Hi Guys, Firstly thanks to the collective community for all the info and open source approach, it's enabled me assemble a pretty effective CNC using the C-Beam XL design (my first CNC experience). I'm hoping people can share their experiences, and whether I'm getting what I should out of the machine. Noting that 80% of use case is for cutting Delrin, 10% alloy and 10% timber. I've began to cut my first jobs (e.g. MDF wasteboard grid holes and some acrylic gears) and just wanted to get a sanity check on expectations for tolerances. I've done a linear direction test with a dial indicator and am measuring backslash as below. Continuous directional movement is consistently 0.01-0.02mm for all axis. Backslash: X = 0.12mm (this is my primary concern, I've tightened grub screws and lock collars are tensions in the end blocks, but can't seem to reduce it. I also note is the longest axis, not sure if that matters) Y1 = 0.06mm Y2 = 0.07mm Z = 0.08mm On actually cuts, I've seen: - Roundness deviate on 7mm hole, by 0.2mm (Acrylic, Fusion 360 Bore) - this is a concern to me, outside my acceptable part tolerance - Roundness deviate on a 8mm hole, by 0.1mm (MDF, Fusion 360 Bore) - Roundness on a 40mm circle, 0.1-0.2mm (Acrylic, Fusion 360 2D Contour) I note different speeds and feed for the above, which likely accounts for variation in 7-8mm holes. I'm still working out that stuff and will run more tests on weekend. The machine is very consistent though; spitting out the same part with same error if that makes sense. Thanks in advance. Damien
And make sure to measure the endmill too. That 1/4inch may have been ground down to an actual 6.27+-mm diameter instead of 6.35mm in the factory (or worn down) for example - especially on cheaper endmills we often see the quoted size and actual size differ
This may be a dumb question, but you said you did linear direction tests to measure backlash. Did you also calibrate each axis?
I was all over the shop to be honest so can't really say. I'm planning to sit down tomorrow and use science to cut some stuff. I haven't yet moved machine to garage (from living room table), so it's a bit messy to run in the house.
I did have that issue, my 6mm end mill was actually 5.8mm, but I realised that and the reported numbers were with the correct tool dimensions.
I did calibrate the axis, though from what I can tell the continous linear direction is very good. From what I can see the calibration doesn't take into account backslash as it asks you to move positive, mark and then move positive again, never goes negative.
I measured direction and backslash with what appears a decent dial indicator with 0.01mm precision. Indicator was has mounted to the axis being measured, and measured against the spindle mount. It was pretty square too.
There's no reason for the calibration to ask you do measure backlash, as grbl doesn't have backlash comp built in. All backlash mitigation must be done mechanically, or move to a different controller (realistically I doubt it's high enough to worry about though). All parts produced in any kind of volume have to be adjusted for deflection error on the machine though, even on 6-figure machines. That's to be expected.
Calibration should be both positive and negative. So, when calibrating, you command it to go as far as you can so the error has more chance of showing itself. Record the measured distance vs the commanded distance. Send it back to the start. Change the steps per mm to what they should be. The change to the steps/mm can then be calculated like this: (Commanded Distance/Measured Distance) * (Calculated steps/mm) = New Steps/mm. Then do it again until it is spot on. Using this method, I can send my machine 1000mm and it goes 1000mm. It then returns back to its exact starting point. Maybe you did this, but to me it was not clear by how you described your calibration method. Disregard if you did it this way. gnea/grbl
This is a sign of very low backlash, not a sign of good calibration- even if your steps/mm were off by a factor of ten, if you have good mechanical backlash control you should return to the exact starting point. The positive-mark-positive system that I'm guessing is built into CONTROL is to eliminate backlash within the measurement by ensuring that the screw is travelling in the right direction for calculation purposes. If you move slightly negative, mark, then move negative by the commanded amount, it should produce the same result.
I followed the OpenBuilds CONTROL calibration wizard, which I guess follows a positive-mark-positive model. I did a few test parts today, with some mixed results (all 2.5mm acrylic, 1/8" 2 flute end mill, measured at 3.12mm and actual dimensions set in Fusion). - Rectangle 33x20mm, 1/8" 2 flute end mill, 2D Contour 18000rpm, 1830mm/min, 1mm stepdown, within 0.02mm tolerance, very consistent across 2 parts. - Circle OD 40mm, 2D contour climb, 18000rpm, 1000mm/min, 1mm stepdown, with 0.2mm off round - Circle OD 40mm, 2D contour conventional, 18000rpm, 1000mm/min, 1mm stepdown, with 0.05mm tolerance - Same circle above with ID 8mm, Bore conventional, 18000rpm, 1000mm/min, hole well undersize, by 0.3-0.5mm undersize and off round I then ran a series of 8mm holes using a spread of speeds, feeds, climb, convention, etc..., all were consistently 0.3-0.5mm under size and off round. I'm a bit lost on the 8mm hole issue to be honest given the other seemingly very accurate parts coming off the machine. 05 Is the tolerance setting of 0.01mm what I should be using? thanks, Damien
Is smoothing turned on in Fusion? Ideally you want more G2/G3s and fewer G1s for circular toolpaths. grbl may be getting swamped by too many short lines, hard to be sure. I suspect more that you're seeing a lot of deflection under the shorter, higher-acceleration toolpaths. How much can you flex your machine if you push on the tool?
