|Figuring out requirements and workflow restrictions for a CNC router|
|User:Ivom, Jeffrey, Yashiko, Jobj, Yvanj, tomvdz|
- Milling regular PCBs with non-SMD IC components should be possible with are fair margin
- Bits for milling MDF, stryrofoam HDPE and nylon should be possible
- Sub 1000 euro budget limit, including a set of drill bits matching our initial goals
5 crucial project ideas to pull this off
- cutting PCB circuit boards (resolution limits between two layers of 0.2 mm and 0.8 mm)
- prototyping custom boxes for the Nanode and Arduino kits people own
- engraving nametags, labels, equipment
- the Voidwarranties logo is the first to go: online barcode creator,
- transform the image to dxf
- mill it on acryl add some neon or LED-light
- mount it outside the josto-building
- cutting out mechanical parts for robottics.
- for furniture restoration: 3D scanning woodcarvings and recreate missing pieces.
- making models, molds, spare parts wich could be hard and/or expensive to come by.
- miniature furniture for the doll house, then scale it by a factor 10 and, goodbye IKEA shopping horrors.
- de houten spatel van de space is serieus aan vervanging toe en mag gelijk van het VW logo voorzien worden. Als het werkt millen we er een 20-tal.
Oops that make 8 :)
Meta project level
- the shear pleasure of learning new skills and the satisfaction to make it
- actually work and implement your design (-> insert education workshops here)
- researching the feasability to convert the cnc router to a touchsensor 3D scanner
Please add additional thoughts on Talk page
Selection: MiniCnc, made at TimeLab Gent
- Q: Minimal distance between tracks when milling PCBs?
- A: About 0.2 mm
- Q: What host software is used to transform CAD designs to G-code?
- A: Replicatorg with skeinforge as plugin software and a custom http://www.repairablemachines.com/usingcadpy.html
- Q: What kind of PCB design is up for testing
- A: find the answer with the picture on the right
Research by Jeff
- I found some interesting links about milling PCB's:
The last days we've discovered some problems when milling PCB's. The accuracy of the CNC router seems to be pretty good, however it seems that our current milling bits are not suitable to perform the job. The ball point milling bit gives smooth tracks without sharp edges, but the milling width is heavily influenced when the PCB surface isn't flat. The other fine milling bit could not deliver smooth tracks due to sharp edges.
Another problem is the production of fine dust (and noise). It's certainly not healthy to breath in small copper particles so we really need to make an isolated box asap.
Thanks to the great insights of Kris Wauters I suggest the following todo's:
- Buy/order better milling bits (done - Jeff)
- Buy/order new clean (and straight) PCB's (done - Jeff)
- Make a flat bottom surface. Please notice that 'flat' is relative to the CNC machine & milling spindle (Dremel). So the best way to do this is placing the CNC router on a new MDF panel (18mm minimum) and mill about 1mm of the surface away. Then we are sure that the relative height/depth is the same for each possible coordinate. We need the following:
- MDF panel (700 x 700 x 18 mm) (or thicker)
- milling bit for MDF, flat top (10 a 20mm milling diameter)
- Find a better way to fix the PCB's. The best way seems to be using double sided tape, egally applied to the bottom PCB surface. Then the PCB is attached to the flattened (by milling) MDF surface. The tape will help in straighten the PCB surface by pulling it against the MDF surface.
- Buy strong double sided tape (not very thick)
- Design wooden box to place over the CNC router. The box should not be attached to CNC router itself, but I guess we can put it loosely on the MDF bottom plate. We can use small wooden beams on the bottom plate to assure that the box and the CNC router are always be placed on the same position.
- design box
- buy MDF plates
- buy wood glue
- buy wooden beams (10x10 or 15x15 mm)
- use plexiglass as lid? ("deksel")
- buy screws?
- design a panel to prevent dust from milling reach into electronics area
- measure and design panel (done - Jeff)
- create panel (planned - Jeff)
- use some kind of cleaning device to remove possible toxic dust
ivom, tomvdz, sorry-forgot-your-name
- checked and fixed all wiring
- first attempt to draw a square
- the beast works!
- Installed the Dremel
- Calibration night:
- The square wasn't a square
- tweaking parameters in the machines/minicnc_rep25.xml file
- calibration done, error deviation is about .2mm
Jeffrey, tomvdz, ivom
- accuracy testing on pcb milling
- test 1
-spheric milling bit, 1mm -speed F66 -depth -0.1mm (0.2mm will be fine too) -accuracy: very fine to 1mm distance between lines -no rough edges
- test 2,3,4
-pointed drilling bit -speed F66 & F30 -different depths: from 0.2mm to 0.4mm -accuracy: +- 0.25mm, but edges are too rough and unclean
- used G-code
G21 ; millimeters of course! G90 ; absolute position, no opto-end stops... ;G91 ; relative position, no opto-end stops... G92 X0 Y0 Z0 ;huidig pos = oorsprong G00 Z05.000 ;boor omhoog G1 X10 Y10 F66 ;begin pos 1ste lijn G1 Z-0.4 F66 ;boor omlaag G1 X30 Y10 F30 G1 X30 Y20 F30 G1 X10 Y20 F30 G1 X10 Y27 F30 G1 X30 Y27 F30 G1 X30 Y32 F30 G1 X10 Y32 F30 G1 X10 Y34.54 F30 G1 X30 Y34.54 F30 G1 X30 Y37.08 F30 G1 X10 Y37.08 F30 G1 X10 Y38 F30 G1 X30 Y38 F30 G1 X30 Y38.75 F30 G1 X10 Y38.75 F30 G1 X10 Y39.25 F30 G1 X30 Y39.25 F30 G1 X30 Y39.5 F30 G1 X10 Y39.5 F30 ;G1 X10 Y110 Z-1.0 F66 G1 Z+5.0 F66 G1 X0 Y0 Z+5.0 F66
- Build a wooden case
- Build a seperator to insulate the electronics from dust
- measured the maximal range (X;173mm, Y:179mm)
- recalibrated the tension on the belt, the axis weren't true relative to one another
- tried to fix the third motor controller above the second one using 'draadstangen'
- mounted the third motor controller using draadstangen
- tried to engrave tux in scrap wood, which failed miserably
- increased the tension on the belt
- while doing the former we adjusted the way the tension bolts are attached
- increased the voltage on the Y axis motor
- second try at engraving tux, gcode is at http://pastebin.com/g6gK413Y#
- got loats of errors about a feedrate of zero, this is normal acording to https://groups.google.com/group/ultimaker/browse_thread/thread/6892f2b14d11b361 becaus of no feedback mechanism
- We didn't have enough time to let it mill the whole image, but we still saw that the result was quite accurate, exept for some strange errors. These were error always faded in or out, indicating variance on the Z axis. We suspect that the mill doesn't wait for the drill bit to reach traveling height before moving to another X,Y coordinate.
- This is probably caused by errors in the gcode outputted by cad.py (version 16/5/11 kurt)
- The error seems to be a missing G01 statement when raising to travel height:
G01 X47.4486 Y61.9125 Z-0.3000 Z1.0000 G00 X51.6819 Y72.4958 Z1.0000 G01 Z-0.3000 G01 X52.3875 Y72.4958 Z-0.3000
- Third attempt at engraving tux in scrap wood, gcode at http://pastebin.com/ee8g8e0T
- Gcode was produced by cad.py (16/05/11, kurt) and corrected by the sed line `sed 's/^ Z/G01 Z/'`
- Engraving details: used a 1mm diameter round milling bit; dremel set a speed 4; depth 0.7mm; distance between paths 0.5 times the tool width
- both the gcode and the machine worked fine, leading to satisfying results, however some lesser issues were apparent
- the engraving depth was deeper at one side than at the other, which means the base plate is not parralel to the X,Y plane. There seems to be some glue residue on the base plate which we should sand off, and test afterwards if both planes are parralel
- paths need to be closer together than 0.5 times the tool width to get an even engraved surface when using a round milling bit to get an even engraved surface (striping was visible
- photographs of the last 3 attempts were taken, and should be uploaded soon
- started a test run of the VW logo in foam. As the goal is to engrave it 300 mm wide the design was sliced in two. the left part was converted to gcode using cad.py and corrected with sed. We used a 1mm engraving bit. The left part took 94 minutes to engrave (and counting)
- we will need real endmills to do any serious work
- cad.py doesn't optimize the order in which the paths are engraved, which makes the engraving time a lot longer than it needs to be
- I looked into pycam, a tool which is looks promissing.
- Found a nice guide made by a guy who makes robotics components in his garage. He mills out the parts, makes a silicone mold and then casts the parts in PU. Buildlog: http://www.flickr.com/photos/candy-bottoms/sets/72157625850569320/detail/ Guide: http://lcamtuf.coredump.cx/guerrilla_cnc1.shtml and http://lcamtuf.coredump.cx/guerrilla_cnc2.shtml