I completed the build of my MPCNC a week ago. I drew the crown and it looked good so I moved on to cutting some stuff. Originally I was using the smallest bit I had (1/4" straight flute router bit) and once I got the steps/mm dialed in I was able to cut 1" test squares that were consistently ±0.010". About half the time they would even be ±0.005". I was happy with that so I moved on.
I got some new bits and today I was using the smallest bit (1.5mm spiral up 2 flute) and just messing around. I tried to cut some finger joints that came out a bit loose. The fingers were about .020" undersized. Out of curiosity I expanded my digital caliper and set it on the table with the head butted up to the cutter bit (cutter was not running). I zeroed the caliper and used the manual controls in repetier to jog the head 10mm over toward the caliper. I did this a number of times (zeroing the caliper each time). When telling the machine to move the head 10mm I was getting anywhere from 9.90mm to 10.29mm.
I tightened up the belts because the seemed a little bit loose and tried cutting some more 1" test squares. They were also about 0.020" under sized. I adjusted the steps/mm again (added 3.5 steps/mm on x and 4.5 on y) and was able to get them consistently ±0.005".
After all of this, I tried checking the motion with the digital caliper again. Now when telling the machine to move 10mm I am getting anywhere from 9.91mm to 10.19mm. It is slightly improved but I would assume that with almost no load (just enough to push the head of the caliper) the motion should be more consistent.
Am I expecting too much or is a variance of ±0.014mm per 1mm of programmed movement normal?
You can count the length of something like 100 teeth. It’s a pain to count them, but that was the smoking gun with the other case. It ended up being close to 10% off.
How about this. Can you make a mark at the bit. Then move it 200mm and then measure that. If it’s still off by 0.02mm (which you probably won’t be able to see) that would be different than if it was off by 2%.
Is your inaccuracies in all directions? If you draw the same object over and over again offset by 1 cm in both x and y you would get a feel for which axis has the errors.
-increased the current on the stepper drivers (A4988s) to max
-tried out full stepping and then went back to 1/16th micro-stepping
Now Jamie’s test pattern will draw a 200mm ruler that is dead on 200mm long. That being said, the individual mm markings are still not perfect. They will vary by as much as the width of the line. Sometimes a bit more, sometimes a bit less. In the end, the whole thing ends up being just the right length. I will try to get some pictures that show exactly what is going on to post.
The first photo with the red and green marks shows the starting end of a 200mm ruler drawn with Jamie’s tool. The pen is mounted to the mpcnc with this (https://www.thingiverse.com/thing:1612207) mount. The large red marks indicate obvious misalignment between the top and bottom row. The small red marks indicate places where the spacing is obviously off from the lines around it. The green marks indicate the fact that the starting line of the bottom ruler and the ending line of the top ruler are exactly inline.
The second photo shows the right-hand end of the same ruler. The red dots are all exactly the same size and just serve to highlight the differences in the spacing of the lines. The middle dot is between two lines with a tiny bit of white space on either side. The left dot is placed with a tiny bit of white space between it and the line to its right and you can see it overlaps the line on its left. The right dot is placed evenly between two lines and it overlaps both.
The pulley has exactly 16 teeth. The motor has exactly 200 steps/turn. The driver doesn’t make mistakes when counting so it is exactly 16 microsteps/step.
The belt is a little stretchy, so that can add some error.
The pulley may slip on the motor shaft, causing either huge errors, or backlash errors.
The motor may skip steps, but they will skip whole steps, and it’s pretty obvious.
I am bringing this up because if you “calibrate” your steps/mm to be 102 when they should be 100, then there is some other problem that is getting masked by the calibration. And when you measure for the calibration, it might be at 200mm, but the error might not be 1%, it might always be 2mm. So when you make a square 20mm, it will be 22mm.
That’s not very helpful, except to say that the pattern generation without changing the steps/mm might be more helpful.
Can you go through how the motors moves the machine, all the way to the pen, and look for where it might be getting off? There should be no resistance at all from a pen, so I would think the errors you’re seeing are more of something loose than something flexing.
The uneven spacing is troubling. Do your rollers slide easily and do they slide smoothly?
If the rollers are too tight then they can create resistance that can appear as a small amount of backlash. Also if the tubes or bearings are not smooth (they usually aren’t) and your rollers are tight, then the resistance can be lumpy, which can produce inconsistencies. It is easy to accidentally tighten the rollers too much, so I’d say it’s worth checking.
Those pics were taken after it was set back to 100 steps per mm. Prior to that it was set to 101.5 and a 200mm ruler was consistently 3mm too long. I’ll try loosening all of the tension bearings a bit and do some more testing.
So it turns out that the bearings were most likely the culprit. This build was actually started by a friend of mine that printed all of the parts, bought the lead screws and bearings, and then got busy with other stuff. He gave me all of the parts he had so that I could build it. Well, it turns out that the bearings he had purchased cost about $0.27 a pop from China. The end result being that when I disconnected the belts and went to move the gantry by hand it would stick and move and stick and move… like if you had gum stuck to your shoe… or a tire… or whatever. I took everything apart and started removing bearings and some were still reasonable but some had massive amounts of play in them. I replaced 32 of the bearings (basically everything minus the Z axis) with ABEC 9 swiss made fully sealed bearings that I probably paid wayyy too much for but they roll sooooo smooth.
While I had everything apart I decided that the EMT should probably be a bit smoother too. So I chucked each piece in my lathe and went through 3 steps of sandpaper followed by 0000 steel wool. They are not perfect but they are a dramatic improvement.
After all of that, I put the machine back together and moving the gantry by hand was a night and day difference. Without the belts attached, I could push the gantry around with my {{insert small weak body part here}}.
I ran the 100x100mm ruler program again and it came out dead on 100mm and the spacing between the marks is much more consistent. The end of the marks on the way back is slightly off from the original starting mark so there may be some backlash but this is a massive improvement.
They weren’t just bad. Some were really really bad. I have had this together for maybe a week and done minimal cutting and some of the bearings had as much as like .015” of axial play. I am shocked at how terrible they are. I knew when I was putting it together that they were cheap Chinese bearing but I didn’t think they would be this bad.
thanks everyone for your help in figuring this out.
