I Revived 3-Axis CNC Mill G-Code Simulator
(github.com)142 points by harvie 10 months ago
142 points by harvie 10 months ago
In general, FreedCAD has 3-axis tool path support, and I've met the kind local guy who started the original Path toolbox.
For 2.5D there are several engraver options with the inkscape mightyscape plugins (scorchworks etc.)
YMMV with the island-routing and surface probing routines:
https://github.com/pcb2gcode/pcb2gcode.git
https://github.com/pcb2gcode/pcb2gcodeGUI.git
Also, we are currently evaluating: viaconstructor, inkcut, GCAM ( https://github.com/blinkenlight/GCAM.git )
If you want 3D contouring operations, than you could also try:
git://pycam.git.sourceforge.net/gitroot/pycam/pycam
And note, if you patch CAMotics to compile on ubuntu 24.04.x it still has a number of Qt5 GUI problems (not entirely unexpected with Qt impact on LTS programs.)
There are also direct gcode generator macros that support the LinuxCNC/emc language extensions. This is the most accurate methodology for turning and milling ops.
Best of luck, =3
Opensource options include:
- Solvespace --- limited to 2D last I checked
- FreeCAD has a workbench for CAM/toolpath generation
- BlenderCAM is a plug-in for Blender which is well-regarded, and together w/ CADsketcher/BlenderCAD works well for some folks
- Kiri:Moto
- pyCAM --- a venerable option, it worked well ages ago when I used my son's gaming computer to make toolpaths.
Rather rough (possibly outdated) list at: https://old.reddit.com/r/shapeoko/wiki/cam
I've been long-term contributor to bCNC https://github.com/vlachoudis/bCNC/ But FreeCAD seems to be interresting choice lately as well...
Very cool.
It seemed to struggle a bit with a more complex example.
N30 PAN=1 ST1="16MMRH" L=PCUA
N40 Z=PRK
N50 X1846.942 Y1157.433 G42 F=1.0 S20000 VF=1.0 AX=X,Y,Z PRF=-12.7 TP=1 L=PON M55
N60 G1 X1832.8 Y1171.575 Z=-LPZ+-12.7 F=1.0
N70 G1 X1832.8 Y1187.575 Z=-LPZ+12.2 F=1.0
N80 G2 X1807.8 Y1218.8 I=1839.8 J=1218.8 F=1.0
N90 G1 X1.2 Y1218.8 Z=-LPZ+12.2 F=5.0
N100 G1 X1.2 Y757.2 Z=-LPZ+12.2 F=5.0
N110 G1 X1807.8 Y757.2 Z=-LPZ+12.2 F=5.0
N120 G1 X1807.8 Y1218.8 Z=-LPZ+12.2 F=5.0
N130 G1 X1799.0 Y1218.8 Z=-LPZ+12.2 F=5.0
N140 G2 X1767.775 Y1243.8 I=1799.0 J=1250.8 F=5.0
N150 G1 X1767.775 Y1243.8 Z=-LPZ+-12.7 F=5.0
N160 G1 G40 X1778.481 Y1260.693 Z=-LPZ+-12.7 F=5.0
N170 G0 L=PSU
N180 X1846.142 Y1156.633 G42 F=1.0 S20000 VF=1.0 PRF=-12.7 TP=1 L=PON M55
N190 G1 X1832.0 Y1170.775 Z=-LPZ+-12.7 F=1.0
N200 G1 X1832.0 Y1186.775 Z=-LPZ+12.2 F=1.0
N210 G2 X1807.0 Y1218.0 I=1839.0 J=1218.0 F=1.0
N220 G1 X2.0 Y1218.0 Z=-LPZ+12.2 F=5.0
N230 G1 X2.0 Y758.0 Z=-LPZ+12.2 F=5.0
N240 G1 X1807.0 Y758.0 Z=-LPZ+12.2 F=5.0
N250 G1 X1807.0 Y1218.0 Z=-LPZ+12.2 F=5.0
N260 G1 X1798.2 Y1218.0 Z=-LPZ+12.2 F=5.0
N270 G2 X1766.975 Y1243.0 I=1798.2 J=1250.0 F=5.0
N280 G1 X1766.975 Y1243.0 Z=-LPZ+-12.7 F=5.0
N290 G1 G40 X1777.681 Y1259.893 Z=-LPZ+-12.7 F=5.0
N300 G0 L=PSU
N310 G0 L=POFF
N320 PAN=1 ST1="FINTOOL" L=PCUA
N330 X1817.607 Y747.3934 G42 F=3.0 S18000 VF=1.0 PRF=-12.7 TP=1 L=PON M55
N340 G1 X1807.0 Y758.0 Z=-LPZ+-12.7 F=1.0
N350 G1 X1807.0 Y758.0 Z=-LPZ+12.2 F=1.0
N360 G1 X1807.0 Y1216.5 Z=-LPZ+12.2 F=3.0
N370 G3 X1805.5 Y1218.0 I=1805.5 J=1216.5 F=3.0
N380 G1 X3.5 Y1218.0 Z=-LPZ+12.2 F=3.0
N390 G3 X2.0 Y1216.5 I=3.5 J=1216.5 F=3.0
N400 G1 X2.0 Y758.0 Z=-LPZ+12.2 F=3.0
N410 G1 X2.0 Y758.0 Z=-LPZ+-12.7 F=3.0
N420 G1 G40 X-8.6066 Y747.3934 Z=-LPZ+-12.7 F=3.0
N430 G0 L=PSU
N440 G0 L=POFF
N450 G0 Z=PRKhttps://nraynaud.github.io/webgcode/
I made one too, a long time ago.
https://nraynaud.github.io/webgcode/webapp/visucamTest.html
I did some code generation too, but it's bit rotting.
I have a post-processed NC file from Fusion 360...which I think is just gcode. How can I transform that into something this can simulate? My hacky attempts have failed.
This is awesome. I learned CNC & manual G-Code on a Rockwell 3 axis in ~2000 and I am taken back tinkering with this.
The big problem for contemporary hobby CNC is that the G-code interpreters for small single board computers such as Grbl running on an Arduino are extremely memory-constrained:
which means:
- no loops
- no branching
- no variables
One work-around for that is:
https://github.com/NRSoft/GSharp
(I took a very different approach in my project linked elsethread)
I don't understand this comment- gcode never had loops, branching, or variables. And most folks use more capable controllers now based on ESP32 or whatever.
Update: I guess gcode does have loops, branching and variables: https://www.fictiv.com/articles/cnc-machining-macros-subprog...
Actually, in LinuxCNC dialect the gcode does have many features most machines will not understand.
This is why these machines can drill and helical-tap plates with thousands of holes with ease. Also, some folks will cut gear-profiles with a slitting saw, and 4th rotary axis...
https://www.youtube.com/watch?v=HI27vSoxCeo
Trying to pull these things off in a CAD/CAM package is usually a fools errand the old guys prank kids with... oh that sounds easy... lol =3
I used to have some machines with Fanuc 6 controllers circa 1981 - 1983. They had variables, macros, branching, etc. I haven't been able to find what processors were in those old controllers, but I'm certain the single board computers of the last decade have more ram, more flash, and way more processing power. If GRBL can't do it, it is a software issue, not hardware.
They are the high-end/legacy-end of hobby CNC.
The Grbl/Arduino stuff seems most prevalent/commonly sold.
That said, I was remiss in not mentioning LinuxCNC, &c., and as you note, they do have a more traditional G-code implementation w/ loops and variables and branches.
But, I don't think that this simulator supports all of that --- could someone post a sample program which does that sort of thing and test it?
One thing I'll suggest - if you ever want to cut anything metal, you'd probably do well to learn how machining works in general and by that I mean on a manual lathe and mill. The coding part is going to be easy for you (and probably anyone else reading this), but knowing what you want the thing to do is a MUCH different task than being able to tell the machine what to do.
Everything about how you move that tool through the work matters. Materials, workholding, spindle speeds, tooling, tooling coatings, feed rates, plunge speeds, all of these things need to be worked out and they all change when you swap a tool or a part etc. Speeds and feeds are dynamic, you need to learn about conventional-vs-climb milling, you'll need to know about cutter geometries and the materials the cutters are made of, when and how to deploy coolant, the list just never ends.
GCODE is, by far, the easy part for a programmer. It's a very simple language that hasn't changed much in the past several decades. "Knowing machining" is that tougher hill to climb.
My recommendation? Check your local community colleges/trade schools! I took a couple semesters of machining classes as a 40-something grown-ass dude and I learned a huge amount of stuff that would have been expensive or dangerous to learn by trying. You'll get hands-on access to big machines with someone over your shoulder helping to make sure you keep your limbs attached to your body.
Second this. My work occasionally deals with Gcode files. The coding is simple, knowing what to do (or not to do!) is harder. My first foray into the world of Gcode I named "Smokey" because it was a quick fix for another program generating bad Gcode that quite literally caused smoking on the work surface and could have lead to a fire had the operator not been paying attention.
And note that while Gcode is consistent it's meaning isn't always--is Z positive going into the work surface or going away from the work surface?
One would think so, but apparently it varies between manufacturers. Our machines are all carving up a flat surface resting on a table. I think what's going on is that functionally that means that the working head is permanently upside-down. Is Z measured in the context of the whole machine, or is Z measured in the context of the cutting assembly?
If you literally want to emboss hello world on a piece of foam, you can buy cheap CNC kits on Amazon like https://www.amazon.com/dp/B08HRLGL6C/ which will execute G-code.
They're weak and imprecise by the standards of CNC machines but if you're only looking to learn a bit and cut foam? That's a low price.
That's an excellent question.
Also, how are tool definitions to be handled? Each program seems to have their own approach, though some will place the tool description in comments. CutViewer is notable for doing this, and is the example I have been following in:
https://github.com/WillAdams/gcodepreview
(which is quite a different project)
EDIT: Note that the comment system which Cutviewer uses seems to be something of a de facto standard --- Carbide Create uses this same scheme.
FreeCAD and its CAM Workbench?
For the light version maybe Solvespace?
I'm still amazed by Dune 3D: https://dune3d.org/ where the Github page has the footnote:
>I ended up directly using solvespace's solver instead of the suggested wrapper code since it didn't expose all of the features I needed. I also had to patch the solver to make it sufficiently fast for the kinds of equations I was generating by symbolically solving equations where applicable.
Dune 3D is made by the same guy who did Horizon EDA (https://horizon-eda.org/). A really talented and productive person.
Nice and Thanks! Was looking for something like this with minimal dependency.
By the way, changing options updates browser history every time. You may want a debounce. Also, you could remove jQuery and make it even more leaner.