Hole diameter smaller and holes are not round. (Replicator 2)

[Björn Syse]

Hi,

I'm trying to create taps that fit into holes. When I create a circular hole with a diameter of 4.7mm for example, it doesn't come out like that.

It comes out somewhat elliptic and has a varying "diameter", of which the longest part does not even reach 4.7 mm.

I made a few tests with varying hole diameters, and this is what I got:

The top row shows the diameter in the CAD file and the bottom is what I measured from the print. 

Is anyone experiencing the same thing? 

[Eighty]

You could have two issues - one mechanical, and the other software-related.

 

If you're getting an elliptical shape, then you may be experiencing some issues on one axis.  Pictures of the printed object would help.

 

The holes not being properly sized is more of a slicing/calibration issue.  Have you printed a calibration cube to see if you're putting down the proper amount of plastic?

[Björn Syse]

Hi

I see. Here's a picture attached, It is not super obvious but.. 

No, I haven't printed a calibration cube, how is that done? please point me in the right direction.

[Eighty]

From makerware, go to File->Examples->20mm_calibration_box.stl.  Slice it with whatever settings you want (but use 100% infill), and print.  Then measure your box with calipers and see what dimensions you have.  Too big = too much plastic.

Try that first, and see what you get.

[Björn Syse]

Hi again

Actually, I did print a few of those but It never occured to me what I could do with the information. Here are my measurements:

X width: 20.03 - 20.30 depending on where I measure. That is, if I measure only the top half or if the bottom half is included. (it seems the first layer is a bit wider, and then the whole box tapers slightly, but as I understood that's part of the limits of this manufacturing technique?)

Y width: 19.84 - 20.14 with the same reservations of where I measure. 

Z width (center): 9.80 - 10.10 (the bottom is very curled. The corners have lifted from the build plate like a pillow)

What can I do with this information?

[Eighty]

Well, those results sound ok.  This indicates that you don't have any major calibration issues with the amount of plastic being extruded.  So you can rule out some troubleshooting steps.

 

The curling means you're not getting good adhesion.  You might try cleaning your build surface.  Are you printing on blue tape, or directly on the acrylic?  Slightly lower temperatures might help reduce the curling.  I usually use 220 degrees.

 

 So now we look at other possibilities.  First, your holes are pretty small (~4.5mm diameter).  Have you viewed your Gcode output in a viewer (such as Repetier-Host) to see if it was sliced this way?  Sometimes the STL file adds facets that directly impact your product.  Other times, the slicer itself adds weird stuff that end up in your print.  For example, the attached picture shows an odd "dent" in an item I printed.  I thought it was a printer issue until I checked the gcode visually.  Turns out, Skeinforge sliced it like this.

 

That would be my advice next.  Find a way to view your gcode, to see if it was sliced like this.  I like Repetier-Host personally, but there are other viewers as well.

 

Another option would be to scale up the print a bit, and reslice/reprint.  Does it still look like this?

[Joseph Chiu]

Hi Bjorn,

The elliptical shape of the hole may be indicative of a printer that has too much backlash -- tightening up the belts can help with that.  (I don't have an R2, so I don't know the steps.) 

As for the overall size, if you are getting within a nozzle-width of the target size, that's you've achieved the easily achievable tolerance, and you're now in the realm of needing to tune thing more carefully.

You might want to read about the stretch plugin for SkeinForge and consider using it.  Unlike standard mills, our printers do not have tool-diameter compensations for arcs.   Moreover, you have series of line segments approximating the curve, and that leads to a smaller "circle" as well, depending on the number of segments used to model the circle.

And then, you'll have to fight the shrinking of the plastic.  Printing slower will help here.  With PLA, this might not matter as much, but it still does happen in my experience.

The way most people seem to approach this is to make series of test prints and fudge the values.

Good luck.

[Dan Newman]

And this is a worthwhile read,

http://hydraraptor.blogspot.com/2011/02/polyholes.html

Dan

> --
>
>
> <Screenshot 2013-01-08 01.jpg>

[Eric Cheung]

I did some test exactly like this, have stored up info here: https://docs.google.com/spreadsheet/ccc?key=0Au2joJ_XuYbOdHhkOXdRaXVMdG01MTZCU1Y5NjY4VkE

In general, the dimension are all over the place. Exterior dimensions are probably slightly better than interior (holes/slots). Error along different axis are different. Errors are not stable.

And of course, the small holes (4.7mm) you were making, makes the elliptical-ness even worse.

By the spec, the nozzle diameter is 0.4mm (rep2). Then, your dimension error would be 0.2mm+ on each edge, with oozing. If you have a hole, your diameter error could be as much as 0.5mm (I had printed a 4.4mm hole with 5mm diameter in CAD).

[James McCracken]

That was going to be my suggestion.

Even if you get your printer tuned 100% perfect, you still need to follow the advice at that link to get small holes printed properly.  On my cupcake CNC, before I upgraded to 3G 5D, it was the only way to get well sized small holes.

One way to think about it is; you're approximating a circle using a polygon.  The way most CAD software works, it's going to draw the largest polygon that is still inside the circle.  When you think about the physics of printing a hole, you actually want the smallest polygon that is completely outside the circle...

--

[whpthomas]

Shark FX has adaptive facets in the export, which means that when the curves are exported, the points are outside the circumference, and the lines cross over to the inside. If I print anything down to about 5mm generally it will insert and turn with a snug fit. Another trick I use is to change the speed module in replicatorg and slow the perimeter right down to 0.2 - 0.3. This really improves the accuracy of the outside skin when you want moving parts to mate properly. If you want a loose fit with little play subtract 0.5 or 5% off the diameter - whichever is smaller.

[Björn Syse]

The curling means you're not getting good adhesion.  You might try cleaning your build surface.  Are you printing on blue tape, or directly on the acrylic?  Slightly lower temperatures might help reduce the curling.  I usually use 220 degrees.

Yes, I'm on blue tape. But you're right, It's a bit worn, and I haven't bothered to change it because I don't want to run out of it :) And aswell, I'm ordering a glass plate any day now so that's also a reason I haven't changed it. I've wiped it off now anyhow, but it's good to know that the curling is an adhesion problem. Just to be clear since english is not my first language  when I say Curling, I mean the corners of the piece lift from the bottom plate.

 

 So now we look at other possibilities.  First, your holes are pretty small (~4.5mm diameter).  Have you viewed your Gcode output in a viewer (such as Repetier-Host) to see if it was sliced this way?  Sometimes the STL file adds facets that directly impact your product.  Other times, the slicer itself adds weird stuff that end up in your print.  For example, the attached picture shows an odd "dent" in an item I printed.  I thought it was a printer issue until I checked the gcode visually.  Turns out, Skeinforge sliced it like this.

Thanks for that tip, didn't even occur to me at first. I did have a look in both Skeinlayer and Repetier host now, and the holes are circular allright.  No ellipticalness going on. The also have pretty many facets (polygon with 31 points) since I meshed that STL quite fine. 

[Björn Syse]

Interesting, I'm trying the perimeter speed reduction right now. Will get back with results. 

If you want a loose fit with little play subtract 0.5 or 5% off the diameter - whichever is smaller.

Just to clarify, You mean if I want a piston to fit into a cylinder of 10mm diameter I would make the piston 9.5 mm diameter, and if the hole was 5mm I would make the piston 4.75 mm ?

Regards

- Björn

[Björn Syse]

So, I've made a test using the perimeter speed reduction, but the results are similar or even smaller. Very weird.

4.7 hole turns out to vary around 4.17-4.40

4.8 hole varies around 4.35-4.55

What also interesting, I tried a hexagonal hole with 4.6 mm between two opposing faces. These are the measurements from that hole on the print (attached)

[Björn Syse]

On Tuesday, January 8, 2013 7:41:45 PM UTC+1, MerlinJim wrote:

One way to think about it is; you're approximating a circle using a polygon.  The way most CAD software works, it's going to draw the largest polygon that is still inside the circle.  When you think about the physics of printing a hole, you actually want the smallest polygon that is completely outside the circle...

This makes total sense. But in this case I was using a polygon with 31 points, which should've come close to that circle then measures. Also, my hexagonal test down below doesn't print well either.

But thanks for the link dnewman and MerlinJim, that was a very comprehensive breakdown of the problem. I'll study it again and see if there is anything else I can respect.  I don't know if segment pausing could be a part of the issue. I'm on a Replicator 2 running from the SD with firmware 7.0. Could it be an issue? 

[Björn Syse]

So I tried another hexagonal test with both a piston and a hole. Both have a distance between opponing faces of 4.6 mm. 

The piston grows and the hole shrinks with about the same distance, 0.25 mm, which creates a worst possible fit :)

[pfinucan]

It looks like you are over extruding a little bit.  Try a test print and INCREASE your filament diameter setting by 0.1mm or 0.2mm.  That will cause you extruder to extrude a little LESS plastic, making each line/pass thinner.  Which in turn, will make you holes larger, and your piston smaller.

[Björn Syse]

Hi, I was curious and hopeful to try this. I had my filament set at 1.74, but now I tried 1.84 and printed the same hexagonal shapes. Things changed slightly, the pillar grew slightly smaller, but the hole did aswell. Very strange. 

The surface quality seems somewhat worse at one point, and better at other aswell. 

What's the best way to calibrate what filament diameter I should have set? Any test pieces that are suited for this?

[Shawn]

Sounds like you are getting too much (or too little) plastic per print.
I know of 3 ways to deal with this.

1) Adjust the filament diameter on the slicing dialog until you find the
magic number. This is the easiest solution, but the most work too in
the long run - that magic number will change over time, and probably for
different spools of plastic.

2) Adjust the packing ratio in Skienforge. Then enter the measured
filament diameter. Again, this will be trial and error to get it
matching your specific environment and is likely to change over time.

3) Adjust your stepspermm setting in the machine's xml config file.
This is a little more daunting, but easy if you are familiar with XML.
This tells the printer how many steps of the extruder motor are needed
to extrude 1mm of plastic. This value seems to not care about the
environmental changes or packing ratio as much, and results in you
entering the measured filament diameter. So it seems you set this once
for your particular bot/motors and then forget about the issue. (at
least that's been my experience thus far...)

If you opt for choice 3, search the group (and/or the jetty-firmware
group) for the "stepspermm" phrase. This has been covered in some
details a few times.

Hope that helps.

Shawn

On 13-01-11 12:18 AM, Bj�rn Syse wrote:
> Hi, I was curious and hopeful to try this. I had my filament set at
> 1.74, but now I tried 1.84 and printed the same hexagonal shapes. Things
> changed slightly, the pillar grew slightly smaller, but the hole did
> aswell. Very strange.
>

> <https://lh3.googleusercontent.com/-CRMQoRjfVH8/UO-8jJD_d5I/AAAAAAAAAEk/zhk94rtOcTU/s1600/Screenshot+2013-01-11+00.jpg>

>
>
> The surface quality seems somewhat worse at one point, and better at
> other aswell.
>
> What's the best way to calibrate what filament diameter I should have
> set? Any test pieces that are suited for this?
>
> On Wednesday, January 9, 2013 6:42:48 PM UTC+1, pfinucan wrote:
>
> It looks like you are over extruding a little bit. Try a test print
> and INCREASE your filament diameter setting by 0.1mm or 0.2mm. That
> will cause you extruder to extrude a little LESS plastic, making
> each line/pass thinner. Which in turn, will make you holes larger,
> and your piston smaller.
>

> --
>
>

[Björn Syse]

Thanks you shawn, very comprehensive!

I'm just thinking, by calibrating this, what should I expect? Should I keep on calibrating until 4.6mm in the CAD shows up as 4.6 in the print, or is that too optimistic?

Also, I'm experiencing a bit of a different sizes on different sides of that hexagon, as if there is a difference between X and Y.

3) Adjust your stepspermm setting in the machine's xml config file.
This is a little more daunting, but easy if you are familiar with XML.
This tells the printer how many steps of the extruder motor are needed
to extrude 1mm of plastic.  This value seems to not care about the
environmental changes or packing ratio as much, and results in you
entering the measured filament diameter. So it seems you set this once
for your particular bot/motors and then forget about the issue.  (at
least that's been my experience thus far...)

I'm curious about this setting. The number of steps the motor needs to extrude 1 mm of plastic, is that really dependent on the plastic at all? That should just be a function of the extruder cogwheel radius, am I right? And that should be the same for all MK8 extruders, no? But are you saying there are other parameters that affects this also?

[Shawn]

The common approach is to print out 20x20x10mm cubes for calibration. At
100% fill, with 0 extra walls, and anything other than hexagon fill
pattern selected (there's a known bug there with regards to 100% fills).
When that cube looks right, you are pretty much set to go (until next
time).

By "Looks Right", there's a few things to watch for.

1) Are the sides building straight up/down. If not you have a shifting
issue - either slippage, skipped steps, or some other mechanical cause.
2) Is the top flat. If it curves up you are getting too much plastic,
curves down indicate too little.
3) Look at the individual strands of plastic. Are they spaced right and
evenly. If you can see through to the layer below, you are not getting
enough plastic. Or maybe you are printing too fast for your extruder
temperature.
4) Do you have smooth walls. If you are getting too much
waves/zits/etc. you have other issues to deal with. Too much plastic,
too fast, etc.
5) does the cube measure 20x20x10? Each measurement may be off by
+/-0.1mm or so due to the tolerances of the machine (I think that's the
right number). If you are seeing something noticeably more, you should
continue tuning - chances are one of the other issues is occurring as well.

This isn't a definitive list, just my experience. Also, I print with
ABS on a Rep1, so your results may be different. Look for the Sailfish
tuning wiki page - it has some tips in there that greatly help as well.
There was also a page somewhere for the Jetty Firmware (pre-sailfish)
that discusses how to recognize the different problems - I'm not sure
where that went with the recent re-org of the wiki content.

> --
>
>

[66tbird]

After a week of chasing this issue months ago I final just compensated in the drawing. I'll do some test fit only sections then when all is good I'll dynamically dimension it in a copy drawing used just for printing. Seems to work because each roll is different. So I'll end up with a few files, one for each filament size and right or left extruder. I'm getting closer at just knowing how much to compensate.

[whpthomas]

If you want really precise holes - in the order of +/- 0.01 mm I would recommend investing in a set of reamers. This is a set of German made dormer reamers in standard metric sizes.  

[Björn Syse]

Hi,

You're right but that kinda defeats the smoothness of 3d printing in the first place.. Somewhat at least. I was mostly concerned with trying to create axis that other parts can rotate around etc. but sure in some cases where exact diameter holes are needed the reamer is probably the way to go!

-Björn 

--

Björn Syse 

bj...@syse.se

+46 706 281244

On 15 jan 2013, at 09:15, whpthomas <m...@henri.net> wrote:

If you want really precise holes - in the order of +/- 0.01 mm I would recommend investing in a set of reamers. This is a set of German made dormer reamers in standard metric sizes.  

--
 
 

<DSC00293.JPG>

[whpthomas]

Yes but even if you are running a CNC mill you will tend to ream out bushings when you want +/- 0.01 tolerances - unless you have a really high end setup. Its just what you do.

So to expect more from a printer that is squirting out plastic through a 0.4mm nozzle onto a 0.1mm layer on a build platform that is not at all ridged - even if the stated x and y accuracy is 0.01 is just a bit unrealistic IMHO. Better to spend the money and devote your time and attention to outcomes that are more realistically achievable. This is not to say you will not get good results, but for small mechanical parts I would ream all the holes and print your shafts along the XY plane and not vertically up the Z axis.

[Wingcommander whpthomas]

Now that I am actually working on these again I will give an example of what I have been doing that is working.

I tend to supersize axles in models, so if the axel should be less then 8mm I make it 8mm.

In the current model the axels are 10mm so I use that dimension for the holes and ream them out. For the axles I make them 10mm if I want a firm fit with no play and 9.9mm if I want them to turn freely. I print the holes with the axis aligned with Z, and axles with the axis aligned with X or Y.

For threads, working with PLA I have found if you halve the thread pitch and subtract that from the hole diameter you can self tap without too much force.

[sciencemike]

Others have discussed the polygon issue.  We are mating our Rep 2 parts onto machined surfaces.  The way that we consistantly get the parts to work is to turn up the number of polygons so they do not fill the hole up.  We have found that we need to use <5% angle tolerance to get the surfaces to be nicely fitting.  2.5% angle tolerance is even better, but the files get pretty large.

[Björn Syse]

Hi again Henry

Sorry, I thing I misunderstood you about the reamers. English is not my first language and I have no background as a machinist, but I'm not sure what the reamers are? I first though you meant I should make the holes using a cnc milling machine with little end mills. What is a reamer? Can I attach them to and hand held battery screwdriver you reckon? Or are there hand cranked ones aswell?

Sorry for the confusion. 

[Wingcommander whpthomas]

Sorry, I thing I misunderstood you about the reamers.

No problem, reamers is for most people a rather specialised bit of kit. There are probably three general types or reamers. Tapered hand reamers which are good for reaming out holes in thin sheet metal panels. Adjustable reamers which are probably best for use in a milling machine or lathe, and finally fixed reamers which will likely be the most useful for your purposes. Do a search for 'metric' or 'imperial reamer set' on ebay - you should find one.

You can use a t-handle tap wrench work these by hand.

[Björn Syse]

Thank you, I've bought a couple at ebay now (couldn't find a collected kit like yours unfortunately) but still. Also bought a tap wrench. 

I was just curious, in your extruder upgrade modification you mention one should ream out the holes. I tried to measure the holes on the step, and one seams to be 6mm, but the others are not even on the mm, that is their something like 2.2 mm or similar. What reamer size should I use here? Is this perhaps an effect of you using imperial measurements and me metric btw?

[Jetguy]

When you ream metal, yes the hole size is the reamer size. When you
ream plastic, the plastic deforms and thus often and especially on
small holes is smaller than the reamer. You also can stand for the
hoels to be sloppy (compared to the screw or whatever) since it's
often more than one hole that lines up to another part. We don't need
fractional precision everywhere. You could waste a lifetime worrying
about things that don't matter.

On Jan 29, 11:01 am, Björn Syse <bj...@syse.se> wrote:
> Thank you, I've bought a couple at ebay now (couldn't find a collected kit
> like yours unfortunately) but still. Also bought a tap wrench.
>
> I was just curious, in your extruder upgrade modification you mention one
> should ream out the holes. I tried to measure the holes on the step, and
> one seams to be 6mm, but the others are not even on the mm, that is their
> something like 2.2 mm or similar. What reamer size should I use here? Is
> this perhaps an effect of you using imperial measurements and me metric btw?
>
> On Tuesday, January 29, 2013 11:23:18 AM UTC+1, Wingcommander whpthomas
> wrote:
>
>
>
>
>
> > Sorry, I thing I misunderstood you about the reamers.
>
> > No problem, reamers is for most people a rather specialised bit of kit.
> > There are probably three general types or reamers. Tapered hand reamers
> > which are good for reaming out holes in thin sheet metal panels. Adjustable
> > reamers which are probably best for use in a milling machine or lathe, and
> > finally fixed reamers which will likely be the most useful for your

> > purposes. Do a search for *'metric'* or *'imperial reamer set'* on ebay -
> > you should find one.
>
> > You can use a t-handle tap wrench work these by hand.- Hide quoted text -
>
> - Show quoted text -

[Björn Syse]

You could waste a lifetime worrying 
about things that don't matter. 

Haha,  that's so true. Thanks for the awakening bitchslap ;)

 

On Tuesday, January 29, 2013 5:55:47 PM UTC+1, Jetguy wrote:

When you ream metal, yes the hole size is the reamer size. When you
ream plastic, the plastic deforms and thus often and especially on
small holes is smaller than the reamer. You also can stand for the
hoels to be sloppy (compared to the screw or whatever) since it's
often more than one hole that lines up to another part. We don't need
fractional precision everywhere. 

[Jetguy]

What I was implying was your questioning on the holes of the printed
extruder drive. You're worried about a fraction of a mm on a hole that
as long at the screw head doesn't pull through doesn't matter. Say the
screw diameter is 3mm, well with threads, you probably want 3.3mm
(10%) to slide the screw in, not have it bind and line up with the
existing threaded hole in the motor. Hell, the hole could be 4mm at
long as the head doesn't pull into the plastic too far. It's a non
critical application.

This is the problem, they cannot determine a critical part from a non-
critical part. If one of the 2 parts has adjustment, then the other
one is a not critical part.

A fine example is a door hinge on a typical inside door. The ones I
see have 6 screws, 3 per side of the hinge 3 hinges total. The
handyman measures for one hole, attaches the hinge and uses that known
hole to locate the other proper screw positions.

The same door being hung by an engineer does measurements for 2 days.
He looks up the factors for expansion and contraction, does pages
worth of math. Finally, he breaks out a CNC mill to drill the pilot
holes for the screws. He realized he forgot to measure for the
thickness of paint and scraps the door 3 times because of such tiny
errors. Finally, 2 weeks later the door is mounted and the screws are
torqued to spec. The engineer then realizes the door swings the wrong
way, the lock is on the outside, and they are now locked in the room.

Don't be the engineer, be the handyman.

> > > - Show quoted text -- Hide quoted text -

[Joseph Chiu]

> Don't be the engineer, be the handyman.

Love it.  This is going to be in my head for the rest of my life!

[Wingcommander whpthomas]

The same door being hung by an engineer does measurements for 2 days.
He looks up the factors for expansion and contraction, does pages
worth of math. Finally, he breaks out a CNC mill to drill the pilot
holes for the screws. He realized he forgot to measure for the
thickness of paint and scraps the door 3 times because of such tiny
errors. Finally, 2 weeks later the door is mounted and the screws are
torqued to spec. The engineer then realizes the door swings the wrong
way, the lock is on the outside, and they are now locked in the room.

It sounds to me like the engineer you are describing has never actually made anything - just designed it in a 3D CAD program perhaps - is that a real engineer?

The engineers I know (who I would consider real engineers) are more like the handyman you describe - that is the whole point of things like specifying tolerances - sometimes you need things to be +/- 0.01mm, like when you need to press fit bearing housings - but then you weld it to a weldmesh and its going to get all pulled out of shape and be +/- 0.1 - So a bit like hanging the door - you start with the things that need to align and cut and shut back from there.

A door hinge would never be +/- 0.01mm, so you would never measure the paint - but prime the inside of a bearing housing with 2 pack enamel and the paint can be an issue - nothing a 12 ton press can't fix though ;)

[Wingcommander whpthomas]

I was just curious, in your extruder upgrade modification you mention one should ream out the holes. I tried to measure the holes on the step, and one seams to be 6mm, but the others are not even on the mm, that is their something like 2.2 mm or similar. What reamer size should I use here? Is this perhaps an effect of you using imperial measurements and me metric btw?

I just reamed the 6mm and 8mm sides of the pin and cleaned up the countersunk recess with the 6mm -  the other holes are designed to be self-tapping using the M3 countersunk cap screws.

[Björn Syse]

aha, thanks!