Oh ok, good then. It’s unlikely that the belt calibration from a week ago would not be good enough.
Quick test you can make: if you have a dial indicator on hand, install it against the shaft of the endmill, try jogging manually by 20mm, and measure how much the bit actually travelled. A caliper can be used too if you figure out a way to hold it. Or whatever method you used for that calibration last week, as long as it is precise enough
If the value is off doing this “air” move, then you’ll know something needs to be (re)tuned in the calibration.
If the value is spot on, then it points to something during the cut:
either the bit is not exactly the diameter you think it is (unlikely for an Amana bit, but who knows)
or there is tool deflection (but this is unlikely at 1mm DOC in ply on a 1/4" tool)
I will rerun the belt stretch calibration, previously I used just a hole in MDF, but your method involves 3 square pockets which will be easier to measure between the squares. Before I was getting very small deviations, which I accounted for as the accuracy of my calipers. But I will run this, change settings even if it is slight and see if I can get more accurate dimensions.
Adaptive is not a finishing strategy. Run Adaptive rough with stock to leave then contour and dial in with stock to leave. Your tolerance and smoothing values matter.
With steel core belts I don’t run a calibration and everything is compensated for in cam. The Shapeoko is more repeatable than it is accurate
For the hole, adaptive rough with stock to leave then come in with a Bore op and run it with a spring pass, adjust stick to leave for final dimensions.
Your dimensional error there seems to be quite similar to what I see on my XXL with the kevlar cored belts and everything else nice and tight.
I see about 0.1mm of lost motion (backlash) in both X and Y axes and have to add this in as an offset on cuts when modelling up the CAM. Not sure if everyone sees this level of backlash on the longer axes but I’m unable to rid myself of it.
Your outside profile cuts are coming out approximately right, but the inner profile, the hole seems to be about 0.3mm down? That’s not inconsistent with the +/- 0.1mm I have to add for “backlash compensation” to take account of the belt stretchiness. In fairness to the machine, this is pretty good for belt driven and the repeatability is very good once you’ve got a set of dimensions dialled in. What I normally do is put a parameter “backlash” and frequently also one or more for “clearance” into Fusion and then use a tool like offset face to apply these compensations to the parts prior to the CAM.
If you want to check backlash for yourself, if you have a dial indicator, set it up and jog the collet (or Z carriage) forward to the dial indicator (with the router off preferably) and then zero the indicator, jog forward another 1mm and you should get approx 1mm of movement, jog back and you’ll get something less than 1mm of movement, that’s the backlash. I don’t think you’d get a useful measurement at this scale with calipers.
Calibrating the belt tension / distance is absolutely worthwhile, Mike’s caliper method seems to be the easiest, measuring over 100mm is much easier than 10mm and takes you well outside the backlash / step compensation scale.
If you’re calibrating over a shorter distance (like 10mm) then step in one direction only whilst measuring to avoid the backlash, e.g. step into an X measurement going right, then move right to measure.
@Vince.Fab I am using adaptive with stock to leave, then running adaptive again to get to the final dimension. I can change that finishing pass to a contour and see what I get. Right now, I am just working on getting the part within dimension, then was planning on optimizing tool paths, feeds and speeds, to get a better finish. This is why I left is as adaptive for now, but will change to contour.
@LiamN I Expect there to be some dimensional error, due to the nature of steppers, and belts. I am used to designing for this, as I have been 3D printing for about 7 years now and that also requires adjustments as plastics shrink when cooled. Again I just want to get something that is within 0.1-0.2 mm and is repeatable. I will run a belt calibration and test for backlash shortly after. I can feel some play within the stock Z axis, not with the eccentric nuts, but a slight wobble between the plates. I am looking for an HDZ and am on the list to be notified when they go on sale.
Yes, shrinky 3D prints are a very similar exercise, not your first rodeo
As Vince says, the repeatability is very good once you have your machine and toolpath understood. I am cutting dovetail joints in multiple fixturing with gaps of well under 0.1mm, with the appropriate backlash compensation applied in CAM.
Moving to the HDZ made a substantial difference to the machine when cutting, the Z plunge depth is much more predictable now and Z backlash is small enough for me to have trouble measuring it and not worry about it. I don’t think the HDZ made any difference to my dimensional accuracy on a finishing toolpath though as those are deliberately a very light cut and I frequently check the “repeat finishing pass” option in Fusion, this effectively removes the cutting force deflection on the final pass.
There are various sample shapes out there to mill for testing calibration and backlash but I was never quite able to figure out which error was which. I have found a dial indicator (one of the cheap end of the Mitutoyo range) and a basic magnetic stand has allowed me to obsess over small measurements in much greater sleep denying detail than I could previously when I didn’t know where the dimensional errors were coming from.
Well just my luck, I cut a belt stretch calibration piece out of MDF and somehow during their process a screw was left in the MDF below the surface. Well there goes that Amana endmill, I reordered another so I cannot retest the cut until the end of the week.
I set it up, to have 60 mm between squares, and these are the results I am getting on my calipers.
60.07 mm, X-Axis
60.05 mm, Y-Axis
I will adjust the stepper values, change to a contour for the finishing pass and recut when the new end mill arrives.
I ordered the HD eccentric nuts and are waiting for them in the mail. My goal is to use this machine to produce a good amount of products, so I need repeatability and reliability.
So, I ran two tests today. Please note, instead of the straight cutter used previously, I used an Amana Compression Endmill (46170), as I am still waiting for a straight cut in the mail.
First, I ran the same cut and only changed the adaptive toolpath, to a contour toolpath. This was in hopes of trying to isolate where the issue was coming from. Here are those results:
Left Arm: Goal 16 mm, Actual 15.95 mm
Right Arm: Goal 20 mm, Actual 19.91 mm
Center Hole: Goal 21.7 mm, Actual 21.5 mm
In Fusion I changed the dimension of the middle hole from the previous cut, but this change reduced the 0.3 mm undersized cut to 0.2 mm, so an improvement.
I then used my calibration data, sent the adjustments to GRBL, using Carbide Motion and using a contour tool path. this is what I am measuring.
Left Arm: Goal 16 mm, Actual 15.88 mm
Right Arm: Goal 20 mm, Actual 19.88 mm
Center Hole: Goal 21.7 mm, Actual 21.49 mm
It appears the difference in the center hole between a belt calibration and the stock GRBL settings, did not change the value significantly. However, the dimensions on the left and right arm, changed with the belt calibration. I will probably run another calibration test in something better than MDF, as the edges of the cut out’s may have been changing while measuring as MDF is rather soft.
If this is as good as I can get, then I can deal with that and can account for it within design. I just wanted to narrow down on any possible errors, or mistakes on my part, such as the use of adaptive as a finishing pass.
Arguably the best way to arrive at a spot on dimension is to sneak up on it by cutting with a suitable roughing clearance, measuring, adjusting the file and recutting and repeating as necessary.
You might want to try a different method of calibration, which can also tell you exactly where you have backlash, along the way if you’re looking. You’ve probably got all the stuff needed already if you’ve been doing 3d printing.
@WillAdams I am not seeking perfection in anyway, I was just confused as to why the majority of the dimensions on this cut were within +/- 0.05 mm while the center hole was off by 0.3 mm.
@mikep I will look into that calibration, thank you for posting that.
