[corneliusbrown]

New to CNC stuff.

The output of a g-code file in mach3 does not reflect the original dimensions as specified in cambam.

I'm pretty sure everything is setup in MM on both ends.

It seems like i read somewhere that that some feedrate (or other) settings need to be the same within both programs but I havent figured out which settings need to match yet

maybe my steps per inch setting (in mach3) is incorrect?

Any help would be appreciated.

[10bulls]

Hello Cornelius,

My suggestion is to first verify that you have Mach3 calibrated correctly by doing some simple tests.

In Mach3, zero your DROS, then go to the MDI window and enter some simple gcode commands like:
G21 (this will put you in MM mode)
then
G1 X100
Then measure the distance travelled with a vernier to confirm that it is indeed 100mm.
(you might want to get ready with the e-stop button if your machine tries to head for the hills ;D )
Then try your other axis (G1 Y100, G1 Y100).

If this doesn't do what you'd expected then you need to change your motor tuning parameters. There is also an automatic calibration routine in the Mach3 setup screen you might want to check out.

If the above all checks out, you could then create a hand coded gcode file using a text editor. Something like:

G21
G0 X0 Y0
M5
G1 Z-1
G1 X100
G1 Y100
G1 X0
G1 Y0
M30

Try cutting that in some foam or MDF and measure the resulting square.
Don't forget to take into account the cutter radius. So if you used a 6mm cutter, the above square should be 94 x 94 mm.

If that all checks out then it's time to start checking things in CamBam.

I hope that helps and good luck!

[corneliusbrown]

Will try this and report back. Thank you much for the info

[Fishnlight]

Hello. New to the forum. Having an issue after a controller retro replacing analog drives with Viper75, PC and went to Mach3 (licensed). Machine is a Precix 4x8 router. Previous controller worked well in DOS, but the drive controller died, could not find a replacement, and Mach had more features so I felt it was time for an upgrade. My issue: I can post a drawing (with Cambam, also used with previous controller) that contains a circle (4") with a square in the center (2"x2") and 8- .125" parameter holes spaced equally i.e. 9, 10:30, 12 etc. around the inside of the circle. When cut (.125" end mill) my circle would be perfect but my square could be -.040 on X. If I recalibrate steps in Mach I would get the square dead on, but my circle and parameter holes are out in X. I can't get both dimensions correct at the same time. I have tried different IJ settings in Cambam and Mach along with changes to velocity and acceleration with no mentionable difference. The machine is tight and cut fine with previous controller. I have done XY ocillation test (40 reps) with a dial indicator and test allways ends up within .001 for both axis. After all these test, I wrote a program for a .125" end mill for X and Y to drill 6 holes for each at 1" apart. Oddly enough they measured approx +.080" at 6" and the incrementally closer at 1" (1" +.008), even after the dial indicator test? Recalibrated Mach and got the 6 hole measurement perfect only to find the circle test previously mentioned is way out, +.050". Uninstall/reinstalled Mach 3 times with crucial reboot. Any help is appreciated. Sorry for the long wind.

[10bulls]

Hello there!

I know you said the machine was tight, but it sounds like a backlash problem.
A repetition test wouldn't pick this up.

My favourite way of testing backlash is to set Mach to a fine jog increment (eg .001mm or 1 thou ). Then jog an axis up against a dial indicator until it moves a bit. Zero your Mach DRO, then fine jog the reverse direction until the dial indicator starts to move again. The number in the DRO is then your backlash measurement.

I am a bit wary of calibration routines. If you know your lead screw pitch, stepper driver microsteps and any driver gearing due to belts etc, you should be able to calculate what the steps per unit should be. If this is different to what the calibration routines give there might be a problem somewhere.

I hope you get it sorted. Good luck!

[Fishnlight]

10bulls. Thank you for quick reply. I did as instructed. The DRO reads X at
.001", Y .002". Do you believe this small amount will give the results I been getting? I will set in backlash and run a part again later today, then post results. What backlash speed do you recommend? Thanks again.

[Fishnlight]

Hello again 10bulls. I ran 12 test cuts setting backlash with X at .001" Y .002" at 100% speed. The results were the same as no backlash. I increased both each run all the way to .015". The overall dimensions decreased proportionately till I stopped at -.060" from the starting point. I could have stopped sooner as I saw the changes heading the wrong direction, but wanted to see the effects. I'm back to original starting point scratching my head. My 8 parameter holes previously mentioned measure perfect (as to the drawing measurements) as the circle X is +.021", Y+.011" and the square X-.003" and Y +.026". Do you have a circle with square test code I could try? Thanks again.

[10bulls]

Hmm... the backlash doesn't sound too bad so back to head scratching as well I'm afraid.

So, without cutting, can you make a move by just entering a G1 command in Mach3 MDI window and measure how far it actually moves.
Can you fix a vernier to your bed somehow and drive the head against that to measure?
I would have thought if you can get the calibration right so that what you measure is the distance you tell it to move, that would be a good start.

If the test files are still out then, I think you should leave the calibration as is and...well...maybe all this head scratching between us will loosen something and we'll have an insight.

[Fishnlight]

Hi 10bulls. I did my calibrations in this manner along with the ocillation test mentioned previously. MDI 1" = 1" within .0005" X & Y. After seeing these numbers I believed the machine would be better than ever (errrr). One thing I failed to mention before, is one quadrant of the circle had a protrusion at the 10:30 position of +.020 when the rest of the circle was equal. I came across an article with respect to this issue. Perhaps you are familiar with it (link below). I will review and re-think. Maybe that proverbial light will come on. Let me know if yours does. Thanks!
http://findarticles.com/p/articles/m...6/ai_14335487/

[10bulls]

Well it sounds like your calibration is OK.

Being a struggling programmer, I mostly stick with steppers :violin:and haven't had much servo experience I'm afraid. Have you tried on the Mach forums? Either Here:
http://www.cnczone.com/forums/forumdisplay.php?f=7
or here
http://www.machsupport.com/forum/

I know you mentioned IJ mode, but just to be double sure, you are using the same ArcCenter mode in CamBam as in Mach?
Constant velocity mode might be another one to check. Try setting CamBam's velocity mode property to ExactStop or change your gcode to use G61 rather than G64. Not sure if this will help, but it may help rule out another possibility.

Last thought for now... what Mach kernel speed are you running? For servos you should be running at as high a kernel speed that your PC will reliably support (as far as my limited understanding goes).

Hang in there... I'm sure you'll crack it!

Hi 10bulls. I did my calibrations in this manner along with the ocillation test mentioned previously. MDI 1" = 1" within .0005" X & Y. After seeing these numbers I believed the machine would be better than ever (errrr). One thing I failed to mention before, is one quadrant of the circle had a protrusion at the 10:30 position of +.020 when the rest of the circle was equal. I came across an article with respect to this issue. Perhaps you are familiar with it (link below). I will review and re-think. Maybe that proverbial light will come on. Let me know if yours does. Thanks!
http://findarticles.com/p/articles/m...6/ai_14335487/

[Fishnlight]

10bulls. I thought maybe something was wrong with PID settings, so I spent the better part of the day checking and changing parameters. Found nothing wrong there, and no change on test cuts. I've attached another cut that also has me baffled. All dimensions within red measure exact to drawing dimesions, but the area in black is +.080" out when cut. These holes measure 5.175" outside, and if you recall from my first post, I did a 6 hole test X Y in 1" increments that I got to measure exact. It appears when circles or a radius is apart of linear moves the program or hardware has issues. I did as you suggested and changed CV to exact stop and no change occured. Have you ever heard of encorders causing this type of issue. One would think all coordinates would be varying. I can cut 100 pieces and all measure the same. I have to get operating ASAP as production is needed. Thanks for hangin in there.

[Fishnlight]

10 Bulls. Update: Went back to a 2"x2" square with 2"x2" diamond inlay code that I created, and all dimensions were correct. I went to Cambam and changed a few things on the drawing from last post. I made a points list on the 2 holes that were .080" out and allpied drill profile first then pocket, and everything is prefect now. Never had to do this before, so don't kow what would change it. I believe the code being generated from Cambam is the culprit for my circle square inlay test as well. I like the ease of Cambam and would like to continue to use it if I can get the issues resolved. I will try orther things for circles later and keep you posted, but for now must run parts I'm way behind on. Thanks

[10bulls]

When you get time, and if you still have a copy of it, can you upload the .cb and gcode files that you were using for testing? ...the ones with the incorrect dimensions that is. Then I can take a look to see if I can spot anything.

[Fishnlight]

10bulls. Sorry for the delay. Still perplexed as I have tested pulse stream with O-scope, all looked good. All motor signals looked good and driver test was excellent to 100K. Then tried another computer with same results. I tried lowest possible settings for vel and acc with no change. I will try to send code tomorrow. I read a forum stating that a Dell (which I was using) produces bad PP signals. Second computer (still connected) is an HP. Any insight on this? Thanks

[Fishnlight]

Hi 10bulls
The attatchment is not what I reffered to previously as I don't have access to that file right now. This one is just as inconsistant on Y. The hole diameters, rectangle width and length are correct. When the larger holes on either side of a rectangle are measured by dragging the curser around both and checking measurement extreme in Cambam, I get 1.207" OD. This measurement when cut is correct. When I drag my cursor around the larger holes between the rectangles and get measurment from CB, it reads 5.327". When cut I get 5.390"? When the triangles are cut between the holes, they are +.030" on Y. The rectangle at Y 38.35" is out .047" and the rectangle at Y3" is out .082". I also did an additional test I read about that has you jogging axis while watching the feed DRO for erratic behavior. This test too was good. I'm using a CNC4PC C11 BOB. Maybe I should have gone with the smooth stepper? Hope I explained it well enough. Thanks

[Fishnlight]

Sorry 10bulls. I see that I can't send the g-code as attatchment. Will have to do on Mon. Thanks

[ger21]

Just change the extension to .txt and you can upload it here.

[Fishnlight]

Hi guys. Sorry for the lengthy delay. I was able to get the machine to make the cuts needed to complete my projects, but would like to correct the problem. I sent a jpg. drawing of a circle and 2 squares. I have seperated the art to allow better viewing. The parameter holes are within .003" of each other with the exception of the 10:30 position. It's out +.020". The bad coordinates all increase and decrease with different diameter size cuts. Run the code from the text file if possible and let me know what your findings are. Thanks again for the help! FYI: I ran another dial indicator test today and machine is within .002" XYZ.

[blowlamp]

Hi there.
It would be much better if you can post the .cb file that you're using to produce the GCode. Zip it up by right clicking on the file, >- Send to >- Compressed (zipped) folder.

Martin.

[Fishnlight]

Hi blowlamp. Attached is the cb file. Let me know if the file works. Thanks for the help.