Daid published a new build:

Read more about this build...

 

Replaced all the wooden v0.1 parts with v0.2 parts yesterday. The Z axis is more rigid now, as it has 6 wheels instead of 4. Preventing a lot of flex in the plywood plate in which they are attached.

Also I can tighten the X wheels on the aluminium extrusion now, as the eccentric spacers are now at the top. So that has also increased rigidity as well.


Bad thing, I forgot to disconnect the power from the electronics while working on it, and the UI board touched the main board at the 24V power, frying the electronics. I won't be able to get a new one till Tuesday.

 

Finally found some energy to do something again in this heat.

Replaced the electronics, and did my first cut. Made a 100x100mm square. It came out 98x98mm, so something is off. Not sure if it's backlash, but that is most likely. Something not stiff enough causing backlash. Could just be belt tension, as I didn't tension those properly yet.
My feedrate was also quite high compared to the spindle speed. Still getting used to the different world of milling compared to 3D printing.

 

Well, as I used motors, belts and pulleys from Ultimaker. I'm pretty sure I'm spot on with 160 steps/mm (400 steps per revolution motor, 16 microsteps, 20 tooth GT2 pulley, GT2 has a 2mm pitch, (400*16)/(20*2) = 160 )

It looks like it's a buildup of play in various areas. The belts where not tight enough, with 1.5m belt, that gives quite some stretch potential (guess what, belts are usually not tight enough, GT2 apparently needs 70N applied to them, according to a hardware engineer at the office)
The front and back of the X carriage where not tight enough together, so the front could move in relation to the back, this seems to be pretty big part of the play causing problems in both X and Y directions.

There is also some play in the whole Y arm, not sure where it comes from yet. Or how large it is.

So, need a better solution to tighten the belts. And tighten every screw in the machine.

 

I'm not sure how to respond to that. But let's start off by saying that you come across as needlessly aggressive.

I also think you are wrong. A 2% difference isn't due to some tolerance difference. As the whole positioning comes from the belts, it's not like the steps/rev and tooths on the pulley will have some tolerance on it that effect total distance. Only linearity.
And the belts won't show a difference of 2% (unless you have different belts then expected. MXL vs GT2 is about 2%) Sure, there is some difference in tolerances in those belts. But nothing near what I'm seeing.

Just to extra prove my point, of the 20.000 Ultimaker2 produced with the same belts, we never had a single problem with X/Y movement distance. Without calibration. But I might measure to see if my cheap Chinese 1.5m belt is accurate compared to the official GT2 belts I have from Ultimaker.


Finally, I can just feel the play in the X/Y direction with ease. And one side of the square has "bites" out of it indicating that the router bounced on that cut.

 

Not worked on this for about a month. But, yesterday and today I made 2 actual cuts.

First I cut a small piece so I could tension the right side Y belt better. The other part is 3D printed, belt is only attached to the 3D printed part, and the M5 bolt can be used to screw the piece closer to the wooden part to tension the belt.

So the wooden part was cut on the router itself, with a temp fix for the belt. Used a 3mm endmill, and 6mm plywood.



Next I cut something more complex, figured I try to make a toy car, own design. Most of the flaws you see in this cut are more down to the fact that I didn't have enough Z travel with the tool, and while adjusting this after making the cut 80% down, I slightly moved the router, causing a shift. And a piece broke off that was simply too thin.
And this was done with a 8mm straight flute end mill, as that is the only endmill that I currently have that can cut deeper then 20mm, this is 28mm deep. The wood is left over parts from a wooden cabin.


Overall, I'm very happy. Accuracy seems a lot better with proper tension now.

Note: Used my own firmware, and self created CAM software, so, that's starting to work as well.

 

Ok, back from the depths. Didn't do much during the cold days, due the machine being in the garage and it being very cold there, as well as limited time.

Good news, I'm still working on this. Or actually, it works now. Pretty well the last 2 cuts I did. Long story short, looking for the play, I looked everywhere, except on the right spot. When I took a good look at the machine while it was cutting, I noticed the whole spindle would just tilt, the single bracket that was holding the spindle simply wasn't stiff enough. Added a 2nd bracket at the top, and suddenly my accuracy and reproducability is where I want it to be.
(I also forgot to set proper motor current, causing skips, but that was just a stupid software mistake)

Also, my CAM software is starting to actually become usable, all my cuts are made with this CAM software.

It follows a less=more mentality. It still lacks a few things that I deem essential for most people. The the core concept is that you have tools, tools have operations (like cutting outside of the line, or pocketing) and that you assign those operations to different parts of the drawing (on color/layer). There won't be any drawing/editing tools, as I want to keep that on the CAD side. Once setup properly, I hope I end up with a process that reduces my time in CAM to zero.

But once I have a showable version of the software, I'll make a seperate topic of that.