Welcome

This page is about my interests, projects, and profession (technical artist in the video games industry).  Most of my hardware\software projects are coded in PythonProcessing, & Arduino.  I also enjoy 3d printing, you can find my designs for download over on Thingiverse.  More pics over on Instagram.

Find Processing\Android\Python programs\apps I’ve developed via the above title bar.

I also maintain several wikis on Maya\Python\Pygame\Processing that I update far more often than this blog.  See them on their page.

All information on this site is copyright under the Apache Licence v2.0, unless otherwise noted.  Which means you can pretty much use the information here for whatever you like, but I always appreciate credit where applicable.

Have a look around.  Thanks for stopping by.

— Eric Pavey

CNC The Bay

I bought & assembled my X-Carve cnc during Christmas, 2015.  Ever since then, I thought it would be perfect to do topo maps.  I had great success with Denali & Lake Tahoe, but always wanted to do something bigger.  In fact, a year before the X-Carve’s arrival, I’d 3d printed the (lower half) of the San Franisco Bay Area, which I felt turned out well.

Some friends and I took a trip to Firewood Farms in Half Moon Bay, where I picked up a couple slabs of redwood, without really knowing what I was going to do with them at the time.  Then the stars aligned, and I realized one was the perfect shape to do a cnc topo map of the whole bay area.  “The Bay” is the result:

uberbay_final_web

uberbay_persp_web

uberbay_closeup_web

More pics below.

Info on it:

  • The piece measures 17×30″, and about 1.75″ deep.
  • Rough cut was 3h, 10min.  Finish pass was 6h, 20min.   Total was 9.5h (my whole Saturday…).
  • Used a 3-flute .25″ ballnose for both passes, DeWalt router on speed 1.
  • Roughcut feed for X/Y was at 150″/min, finish pass was at 210″/min.
  • Roughcut feed for Z was 20″/min, finish pass was 40″/min : I had to modify my firmware to get these speeds.  I tried a faster z-feed, but the machine couldn’t raise the spindle that fast, and chaos ensued.
  • Very little sanding was needed, based on the 10% stepover during the finish pass.
  • Used terrain2stl to get the topo data.
  • Used Autodesk Maya for all the mesh modification, and text creation.
  • Used MeshCAM for the toolpath generation.
  • Used Universal GcodeSender to send the cam to the machine.
  • Applied a ‘natural’ Minwax stain, and “Blue Lagoon” satin paint.

Issues:

  • I think the poor Arduino Uno couldn’t handle the fast feedrate during the rough terrain parts:  The whole thing would start & stop over and over (giving me an anxiety attack at first), while the buffer caught up with the operations.  Maybe it’s time to get some new electronics?  That really slowed down the finish pass.  Didn’t have any trouble during the rough.
  • I think I could speed up the rough-cut even more by not using MeshCAM’s ‘Use 3D Roughing’ option : while it makes a nice looking roughcut, who cares, it’s all getting cut out anyway.  The result was a lot of small up/down z-travels, that really slowed it down overall.

Here’s a series of photos showing the process:

A shot in Maya, while I got the text arranged:

uberbay_maya_web

1.15 million triangles:

uberbay_maya_polydensity_web

The very first cuts:

uberbay_initialcut_web

Roughcut complete:

uberbay_roughcut_web

The final piece, before stain and paint:

uberbay_unpainted_web

In conclusion, I found this to be really satisfying project, and it looks great hanging on my living-room wall.

 

A comparison of Simplify3D and Slic3r Prusa Edition

allBenchies_web

I printed so many 3dBenchies for this comparison….

Preface

I’ve been a 3d printing hobbyist for right around five years now:  I purchased my first printer, a Makerbot Replicator 1 (which I still own, still prints great) in March of 2012.  Since then I’ve used multiple slicers to generate my gcode, but mainly (and in order):  ReplicatorG, MakerWare (now MakerBot Print), and Simplify3D.  And for the time I used them, they’ve all been great.  I’ve also built my own printer a few years ago.

However, and as much as I like Simplify3D:

  • Despite it does so many things so well…
  • The fact that I paid cash money for it.
  • And that their updates are so few and far between…
  • And the updates don’t include fixes to some major issues I feel it has…

I thought I should revisit what slicer software 2017 has available.

I know that slicer software is like religion, politics, and CAD software:  People can get very attached to them.  And there’s a good chance someone out there is going to disagree with my findings : That’s great, I’d like to hear your perspective, leave a comment below.  Everyone uses the software differently, and the things I find important (or not) could be polar opposite of someone else.

Based on a general poll I posted, it looks like these slicers are out in the wild and commonly used.  Yes, there are more, but these seem to be popular:

After reviewing all of them, and for reasons listed below, I decided to investigate Slic3r Prusa Edition.  And mainly that edition over vanilla Slic3r because vanilla Slic3r hasn’t been updated since mid 2015, while the Prusa edition is being actively developed (+ extra fancy features).  I know vanilla Slic3r is still being developed, it’s just been a while since an official release has gone out.

Update:  As it turns out, dev releases of vanilla Slic3r are generated often, so it’s pretty easy to be on the cutting edge, find downloads here.

This is not a post designed to bash Simplify3D and espouse the values of some other slicer:  My goal is to make it an honest comparison of the two software.  It should be noted however, I have over two years of experience with S3D, and only a few weeks on Slic3r, but so many of the concepts of the software overlap, I feel I have a pretty good grasp of it.

Finally, why am I writing this? It’s taken quite a while to do all the research!  Mainly for myself actually:  I wanted to get all my research down on paper to see if I should make a slicer switch or not.  There are so many facets to all of this, I wanted to get it in one place to make sure I wasn’t missing anything.

I will continue to update it over time, see bold “Update” text below.

What I used for testing

  • 3D printer hardware:
    • My custom core-xy C-Bot : It has a 12x12x21″ build volume, auto-leveled on a heated, removable, mic6 aluminum bed covered in a PEI sheet.
    • Runs RepRapFirmware on 32-bit RADDS/Due electronic.
    • All steppers set to 1/128 microstepping using SD6128 drivers.
    • I currently print mainly MakerGeeks PLA out of a .6mm E3D-v6 Volcano nozzle at 230 degrees on a 60 deg bed, with active PLA cooling from a ducted 24cfm fan.
  • 3D Model:  I used the good’ol 3dBenchy as my test model for all comparisons, except the thin-wall compare, it’s linked to below.

So let’s get into it…

Simplify3D

Version Tested: 3.1.1

Unique things it does well

These are things I’ve found it to do, that other slicers may not.  Or it does better.

  • Multiple ‘Processes’:
    • A process in S3D is basically a print preset.  And you always need at least one:
      • s3d_process x
    • The power comes into play however, that you can assign them per mesh to print (if there are multiple being printed):
      • s3d_process_permesh
    • Or by layer height:
      • s3d_process_heights
    • You can even ‘group’ processes so as to link their similar parameters together, so changing one changes all others (see how this works here).  This is extremely powerful, and arguably one of the best parts of the software.  And when I need it, it’s great.  But honestly 95% of the time I don’t need it.
    • Not that it effects me so much, but it has pre-configured profiles for many printers, allowing a larger market-share to have success with the software sooner.
    • I should note that I’ve found no other slicer that treats “processes” as well as S3D.  This is a major selling point, and brings much power and convenience when needed.
  • Print options:
    • In its ‘Advanced Mode’, there’s a knob to tune just about every aspect of your print.
  • Post processing of gcode:
    • You can modify the output of the gcode via S3D’s own custom syntax.  You can see this post on an example of how to do it.
  • Printer communication & machine control panel
    • S3D has a robust “machine control panel” for operting your printer, issuing gcode, printing gcode either over USB, or via SD.  It also allows you to make your own macro’s that show up in the UI.
  • Manual supports:
    • An area where S3D really shines, is how easy it allows you to place and generate your own support material (vertical columns).
    • Even in ‘manual’ mode, you can place them ‘automatically’, then add/remove them as you need.  A nice feature is you can toggle on/off ‘from built platform only’, so you can generate automatic supports for everything touching the bed, the later add in manual supports as needed elsewhere.
    • Support removal has always been great.
  • Technical support:
    • Yes, they actually get back with you if you email them with a problem.
  • Active forums:
    • Linked above:  A very active community.
  • Gcode previewer:
    • Does a great job displaying gcode through a variety of filters (print speed, feature type, etc).
  • Slicing speed:
    • Fastest slicer I’ve ever used, 3DBenchy sliced in 1 second.
  • Print results:
    • There’s no question that once you get your settings tuned right, S3D produces great results.
  • Object transformation:
    • In addition to translation\rotation\scale of parts on the build platter (either via a manip, of numeric entry), you can also specify dimensions, and it will scale to match.  Handy.
  • Multi OS support:
    • I’ve used it on powerful PC’s, and pretty old Mac’s, and it’s never had a problem.

Things it does not do well

This list is the main reason I started looking for other slicer options.  The “Major Issues” are things that I personally feel really, really need to be addressed.

  • Thin wall detection:
    • Major issue“: If you have a wall that’s .5mm thick, but you’re printing with a .6mm nozzle (or extrusion width), S3D won’t print it.  This drives me (and many others) nuts, considering old slicers like Makerware handled this just fine.  Simply having the ability to ‘opt in’ would be great.  Since I print with large (.6 -> 1.2mm) volcano nozzles, it means many files are simply unprintable.
  • Automatic supports:
    • Major issue“: S3D does a great job with manual supports (that you place by hand, which can be accurate but time consuming), but it’s automatic supports (most commonly used) are some of the worst I’ve seen, because… it doesn’t support any type of stalactite (or icicle) like feature.  While the supports it creates are fantastic, and remove great (better than any other slicer I’ve used), when it fails to create them at all where it should ‘automatically’, that’s a major failure.   I have an ongoing thread here with plenty of pics showing this problem.  This is a straight up bug that other slicers don’t have any problem with.  In fact, Simplify3D, KISSlicer, (Update:) CraftWare and ideaMaker have this problem:  Cura, Makerware, Slic3r, and ReplicatorG all handle this just fine .  Here’s an informative pic with the problem.  Just imagine how this can make your prints fail, I’ve had many wreck because of it:
    • overhangCompare
  • Cost, no trial version:
    • That about says it all.  I’ve been told they’ll refund you if you don’t like it after you buy it.  But seriously, there should be a trial for software like this for new people
  • Timely updates
    • They have maybe one update a year… and since it’s closed source software, you have no idea when its going to be released, or what will be in it.  And the last update was pretty lackluster as far as major bugfixes (anything on this list) or improvements.
  • Listening to their community:
    • While they do have tech support (that does respond), they seem to be completely absent from the forum.  Which I personally find quite questionable, since it gives the impression they don’t care about their user base.   There can be a certain topic that has 5 pages of people voting it up, but you never hear if the developers know, or care.
  • Toolpath simplification of high res mesh:
    • Major issue“: This is another major gripe of mine I’ve talked with tech support about, and they seem to have no want to address:  If you pass in a very high-res mesh from zbrush (for example), that may have a much higher poly density that what can actually be printed, the software will do no sort of simplification on the tool path.  The result is that it still tries to print all the detail, and the fallout is your print having a surface quality of oatmeal.  I have example prints that show this off pretty well, as they transition from low-poly to highres mesh.  Their suggested fix is that you manually decimate all your models.  But this means any time I scale a model, I have to re-decimate it for that particular scale (and how do I know how much to decimate it?).  That’s horrible.  Considering other slicers, like ancient Makerware or Slic3r do this for you.\
    • UpdateThis post of mine talks about the problem more, but here are some pics showing it off.  If you enlarge the pic on the right the quality difference is pretty clear.
    • resCompare_s3d_web foot_web
  • Lumping printer settings, print settings, and filament settings into the same preset/Process.
    • Not the end of the world, but since all these things are saved together, it means I need to have… a lot of processes for every speed, resolution, and filament type combo.  Splitting these settings up (like Slic3r) would only make data management easier.
  • Process Storage:
    • Even though S3D does a great job with it’s ‘processes’, it took a step backwards with it’s v3 release:  Up until then, when you’d save a process, it would save on your hard drive, allowing for multiple people to easily share process values.  But with v3, all process are saved “internally” in the software, making it very hard to do any sort of distributed process sharing between employees.  Yes, you can export them out to your HD, but this creates an extra, unnecessary step.  The issue is discussed here.
    • Update:  In addition, if you ever ‘reset S3D to factory settings’, it will wipe all your custom process and firmware setups without warning.  Which would be a non-issue if theses were stored on disk.  I’ve been bit by this in the past, its painful.

Slic3r Prusa Edition

Version Tested: 1.33.8-prusa3d-win64

Unique things it does well

These are things I’ve found it to do, that other slicers may not.  Or it does better.

  • Splitting printer settings, print settings, and filament settings into different presets.
    • slic3r_printsettings_tabs
    • I love this feature.  I can mix and match printer settings, print settings, and filament settings as needed.  Cuts way down on the number of ‘profiles’ needed vs say, S3D.  Plus they’re all saved on disk, easily shareable.
  • Toolpath simplification of highres mesh:
    • You can specify a ‘minimum detail resolution’ that will help it print higres mesh better.  Great feature.
  • Thin wall detection
    • If two walls are close together, they can be collapsed into one, so as to not cause an overextrusion problem in that area.  In addition, within reason (and discussed below in more detail), it will print a collapsed single wall much thinner than your nozzle width, which is desirable.
  • Variable layer height
    • Powerful new feature allowing you to smoothly adjust the print resolution over height.  See an informative post here.
  • Unique infill:
    • In addition to what you’d expect out of a slicer (hexagona, triangluar, grid, etc) Slic3d has additionl types like cubic and 3d honeycomb.
  • Unique top layer infill:
    • Whoe doesn’t love a hilbert curve for their top layer, eh?
  • Notes
    • Simple, but handy:  There’s a notes page for your prefs, to jot down extra info.
  • Max volumetric speed:
    • A handy way, mainly when dealing with bigger volcano nozzles, to limit how much material you extrude.  Based on resarch, the common thought is a volcano heater can melt about 30 mm3/sec, so you can actually enter that in, to make sure you don’t try to extrude more material than you can actually melt.  Which I’ve done before, it’s not good.
  • Verbose cooling description:
    • A little thing, but they have a human-readable paragraph describing how your print cooling fan is going to work, based on the values you set.
  • gcode postprocessing via scripting language of your choice:
    • This is pretty great:  You can pass in code from any scripting language (Python, Perl, JavaScript, etc) to postprocess your gcode.  I found this valuable when I had to modify the gcode output to support my RepRap Firmwares, by writing a custom pos-tprocesser in Python.
  • Slicing speed:
    • Not as fast as S3D, but still pretty fast.  3dBenchy sliced in 4 seconds.
    • Update:  Based on the v1.34 release, which the release notes state moved to a parallelized slicing system, dropped the benchy to a 2 second slice time.  Nice!
  • Print options:
    • In its ‘Advanced Mode’, there’s a knob to tune just about every aspect of your print.
  • Update: Octoprint connection:
    • It can connect to, and upload gcode directly to Octoprint.
  • Update:  Run it from the command line.
    • Fully scriptable from the command-line, no window needed!  That’s great for custom toolchains.
  • Update: Object Modifiers
    • Object Modifiers allow you to use stl files to define volumes of your print for custom print settings.  This is similar to S3D’s ‘processes’, except you can use an stl to define a volume to define the custom settings, rather than just a layer height.
    • You can use your own stl files, or it provides primitives like box, cylinder, sphere, and slab.  With the later you only have to define a height, so it acts a lot like defining a S3D process height.
    • In this example, I have a helix for print, and two cubes defining volumes for custom print settings:
    • slic3r_modifiers

Things it does not do well

  • Printer communication
    • Vanilla Slic3r  doesn’t have this option at all (yet, I hear it’s in the works).  It just means you need some other sender software like Printrun or Repetier Host (or, ironically, Simplify3D).
    • The Prusa edition has the option to connect directly to your printer over a serial/usb connection, but I’ve been unable to get it to work.
    • The current dev (non-release) version of vanilla Slic3r does have this option however, and I’d tested it successfully.
  • Thin wall detection:
    • This is a pro above, but I also found it to be a con:  There were issues on my 3dBenchy tests (mainly the flag-pole hole on the stern) where it would collapse thin walls, but not replace them, causing gaps to form in the walls.  Disabling this option caused noticeable over-extrusion in some areas.
    • Update:  You can track this issue here.
  • Mac support:  It crashes a lot on my (old) mac.  No problem on my PC.
  • Update : Adjusting settings print settings per model:
    • Slic3r only lets you modify the ‘print settings’ per model, you can’t adjust printer or filament settings, which would help greatly.
  • Update : x3g support:
    • This only matters if you have a Makerbot printer (like myself), but while you can export a Makerbot-flavored gcode, you’ll still need some external tool (like gpx) to convert the gcode to x3G.  Technically since Slic3r supports postprocessing of gcode, it could call out to a x3g converter via a scripting language of your choice as well.
  • Update:  Saving/loading ‘scenes’:
    • It seems a bit awkward to save and load a ‘scene’.  To me, a ‘scene’ is  the current 3d mesh on the platter, plus all the current machine, print, and, printer configuration states.  You can do it, but (instead of the File menu) via “Platter -> Export Plate as AMF…” (Additive Manufacturing File Format).  you can later “Add” the xml file that process generates, which has the effect I’m after, it’s just a weird/clunky process.
    • Update:  I’ve read that per v1.34.1 it now supports a .prusa format that should handle this… but I can’t actually find that functionality in the software.
  • Update : Undo : Slic3r hands undo ok inside the print\printer\filament settings, but it doesn’t work on the platter:  I accidentally hit “Delete All”, and I was unable to get my data back :(

Comparing Simplify3D & Slic3r’s successes & failures

  • Print Quality
    • They both print great, in my opinion, see the below section with all the pretty pictures.  After setting up similar profiles in both slicers and printing 3dBenchies, the results look different, but great.
  • Overall level of print settings:
    • They both provide a similar level of knobs to twist for designing your print settings.  A wash.
  • Setting organization:
    • While Simplify 3D’s ‘processes’ are great, based on the fact that I use multiple ones so infrequently, I give Slic3r the win here, since it allows you to split you printer settings, print settings, and filament settings to separate presets, while Simply3D lumps them all into one.  Plus Slic3r’s are saved on disk, while S3D hides them inside the software.
  • Gcode previewer:
    • S3d’s better than slic3r for the fact it has more ways to preview the gcode.  However, both of them physically render the gcode about the same.
  • Gcode postprocessing:
    • Slic3r wins here, with it’s ability to use any scripting language.
  • Printer Communication:
    • S3D does a great job with its Machine Control panel.  Update:  Slic3r’s only seems to currently work in a dev branch.
  • Supports:
    • The supports that S3D actually creates (manual or automatic) are fantastic, but when it completely fails to ‘automatically’ support certain features that cause print failure, it’s just sad.  Slic3r’s supports are pretty standard, but don’t fail when they shouldn’t, so it gets a win for just being status-quo.  Both software could learn from the other.
  • Object Transformations:
    • S3D wins here, providing more tools that are easier to use.
  • Tech support & forums : S3D for the win, just a bigger community (or better communication platform) it seems.  It’s not that Slic3r doesn’t have that, S3D’s just seems more active.
  • Thin wall detection : Even though Slic3r has some problems, the fact that it recognizes them at all (compared to S3d) is a win.
  • Cost : You can’t beat free, Slic3r FTW.
  • Software Updates:  The Prusa Edition of Slic3r seems very actively developed… who knows when S3D will get another update, or what it will even have.
  • Highres mesh toolpath simplification:  Slic3r clearly wins here, realizing this is an important aspect of any highres model.
  • Setting storage:  Slic3r for the win here:  All if it’s printer, print, and filament settings are stored on desk in ini files, easily accessed by others, while S3D hides all the settings in the software itself.
  • Update: x3g support:  S3D wins, supporting it natively, with no extra steps needed for export.
  • Update : Slicing speed:  S3D wins, but with the latest Slic3r release I tested, the gap got a lot smaller.
  • Update : Scene saving/loading : S3D wins, it behaves like you’d expect any software too.  While you can do it in Slic3r, it’s just a bit clunkier.
  • Update : Overall experience:  S3D feels like a more polished piece of software, while Slic3r sill gives me that ‘shareware’ vibe from the 90’s.
  • Update: Setting customization per model / height:  This is a tossup:  S3D makes it easy to define processes by height, and you have full control to all print settings within.  Slic3r allows you to create robust Part Modifiers, but it takes a bit more time, and not all options are available within.
  • Update:  Command line api access:  Slic3r has it, S3D does not.
  • Update: Undo : S3D handles undo great, Slic3r is missing it in some areas.

Things I didn’t test

  • Supports : Since I did this entire test with the 3dBenchy, which requires no supports, I didn’t feel it was important to test them.  But I’ve used plenty of supports generated by S3D in the past, and they were always easily better than the older slicers I used like Makerware or ReplicatorG.  I’ve actually not ever printed supports in Slic3r, just looked at the gcode output.
  • Other slicers :  I was considering mixing ideaMaker, Craftware and Cura into this test… but I’d probably never have finished it.

3dPrint Comparison

I should be noted that all prints have had zero cleanup work done to them.  This is to show all possible flaws.

Print Stats

I did most of my testing with the 3dBenchy model, trying to make sure both slicers had similar values.  This is what I printed with:

  • 280 micron z layer heights (‘medium res’ for a nozzle this size, set to full z-stepper steps).
  • .6mm E3d-v6 volcano nozzle, extruding at .72mm.
  • 2 shells, 3 top and bottom layers.
  • Triangualr infill @ 30%.
  • MakerGeeks Urban Camo PLA extruded at 230 deg, with a .97 extrusion multiplier.
  • Printed at 60mm/sec with active filament cooling, and slowing layers down for details like the smoke stacks.
  • The gcode for both slicers were sent over USB, via S3D.

Printer issues:

  • It should be noted my bot has a slight z-wobble I’ve been working on addressing.
  • Even though I think I have it tuned in for that filament, I feel like my tests were slightly over-extruding.

Print time and weight:

  • Both slicers generated prints weighting 15g.
  • Both prints took exactly 41 minutes.

Visual Gcode Comparison

On the left is Simplify3D, on the right is Slic3r

s3d_bency_gcode slic3r_bency_gcode

You can note the nice on-screen legend that S3D provides, not to mention it has many other display modes.  With Slic3r, on the right is what you get.  Visually, they’re about the same aside from the colors used.

Poll Results

I know I’m biased about what I think are good/bad print results.  I though it would be interesting to see what others thought, that really have no experience with 3d printing.  I did a poll at work where I brought in benchies sliced in both software(shown in the pics below), put it in the kitchen, and over the course of the day let people vote on which they thought was ‘better’.

I had a total of 57 people vote, and by 2:1, they chose the Slic3r version over S3D.  I didn’t tell them anything about how they were made other than “I’m comparing two different prints, what do you think is better?”.  I should also note that a number of people abstained saying “I don’t see any difference”.

Do I feel Slic3r’s visual results outweigh S3D’s by 2:1?  Not at all.  But I still find the results interesting.  Take them for what you will.

Visual Print quality comparison

Click on images to expand.

Side By Side

Beauty shot, side by side, S3D on the left, Slic3r on the right:

s3d_slic3r_compare_web

Pretty darn close!  Slic3r has slightly more stringing, but at a glance they look like twins.

Individual Simplify3D & Slic3r closeups:

s3d_closeup_web slic3r_closeup_web

Thin wall comparison

Since the bencny wasn’t the best test for this, I down laded a “thin wall test” from Thingiverse, and ran them through both slicers.  I didn’t print these, but I didn’t need to:  The gcode preview tells enough.

FYI, the walls on the top start at .05mm (on the left), and end at 1.0mm (on the right).  The walls on the bottom start at 1.0mm (on the left) and end at 3.0mm (on the right).

Again, I generated this gcode with a .6mm nozzle, but with an extrusion width of .72mm (20% over).

Simplify3D

Did not do well at all.  It didn’t start printing until the wall with was greater than my extrusion width, so it left out all walls under .75mm.  Kind of hard to print anything architectural related at scale with this limitation.

s3d_thinWall_lines_web

Slic3r

Did great:  It was able to detect and deal with walls almost 1/3 the width of my nozzle, the first one being .25 mm wide.  Note, this was with “Detect thin walls” turned on in the settings.  With it turned off, it had the exact same results as S3D.

slic3r_thinWall_lines_web

Slic3r ‘Detect thin walls’ issues

Interesting findings:  As mentioned above, Slic3r allows for thing wall detection and collapse, S3D does not.  The above benchy pics of Slic3r have the “Detect thing walls” feature turned on.  But I figured out, it was collapsing more than it should, in the case of the flagpole : There’s a small hole in it, which is also shown in the gcode itself:

slic3r_thinwall2_web slic3r_flagpole_hole

Turning this feature off fixed the flagpole, but caused a slight over-extrusion in other places, like where the bow meets the sidewalls, visualized in this comparison:

slic3r_thingwall_compare_web

Note on the right benchy, the line running parallel to the bow, below it.  But also go look at the top/first pic of the S3D/Slic3r compare above:  You can see the same artifact in the S3D print as well.

The thin wall collapse does help give better definition to features though, compare the portholes in these to Slic3r prints:

slic3r_thinwall_compare2_web

Note how the porthole on the left has better definition.  It looks better than S3D’s as well.

In Conclusion

They’re both great pieces of software, and they both a have many features in my ‘pro’ sections.  But based on the cost involved, how Slic3r has matured, and how few cons Slic3r has in comparisons to S3D’s cons, I’d have a hard time recommending S3D off the bat to someone new to this:  Give the free software a shot.

For myself, I’m definitively going to give Slic3r Prusa Edition a look:  Considering how it addresses all my ‘major con’ issues with S3D (correct automatic supports, properly handles highres mesh, deals with thin edges), I’m going to invest some serious time in it.  I think sometime in the next few months S3D should have another release if they follow years past, I definitively look forward to that will bring.

Finally, I’d like to give some suggestions to the S3D dev’s if they ever end up reading this:

  • Regularly comment on the forums.
  • Expose upcoming features & bug-fixes that are in development:  If I knew all the problems I discussed above would be fixed in an upcoming release, I’d be far less likely to spend the hours I have looking at other slicers.
  • Release more than once a year.
  • Basically, stop being such a black box.
  • Charge a reoccurring fee rather than a single upfront lump-sum:  I’d happily pay, say 5$ a month for an ongoing subscription where the above issues were met.  The OctoPrint Patreon subscription model (that I contribute to on a montly basis) is a great example of this.

Using Autodesk Fusion 360 with the X-Carve CNC & 3D Mesh

Overview

I’d been stuck on an old Macbook Air for all my CAD/CAM work for the past number of years.  And because of that, I was limited to MeshCAM (no disrespect, I quite like it, but one of the few mac apps for CNC I could find) for all my 3d-mesh carving.  I’d dabbled with Autodesk Fusion 360, but the ol’ Mac wasn’t powerful enough to handle it well unfortunately.

Finally got a new PC, and can start to branch out into other CAD/CAM apps.  I’ve been using my X-Carve CNC for just over a year now, been a great learning experience both with its hardware, and the CAM software that generates the toolpaths for it.  And as much as I appreciate how much I’ve learned from MeshCAM, I felt like there could be a more robust solution, and that’s where F360 comes into play.  I am by no stretch any sort of CNC expert, just an avid hobbyist, but I was really surprised how much the F360 team has put into its CAM solutions.

The below post is an overview on my first steps to get it working, specifically for cutting 3d mesh (rather than parametric solids) so I don’t forget what I did in the future.

I should note though, if you don’t care about cutting 3d mesh, this tutorial is still a valid intro to generating toolpaths with F360.

Disclaimer : Not a CNC expert.  Not a F360 expert.  What I list below are ways I figured out how to do things by looking through the docs, and a lot of forum posts.  There may be better ways out there to do any of this.  If so, please let me know!

Note, I use the term “RMB” a lot below.  That means “RightMouseButton”.

Updates:

  • 2017-04-23 : Based on the 2.0.2989 release,  they made a huge improvement to importing mesh for CNC cutting.  While I’m not going to go through this whole post updating it, things worth calling out include:
    • It doesn’t have to be a solid body.
    • You no longer have to convert your mesh to a BRep to cut it.
    • It does seem a little twitchy:  Some mesh will come into the CAM workspace fine, others does not.

Fusion 360 CAM Docs

These are the docs I got the most out of when it came to learning its CAM capabilities.

And an informative tutorial:

F360 Installation

Really nothing to report here:  Install went off without a hitch.  It should be noted that F360 is cloud-app, meaning everything you do gets uploaded to the cloud.  I remember back in the day when it first came out there was a way around this, to only ever work locally.  I’ve not looked into that recently though.

Importing your mesh

F360’s CAM package can’t deal with 3d mesh natively (at least, I’ve been unable to find a way) : You first need to convert it into a solid it can deal with.  And in this case, it needs converted to a ‘BRep’ : I’ve been unable to actually figure out what the definition of a BRep is, but I’m guessing it’s a (solid)BodyRepresentation.  This process probably took me the most time to figure out, not having a broad knowledge of how F360 works.  The trick is, you can’t have any ‘history’ enabled when you do the conversion : The menu won’t even be there if that’s the case.

Note if you don’t care about 3d mesh, and just want to cut your solids, you can just skip this section.

Prep the mesh for import

Before you import/convert your mesh, you should be certain of a few things, based on the way you want to import/convert the mesh:

  • There are no holes in it.  It should be a solid mesh, no non-manifold geometry… quality solid geometry.

Triangulated:

  • It should be triangulated:  While F360 accepts quads or n-sided polys, I found that it’s internal triangulation function, which appears to be auto-ran when converting to a BRep, sucks, causing poorly converted mesh. Pre-triangulate your mesh.
  • F360 will complain (but still do the conversion if you tell it to) if the mesh is over 10k tri’s, citing poor performance to BRep’s.  I presume this is for good reason.  Right now, I just auto-downres my mesh via Meshmixer or Remake.
  • Note, if you want to use this method on mesh that is more than 10k tris, you can cut your mesh up into 10k or less chunks via software like Meshmixer, export all those chunks, then import them all in one at a time via the below instructions.  Super clunky.

Quadrangulated:

  • If you want to get passed the 10k poly limit, you can instead import quad-based OBJ data.  The trick is to first get it converted to a quad obj:  I learned most of this from this video.
  • Very important:  Make sure in your F360 Mesh Preferences, the “Triangulate Mesh Polygons” option is turned off.
  • Based on what I’ve read, you can do this via Remake:  First you should still decimate the model to some reasonable amount (via its Retopologize menu).
  • Then, via its Export Model windows ‘advanced’ tab, you can choose OBJ(Quads) as an export format.  You can additionally reduce the poly count here, but remember the tri count will be 2x the quad count you’re putting in.
  • Also note, this tool likes to blue-screen my PC from time to time :(
  • I’ve tried to get software like Meshlab to convert to quads and export as an OBJ, but F360 doesn’t seem to like that output (maybe I’m doing it wrong…).

Import into F360

Initial import steps:

To import in your mesh, regardless of it being triangles or quads:

  • From the Model Workspace, access Create -> Create Base Feature
  • Insert -> Insert Mesh
    • Browse to the mesh on disk, hit “Ok” in the “Insert Mesh” dialog

If you have triangulated mesh:

  • Once inserted, access the RMB marking menu on the mesh and “Mesh to BRep”
  • Finish Base Feature via the top menubar.

The imported mesh:

meshImport

Via either method, at this point you can delete the original mesh from the Browser.

This will now give you a ‘body’ the CAM package can deal with.

f360body

If you have quadrangulated mesh:

  • Press the “Finish Base Feature” after the mesh is imported.
  • Press the Create -> Create Form button:  This will open the Sculpt Environment.
  • Utilities -> Convert
    • Choose : Quad Mesh to T-Splines
    • Select the mesh
    • Hit OK
  • This will create a new T-Spline body.
  • Press “Finish Form” From the top menubar:  This will auto-convert the T-Spline into a BRep (Body1), ready for CAM.
  • And again, here’s the video link I learned most of this from.

Dealing with scale

Note, that based on where you authored your mesh and the units it was in, you may need to scale it up.  For example, I work in Autodesk Maya in cm units.  I need to scale that mesh up by a factor of 1o either before I export, or once in F360, to get it in mm correctly.

Setting up CAM

First, switch to the CAM workspace in F360.

Manage your tools

Before you can start making toolpaths, you need to tell F360 what cutting tools you have:  Via the CAM workspace, access MANAGE -> Tool Library.

Uncheck all the libraries on the left except “Local”/”My Tools”.  Select “My Tools”.

Press the “New Mill Tool” button on the top bar, and start filling out the info for all the bits you own/want to cut with.

Nice page showing what all the values mean:  “New Mill Tool Reference

But in general the below values are always equal to or less than the one that came before it, from the “Cutter” tab:

  • Overall Length:  Total length of the tool,  say, 2.5″ for a fictional bit.
  • Body Length : How much of the tool sticks out of the chuck, or “Holder” in F360 terms. Obviously variable every time you insert it, but say 1.5″ here.  That leaves 1″ inside the chuck.
  • Shoulder Length:  From the tip of the bit to the last part of the thread, even the non-cutting part of the thread, say 1.25″.  That leaves .25″ between the end of the shoulder, and the chuck.
  • Flute Length:  Length of the cutting surface, like 1″:  There’s .25″ of non-usable cutting surface.

I should also note that with the 1/4″ chuck on the DeWalt611, it has 2″ of internal space for bits.

You can also make “Holders” for your different chucks:  Doesn’t seem necessary, but I did it anyway.

Create a Setup

Setups collect the rules that define the the stock to cut, and the toolpaths to cut in them.

Via the SETUP menu: New Setup : This will immediately open the SETUP dialog.  Below are good defaults I start with, but obviously every situation is different.  I skip over values I don’t change.

  • Setup Tab
    • Setup:
      • Operation Type:  Set this to Milling.
    • Work Coordinate System (WCS):
      • Origin: Selected Point : I pick the point on my mesh in the top of bottom left corner, which is the usual (0,0,0) position for the X-Carve:
      • stockZero
  • Stock Tab
    • Stock
      • Mode : Relative Box Size : This creates stock relative to the size of the model you’re working with.
      • Stock Offset Mode : No additional stock:  The stock is the exact size of the model.

This creates a new Setup for the given solid model.  If you hide the body via the lightbulb in the browser, you’ll see a ghost of the resultant stock:

stockGhost

Create a rough clearing operation

Now that there is a setup which defines our stock and the WCS origin, we can start adding toolpaths.  The order is important, since F360 cuts them in that order in Simulate mode.

Note, F360 has great tooltips, many with pictures:  Hover over the given field and see what pops up.

Via the “3D” CAM button, there are two clearing options available:  “Adaptive Clearing” and “Pocket Clearing” :

  • Pocket clearing is closest to what I’m familiar with in MeshCAM:  It removes material layer by layer, which is good for a light-weight machine like the X-Carve.
  • Adaptive Clearing will attempt to burrow down to the full length of your tool, then start cutting material against the full length of the tool. While this sounds great, I’m not sure how well the X-Carve would handle it, and have yet to test at this point (probably would need to cut very slow…).

Choose the “Pocket Clearing” operation:

  • Tool Tab
    • Tool : Select a tool you defined above.  The below “Feed & Speed” section will be auto-populated with that tools defaults.
    • Feed & Speed
      • Adjust as necessary based on the type of material you’re cutting.  How do you know what to set?  Based on my hobbyist level experience it’s all about learning what others have done, + trial & error.
  • Geometry Tab : Controls what area of the mesh will be machined.
  • Heights Tab:  A nice graphical way to set the vertical areas to be machined, and how far the tool-head should retract.
  • Passes Tab
    • Passes
      • Manual Stepover : If you want direct control over this, you can check this and set the min/max values.  Note that F360 can either use hard-coded values, or expressions for the fields : RMB on the “Maximum Stepover” field and “Edit Expression”:  You’ll see something like this pop up:
        • Math.max((tool_diameter – (2 * tool_cornerRadius)) * 0.95; tool_diameter * 0.20)
      • Change it, edit it, etc, based on your needs.  Here’s an example video on how to do that.
      • Direction : Climb by default, but machines like the X-Carve should have it set to “Conventional”, so the cutting edge bites into the material with the direction of toolhead travel.
    • Stock to Leave:  Here you can set how much stock is left over for the finish passes.  How much should you leave?  I generally make this a percentage of the diameter of the finish bit I’m using.
    • Smoothing : In general it sounds like you want this checked on, check the tooltip.
  • Linking Tab
    • Linking
      • Retraction Policy : I’ve had good success with “Minimum Retraction”.
    • Ramp
      • Ramp Type :
        • If you’re cutting wood, “Plunge” seems good.  If you’re cutting metal, then “Helix” is the way to go.

There are obviously many more options, but the above got me started well.

After you hit “Ok” in the dialog, you’ll see the new Pocket operation in the browser show up : While it’s calculating the toolpath a % value will be visible.  After a few seconds, the toolpath should show up, presuming you have the pocket operation selected.

roughPocket2

 

Create one or more finish operations

F360 comes with quite a few finish operations.  Technically you can use can combine as many as you want, but again, build them in order of operation.  In this example, I’ll use the “Parallel” operation, because it most closely emulates what I’m used to in MeshCAM.

Via the 3D  Menu access “Parallel” operation:

  • Tool Tab
    • Tool : Select the tool for this pass. But default it’ll choose the tool used in the operation before it. Note if you choose a different tool, obviously you’ll need to go through a toolchange operation (which on the X-Carve is a very manual step), and need to save out your gcode as multiple files.
  • Geometry Tab : Controls what area of the mesh will be machined.
  • Heights Tab:  A nice graphical way to set the vertical areas to be machined, and how far the tool-head should retract.
  • Passes Tab
    • Passes Options to control the angle of the surface to be machined, the stepover, etc.
      • I like to set my stepover to be 10% of the bit width to give a nice, smooth finish, which you can set as an expression by RMB on the “Stepover” field & “Edit Expression”:
        • tool_diameter * .1
    • Smoothing : Like in the rough cut, this seems good to check on.
  • Linking Tab
    • Linking
      • Retraction Policy : I’ve had good results with “Minimum Retraction”.

After you hit “Ok” in the dialog, you’ll see the new Parallel operation in the browser show up : While it’s calculating the toolpath a % value will be visible.  After a few seconds, the toolpath should show up, presuming you have the parallel operation selected.

parallelFinish

Like mentioned above, you can add more finish passes using different techniques here as needed.

Name your setup and operations

It’s a good idea to name your setup and operations, so you know what you were up to months later.  Click once on a name in the Browser, then click again after a second: It’ll let you rename them:

passNames

Save operation templates

Presuming you get values you like and want to resuse thes on other cuts later, there are two ways to access the data:

  • You can open a previous cut, then RMB on any operation and “copy” it.   You can go to your other cut’s setup, and “paste” it.  Works, but clunky.
  • Better, is to select one or more operations in the browser, and RMB on them -> Store as template : This will give you a broswer to your local HD for storage, and later import via the “Create from Template” Setup RMB menu.

Simulate / preview the cut

This is an incredibly powerful part of F360:  You can pre-visualize your cut.  Either select your setup folder in the Browser, or multi-select all the operations you want to preview (basically, whatever toolpaths are shown will be simulated), and press the  ACTIONS -> Simulate button.

In the SIMULATE popup, check on the “Stock” box.  This will show you the uncut stock.  You can uncheck the “Toolpath” if you don’t want the lines to obscure the view.  Then simply press the > Play button at the bottom of the screen:  All the selected operations will begin previewing their cut:

cutPreview

Green is the roughcut, blue is the final pass.

Using the correct postprocessor

For F360 to generate the correct gcode/nc data, it needs to be post-processed for your given machine.  Since I’m using an Inventables X-Carve (grbl-based), I had to do some searching.

This section has been edited since the original blogging:  Originally I’d found a couple postprocessors online, and they “worked”, but had a “bug” where at the end of the cut, they’d send the machine back to the “machine home” (where the toolhead was when the machine was powered on), not the “work home” (the zero/home position of the cut).  This caused two failed cuts for me as the tool-head ripped into my stock trying to go somewhere bad.

After troubleshooting and posting on the forums, I was made aware that F360 comes with a ‘generic grbl postprocessor’, that needs only one small adjustment.  I’ll show that info below.

Generate or “post” the gcode/nc data

Once you’re happy with your toolpaths, you can ‘post’ the data for your CNC to use.

The big decision here is how you’ll save the operations you’ve made previous:  If they all share the same bit, you can select them all at once in the Browser, and save a single file for all of them.  However, if toolchange is needed, you should save each operation as a separate file, selecting one at a time and going through the below steps.

Via the ACTIONS -> Post Process Menu:

  • Setup your postprocessor.  Since I’m using an X-Carve that runs grbl:
    • Setup : Use Generic Posts
    • In the Post Configuration section, select “grbl.cps – Generic grbl”
    • In the “Property” section in the bottom right, make sure that “useG28” is “No”.  If this is “Yes”, it’ll return to machine home (instead of work home, where you zeroed the cut) upon completion of the job, discussed above.  If this is “No”, it simply raises the toolhead and leaves it there.
    • grbl_post
  • Define the Output folder where you want the .nc data saved.
  • Under Program Settings, set the name of the file under “Program name or number”.
  • Press the “Post” button:  The .nc data is saved to disk.

Cutting with the data

There are number of sender software available to send the .nc data to your machine:  Easel (browser-based) can do it, I’ve had success with Chlipeppr (browser-based), I hear good things about LaserWeb/CNCWeb (browser-based, not yet tried), but currently use Universal GCode Sender (Java applet)  for all my work.

The F360 data didn’t cut any differntly in UGS than MeshCAM data, and had great result:

finalCut

In fact, the few hard-edges around the hemisphere are due to a pencil-cut I was toying around with as an extra pass.

In Conclusion

I feel F360 provides a tremendous amount of power to the hobbyist CNC user, especially considering the price (free).  The learning curve is a bit steep, but it has a large amount of helpful documentation, and videos available.  It’ll be my go-to piece of CAM software in the future, hands down.

 

Customizing Oculus Medium

Over the holidays I picked up an HTC Vive, and the first thing I did was jump into Google Tiltbrush and Oculus Medium : Both awesome VR painting apps, both very different.  This post will collect my (ongling) notes on Medium.

First tests

Here’s a couple of kitbashed robots that took me about three hours.  So much fun!

kitbash_robo_1 kitbash_robo_2

Installation

To get Medium working on my Vive, I had to setup an Oculus account (+ purchase Medium) and install “Revive” : Both painless.

Documentation

Mapping the Oculus controllers to the Vive

At first I had a lot of trouble trying to figure out which Vive buttons mapped to the virtual Oculus controllers.  Couldn’t get anything to reliably work.  I finally read you just touch on the thumbpads where the corresponding Oculus buttons are, and it ‘just worked’.  Suddenly became usable.

Installing custom brushes

You can install your own custom brushes in Medium. Here are the rules (many pulled from the above FAQ):

  • Must be saved as .obj
  • Must be triangulated.
  • They need to be saved here:
    • C:\Users\USERNAME\Documents\Medium\_Import\Stamps
  • In Medium, open the Main Menu, navigate to the personal tab, and select Stamps. At the bottom of the window, you’ll see an import button.
  • They’ll then show up in the ‘Custom’ stamps section.
  • Should be as low-poly as possible.
  • Scale doesn’t appear to matter:  I made test with different sized models.  My guess is that Medium auto-scales them to fit some standard sphere that is used on the brush.

Medium’s coordinate system

I did some test by bringing in three different cones into Medium each build along a different axis.  I was able to then reverse-engineer the coordinate system relative to the sculpt tool.  You can use these notes when building your own brushes, to make sure they show up in Medium correctly.

I made a handy infographic:

mediumCoordSystem

X is towards you, Y is towards the controller, and Z is up.

This also means that when you create stamps that you want to project onto another surface, you should orient them so that their ‘up’ vector aligns with the +Z worldspace axis in your 3d software.  This will thus orient them properly onto the surface in medium.

Performance Issues

I’m running a GTX 1080 GPU, i5 CPU, 32 gigs of ram : The software will definitely stutter if I paint large items, or paint quickly.  If I paint one stamp at a time it doesn’t seem to be a problem.  Also seems to loose tracking far more often than other software:  I can watch the sculpt hand just ‘float away’ sometimes.

New CNC Cut : HexBeam

Continuing to play more with MASH in Autodesk Maya, I came up with this experiment:  I used a ramp node with a wave to mask where the hexagons are placed, then randomized their scale.  Applied a dark stain on the uncut top, and a natural stain on all the hexagons.  Material is a reclaimed redwood beam.

hexbeam

About half an hour of modeling in Maya, and 1h45min on the X-Carve CNC with a 1/8″ 1-flute upcut endmill.  Probably another half hour of sanding.