Hi there,
my MPCNC is on its way to preliminary completion. Today, I measured the tool offset towards the table, I used a 100mm raster, thus getting 48 points on my more or less 500 x 700 mm² table. I entered these in a simple Excel table and fiddled around a bit to achieve a nice 3D representation.
The scaling in the vertical axis is 100ths of a millimeter, showing that the error over the whole working area (in the uncorrected setup) is not worse than half a millimeter. Not bad. There was no tool mounted on the Z-carrier yet, so, this is the pure machine geometry showing here.
The next step will be to raise the three lower legs to a level that lets them all have the same offset.
Of course, this does not say that the achieved tool plane at a given Z value will be an ideal plane. The irregularities of the ply board are not corrected then, the elasticity of the tubes, once loaded with the weight of the spindle (a RT 0700 right now here) will have to be taken into account, but the corner points at least will be leveled. After adding the necessary pieces of feeler gauge blades I’ll measure again and will report, if there is an interest.
Here is a new measurement after adjusting the height of the four corner crossings. The area shows a pretty large practically even plane with up- and downward slants at the borders.
This is still measured without any tool mounted, i expect a hang-through of siginificantly more than the error in this idle state.
Even adding support pylons under the frame rails would not compensate that totally since the tube cross carrying the gantry remains in free flight. Pretty complex if you break it down. I’ll take the remaining error as a given property.
Even further compensating the asymmetries in the four corner points (the only really pretty fix points) would worsen the now existing even plane again.
And: Further equalising the distance between the base board and the frame plane does not mean that the tool path would run on a flat plane but just in a constant height over the base. Like sea level and mountains.
At least, the total variance became smaller, not worse. Real world success.
Likewise, planing a sacrificial bed will just create another big flat saucer. Limits of the expectable.
HTH.
So that graph is kinda upside down correct? The blue area is a low spot? So you have a maximum deviation of .440mm, and that is from opposite corners. A bit of a shim will not clear that up some, or maybe you have drag chains, vac hose, or something tugging on it?
I think .4mm deviation is well within everyday usefulness but you have the measuring tool set up I would be getting out sheets of paper to shim it up!
The question is if the bed is the reference or the rail level. Whatever the convention is, it seems that on the first measurement, the foot-tubes may not all have settled on the bed. Even now I am not quite sure, I will loosen the clamping screws holding the feet on the tubes.
I even use thin metal shim stripes to level the feet, and I will probably start over once more from scratch, making sure now that all tubes and structures are sitting on the base at their lowest possible position. A bit like trimming a rig on a sailboat. Or tuning a piano…
Thanks for your motivation
You are right, sort of a chicken and egg situation. I would guess the only errors I would like to see are in the middle. If the table is warped the legs should follow the warp and be your Zero points. I have never really tried though, if I get it under a millimeter with a tape that does it for me. Most everything I do is through cuts.
Same here. Through cuts or pockets where it does not depend on a tenth of a millimeter.
And I won’t start thinking about a diagonally folded table and such… Looking at what others use, my efforts look like a big chunk of overshoot, but …
And for keeping a drag knife at constant height over a surface there are other options.
It would be interesting to surface the spoil board and measure again. The theory is that you can easily flatten 0.4mm, but the trick is, how consistent is it? I am guessing it is very consistent, but there will be some error that is just random. Thermal expansion, binding, wires, they will all add up to odd forces that will hopefully be very small.
Manually. Old school mechanical gauge dial. Using LinuxCNC, I just push the instrument around in a raster, making sure it has moved in the same direction before reading it. Alone the friction forces of the probe on the surface cause some bending (invisible to the unsuspecting eye, but apparent in the data).
Using Marlin, I even could use automatic bed leveling, like I did on my first 3D-printer, the famous OrdBot Hadron (sic!) by Bart Dring. See here. Five years ago again already, sigh…
Highly effective, but not directly applicable for a milling machine.
The highest (verbally) impact will be the weight of the router/ spindle AND the cutting forces. A spiral bit can develop very much power, and not every material leads itself to the use of twistless flute bits, like the 2-flute cutters made for wood by Sorotec.
Flattening the spoil board, well, yes. But if your swinging machine sinks down in the middle, you produce saucers.
Here are the latest measurements. Feeling motivated by Ryan’s suggestive remarks, I tore out all previously set shims and started from scratch, first making sure that the feet tubes sat really firm on the base board. Then I did my best to approximate all four corners to an even level. One remained an exception since it showed during the measurement already that the base board had a significant fall off in a very limited area, close to the corner. The good news is that that will be the least used corner anyway, and if, only for 2D through-cuts. The rest of the board has a few dents and bumps, but in the end, it is a natural material, subject to air humidity, temperature changes, ion storms and volatile moods.
When replacing the outmost corner value with a coarsely extrapolated 90 (meaning an offset of 0.1mm!), it looks even better. I left the original value in the table, though.
The unit of the numerical values in the table is 0.01mm.
Thanks, but still the native, unloaded system. Nevertheless, before I started my upgrade, you said the new design is a totally different story (related to the first one). It is. Cudos.
Oh man, I woke up this morning just absolutely dying to work on the new new machine. I am feverishly packing boxes to try and keep some of this excitement going for the new one. Doing my best to one up myself. I have no idea what happened but today my brain is going 1000miles an hour with ideas for every single machine. I seriously might skip a day of shipping just to bury myself in CAD! So by the time you are ready to upgrade again maybe it will actually be ready!
If the machine sunk in the middle, by 0.4mm and you flattened it, the material you’re trying to cut won’t be able to resist confirming to the saucer spoil board. I think you’d still get great results.
