Machsupport Forum
G-Code, CAD, and CAM => G-Code, CAD, and CAM discussions => Topic started by: lukeroywhite on October 16, 2012, 07:52:05 AM
-
Hi,
This is my first post on the forum so i hope I have put it in the correct place!
I have been struggling with the attached set of programs. The show the outline of a guitar body with 4 symmetrical tabs at its extents. When the 'outline' code (attached) is machined the body it creates should fit into some MDF blocks that are mounted to the bed and then trimmed down using the 'location' code also attached.
When machined the body fits exactly one way round but when flipped won't which i guess shows asymmetry. The CAD model has been checked and i have printed a full size outline to double check which all measures up perfectly symetrical.
The code looks symmetrical on the mach 3 toolpath screen, & i have also overlaid the 'outline' code with the same path but machined from the other side and these join up on the toolpath screen.
Can anyone suggest where this inaccuracy could be coming from? The inaccuracy is in the region of 1-2mm & it is cut the same every time whether cutting through 5mm of wood or just marking the surface of the MDF bed.
I have also attached the xml i am using.
Any help is greatly appreciated!
-
Could be inaccuracies in your machine.
Gantry or bridge type router ?
Out of square ?
Backlash in any axis ?
Can you post the DXF ?(it's probably OK)
More details about the machine would help.
Russ
-
Hi russ,
Thanks for the input. The machine is basically as follows just slightly larger.
http://www.worldofcnc.com/3-axis-a3-ballscrew-assembled-cnc-router-package-p-1318.html
I have had good sucess machining other items acccuratly.
I don't have any allowance for backlash, and don't know how to measure if this is a problem can you suggestr soemthing?
A dxf of the outline is attached
Thanks very much
Luke.
-
The DXF is symmetrical.
Backlash is probably not the problem, it would be minor.
If it shows up with light cuts, it's probably not tool deflection.
I would double check for squareness.
With a precision square or similar to what you have already done.
Mill out a perfect square block, measure it, mill a matching square pocket, then flip the male part and see how it fits in the pocket.
It will show any misalignment.
You can also clamp a good square on the table, indicate one edge true to an axis, then check the other side with the indicator.
Russ
-
Hi Russ,
I have checked for squareness and machined a block that flipped perfectly.
It is probably also worth mentioning that this has happened across two very similar machines (from the same manufacturer) Running virtually identical mach 3 settings (only the bed dims are slightly diffrent in softlimits)
Could it be something to do with the mach 3 to controller box interface?
Luke.
-
If the machine is square and the files are correct, you might verify that you are not loosing any steps during machining by checking that the axis' return precisely to a known REF point after the file runs. If all is good there (likely so as it is repeatable between 2 machines) then I would look more closely at the 2 cut files. Some here can back plot your .tap's or you can post the other DXF so we can check for any anomalies.
Russ
-
Hi Russ,
I don't believe that any steps are being lost as no matter how i try and cut the location feature (contours or pockets) they line up after having returned home... is that correct or could this still not be showing up?
I have atatched the DXF that would be used to generate the code for the location features.
Thanks.
-
It could be that the CV function in MACH3 is altering your code just a squeak. You r ACCEL rate may be a bit to slow and mach3 cannot create the precise contour needed while in CV mode.
Try cutting it in exact stop mode to verify.
(;-) TP
-
Hi Luke,
The DXF's look to be fine (other than they show up one twice as large as the other, and very large dimensions)
Noticed also that both of the .tap files have the exact same coords at the outer flats of the tabs indicating there is no offset to allow for the cutter dia.
Not sure where to look now.
-
Hi TP & Russ,
I have tried to verify the cut running in exact stop but still following exactly the same path!
Is it worth playing with the acceleration and velocity settings beyond this?
Also looked at a PDF about cv mode and my CV dist tolerance is 180units is this high would it have any effect?
Any other ideas?
-
I checked teh DXF file it is fine , I checked the Gcode file it iooks ok, I converted the Gcode file back to a dxf and verified it is fine as well.
That eliminates the drawing and the Gcode.
That leaves the machine and mach3. Mach3 is NOT known to be wrong in its geometry. Are you running any FORMULAS in MACH3??
That leaves out of square machine, lost steps, Cutting deflection, Slow ACCELL for feedrate causing CV to compensate.
BUT you say you have veried the square and deflection and exact stop cuts the same.
NOW they ALL can't be right and it still cut wrong (;-). I would go back and check everything machine related step by step until you find the problem.
(;-) TP
-
I have re checked square and all is ok. I don't know how to find out if i am running a formula, could you advise...
-
Check here.
-
Please explain HOW you checked square.
(;-) TP
-
Hi,
No formulas enabled. I checked square by machining a cross into MDF bed that runs the full length of x & y then putting a straight edge along one of the lines and a square against the straight edge, lined up with the machined line along the other axis. We are having the same problem across two machines bought 2 years apart so would not expect the same machine build problem to have gone unnoticed (hopefully!).
I was thinking about the loosing steps comment earlier. When i have machined something and returned to home and machined again it has followed the same path does this verify that this is not a problem? - it follows the same path what ever shape i cut and whether it is from the mahine home as x 0 y 0 or an offset somewhere else on the bed...
thanks
luke.
-
IF you have home switches then after a program run you can do VERIFY and mach will return to the home switches to verify you have not lost steps.
HOW fast do you normally cut ?? What are your Motor tuning values Vel? accel? We are down to MACH3 CV altering the path due to slow acceleration vs Feedrate. It will be MORE prevalent on curvy stuff rather than Straight sided stuff such as squares.
HOW far OFF is the cutting? tenths, hundreths, thousanths ?
IS it always to BIG or to small?
BUT let me re verify the problem. You cut the guitar body out then Flip it over and set it into a fixture to cut other features and it does not fit in the fixture ? How did you cut the fixture to hold it?
SHooting in the dark, (;-) TP
-
OK will try that.
Normally cut around 1000 mm/m up to 1500.
Motor tuning values as shown in attached pic. (velocity 1500 acceleration 750)
when we try and flip it fits perfectly one way but the other way up the part is between 0.5mm & 1mm out from fitting always in the same place.
To verify, we cut the body. then fix 4 MDF blocks securly to the bed and run another cut which allows for the 4 tabs around the body to slot in which shoudl locate the body facing up or down. do you want me to put a pic up of the process?
Thanks
-
Just ran a cut around the outside of the guitar body and the VERIFY gave these readings
X = 0.02000 Y = 0.0000 Z = 0.0160
-
Hi Guys,
I have just re modelled the guitar body with a slightly different outline but with the tabs that locate it in the same place these are the same size as before. The cut followed the tabs as before and the curvy part of the body outline changed as it had on the new model.
Any furthur input would be greatly appreciated.
-
I don't have any allowance for backlash, and don't know how to measure if this is a problem can you suggest something?
Luke.
Back to the backlash, I'd check to see if there is any on the X and Y.
Several methods are posted, a search will reveal.
Russ
-
This is probably the most accurate method posted:
http://www.machsupport.com/forum/index.php/topic,15902.msg107130.html#msg107130
Thanks Jeff,
Russ
-
Just a pure shot in the dark here…
I am not sure about the MarchantDice machines but some Chinese machines are not fitted with proper, preloaded, thrust races on the ball-screw at the stepper end. They rely on the stepper to take the load which, as we know, usually has a spring washer mounted bearing and thus it’s shaft can move axially under load.
This problem does not show itself as ‘backlash’ because the movement only occurs under reasonably high load – could this be the issue ??
Tweakie.
-
Hi Guys,
So we have done the backlash test as described the results are as below:
Y axis
we asked it to move 0.004mm (4 steps at 0.001mm)
on the fourth step the dial gauge moved about 0.001 does this mean there is a backlash of 0.003? is this enough to notice the sort of inaccuracy we re seeing in the part (about 0.5mm over 350mm).
X axis
we asked it to move 0.003 (3 steps at 0.001)
on the third step the dial gauge moved 0.001
Thanks
Luke.
-
Hi Luke,
I'd say the backlash is OK. Especially if you are in MM and not inches. Resolution plays in but you should be good there.
Next thing would be to verify squareness. Indicate one edge of the most precision square you have true to x or y and clamp it to the table, then check along the other edge by manually jogging.
Or, if you have a good straight edge, you can place 2 sheets of material on the mill (as large as practical) atop of one another, mill 2 adjacent sides untill they clean up. Un-clamp and flip the top piece (book-match) and check the other 2 milled edges with the straight edge. Done carefully, a good straight edge will display squareness to within the amount of error you are seeing.
A good (large) machinist would be best though.