[racedirector]

I need a little guidance on where to start looking for the solution to this problem...

I started to cut out my machines replacement and found a problem with the finished result. The file was to cut out 2 identical parts, each part was orientated so it was on opposite sides of the table in one sheet 1200x600. The part that came out correctly was furthest from 0,0 side and the other was out by 1 -2 mm in sizing, in the positive direction (that is larger than the first), this part was closest to the 0,0 side of the table.

Both parts are identical in size in the dxf, infact that are mirrors of each other.

I have been racking my brains to try and figure out what may be causing this but am at a loss and I am hoping someone will turn a light on in my brain so I know where to start.

I am using Mach3 in MM mode, the drawings were converted from imperial by scaling them by 25.4. This resulted in some measurements being like X345.88798 if that makes any difference.

I am going to cut the same file with LinuXCNC to see if it makes any difference, clutching at straws I know......


Anyway, any ideas or pointers would be greatly appreciated....

Cheers
Bruce

[Walky]

Maybe your X/Y axis are not perfectly perpendicular to each other? This can cause problems with mirrored parts (Actually, all parts would come out slightly deformed, but it's usually easier to notice if you try to match a mirrored pair) .

[racedirector]

Hmmm, thanks Walky. Will check that out before firing up LinuxCNC

[racedirector]

Ok, checked everything out, made a few adjustments and cut the panel out again. Got exactly the same result. I now have 2 matching pairs of parts but the pairs are different from each other! They *should* all be the same I can this driving me absolutely nuts!

[OCNC]

Ok, checked everything out, made a few adjustments and cut the panel out again. Got exactly the same result. I now have 2 matching pairs of parts but the pairs are different from each other! They *should* all be the same I can this driving me absolutely nuts!

Can you post the g-code file?

I can run it in NCPlot and see if there is anything going on in the g-code that is causing this. That way you can pin it down to either the machine or the g-code generator.

Chris

[racedirector]

Thanks Chris, gcode attached. Had to zip it due to invalid file message. This was produced by CamBam 0.98L

My sanity is at stake here

Cheers
Bruce

[OCNC]

Bruce,

I ran your file in NCPlot and I didn't see anything that would indicate a g-code issue. I also exported the g-code from NCPlot as a dxf file that I then opened in Rhino. In Rhino I measured the length and width of the tool paths for the mirrored parts and they are identical to a 100th of a millimeter.

What is your machine like? Drive method, bearings, etc.

Chris

[racedirector]

Hi Chris

Thanks for doing all that, I really appreciate it. Bummer of sorts that you found that out though I was sort of hoping it was a processor issue.

My machine is probably not the most accurate in the world, was hand built from plans here. The drive mech is 1/2" acme 1 start 6tpi.

I have the settings in Mach and LinuxCNC set at 378 steps / mm using 8th microstepping on Keling 425oz-in motors. These are run from 3 Slamstepper controllers with 36V 32A Server power supplies through one of Romans 4 axis breakout boards.

Linear motion is skate bearings on steel pipe. Motor connectors are LoveJoy and anti backlash nuts are self made with an acme tap.

It would seem to me, the uneducated, that I have a problem on the Y axis more than the X. I will do some more measuring today to see if anything is there although I am probably looking at culmitive errors over 600mm of 1 mm or 2. I only have an inch dial indicator at this stage so will see what I can find by converting the readings to metric, that is until I buy a metric one this week....

Cheers
Bruce

[OCNC]

....although I am probably looking at culmitive errors over 600mm of 1 mm or 2.

You might want to try making each part as a separate file. Load the blank and then cut the first part. Reset the zero on the remaining uncut portion of the blank and cut the second part from that reference. If the problem goes away (both parts match) then it seems like it would be cumulative error (in other words too many steps in one program). If the problem is still there then there might be a variation in the lead screw that needs to be compensated. Just a guess though.

Chris

[racedirector]

You might want to try making each part as a separate file. Load the blank and then cut the first part. Reset the zero on the remaining uncut portion of the blank and cut the second part from that reference. If the problem goes away (both parts match) then it seems like it would be cumulative error (in other words too many steps in one program). If the problem is still there then there might be a variation in the lead screw that needs to be compensated. Just a guess though.

Chris

Thats a great idea. I'll go one step further and do each operation as a separate file as well.

Hmmm leadscrew compensation... I have read about that but never did anything about it. Might have to look into that too.

Good thing is MDF is cheap, wasted a sheet so far but no big loss apart from heaps of dust!

I'll report back in a while with any results I find.

Cheers
Bruce

[louieatienza]

I would find a bit hard to believe a lead error of 1mm (.039") as the specs I've seen for worst-case-scenario at .009".

Here's my list of things to check:

Are you cutting the same direction with both parts? If you climb cut (clockwise for outside profile) it would tend to push the bit away from the cut line, and vise versa with conventional cutting.

You mentioned LoveJoy couplings. Check to make sure there is no significant play between the two parts. I had an issue on my first machine with the rubber spider being a tad loose. This combined with my rails that were tigher in one side than the other, which made parts slightly different on both sides of the machine.

Back to the rails; you can try to disengage the screws and make sure your gantry slides evenly as possible along your rails. Related to this, it's possible the leadscrew may be slightly out of alignment in one side causing tightness at the nut, worth a check.

The only way to know for sure is to run a part, zeroing out at the leftmost bottom part of your table, then do the same part zeroing at center bottom, and check those two pieces. But I'm thinking you might have the toolpaths as well as the parts mirrored, causing a climb cut in one part and conventional in the other.

[racedirector]

Hi Louie

Great checklist, thanks!

Cutting: I am cutting the outers and inners with a climb cut and I never ever thought of the impact of mirroring parts and cut direction. I was trying to conserve material when I laid out the file so I might just flip the parts the same way and try it out.

I am only using LoveJoys as I couldn't get the solid couplings when I set the machine up. It never occurred to me in all the excitement of actually cutting something that I should replace the Lovejoys even though I had read they can introduce unwanted backlash due to the spiders. First things on my list when I get paid is solid couplings and DumpsterCNC anti backlash nuts so I can totally discount those two areas.

As to the rails, knowing my expertise (not!) in manually cutting parts I suspect there are variations I was not aware of, variations that can impact the process while not being entirely apparent. Same goes for the alignment of the leadscrews they too could be out enough to screw things up. I recently replaced all the bearings with abec 7 bearings due to a little roughness. For my replacement machine the bearing guy I buy from recommended I get *real* Japanese bearings instead of the cheapy ebay versions. At about $4.50 each I might just do that.

I will do another cut once I am finished cutting out a replacement dust shoe, getting real tired to my router blowing all the shavings away before the vac can get them

Cheers
Bruce

[OCNC]

I would find a bit hard to believe a lead error of 1mm (.039") as the specs I've seen for worst-case-scenario at .009".

I agree that this is probably not the problem.

Are you cutting the same direction with both parts? But I'm thinking you might have the toolpaths as well as the parts mirrored, causing a climb cut in one part and conventional in the other.

I looked at the profile toolpaths again in NCPlot and they're all being cut conventional.

Chris

[racedirector]

I looked at the profile toolpaths again in NCPlot and they're all being cut conventional.

Chris

Damn, I was sure I changed them to climb cut, was wondering why the edges didn't look any different. I only learnt about climb vs conventional a little while ago and wanted to try out the difference on this cut. Oh well, when I redo the panel I will make sure of two things... I have actually set it to climb cut AND that both parts are the same orientation