I have been trying to do a double-sided flip job using MeshCAM on my 3 XXL. The MeshCAM software is quirky at best, but I think my main problem is that if I machine a hole from both sides of a piece of stock, almost always there is a lip in the middle, indicating that the alignment has failed. Tomorrow I will do another run, this time initializing the 3 XXL after every tool change, on the theory that when I change a tool I jog the machine enough to affect the position.
Is it reasonable to expect a 3 XXL to create accurate double sided jobs? Would the 5 Pro provide any improvement in such jobs? What accuracy would you expect from each?
What if you machined the hole all the way thru the stock for alignment? Assuming this is possible depending on the stock your millling. I’m not an expert on the machines but data does point to a lead screw unit holding accuracy better than belts. Now, does this come into play with normal woodworking as most of us do? Probably not.
Even better yet ballscrews assuming a proper preload can hold tolerances in the tenths of thousandths. (Machine uses 16mm dia 10mm per revolution ballscrews) There have been talks about how the 20mm ballscrews work out to be quite a bit better and more rigid but if it isn’t the weakest point why worry (the weakest point is the bed followed by the gantry extrusion)
That said it is a bit of a proprietary ballscrew as very few make a 16mm ballscrew in the 10mm per rev formfactor. Which makes it a go through C3D thing if there ever was to be damage. Additionally there is no standard for manufacturing of ballscrews and so you would want to replace the screw and nut at the same time if it were to ever come to that.
Finally the 3 uses V rollers which are in my opinion the least usefull form of travel guide. Any little dust or bump in the track can be transmitted to the workpiece especially if the rollerblocks aren’t set up correctly.
I run one job to machine two registration holes in my waste board. I then position the stock and use a similar job to machine the registration holes in the top half of the stock. I then flip the stock and position it on the waste board using registration dowels, and machine the registration holes in what was the bottom of the stock. At this point, I expect the holes drilled on each side of the stock to line up, but they are off by perhaps 0.150".
I have only use registration dowels by setting the stock down, drilling the holes thru the stock and into the waste board, using center of the stock for X/Y.
The idea of making a single hole through the stock and the waste board is compelling, but I think it misses the point. My process involves 10 steps (separate Carbide Motion “jobs”):
1 waste board reg hole roughing
2 waste board reg hole finishing
3 stock reg hole roughing (top half)
4 stock reg hole finishing
(flip and attach stock to waste using dowels)
5 stock reg hole roughing (bottom half [now the top])
6 stock reg hole finishing
7 top roughing
8 top finishing
(flip the stock)
9 bottom roughing
10 bottom finishing
If the machine gets out of registration at ANY point during these 10 steps, then the entire process is ruined. Sure, it can be off by 0.010 inches, but not by 0.100 inches.
The reason to re-initialize the machine at the start of each step (after loading the cutter) is to attempt to ensure that the same zero is used throughout.
If re-initializing produces a different Zero each time, then I am really screwed. I am COUNTING on the machine initializing to the same state each time.
But, of course, I am hoping that the community can guide me to a understand what the two machines can do, and what best practice to employ.
Yes, the SO3 should be able to do two-sided flip machining.
Yes, the SO5 Pro will do better — in addition to the rails and ball-screws, the steps/mm are 4 times finer, so positioning should be that much more accurate, see:
Yeah it’s a bit hokey and it’s kind of left up to the consumer on how to do it but when you learn the right ways it does end up working 99% of the time. Also dowels should be something that doesn’t deform over time or with heat etc. I use PG 1/2" dowels
Interesting process. I can see the reinitialization returning the machine to the same “state”, however that does not mean it will get back to exact zeros each time. As you stated, perhaps it can be off by 0.010. Maybe that’s the delta your seeing now?
We could probably calculate the maximum offset by checking RPM in the homing process, multiply by microsteps that is how many microsteps per second the steppers are doing and then divide by the polling rate of the 8bit microcontroller. It will still be ± but it would give you an idea of how much it could be off.
@MindlessCorpse: good question about the orientation of the lip. Because I keep flipping the pieces, it is not always easy to figure out whether the orientation is uniform. And since the lip represents an offset, the lip appears on both sides of the piece. I would estimate that the lip is about 0.10 inches (hard to measure something at the bottom of a 3/4" hole with diameter 0.3"). But my recollection is that the offset is front-to-back (in Y)
I can find almost nothing on the web about how to assess accuracy or repeatability for a CNC machine. I have found many, many articles (often copies of each other) about the terminology, but no NC files or procedures or statistical discussions. This is astonishing to me.
It is our belief that the 3 XXL gets progressively less accurate as a long (2 hour) job proceeds. We see evidence that the 3 XXL has a different Zero end of the job than at the beginning. At this point, we are re-Zeroing the machine after every tool change and before every Job, but this is voodoo engineering, not science.
Is it true that this community has never come up with a procedure to measure repeatability and accuracy, or am I mistaken??
Based on my rather short time of owning and using the 5Pro it seems the issue is not software related with respect to becoming less accurate over a longer job time. I have read numerous posts about all of the C3D machines on this and other forums, the belt machines work very well when tuned up and maintained. This includes V-wheels, belt adjustments etc. It just seems to me point back to that being the issue with your machine. I’m certainly not suggesting you have not maintained it, just the fact of belts vs lead screw tech and improvements made to the machines since the 3XXL.
As for accuracy, I have installed a reference fence on the left (Y2) rail and machined in place. I used a V bit to run parallel with this fence on several of the spoil boards from front to rear. I can now put the same bit back in and go to these coordinates and run the bit right back over them and check the accuracy. I have had no deviations as of yet as well as using the same X/Y homing for projects with repeatability.
The issue with a CNC which doesn’t use closed loop control is lost steps — feeds and speeds have to be managed so as to ensure that each time the controller sends a signal for the machine to move a step, it does so.
My approach for this has always been to avoid slotting and drilling — where possible avoid slotting and add geometry and cut as a pocket
and/or
and consider leaving a roughing clearance and taking a finishing pass.
doing thus, I’ve been able to run quite long cuts, and have done projects where I have re-run a toolpath which has lined up just about perfectly with previous work.
If you’re having trouble, upload a .c2d file, photos showing your stock/part clamped in place, and photos showing how the cut turned out and how you are evaluating it. Note that when measuring, you want to use a tool which is markedly more precise/accurate than you are trying to measure.
