Apparently Impossible Sphere / Week of June 16, 2012

Happy Father's Day to all the geeky dads out there.

As some of you will know, I am a big fan of mechanical puzzles, and got into 3D printing about 4 years ago as a way of producing high quality 3D puzzles without needing the woodwork craftsmanship. So, I have decided to encourage you to make a puzzle for this week's competition.
Many of the puzzles I make consist of basic cubic shapes, which isn't too challenging for our printers as long as there are no overhangs involved.
However, there was one puzzle I made recently that really tested my printing skills and could definitely be improved upon. This was UechiMike's derivative of my Apparently Impossible Cube called the Apparently Impossible Sphere.
It was a really clever idea to turn a cube puzzle into a ball with the same method of construction and the fact that it still works as a puzzle is a bonus. And as UechiMike says "While a Cube is cool a sphere is much cooler".



As you can see, even after clean-up my first attempt was quite lumpy and not a great fit. I'm sure you can do better! The sharp edges, slopes and any tendency to warping will all affect the fit and finish of your puzzle ball.

I would like to see a nice round sphere and a great choice of colours. First prize will go to the best looking sphere, artistically and functionally, which means that you not only have to print the pieces, but solve the puzzle as well. Don't worry if you're not a great puzzle solver - just follow the video from the original cube puzzle page.

Download the two stl files from Thingiverse and print two copies of each file in colours of your choosing (as many as you like).
Use your slicing software to scale the print up by 150% to give a finished sphere of about 4.5 cm.

Good luck! I look forward to seeing your creations.

richgain
microcubology

That's a good one! I'm downloading it now...

---

[richrap.blogspot.com]

Mike has commented that he was unsure how accurate his boolean intersections were so I have had a go myself using Blender to position the cube pieces and the sphere perfectly.
Note that I have also scaled these pieces up by 150% so there is no longer any need to do that in the slicing software. I also removed the bevel which I thought was unnecessary.
In theory, these pieces should have a marginally tighter fit because the scaling is now done before the offset is applied - in practice I suspect that the tuning of the printer will play a much greater role.
Visually there is very little difference so I think Mike actually did a pretty good job but you are welcome to use these models instead if you prefer.

richgain

Doh. I printed the original scaled by 150%. I still got it to go together:
I printed 2 parts on one printer and 2 on the other overnight.

; generated by Slic3r 0.8.2 on 2012-06-17 at 20:57:18

; layer_height = 0.1
; perimeters = 3
; solid_layers = 9
; fill_density = 0.25
; nozzle_diameter = 0.31
; filament_diameter = 2.87
; extrusion_multiplier = 1
; perimeter_speed = 20
; infill_speed = 60
; travel_speed = 80
; extrusion_width_ratio = 0
; scale = 1
; single wall width = 0.43mm


They both took about 3:24 to print.

There was some minor surface blemishes on the faces that faced each other on the bed from when the extruder moved between the parts, but a few scrapes with a hobby knife got most of that off.

A bit of a spray with silicone spray before assembly prevents them from binding. They go together quite nicely.

Great start! That sets the bar pretty high.
Whenever I try to print at 0.1 layer height through a 0.3 nozzle at anything less than 50% infill the filled section eventually degenerates into a spiky mess. What's the secret?

richgain Wrote:
-------------------------------------------------------
> Great start! That sets the bar pretty high.
> Whenever I try to print at 0.1 layer height
> through a 0.3 nozzle at anything less than 50%
> infill the filled section eventually degenerates
> into a spiky mess. What's the secret?

Use hexagon or concentric infill.

---

FFF Settings Calculator	Gcode post processors	Geometric Object Deposition Tool Blog
Tantillus.org	Mini Printable Lathe	How NOT to install a Pololu driver

Sublime Wrote:
-------------------------------------------------------
> richgain Wrote:
> --------------------------------------------------
> -----
> > Great start! That sets the bar pretty high.
> > Whenever I try to print at 0.1 layer height
> > through a 0.3 nozzle at anything less than 50%
> > infill the filled section eventually
> degenerates
> > into a spiky mess. What's the secret?
>
> Use hexagon or concentric infill.

The other trick (better than Hex infill speed wise) is to use infill every n layers.

I did infill every 3. Unfortunately that setting doesnt get recorded in the gcode (athough it should be obvious scrolling through the layers). So basically the infill is printed as if you were printing a 0.3mm layer height instead of 0.1.

Incidentally, I actually have a 0.4+mm nozzle (drilled with a 0.4mm drill bit). I only set the nozzle size to 0.31 because I wanted a bit finer thread spacing than I would have got If I specified my real nozzle size. And with infill every 3 layers, the floating point math decided that 0.1*3 was greater than 0.3 so I bumped the nozzle size to 0.31 and Slic3r became happy.

A few tips for others that may be printing this: Print the parts 1 at a time, or if you do print 2 at once, make the inside of the parts face each other so that the nozzle never crosses the boundary on the surface of the sphere. You may get a few dags on the inside of the part (which will affect the way the puzzle slides together, but you can tidy that up with a knife/sandpaper and it will be hidden when the puzzle is together and you will still have a nice outer surface finish.

Also, If you print in 2 colours, print one of each part in each colour instead of the way I did it (part 1 purple, part 2 silver). I think this may look a bit more funky.

I have to say Greg, that is a lovely smooth print! The infill every n trick is particularly useful too, thanks

These competitions started at a perfect time for me, as i had just finished my printer and done some calibration and wanted to see what it could do. It might be worth building up an assault course of objects (preferably aesthetically pleasing or useful) and detailing issues highlighted by specific objects with some tips, tricks or suggestions to help. I found attempting the competition objects to be particularly helpful milestones (along with 1:24 scale Windsor chairs which is a great test of bridging and correct calibration I think)

For example, something like this set of calibration objects with its calibration suggestions but taken to the next level, making some useful/interesting objects in the end. I got real tired of printing cubes

If something like this already exists, point it out!

Printer working again - so Here is my Entry for this week's competition - and I love this object/puzzle, thanks for designing it.

; generated by Slic3r 0.8.3 on 2012-06-19 at 19:09:57

; layer_height = 0.3
; perimeters = 4
; solid_layers = 2
; fill_density = 0.12
; nozzle_diameter = 0.40
; filament_diameter = 2.86
; extrusion_multiplier = 1
; perimeter_speed = 60
; infill_speed = 60
; travel_speed = 190
; scale = 1
; single wall width = 0.42mm

I didn't check how long, but it was less than about 40 mins for each set of two parts to print.





I cleaned off a few blobs and lined them up and crunched my hand together, they popped into place, a little wiggle and I can just about get them apart.







I'm going to do another one with much finer layers, I'm very impressed with Greg's fine print.

---

[richrap.blogspot.com]

You have an interesting ripple on your prints. Do you use T5 belts that cause that ripple? or is that a speed artefact ringing as it turns corners??

Nice use of colour by the way!

Edited 1 time(s). Last edit at 06/20/2012 04:50AM by Greg Frost.

Greg Frost Wrote:
-------------------------------------------------------
> You have an interesting ripple on your prints. Do
> you use T5 belts that cause that ripple? or is
> that a speed artefact ringing as it turns
> corners??
>
> Nice use of colour by the way!

Hi Greg,

I was just trying to work that out, I have not really noticed it in other prints so I'm thinking of things to try to investigate it -
I'm using 2.5mm belts and metal pulleys with Marlin V1 RC2.

I think it must be speed related, unless my belts are now not as tight as they should be.

My perimeter was at 60mm/sec the same as my infill, I noticed you did it at 20mm/sec for the perimeter.

I use very fast acceleration settings in Marlin on the MendelMax, so it may need tweaking down a little, I'll investigate.

---

[richrap.blogspot.com]

Very nice set of photos, Rich. I really like the rainbow effect and the fit of the sharp angle is looking pretty good too. If anyone else fancies having a go at printing this (don't be put off by the stiff competition - you could get useful feedback on your prints) I would really value photos that clearly show the fit of the sharp point to help me with judging the result.
Thanks
richgain

richgain: I just downloaded your zip and looked at your version of the model, and it looks like you have reduced the poly count. The original model was much smoother.

An openscad version of the model would be nice. That way you could parametrically control the poly count.

Greg Frost Wrote:
-------------------------------------------------------
> richgain: I just downloaded your zip and looked at
> your version of the model, and it looks like you
> have reduced the poly count. The original model
> was much smoother.


I was using the Richgain version, I wonder if the reduced poly count is the cause of my ripples?


> An openscad version of the model would be nice.
> That way you could parametrically control the poly
> count.

That would be great, I would love to see an octagon or star version of the puzzle!

---

[richrap.blogspot.com]

Printed this up at .17mm, 10% fill, 3 'shells' using KISSlicer on Fablicators.

Total print time was about 3hrs, though each part was made on a different printer, and the whole thing took ~45 min

Of course I never bothered reading the updates in the thread before I printed this, and did the original at 150% instead of the zipped file.





Edited 2 time(s). Last edit at 06/20/2012 10:38AM by Andrew Diehl.

---

www.Fablicator.com

Greg Frost Wrote:
-------------------------------------------------------
> richgain: I just downloaded your zip and looked at
> your version of the model, and it looks like you
> have reduced the poly count. The original model
> was much smoother.
>
> An openscad version of the model would be nice.
> That way you could parametrically control the poly
> count.

I didn't deliberately reduce the poly count, but I do sometimes struggle to make Blender do what I want it to do.

Anyway, thank you for the prompt I needed to finally start getting my head around OpenSCAD - it's long overdue.
I have written a simple script that takes the intersection of a parametric sphere and each of the two regenerated 48mm cube .stl files.
Unzip the contents and be sure to keep the .scad file and the two .stl files in the same folder.
Does that work for you?

And here are the two higher-res parts generated by that script (I've exported two parts separately so you can place them where you like or print them separately).

Given the simple geometry, I modelled the parts in openscad and made a super high resolution version.

The openscad is parametric and allows you to specify the radius of the sphere and the gap between the parts when assembled to account for printer tolerances.

The one I exported has 0.3mm gap and 32 mm diameter (the same diameter as for this challenge).
The files are too large to attach here. I have uploaded them to thingiverse.

[www.thingiverse.com]

Edited 1 time(s). Last edit at 06/21/2012 08:45AM by Greg Frost.

richrap Wrote:
> > An openscad version of the model would be nice.
> > That way you could parametrically control the
> poly
> > count.
>
> That would be great, I would love to see an
> octagon or star version of the puzzle!
"
OpenSCAD is awesome! "Apparently Impossible Venus de Milo", anyone?


Only joking!
richgain

richgain Wrote:
-------------------------------------------------------
> And here are the two higher-res parts generated by
> that script (I've exported two parts separately so
> you can place them where you like or print them
> separately).

I processed the script and printed the output, looks good, but still getting some very fine banding, I'll try Greg's new version tonight.

I do like the impossible Venus de Milo - if you post it, I will print it!

---

[richrap.blogspot.com]

Andrew Diehl Wrote:
-------------------------------------------------------
> Printed this up at .17mm, 10% fill, 3 'shells'
> using KISSlicer on Fablicators.
>
> Total print time was about 3hrs, though each part
> was made on a different printer, and the whole
> thing took ~45 min
>
> Of course I never bothered reading the updates in
> the thread before I printed this, and did the
> original at 150% instead of the zipped file.
>
> [i211.photobucket.com]
> Ball1.jpg
> [i211.photobucket.com]
> Ball5.jpg
> [i211.photobucket.com]
> Ball3of4.jpg


Really awesome Andrew! Congratulations for your printing.

; generated by Slic3r 0.7.2b on 2012-06-19 at 18:49:07

; layer_height = .2
; perimeters = 2
; solid_layers = 3
; fill_density = .2
; nozzle_diameter = .35
; filament_diameter = 2.82
; extrusion_multiplier = 1
; perimeter_speed = 40
; infill_speed = 60
; travel_speed = 80
; extrusion_width_ratio = 0
; scale = 1
; single wall width = 0.49mm

I wanted to get some other colors into the mix, but I dont seem to have the other ones as dialed in as I do with the silver. Its really tight and doesnt just slide together/apart without some real oomph. Some sanding might help (might get some of the small gaps a little closer too), but I at least wanted to post pics before I got into any modifying. I was really happy with the way it turned out until I saw some of the other posts! Really wish I could have done more colors, but maybe next time! Always fun!

Rich, I'm dissappointed that I dont see any tiny bearings poking out of your print! JK, glad you got it all resolved!

---

Josh

//++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
//Prusa / RAMPS_1.4 & ATMega_2560 / Slic3r > Pronterface > Marlin / Luck
//++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

richgain Wrote:
--------------------------------------------------
>
> Anyway, thank you for the prompt I needed to
> finally start getting my head around OpenSCAD -
> it's long overdue.
> I have written a simple script that takes the
> intersection of a parametric sphere and each of
> the two regenerated 48mm cube .stl files.
> Unzip the contents and be sure to keep the .scad
> file and the two .stl files in the same folder.
> Does that work for you?

This one was done from the SCAD Script and printed at 0.2mm layer height -



A smoother fit.

I have not had a chance to try Greg's High Resolution version yet.

---

[richrap.blogspot.com]

This is the high resolution model (not scaled up by 1.5) with my original print in the background.
Same settings (0.1mm layers).

Well, it's judging time and I must say I've been very impressed by the quality, if not the quantity, of entries this week. So firstly, my apologies if puzzles don't float your boat, or the longer print times were off-putting. There's always next week!

Well done for having a go at it, Josh, and for getting a good fit on the pieces and some nice sharp corners.
Rich gave us the beautiful rainbow version first and then followed up with a very smooth caramel swirl - yummy!
Andrew, I loved your choice of colours and the pieces were really nice and round, especially considering they came off the lower resolution models. Well done.
But I think this week's winner is Greg. That first picture of the purple and silver ball was simply amazing to see. The smoothness of the print has shown me that I really have to get low layer heights nailed, and Greg's advice should prove very helpful with that.
So, congratulations Greg, and thanks to everyone who had a go this week.
richgain

Well done Greg, that truly is a wonderful print. I'll have to try the new High-res version.

We really did have a fantastic set of prints this week can't wait for the next challenge...

---

[richrap.blogspot.com]

They were all great prints this week. I particularly liked Andrew Diehl's. Truly impressive.

It is interesting that the smaller one that I printed still has banding on it but this time I think it is something to do with the stepper motor precision something like nophead mentions here: http://hydraraptor.blogspot.com.au/2012/04/stepstuck.html