I’ve been trying to install my Neje diode laser in the LR3, and while testing it I’ve been getting a weird artifact where it seems like the x is missing steps (microsteps rather, It’s a very small amount) very specifically and repeatably.
After a full afternoon of tinkering and debugging I’ve come to a conclusion with my LR3:
There’s backlash in the X axis (I’ve measured from 0.12mm to 0.06) and said backlash seems to be related to belt tension (It lowered about half after tensioning).
The same is probably true on the Y axis but It’s set up shorter (800mm working Y) and I’ve tensioned it more agressively, so I’m not really noticing it
This of course makes a lot of sense, my build is the full sheet width (1220 working X) and the belt is rather long, but it got me wondering, Is there a way of determining the required tension, and measuring it to boot so that it can be kept that way?
Also something that has been on my mind thoughout this whole build, does PLA not creep under tension and end up giving after some time and temperature changes?
Yeah, I was looking at those but our belts are wider and having that would only help with repeatability unless there’s one for the exact belts we’re using
On a laser make sure you have overburn on and you probably need to dial in accelerations. A laser build is going to move significantly faster than a CNC build so a acceleration tune will probably make things a bit better.
With that said, 0.06mm is not bad at all. How are you actually testing something that small? Is it actually backlash or is the belt springing?
I’m not sure if it’s backlash, given the length of the belt and the fact that increasing tension helped I would assume it’s probably the belt stretching.
I did it with a dial indicator mounted to the tool head referencing a block of wood clamped in a fixed position, and doing repetitive movements of the same amount in both directions.
By using a fixed center point and moving the same amount to both sides I could always see that the offset from 0 when coming from only one side was consistently that (~0.06mm).
This “backlash” switched sides depending on what the last movement direction was before setting the 0 reference on the dial indicator. I don’t remember what that relationship was but I could repeat the test if it would give any extra insights.
On a laser make sure you have overburn on and you probably need to dial in accelerations. A laser build is going to move significantly faster than a CNC build so a acceleration tune will probably make things a bit better.
The laser is attached with the mount from @DougJoseph on a normal lr3 with a spindle mount. Maybe
I should’ve removed the router to remove some inertia but in any case it’s a neje A40630 so 600-1000 mm/s is the highest speed I’m running at. With overburn you mean overscan option in lightburn?
I’ve now remembered this and thought to maybe enable backlash compensation in the firmware as the machine backlash I’m observing seems very very consistent, but I’m running fluidnc (with a tinybee, can’t wait for a jackpot to install to my LR3) and when looking for backlash compensation support I found this:
It seems there’s no plans but maybe if someone can see that it might be useful to get a bit of extra precision from our awesome machines!
With belts we have little to no backlash. If you have stretch issues tension can be adjusted, but other than that stretch varies by how far it is from the anchor and which way it is loaded. It would not help with our machines.
Then either my belt tension is way off, which I doubt since i tested quite a bit tighter and quite a bit looser and only when loosened the results changed meaningfully (for the worst) or my measurements were somehow flawed
The same method is what i use in my smaller, full metal, ballscrew machine to measure effectiveness of preloading with new balls in the nuts and such and so far it has seemed effective.
I could consistently see a slop or backlash of what probably was the belt tension changing when changing direction on the X axis, it was repeatable and predictable; The belt side that was “loose” was the side where you would lack the movement.
This is something that might warrant proper testing (within my means) and a proper experimental method that we can come to an agreement with (getting and setting the belt tension by frecuency maybe and a standard test-bench and gcode).
Maybe my belts are lower quality, if you can’t replicate I’ll order some from the official store and test.
I’m not saying this machine has SIGNIFICANT backlash or precision issues because well, It doesn’t! It’s pretty incredible for a machine this size built with a 3d printer and very basic parts ( wich BTW, inmense kudos to @vicious1) to show such a degree of accuracy, and it’s daam accurate, but yeah I’m just chasing zeroes at this point i guess, and with a belt there’s some stretching for sure.
The pulleys have those little screws which hold them on the motor shafts. If they are a little loose, the motor shaft can turn without turning the pulley. You get a very backlash-like effect if the screw is at the flat part of the motor shaft, but not quite tight enough. The shaft can turn a little before the screw hits the other end of the flat part, but then turns as normal until the motor changes direction. This is pretty much independent of belt tension.
Grub screws cause a large percentage of the problems with these machines.
Test by pushing the core back and forth with the motors engaged. If it wiggles and then catches…
So just for numbers, say the grub screw allows 2° of movement (reasonable for almost tight enough grub screws) that translates to a little more than 1 full step of the motor, which is 16 microsteps, which is 0.16mm of movement. That’s detectable. Double that and it’s still reasonable for grub screw issues…
I was wondering / thinking the same thoughts.
But thought, I was guessing the wrong thing, and decided not to mention it, until someone with a machine running, confirmed me thoughts.
Hope Dan is correct, as it only helps me learn that I’m starting to understand the machine more, as I’m building mine at a snail’s pace.
I’m gonna go ahead an say that they’re properly tightened and threadlocked, but It’s true that I didn’t use a torque wrench for those. I can test that today and will do.
What I have understood is at 3’ that would be half as much as 6’. A backlash number would not get rid of that, they are a constant (provided you move far enough to trigger it). So if you wanted to add that 0.06mm over 8’, Your small parts would all come out 0.06mm over.
The build is a full 1220mm (4 foot i think) useable X axis build, the Y is now smaller but because I use it in multiple table sizes.
As I understand backlash compensation does only add the backlash amount when changing directions for the relevant axis, meaning it’s effectively not there for the movement of the axis (meaning only when changing directions that extra rotation amount would be done by the motor)
It doesn’t really matter since there’s no feature for it in fluidnc and I’m not planning on changing my firmware anytime soon.
You didn’t directly answer the units question. You are talking about 0.06mm? I would be satisfied with that. Especially on a 1220mm machine.
My issue with backlash compensation is that the tool will have load. That load is going to push inside that backlash. The compensation isn’t going to fix that.
But, with a rough and finishing pass, the finishing pass should be low load and it might work. It probably won’t make it more than 0.06mm worse.
