reinforced plastics cling the most. It's kind of hard to describe with words exactly how the plastic behaves when the nozzle goes around a small hole in the print path, but just watch it really closely and you'll see what I mean.


-Small nozzle
-Slow speed
-Smaller layer heights

These all make your part tolerances more accurate. I just switched to a PTFE magnetic arm system (check the thread in the Mods and Upgrades section) and I can now print at 0.03 layers. The smoothness of magnetic arms mostly eliminates tiny height variations caused by backlash, and those little backlash gaps really made small holes print worse. I'm working on a 0.1mm nozzle for even more detailed prints.
And for dual extruders, google 'Tri hotend'. I might be slightly biased since I'm the one making it.


I think the people who make slicing engines need an option to increase the size of small holes. Most of the designs I make need small, accurate holes, often with threads printed into them for assembly. A 0.5mm nozzle just isn't good enough; 0.35 is the largest nozzle I use for anything that needs to look good or fit precisely.

As for overall dimensions, try adjusting your arm length and radius in the firmware to make your prints slightly bigger. Outside dimensions of large parts should be a good indicator of your size calibration. If they're too small, try decreasing your arm length in the firmware to make your printer think it has to move further than it actually does.

[bdjohns1]

One thing I've started doing in part design is that when I'm going to join two plastic parts together, I steal from the polyhole approach, and I use square pegs / square holes for parts. I can get a snug press fit with holes only 0.1mm larger.

[Captain Starfish]

I think the people who make slicing engines need an option to increase the size of small holes.

Unfortunately, with STL you can't easily determine that there's a hole in the model.

[Polygonhell]

I think the people who make slicing engines need an option to increase the size of small holes.

Unfortunately, with STL you can't easily determine that there's a hole in the model.

Actually that's not very hard, I think Slic3r did try this at one point around 0.7 or 0.8, they took it back out again, the problem was many of them existing models have holes sized such that they work with the intended hardware, and when you start messing with that in the slicer you get unpredictable results.
Not to mention the ammount of shrinkage changes with print speed, because printing faster tends to cause the filament to cut corners more. The same is true for lower layer heights, Under vs Over extruding etc etc.
In the end I'm not sure you can compensate in the Slicer without a lot of tuning for a particular machine.

What most models with tighter mechanical tolerances like the prusa/mendel 90 etc. do is have test pieces that you are supposed to tune your slicer settings on.