YAY! IT was a success. Last Fitness was .5043329.
Thanks for all your help David and team!
So my question is now does it need to be adjusted in settings again when I introduce a work piece sheet of wood?
I must admit the sled did lift a tiny bit closer to the edge of the spoil sheet, but I’m thinking that won’t be the case when we introduce a work piece sheet? Lifted about 2mm on the lower left belt section. Thoughts?
TimmyZ wrote:
YAY! IT was a success. Last Fitness was .5043329.
Thanks for all your help David and team!
great!
So my question is now does it need to be adjusted in settings again when I introduce a work piece sheet of wood?
in a perfect world, yes, every time you change the alignment of the surface you
ride on compared to the anchors, the Z offsets should change.
In practice, it’s unlikely to matter a lot, the further you are from the
anchors, the less it will matter.
but this is again, theory, testing needed.
I must admit the sled did lift a tiny bit closer to the edge of the spoil sheet, but I’m thinking that won’t be the case when we introduce a work piece sheet? Lifted about 2mm on the lower left belt section. Thoughts?
it probably won’t matter in practice, but something to be aware of.
David Lang
Very good to know. As long as you know the dimension with your spoil, it shouldn’t be hard to update (assuming you’re already in the habit of measuring the thickness of the wood you’re cutting).
Bar has accepted the idea of adding a ‘thickness’ variable so that you can enter
one value to change all of them (to adapt to different workpiece sizes without
having to do the math and change 4 variables)
David Lang
I think this question comes
Yeah, I think that makes a lot of sense. Adding a variable would be really helpful and should be pretty easy and quick to program.
So just to be clear… Am I manually setting that for the 4 arm Z settings when I introduce a workpiece? So subtract or add whatever thickness measured from the sheet of wood or as long as the work piece is 3/4 inch thick the calibration presets already accounts for the work piece?
TimmyZ wrote:
So just to be clear… Am I manually setting that for the 4 arm Z settings when
I introduce a workpiece? So subtract or add whatever thickness measured from
the sheet of wood or as long as the work piece is 3/4 inch thick the
calibration presets already accounts for the work piece?
in a perfect world yes.
in practice, the flatter the belts, the less it matters. Again, we don’t have
good testing of this (I think a lot in terms of theoretical problems, so not
everything I spot is a big problem in the real world, and it depends on what you
are doing. Are you doing theater sceanary? or small parts that must slot
together exactly?)
To put some numbers together here to see how much error is being introduced:
if you are near a corner, you have ~2’ of belt extended, let’s say 600mm to have
a nice round number.
If the real Z offset is 100mm, then the actual x/y distance to the anchor is
591.6mm
if the real Z offset is 120mm (~3/4" difference), then the actual x/y distance
is 587.9mm
3.7mm of difference when you add the height
now let’s do a lower arm
If the real Z offset is 38mm, then the x/y distance is 598.8mm
if the real Z offset is 58mm (again adding a 3/4" piece) then the x/y distance
is 197.2mm
1.6mm of difference
if you set things up so the anchor is level with the arm, the x/y distance is
600mm
add the same 20mm and the x/y distance is 599.7mm, a 0.3mm difference
now let’s say you are 4’ from a corner (1200mm)
Z offset, distance
38, 1199.4mm
58, 1198.6mm .8mm difference
100, 1195.8mm
120, 1194.0mm 1.8mm difference
20mm, 1199.8mm .2mm difference from flat
at 8’ (2400mm) from a corner 100mm z is 2397.9, 120mm z is 2397.0, a .9mm
difference
But keep in mind, if you add height to make the belts flat, but your anchors
have 3mm of flex in them under full load, you have added more error than you
eliminate. Flex is a bigger problem in most cases than the angle.
This is actually worse in the corners than I expected for the upper arms, and
this sort of error really causes grief if it happens during calibration, so it
may be that we need to mention setting the Z offset values and being sure the
router is all the way down before doing the calibration @bar, @anna
David Lang
So… the highest arm will probably have the most flex if mounts and arms were the same height, and all arms zeroed put for the “Z”. Maybe I should print some spacers buy longer bolts and 1 mm thick washer and do some real world testing. For the Team. Thanks again David.
Of course after I cut something successfully.
TimmyZ wrote:
So… the highest arm will probably have the most flex if mounts and arms were
the same height, and all arms zeroed put for the “Z”. Maybe I should print
some spacers buy longer bolts and 1 mm thick washer and do some real world
testing. For the Team. Thanks again David.
jsut a note that a stack of spacers is far more likely to flex than those same
spacers glued together.
better wood spacers glued together than metal washers just stacked.
and as you see from the numbers above, small Z differences over long x/y
distances don’t matter much, so you don’t need to try and get too precise. The
arms are just over 20mm apart, so stacks of 3/4" disks will get you pretty
close.
David Lang
I cut some longer pieces of 3/8" threaded rod that I have going down to my concrete anchors. For the two tallest arms (top left and bottom right), I threaded on extra 3/8" couplers (~1.75" long), as spacers, and tightened them down tightly against the floor. For the tallest, I tossed on a couple extra nuts for now to take up a little extra space (also tightened down). With everything going down to the concrete anchors nice and tight, there seems to be practically no movement in the mounts when the belts pull against them (note that this is just an unaided visual inspection; I can try checking again with a straight edge sometime later).
After raising the mounts, I tried calibrating and had significantly better fitness values on 5x5 and 7x7 grids, but whenever I try calibrating on a 9x9 grid, my machine keeps spontaneously stopping mid-calibration (light still blinking) and needs to have the power cycled off and on before it will respond to anything (I don’t think this issue is related to tweaking the mounts). As an additional bonus, after the successful calibrations, I noticed that the frame dimensions produced by the calibration match my actual frame dimensions much more accurately.
I haven’t had a chance to make any test cuts since the changes, but I am hoping that this all will translate to an increase in their accuracy.
Andith wrote:
After raising the mounts, I tried calibrating and had significantly better
fitness values on 5x5 and 7x7 grids, but whenever I try calibrating on a 9x9
grid, my machine keeps spontaneously stopping mid-calibration (light still
blinking) and needs to have the power cycled off and on before it will respond
to anything (I don’t think this issue is related to tweaking the mounts).
can you give any more information on this. Is there any error message on screen?
did you change the Z offset values to match the new mounts?
David Lang
After raising the beltend mounts, I set each of the z-offsets to 0, since the beltend mounts were essentially even height with the arms. I was able to calibrate on a 5x5 grid and a 7x7 grid. However, when I tried to calibrate on a 9x9 grid, the calibration began but stopped after around the 25th calibration point. No errors or anything appeared in the log, but nothing else appeared either. I let it sit for quite a while in case it was processing something, but nothing ever updated in the log. The lights on the machine appeared normal, and the fans were still running as if it was trying to do something, but it never moved and nothing on-screen ever updated. Eventually, I had to power-cycle the machine, release belt tension, and retract the belts. I tried another time, and it did the same thing, but around calibration point 13.
I didn’t measure precisely where the M4 was positioned when it froze during calibration, but it stopped in the bottom-right quadrant (maybe ~18" right and 12" down from center). When it stopped at the 13th calibration point, I believe it was roughly in the same spot as when it stopped at around the 25th point.
This is the weird part to me. It shouldn’t just stop without saying anything 
Please forgive me if this has already been covered (I bet that it has), which firmware version are you running? We had an issue like this early on, but I haven’t seen it in a long time
It currently has the latest release (v 0.84).
I have been having the weird thing where after I update the firmware, the log always says that the FluidNC version is one number down from what is actually installed (I posted screenshots a few weeks ago of where it says that). I decided to try a firmware reset earlier this weekend, in order to see if that would correct the issue of the log saying that the firmware is a version lower than it really is. The issue of the firmware version remained after the full reset. Regardless, given the full reset, I know for certain that the firmware is up to date.
To follow-up, I tried again yesterday and was able to get it to complete calibration with a 9x9 grid and calibration area of 2000 by 1000. Fitness score at completion was 1.028 (rounded).
Cool design. Do you have it posted somewhere?
Hey @Andith what was your score before raising the anchor points? Curious to see how much of an improvement you got. If its significant, would you mind doing a small write up with a few photos (maybe in a new post?) I’m considering doing the same!
Thanks!
-Sam
Before raising the corner anchors, I was getting fitness scores around 0.4 to 0.5 on 9x9 grids. I can do a more extensive write-up later if you’d like, but there really isn’t a whole lot to tell. I basically summed it all up above:
My frame is horizontal on my garage floor. I have 3/8" threaded concrete anchors in the floor, then some 3/8" threaded rod into the concrete anchors. Above floor level, I have a 3/8" threaded coupler (1.75" in length) that I bored out halfway to 10mm for the beltend pins to go into. I basically just increased the length of the threaded rod and filled the space with additional couplers, nuts, and washers, and tightened them all down really tight to prevent the rod from flexing when the belt is pulled. It’s a bit jerryrigged at the moment, but is working. If you have access to a 3d printer, then I think making some feet/mounts/anchors at the various arm heights (like @tzarza has posted above) will likely work better. The height you need to raise each anchor will depend on your setup as well as the thickness of your spoil and ideally also the average thickness of the material you will be cutting.
Thanks. Not yet. Still need to test. In had to adjust settings for each arm to get the right Fitness down.