Calibrate your 3D printer to print parts to fit

by whitemousegary, published

Calibrate your 3D printer to print parts to fit by whitemousegary Feb 26, 2013


I have my Thing-O-Matic for a year now. From time to time, I tried to calibrate it to print parts to fit. Finally, I got it done (See youtube.com/watch?v=lQbvfiZAm-c). It turns out I just had the edge width and the scaling wrong. If you also have the problem to print parts to fit, read this article. It might work for you too. Here are the symptoms of the problem I'm trying to fix.
1) A printed plug is impossible to fit into its printed hole.
2) Small parts (around 1 cm size) are too big, while larger parts (around 10 cm size) are too small.

Recent Comments

view all
I would like to make you aware of a discussion started in another Thing, meant to do calibrate printers too: thingiverse.com/thing:342198/#comments
I am convinced that calibrating extrusion by measuring single-walls is misleading because some softwares (Slic3r) assume the threads are rectangular, but that happens only when they are surrounded by other threads and not in single-walls: the thread of single walls is rounded on the sides. If you calibrate your way, you may incur in a suboptimal thread bonding (on the sides) and weaker prints than they could be. It seems you can compensate in other ways with skeinforge, but it is still good to know.
For some reason i can not delete or edit this comment so please ignore this one. The one below has more information
I am getting a strange space in my prints.http://imgur.com/V1P3lZWWhat might be causing this? I tried messing with

More from 3D Printing

view more


Liked By

view all

Give a Shout Out

If you print this Thing and display it in public proudly give attribution by printing and displaying this tag. Print Thing Tag


Okay okay, let's see what our 3D printers should be offering first (if configured correctly). See youtube.com/watch?v=lQbvfiZAm-c . The video shows two parts, an S-Shape hole and an S-Shape plug thingiverse.com/download:141737 . The plug and the hole are edge to edge touching each other in the STL file. As show in the video, I can actually plug them together just by pushing real hard with bare hands, no tolerance compensations in the design, no drilling or filing, they just fit. The S-Shape is chosen deliberately as it shares a lot of common features with interlocking designs. Designing is fun with interlocking designs. ( A more fancy result, same premise, youtube.com/watch?v=O3dJsjv-8vA )

This article is intended for Skeinforge (SF) calibration, and I did my calibration with ABS plastics, Thing-O-Matic (0.4mm nozzle, firmware 4.1) and ReplicatorG+Skeinforge 50.

**Before the calibration**
Temperature will affect the actual amount of the extruded plastics and, therefore, the realized line width. Though it is not necessary, I strongly recommend you to figure out the strongest bonding temperature initially, see thingiverse.com/thing:35088 . You really need every bit of available bonding strength for printing large ABS objects.

If you are using ReplicatorG, please disable its user interface *"Use Print-O-Matic"*. This *named-funny-UI* is just to override some Skeinforge variables without your notice. It is easier to calibrate your 3D printer to fit without this *named-funny-UI* . You can do it by unchecking the checkbox *"Use Print-O-Matic (stepper extruder only)"* in the *"Generate Gode"* window (See thingiverse-production.s3.amazonaws.com/assets/95/f2/da/5d/af/disable_replicator_setting.png ). If you insist to use the user interface *"Use Print-O-Matic"*, you won't be able to adjust *the infill width* such that you would end up with a fragile printout (See *Section Calibrate infill*).

Here is the list of necessary variables for this calibration to work,
- *Carve/Extra Decimal Places (float)* : change to 5
- *Carve/Edge Width over Height (ratio)* : nozzle diameter/layer height
- *Inset/Infill Width over Thickness (ratio)* : nozzle diameter/layer height (initially)
- *Dimension/Filament Packing Density (ratio)*: needs calibration
(equivalent to the reciprocal of *the extrusion multiplier* in Slic3r)
- *Scale/XY Plane Scale (ratio)* : needs calibration

Before the calibration, you need to increase *Carve/Extra Decimal Places (float)* to 5. This variable is to control the significant digits of the values in a gcode file. There is no point to reduce precision at this point. Precision is king if you want to print parts to fit.

**Calibrate line width**
If you are having problems to print parts to fit, your 3D printer is probably drawing lines wider than Skeinforge expecting. In my case, SF is expecting 0.4mm, while my 3D printer is drawing 0.64mm (See thingiverse-production.s3.amazonaws.com/assets/b9/49/7b/db/d0/IMG_20130225_124259a.jpg ). The optimal line width that your 3D printer can do is the same as its nozzle diameter, so you want to enforce that to optimize its ability to print small features. To calibrate the line width to optimal,
- 1a) set both *"Carve/Edge Width over Height (ratio)"* and *"Inset/Infill Width over Thickness (ratio)"* to *"nozzle diameter/layer height"*.
- 1b) set both *"Speed/Feed Rate Setting (float)"* and *"Speed/Flow Rate Setting (float)"* to the same value.
- 2) print the thin wall model thingiverse.com/download:259710 , and measure the wall thickness with a caliper.
- 3) adjust *"Dimension/Filament Packing Density (ratio)"*.
- 4) repeat Step 2 and Step 3 until *the measured wall thickness* meet *the nozzle diameter*. As a starting point, the *new value* can be estimated by *(measured width)/(nozzle diameter) x (old value)* .

The goal of this section is to fine tune the volume of the extruded plastics to match the expected line width. Therefore, *Filament Packing Density (ratio)* is not the only option. There are a brunch of variables which can alter the realized line width, e.g. e-step per mm , changing the flow rate relatively to the feed rate, etc... . Just pick one and stick to it. It probably can work just fine.

Just to remind you: when the amount of extruded plastics is too little, it will not make the wall thickness smaller than the nozzle diameter. Instead, you will have some spongy like walls with wall thickness roughly the same as the nozzle diameter (See thingiverse-production.s3.amazonaws.com/assets/3a/04/82/c9/c7/wall_calibrate.png ). If it is simply impossible for you to calibrate *the wall thickness* to *the nozzle diameter*, you may substitute *the nozzle diameter* by a bigger value (say 0.1mm bigger) and try again.

**Calibrate scale**
Once you have the line width right, you can go on to calibrate the scale, here is the procedure:
- 1) Print the 20mm test cube thingiverse.com/download:139958
- 2) Measure the size of the cube with a caliper (CAUTION: measurements must be done after the cube fully cool down)
- 3) Divide *20mm* by *the measured size*, and set the value to *Scale/XY Plane Scale (ratio)*.

*The measured size* should become approximately 0.5% to 1% smaller than *20mm* after the line width calibration, thingiverse-production.s3.amazonaws.com/assets/f3/c5/51/32/08/scale_calibrate.png . This difference is probably caused by the plastic shrinkage, which can be fixed by adjusting the xy scale. After the scaling calibrated, the ability of your 3D printer to print parts to fit should have improved dramatically. You can print the test plug thingiverse.com/download:139973 or the S-Shape plug thingiverse.com/download:141737 to verify. ( Note that, the S character in the S-Shape plug is not symetric. Fitting the S character upside down won't fit )

Please be reminded that, even after the line width and the scaling calibrated correctly, smaller holes (diameter smaller than 3mm) will still be too small due to the arc issue reprap.org/wiki/ArcCompensation . The Skeinforge Stretch plugin can handle the arc issue, and gets the smaller holes to fit. The Stretch plugin worths a shot, just enable it to try. The default setting is a little bit conservative. You might need to adjust *"Stretch/Perimeter Inside Stretch Over Perimeter Width (ratio)"*. The default value is 0.32, and I need to increase it to 0.72 to get a perfect 2mm diameter hole.

**Calibrate infill**
If you did follow my instructions and reached here, you might have noticed something undesirable. The 20mm cube (15% infill) becomes very fragile (appliable to any *non 100% infill objects*). We can fix it by decreasing *Inset/Infill Width over Thickness (ratio)*. The new value ...
- *Inset/Infill Width over Thickness (ratio)* : *(nozzle diameter - overlap)/layer height*
In particular, I need 0.01mm overlap to enforce the infill lines to fuse with their siblings for 15% infill (of course, 0mm overlap for 100% infill). Note that, the *ReplicatorG UI "Use Print-O-Matic"* won't allow you to make this change; please, give up that UI.

The reason to the fragile printout is that: when you calibrate your line width, the surface beneath is rock solid; while the surface beneath infill is sparsely filled. Without a solid surface to support, the infill lines become narrower. Filling up areas with lines too narrow, we have the infill lines loosely bonded to their siblings, i.e. fragile.

You might be confused by the suggested change; indeed, you should be confused. The naming of *edge width* and *infill width* are awkwardly misleading. Both named after *width*, but behave in opposite manners. When you increase *edge width*, SF will extrude more plastic to realize the increased width, and spaces the lines accordingly. On the contrary, when you increase *infill width*, SF WILL NOT change the extrusion rate, but it will still do the spacing with the increased line width... (whatever...). Let me translate it for you. *Edge width* means the width of lines literally, but *infill width* means *line spacing* instead. So, if the infill lines are too far from their siblings, we decrease infill width (i.e. *line spacing*).

In case you really need some extra strength, you can consider using wider lines for printing. You can do it by simply setting (no need to calibrate line width again)
- *Carve/Edge Width over Height (ratio)* : *your desired width / layer height*
- *Inset/Infill Width over Thickness (ratio)* : *(your desired width - overlap) / layer height*
Don't worry, using wider lines is perfectly fine if Skeinforge knew it.

**Something counter intuitive about the calibration**
There is one thing worth noting that my 20mm cube before the line width calibration is actually closer to 20mm in size (See thingiverse-production.s3.amazonaws.com/assets/f3/c5/51/32/08/scale_calibrate.png ). This is a little bit counter intuitive. What really happening is that ABS plastic will shrink (a lot) after they cool down. So, if the dimension is right before any scaling, the size of a printed object should always be smaller than we expecting (See the second graph). We can also see why it is so using the first graph. In the ideal scenario, the mapping between the ideal size and the expecting size is a line with a 1:1 slope. *The wider than expecting realized lines* shifts the mapping upward, and shrinkage makes the slope less steep, so we have *a small range of good mapping* near the intersection (See the first graph). ^^"... *a small range of good mapping* is probably the most devastating form of misleading clues ever possible.

**Two popular wrong interpretations**
By the way, I would like to clarify two popular wrong interpretations of the problem. When I look for reasons for *the symptoms above* in the internet (See Thing_Info/Description), I keep seeing people saying it is caused by "plastics shrinkage" or *"the arc issue"* reprap.org/wiki/ArcCompensation . For "plastics shrinkage": its true that plastics will shrink after they cool down, so a printed hole (e.g. thingiverse.com/download:139973) will become smaller. However, its printed plug will also shrink by the same amount. Therefore, a printed plug should fit into its printed hole regardless of shrinkage. For the arc issue reprap.org/wiki/ArcCompensation : it will only affect smaller holes, not the larger one. The author is too conservative when he deduces the implications of the formula in the page. ABS plastics can tolerate a tiny bit of deformation. From my experience, if you try to fit a metal rod into a smaller ABS hole, as long as the difference in diameter is within 0.05mm, a fit will still be feasible (the tightness will vary though). To make it easier to read, I re-parsed the table (see thingiverse.com/download:139948 ). It shows that a 10mm diameter hole will be 0.008mm smaller than it should be, which is not enough to cause a tolerance problem. To a pair of hole and plug, the arc issue will only become a problem unless the diameter is smaller than 3mm.

This is it. The information mentioned here is probably mentioned somewhere else already. I just meant to put them in an organized manner as a note for myself. If I missed a citation, please show me. I'll put it back.

Yours faithfully,

PS: 2013-05-01 - My 3D printer is delivering printouts reliably with precision and strength for quite a while now. It's time for me to say goodbye to test shapes. For the time being, I kept a small portion of the test shapes I printed for some good reasons. This is the last picture of them before I throw them all away, thingiverse-production.s3.amazonaws.com/assets/be/9a/23/9c/e8/IMG_20130501_070417.jpg . Cheers, fingers crossed...

**Troublesome shooting**
Why the wall thickness didn't change after modifing the *Dimension/Filament Packing Density (ratio)*?
>* Did you press "Save all" after editing profile?
* Did you re-generate G Code file after editing profile?
* Are you using the profile you edited in the "Generate GCode" window?
* Are you using Skeinforge 50 and the latest firmware?

Do I need ReplicatorG for this calibration to work?
> No, you don't need ReplicatorG. This article is intended for Skeinforge 50, not ReplicatorG. You just need Skeinforge for this calibration to work. ReplicatorG is not necessary at all.

There are two thin wall models. Which one should I use?
> * Use the one matches your expecting line width. (e.g. 0.4mm line width vs. 0.4mm thin wall; 0.5mm line width vs. 0.5mm thin wall)
* Both should should be just fine initially. However, when you calibrate the infill width to a smaller value, SF might think the thin wall model should be filled (which can mess up your measurement). Using a matched thin wall model can stop SF from filling the wall.

Will this method work with older versions of Skeinforge?
> Cautiously, yes. For the older versions of Skeinforge, the ways to adjust the realized line width is limited to "changing the feed rate relatively to the flow rate" or "messing directly with *e steps per mm*". ... (there is no point to stick to the older versions anyway. why bother?)
I would like to make you aware of a discussion started in another Thing, meant to do calibrate printers too: thingiverse.com/thing:342198/#comments
I am convinced that calibrating extrusion by measuring single-walls is misleading because some softwares (Slic3r) assume the threads are rectangular, but that happens only when they are surrounded by other threads and not in single-walls: the thread of single walls is rounded on the sides. If you calibrate your way, you may incur in a suboptimal thread bonding (on the sides) and weaker prints than they could be. It seems you can compensate in other ways with skeinforge, but it is still good to know.
I am getting a strange space in my prints.http://imgur.com/V1P3lZWWhat might be causing this? I tried messing with
For some reason i can not delete or edit this comment so please ignore this one. The one below has more information
I am getting a strange space in my prints.http://imgur.com/V1P3lZWWhat might be causing this? I tried messing with Infill Width over Thickness but that does not seem to help.
Thanks so much for this guide. It increases my satisfaction with my thing-o-matic by at least 50%. I come back here every once in a while to re-calibrate.
Thanks a lot. By the way, please tell me something about my writing. Is it good enough to you? ( Part of the reason for me to write is to improve my writing. )
The writing was good. I found it very easy to follow.

I have noticed that on my test cubes the sides then to bow inwards a bit, and the corners stick out a little bit. So if I measure the corners the size might be 0.2 mm larger than if I can avoid touching the corners with my calipers. I know this wasn't covered, but maybe you had an idea how to fix this?
I have noticed that since the very first time I tried to do calibration. The simplest way to get around this problem for the measurement is to measure the central part of the cube only.

However , the real solution is to add a fan for printing. With a fan for printing, the problem you described will be no longer visible. Your caliper might still able to pick it up though.
You have to take into account the thickness of the bead layed down by the printer eg nozzle size would calculate how big the bead of plastic layed down is or.. google it :)
Arr... please read thoroughly, it just takes you ten minutes to read... I have made myself clear enough in the article concerning the issue you said. If you think I didn't address your problem, please be more specific.
Hello, i am uploading to simplify3d the test 4mm, but somehow it shows just one wall , this is very strange...
thanks you in advance
hmm... ive found that is an issue with 4mm object that is in the limit of the nozzle, i scaled the object by 1.0001 and now it shows , is this ok to calibrate like this ?
Update it dose not fit to the Wanhao printer, i guess neither to the flashforge
Shame on Wanhao. Manufacturing an open source product for sell is fine. Taking the open source product and claiming its invention?! WTF!!! Shame on Wanhao!!! ( Note: The Wanhao printer is just a direct copy of Makerbot Replicator.)
Hi there, i am a new user and i am trying to do my best to calibrate my printer (wanhao 4x),
I also use simplify3d, can this method be used also with simplify3D ?
"wanhao 4x" looks like a Makerbot Replicator (^^" I should say it is.). This method should works fine on most extrusion based 3D printer, so Replicator is okay. "simplify3D" looks like a forked Slic3r. If it indeed is originated from Slic3r, look for a parameter called "extrusion multiplier". However, to Slic3r, there is no obvious way to adjust "infill width" like Skeinforge does, so you can either have accurate but fragile parts or inaccurate but strong parts.
yes there is the "extrusion multiplier"
Anyone ever had their line width vary by as much as .1mm on each side? I'm trying to get a .4 line width but one side will have .45 and the parallel side will have .55mm
Mmm... I guess you have a terribly extruder. Say... a long long metal tube connecting to the hot end with no fan, a driver gear with its hole far far from center, a printed extruder with long thin moving parts, ... (^^ I can go on for quite a while)
One more piece of information would be useful for beginners--where does one find these parameters? It's easy to find the Generate GCode ones, but the others must be hiding in some file somewhere...
^^", Skeinforge is full of parameters. If the user is using Skeinforge directly, he must have overwhelmed by parameters already. In case of ReplicatorG, user can locate a profile by MENU/GCode/Edit Slicing profiles... .
Thank you my friend! This worked flawlessly and am now printing extremely functioning parts.
This is the most comprehensive calibration procedure I've ever seen. I'm not even done with the full calibration and I can already see a significant difference in print quality. Thanks!
Glad to hear that. Please tell me how's your machine doing after you did the full calibration. It would be excellent for me if you can report flaws in the article.
Just finished 4 Wall-Blocks on my RepRapPro.
In Slic3r i adjusted the extrusion multiplier to change the flow ratio.

0,5mm Nozzle
Test1: Flow Ratio 1 Wall 0,79mm with standart settings (callibrated the Steps/mm of extrusion material before)
Test2: Flow Ratio 0,9 Wall 0,7mm
Test3: Flow Ratio 0,8 Wall 0,63mm
Test4: Flow Ratio 0,6 Wall 0,48mm

I'll have to check tomorrow in the sunlight how smoothe the surfaces came out. the biggest difference yet is the teh stability when squeezing
is that file just for ABS? i am using pla and i print the wall veryvery thick, it is around 1.2-1.3mm thick, my nozzle is 0.4 mm, i dont know what's wrong,

and it seems that no matter what speed i input, theedge width is still that thick
it's not feed rate alone causing the different. The ratio between feed and flow rate does. The Print-O-Matic UI explicitly makes the ratio to be 1. Therefore, with that UI, you cannot change the ratio, which is why all speed s give you the same line width.
I did what you suggested, and things are much better now, but both my pegs and holes are a tad small when measured. However, the larger dimensions are perfect. Also, if I have 2 perimeters, there tends to be a slight gap and poor adhesion between them. Thoughts?
if you can get S Plug to fit, you almost get the most out of you printer already. Scaling is an issue. The brutal truth is we need a way more intensive geometry analysis to enforce extremely demanding precision requirement. There is no way a simple implementation like skeinforge can fulfill it. Switch to PLA will ease a bit the problem.

For the gaps, you may try to decrease infill.

I have the same problem with Slic3r. I'm going to try to follow the idea of this, but does anyone know what exactly to change for the same effect in Slic3r?
slic3r will decide line width on its own upon different circumstances, with good intentions. However, it also make it extremely difficult to have a simple plan to do calibration. I think the best you can do is to follow the calibration process suggested by slic3r. If it doen't work, switch to skeinforge.
After much wailing and gnashing of teeth, I have switched to Skeinforge. Your guide is thus far very helpful for calibration.
i got replicator 2x, i use makerware makerbot slicer and i have the same problem as jad91. for the 20 mm box i must scale x:1.005 y:1.015 z:1.01 to get the good dimension +/- 0.03. i try to edit a new setup in skeinforge slicer in makerware but if i activate chamber to heat my bed, the print fail but if is not activated, it print but don't stick. aaaaaaaaaaaaah!
I'm sorry. I don't have replicator2X, and therefore, no idea about the usage of makerware. If you bought a replicator2X, you should request helps from customer services. You paid A LOT for that.
yes i know...but i can use another slicer too. i look for another printer specialy for nylon. you have suggestion?
Success! My TOM is now making parts that can interface with real world objects. Thank you I can now print useful things instead of just trinkets to clutter my desk.
where you got this green pad from? thx
Try Wallm...
It's worked well for me. I am having some slight trouble though; when I print the wall model, there's a difference in the thickness between the x and the y directions. In reality, the discrepancy is so minute that it shouldn't matter overall for the prints I do. It is just a curious situation, and my OCD is nagging.

Both walls that run from +X to -X (right/left of printer) are perfectly 0.40mm, while both walls in the +Y to -Y (forward/backward) are at 0.43mm. It prints like this consistently, and I've measured it like that consistently.
Since it was so small, of a difference, I tried moving onto the next model, 20mm cube.

The +X to -X directions are 20.10mm and the +Y to -Y directions are 19.90mm.

Like I said, in the overall scheme of things, it probably doesn't matter much, but it'd be awesome to get it to work perfectly.

Any thoughts on what could be causing this before I move on to try to calibrate skeinforge for holes?

Printer: Replicator
Software: ReplicatorG
Slicer: Skeinforge 50
Nozzle Diameter: 0.40mm
Layer height: 0.10mm
xy dimensions are designed to have the same gear ratio, so the same step per mm theoretically. However, nothing is perfect in real world. For whatever reason they didn't match, you can always adjust one of the step per mm to enforce them to match. If you are using ReplicatorG, there is a folder, called machines) under ReplicatoraG folder storing all machine profiles. You can do your adjustment there. You can try increasing "y step per mm" to 1.01 times the original (i.e. 20.1/19.9). It should do fix your problem.
Please leave me some comments. I really want to know how's the other 3D printers doing.