Many weeks later I finally have an update: I surfaced the bed of the Nomad aaaand it’s still not flat…
Before:
After:
You can see that the variation before surfacing was ~48µm, which is also the same as the variation after. All that’s changed is how it’s uneven.
I don’t suppose anyone has any ideas?
Ide say ask support for their nc file for decking. Proven right?
What was your toolpath like? I know some Shapeoko guys run passes in multiple directions to get flat stuff. Isn’t going to be the best looking though.
The toolpath was just a basic climb milling face operation. I didn’t even use CAM for it, I just wrote a macro in G-code directly:
sub face
; X max, Y max
#30 = 210
#31 = 210
G0 Z1
G0 X[#30] Y0
M3 S10000
While [#5002 < #31]
G1 Z-0.02 F300
G1 X-5 F300
G0 Z1
G0 X[#30] Y[#5002 + 1]
endwhile
M5
endsub
It could be that my probing routing is off though. @Vince.Fab, how troublesome is it to remove the bed and reinstall it properly? If I can take it off, I can check it properly with a flat surface and a DTI.
Nice. Writing stuff like that is beyond me.
Bed is easy to remove, bolts and a single locating pin on each side.
It looks complex but it’s actually really simple:
I’m sure there are some folks at SMW that have done this stuff that you could learn from, there are some damn nice machines there to play with. If you end up with a machine that supports macros and want to learn, let me know.
Cool, I’ll give it a go.
This was a long time ago, but with the help of Carbide 3D I discovered a small washer was missing under the bed of my Nomad.
Interesting. About double the tolerance of mine. I surfaced along the x-axis. Climb only. Stubby 1/8" 3 flute TiCN end mill.
How repeatable is this flatness measurement from run to run?
At these dimensions are you down to the level where thermal expansion could distort the bed this much?
What probe do you use to get the surface measures?
Thanks, I’ll have a look when I take the bed off.
Quite repeatable. Here’s another, higher resolution map from a different run:
Hmm, a good question. Apparently the thermal expansion coefficient for Aluminium is 23e-6, so if the temperature varies by 15°C, I’m looking at 0.00345x expansion/contraction. The bed is 6mm thick so if my maths is right, I’m indeed looking at a 20.7µm difference, which might affect the results.
A 3D-Finder.
Okay, I pulled it off the bed and measured it with a DTI and it’s actually worse than the 3D-Finder reports. Running from the left edge of the bed to the right, straight down the center (i.e. along the X-axis), I see nearly 100µm of variation, around 6x what the 3D-Finder reports for that path.
So I suppose my machine isn’t square or the bed wasn’t bolted down properly.
Time for a surface plate so you can tell what is actually flat?
Are the mounting points for the bed flat and parallel? It’s possible it’s pulling out of true when bolted down.
It’s also possible that the X and Y axes are not fully flat, the Y axes may also not be parallel.
Time to get measuring if this delta is going to keep bugging you.
Sorry, I meant to add, I did measure that on a surface plate (well, measuring table but it has 3µm flatness so same thing).
I also have a straight edge, I can try to slip shims in between it and the bed too…
We’ll see, I don’t think I’ll put too much effort into it just yet since there’s a chance I’ll redo the axes in the not-too-distant future. I’ve updated support though, I’ll see what they say.
Not sure exactly what you measured now…
Are the plots using the 3D finder done with the finder in the Nomad collet?
If so how does the collet view of flatness compare with the measuring plate view?
I removed the Nomad’s bed from the machine, put it on the measuring table with a µm DTI mounted and moved the tip from the center towards each of the four edges, recording the difference between the minimum and maximum reading for each edge.
Yep.
The difference between the minimum and maximum reading using the 3D-Finder is ~45µm. With the DTI, it’s ~92µm.
And specifically, the 92µm variation is along the X-axis. Along Y, it’s only 30µm.
Interesting,
Seems like you’re finding the precision limits of the machine.
There’s evidently some flex or movement between the levelling pass and later measurement. This could be deflection in the linear motion parts (leadscrews, linear bearings) or it could be the machine frame moving.
And then there’s the underlying linearity error in the linear bearings, which seems to be concentrated in the X axis. It would be interesting to figure out a way to check the straightness of the X axis, is it high in the centre or at the ends? (wondering if it’s sagging with Z and spindle weight)
Hmm, the bed is bowed in the middle but interestingly, this is present on both sides of the bed, so the bed is actually a bit of a hyperbola along the X-axis:
The only think I could think of that would cause that is if the bed is under compression when it’s mounted to the Nomad, causing it to flex upwards a bit. That would cause a curve on the bottom side and a bulge in the top. By flattening it and removing the “bulge”, I would have created an intentation on the top side but the Nomad would see it as flat.
This theory seems to be supported by my before and after heightmaps. That “bulge” can be seen in the “before” heightmaps but in the “after” heightmaps, there’s no longer a bulge.
So,
What does the bed mount on? Presumably this is not flat if it’s causing additional curvature? Maybe you can flatten or shim this mounting to flat and then be able to deal with the top side?
Some wet abrasive paper stuck to some float glass or your surface plate may help in getting the plate flat, once the mounting points stop distorting it.
It’s interesting to read about making flat surfaces and the need for 3 different surfaces in:
Indeed,
For those who prefer their learnin’ in short video form;
Which even has civil war era cannons in it for those in the southern states…