Problem of accuracy with box joints - too loose

[davidpichevin]

Hi folks,

I am having a problem cutting box joints from Fusion, they are way too loose. I've attached a file, added to it a smaller piece in order to do some testing. There is a cam setup for that one too. That piece is supposed to be 101.6mm (4 inches) side to side, top to bottom. Each of the big fingers and recesses on either side should be 50.80mm. Problem is my fingers always come out too short, and the matching recesses come out way too wide, sometimes by up to 1mm. For the first test, I've reset the step calibration of my machine to defaults. I've attached an image with some measurements. Each direction fell short so I adjusted the stepping. In the second test, I'm getting a total width and height that is very close to the expected 101.6. Now when I look at the fingers, it did make them longer, but still short of the 50.80 by almost 0.3, so that improved, although it should match the same improvement as for the total height. For the recesses, it made them worse, now being about 1mm off on the Y axis and 0.7 on the X axis. For example the one on the left of each piece went from 51.5 to over 51.8. 

I'm not sure what to do about this. My machine is tight, has no discernible wobble or runout, and the slots from the 1/8 endmill are also about 1/8 (just a little under). It doesn't look like I can improve this for all features by adjusting the machine calibration, since improvements in outside contour (the fingers) also bring worse inside contours (the recesses). I was wondering if there's something in the CAM setup that I'm missing and could account for this discrepancy.

Any ideas? 

[davidpichevin]

I think I might actually be getting somewhere. I switched to a climb cut, since it seems to always want to do that by default, and activated the finish passes (why does it do only one even if set to 2?). Reset the machine calibration to stock, re-ran the job, then adjusted the calibration (now the opposite way, down instead of up), and it now seems to be getting the various lengths closer to each other. Now the behavior is reversed, the fingers are a bit too wide, by 0.2mm compared to the expected 50.8, and the recesses too small by 0.4. I'm not sure what to adjust at this point to get the values closer to the 50.8 while still keeping the overall width and height correct, as they are close to 101.6. Also reduced the depth per pass to 1.5.

Do these results make sense?

[daniel_lyall]

That sounds like a mechanical problem with the machine, ever going to fast or the steps are off.

What Machine is it?

what controller does it use?

Nothing looks wrong in the toolpath other than the WCS has not been set in the setup and you are plunging into the material this is usually not a good idea unless there is a hole for the cutter to go into.

I would start by cutting your feed and speed in half and checking the steeps to make sure if you move an axis 10mm or .125in it does move that amount.

[davidpichevin]

It's a Shapeoko 3.

I don't believe there's a mechanical issue. I've been using for over half a year and have done finer work than this, including engraving. Wheels are all tight, belts are tight, no deflection when trying to force move the router or bit. And changing the stepping doesn't seem to be helping to get the proper lengths on the joints. Switching to climb cutting has made a much bigger difference. I'm also going at about 1000-1200 mm/min, so it's relatively slow, and taking only 1.5mm per pass now.

To summarize: if I'm getting a near perfect 101.6mm on the outside height of the piece, why is a 50.80 finger not near 50.80? I'm now getting the exact same measurements on both axes.

I'm re-cutting the whole job now. I've left it to the best calibration results I could get earlier, and added a -0.2mm stock to leave to get the joints to fit. It might be tight, but I think it's going to work.

I just don't understand why I can calibrate to get to overall dimensions on the pieces to be exact but the joints aren't on point.

[daniel_lyall]

What post processor are you using?

[davidpichevin]

I've been using this post processor:

https://github.com/Strooom/GRBL-Post-Processor

[daniel_lyall]

All I can go on is what I see cam is doing and what the file looks like in a back plot.

[seth.madore]

1200mm/min is still rather fast for a light duty machine, it really is. Try cutting it even further, perhaps in the 600 mm/min range.

As @daniel_lyall is suggesting, this is quite likely a mechanical issue. Why do we say this? This issue comes around with regular frequency; "My machine is not cutting to size, what's the reason?"

Every single time it comes up, it is traced back to an issue outside of Fusion. So, as much as we can help on the Fusion side of things, what happens at the machine and control are out of our ability to troubleshoot, without being present at the machine. 

Try this; Reduce the feed by half. Reduce the depth of cut so you are only scratching the surface of the part. Measure it. Run it again, measure it again.

How are you measuring the part?

[davidpichevin]

I'm pretty sure the machine can handle 1.5mm depth, and I ran my first ones at 1000mm/min. It's always pretty slow and conservative. I'm measuring with a couple calipers.

I'm having a problem with the logic here. If my machine is calibrated well enough to consistently cut the piece at 101.6mm overall width and height, just about exactly, why is the 50.80 slot or tenon on the same edge so far from what it should be? (almost a half millimeter is a lot of difference even for a light duty machine). 

I solved my issue this time and ended up having to file the fingers down to fit, since this was just a simple part, but I'll continue working on this so I can get it ready for the next time I do the same joinery. I wanted to ask here in case there's a setting in my cam that could explain this, as my overall sizes are pretty much spot on and everything is tight in the system.

[Stuart-H]

if it’s a ball screw machine it could be end float on the screw and or nut going only one way it would be accurate , in other words mechanical backlash

rack and pinion same as above it could be loose 

belt drive loose pulley 

write some gcode to move the machine in G0 say x100 x0 assuming you zero before you start 

use a dial test indicator to start from does it return to the exact same place to say 0.001mm if not you need to look

[davidpichevin]

I wish it was that, as I checked many times, but I'll do some more testing after I'm finished with this project (the temps here in Tucson make it hard to remain in a garage that isn't cooled). I can say that the new calibration I did yesterday made things worse in terms of overall size of my pieces, so I'll be reverting back, then test some more.

[daniel_lyall]

Try this move the machine from 0 to 100 and back to 0 10 times then check that it is at zero. If it is do a engrave toolpath with 10 letters and set if it returns back to zero.

simple 5 min tests.

[Stuart-H]

To add to Daniels post hope you don’t mind

we are not talking about the dro on the control with these tests you must test with a dial indicator set at zero on the table to the cutter side if it returns to zero then all is good for that axis now do Daniels first test but in y 

you have to completely eliminate any mechanical errors before you look any further 

could you please explain how you are doing the calibration for me it was just a maths problem with the pitch of the screw , etc enter that it was good

Stuart

[davidpichevin]

Thanks for the suggestions. I will recheck when I get a chance. Note that I'm an experienced woodworker, mechanic, and a web developer as my day job. I'm well used to measurements and precision. Still can't explain why an edge is the proper length while a joint on the same edge isn't, or why a tenon on the same edge is too short, and a recess too wide, even though they should be exactly the same size, while the machine's hardware is tight, and errors are consistent. I've been working with this machine for more than half a year. Looseness usually leads to inconsistent results, rather than consistently the same error.

I attached an example of something made recently, where all pieces (including the circles/discs, the letters etc.) fit in with each other as inserts/pockets. Nothing was loose. Only a little bit of sanding required here and there as I didn't build tolerances into the design....but....this wasn't produced or CAMed with Fusion...

I also have some precise drag bit engraving that comes out really well.

I should actually try to import my test file into another program with a CAM section and see how it measures.