Test 2 was without finishing pass. But was just 1 mm DOC.
All other tests were with 4 mm DOC and a finishing pass of 0.5 , 0.3 or 0.2 mm at full depth.
This are the settings in Estlcam for the last test. (the hole in the last picture with the part from test 2 in it.
Also here are some of the specs of my MPCNC:
MPCNC Primo
x=675 y=600 z=81
V1CNC_Ramps_Dual-2.1.1 firmware
Mega 2560 R3
Ramps 1.4
55 mm 500 watt brushless spindle.
do you know the runout on your spindle?
No. Not really. I don’t have the tools to check. I will get does at work in a few days.
As far as I can check now there isn’t anything obviously wrong with the spindle.
Spindle runout is one thing. Some of the cheaper ones aren’t great in that regard, but there are more basic things to check.
Are you sure you have bearing contact with the rails on all bearings on the core clamps? If these have even a small amount of play, it can let the spindle wobble.
Do you have good bearing contact on all of the Z axis bearings? Again a little play goes a ling way here.
Do you have anything pulling on the Z tower at all? The top if the Z tower, where the motor is has a lot of leverage.against the core, which can and will cause wobble. Ryan has been knkwn to advise that nothing be attached here, even cable chains. (The Z motor wires are unavoidable though.) On my Primo, I have the cable chain atrached to the core itself, not the Z tower for this reason.
Weird to ask, but are the spindle mounts tight? Not overtight, but enough that it’s firm.
With all of these, it is really easy to overtighten things, which results in cracked plastic, therefore looseness, so you need to be careful, but I have found places where you can get an air gap in the bearings, which makes things less accurate than they might otherwise be. Light finger pressure on the top of the Z tower in various directions can tell you a lot about where your machine will flex. Any “clicking” noise with that pressure means a bearing somewhere is loose.
Edit: and how much Z travel do you have? The more you have, the greater the leverage the bit has to make the core flex.
Rechecked those and looks good to me. That is where I had problems before and why I printed new clamps.
I thought those were fine too. But after removing the spindle and spindle mounts one of the lower inner bearings didn’t make good contact. I didn’t notice that with just wiggling the Z tower. It took some force to see the movement.
Yep. The cable chain is attached to the top of the Z tower. Already redesigning that. How did you attach it to the core? I’m thinking about using the extra holes in the core clamps but not sure if that is a great solution.
Removed the spindle and clamps to get to the bearings. But didn’t see or feel anything wrong there before I did.
Z travel is the default 81 mm from the calculator. Just added 18 mm to the legs to compensate for the thickness of the spoilboard.
I’m going to do some test again later today.
That is exactly how I did it.
Thanks. That gave me some ideas.
Testing has to wait a little for me. Just managed to fry my ramps board trying to measure the signal from an endstop. The quick test I did this afternoon looked only a little better. Still some more to do.
I already had a Tinybee on order, but it will take a few days to get it up and running on FluidNC.
I got it running again. Plenty of things to tidy up and setup, but it is cutting again.
Went over the core again and there were 2 bearings that could be a little tighter.
With the latest fixes and the ones before that there really isn’t that much difference to the quality of the cuts between now and before I posted here.
I checked the runout and play on the spindle. No play that I can measure and runout measured on the shank of the endmill is a little over 0.01 mm. Would have expected more on a cheap spindle.
I’m now going around the MPCNC to check for other movement of flex.
I was bit worried that the length of the tube would cause this problem.
I took the z axis out and measured the flex on the X tube (the longest axis) with 10 kg placed on the core. Just 0.5 mm of flex isn’t bad I guess.
Can’t find anything really wring with the MPCNC. There is some movement but that is to be expected with moving parts.
Nothing more than 0.01-0.02 mm
For example: The trucks on the rails can move a little along the rails with the steppermotors powered. But only somewhere between 0.01-0.02 mm again. And when I stick an allenwrench in the setscrew in the pulley (to use as a dail/pointer) I can see that the movement is in the steppermotor. With the 1/16 microstepping that I use that is around what one microstep is. I have to use a lot more force to make it skip steps.
Would 1/32 microstepping make a difference here? But even if it does I can’t see how this would make a part 1 mm over or undersize with higher DOC’s.
For testing I have taken the cablechain and vacuumhose off.
No. The full steps send all current to one coil and none to the other. A 1/2 step sends cos(45) to one coil and sin(45) to the other (both are around 52% of max). When 1/2 stepping, the next step is 1,0. Then back to (.52,.52), etc. The numbers are hard for humans when you get down to 1/16 or 1/32, but half of the positions for 1/32 are the same as 1/16. The new positions are in between the other ones.
The whole thing is just trying to encourage the stepper to sit in a slightly different position between steps. You won’t actually get 1/32 accuracy unless you have zero load.
And, you are right that 0.02mm is not causing your 1mm of error.
Have you measured backlash? Does the machine go to the same position if you drive it from the left or right without load? If you have room in your dial indicator, try moving 10mm right and back, then 10mm left and back. Jaime has a test pattern generator that draws a ruler from both directions with a pen too.
https://vector76.github.io/gcode_tpgen/index.html
What about the router? Does it have runout? Is the mount solid?
Try doing your finish pass with step downs instead of full DOC if accuracy is your primary concern, also use the shortest bit you can get away with (eg, if you are cutting 1/8" material, use a bit with a 1/4" doc at most).
I have tried that. And always back to the same stop within 0.01-0.02 mm.
I will have a look at drawing that ruler. I doubt that will show an error as the machine is accurate under low loads. But I’m running out of ideas myself, so will try that.
No play on the spindle and hardly any runout. (a little over 0.01 mm)
I use this spindle mount posted by V1 Engineering. It feels solid to me. I can’t see any movement Printed it with 3 perimeters with a 0.6 mm nozzle and matching top and bottom layers.
At the moment I one have one type of bit. I will get different ones when I get the hang of this milling or when I feel I need some thing different.
I can do accurate cuts, even without finishing passes. But that would limit the DOC to 1-1.5 mm. That will take a long time to make some parts.
I started this topic because when I tried increasing the DOC the accuracy suffered a lot. There are plenty of videos around that show the MPCNC do really hard work.
Ryan has posted a few. I’m sure he has tuned it a lot to get that far and his MPCNC isn’t a big as mine.
I just want to figure out what causes mine to get inaccurate that quickly.
Just rebuild and adjusted the core again. Reprinted it because I worried I had messed to original one up a bit when trying to get the play out of that one. Haven’t had time to test .
Are you using climb or conventional milling? Had similar issue when i had first started.
I have used both. Started with conventional and after reading more here on the forums I switched to climb milling. Both with about equal results.
I did a test cut again to make a video. I will try to post that here or at least part of it.
But I can post a small gif I made of the flex that I see now. There is no play in any of the bearings, but there is some movement somewhere.
With the X or Y close to the min or max there is much less. But the more I go towards the middle of one of the axis that is much more.
That makes me thing it is flex in the longer rails anyway. Those are 25 mm RVS with 2.5 mm wall thickness. The gantry x rail is 945 mm.
It looks really bad in the gif, but the is a 1/8 bit and it took quite a bit of force to move it that much. And with the camera really close.
I’m printing new feel and going to rebuild it on the spoilboard.
That way I don’t have to take the original feet off and can I easily put it back to the larger size.
Building it on the spoilboard will give a small machine with a workspace of 320x320 mm.
That way I will know for sure if the large build is the problem.
Looking at your video, something is loose, bearings not touching etc… i didn’t see if you mentioned the size rails you’re running. Check that all your bearings are touching the rails and pay close attention to the z rail bearings. My lower ones love to loosen up from time to time.
I want to say that that is not te problem. But as I can’t find a clear problem myself, I will have a look again.
Maybe that video wasn’t the best to post. It looks much worse than it really is.
I have now rebuild it with a 30x30 cm workspace. I’m going to do some test cuts this evening. But when I try to flex it by hand, it looks like not much has changed.
I can see some flex between the core and the core clamps. To me it looks like the bolts that go through the core, bearings and core clamps should be much tighter. But I don’t want to mess up the PLA parts (again)
I hope I can get that flex on video tonight.
Try to get the whole core and z axis in the video. It will give us a better idea of what the issue may be. My working area is 600x900 and there is nowhere near that much flex from the rails. The only time i see flex like that is when the bearings in the core have loosened up. Normally flex like that is the z axis bearings after several long jobs. Contrary to popular belief, nylock nuts can and do loosen up. I’d try to figure out the issue you are having instead of tearing it apart not knowing what the actual problem is.
Just my 2 cents…
This isn’t my usual way of diagnosing a problem. But I have no experience at all with the mpcnc or any cnc for that matter. I know I had over tightened some things trying to get the play out, that is why I replaced some parts. And making sure the large size isn’t the main problem crosses that of the list too. The small build did the same. But gave very little room to measure and feel/move things.
It is back to the original size and build according to the manual.
That should be a good base to work from trying to find the problem.
