if the dead zone would be 30 degrees, the framesize should be 2750x3660
and how large does the grid need to be to hit these problem areas with a 8x10’
frame and with a 8x12 frame? (how likely are we to have people with calibration
problems hitting some of these areas) I know this doesn’t explain it all as
people are running into problems with small calibration patterns
and what is the name of the curve that is formed by keeping the angle between
two arms constant and running them along two points? (it’s been WAY too long
since school)
David Lang
you are right, it is not a real ellips, but it gets close. And I don’t know the name of that curve, it might be a hyperbole?
on the 8x10’11" frame, the grid could be 1800x800mm
It could be bigger, but then you should leave the corners out
Arjen
I figured it was something like that, but there is some text in the zoomed in
views that makes it hard to be sure what’s what.
(and it doesn’t allow for rotation of the sled, but you could anchor a belt to
the edge of a swivel area and work with that)
David Lang
note that it would only be the top and bottom dead zones that shrink, not the
side ones.
and they may shrink a little more than my quick check showed. I looked for lines
tangent to the motors to hit the lead screws, but the arms are actually narrower
out past the motors, not wider. But since it’s 2am for me, I didn’t go out and
try to actually measure the arm 
David Lang
good point! would it not be possible to shrink the side dead zones too?
I don’t have my machine yet, so I am deducing sizes by looking in the assembly instructions and importing the beltguard stl into my CAD.
Here it is 11PM and I really should get to work 
for a 8x10 frame, the grid should not be greater than 800x1600
I’ll continue with this tonight,
Arjen
arjenschoneveld wrote:
you are right, it is not a real ellips, but it gets close. And I don’t know the name of that curve, it might be a hyperbole?
I don’t think it’s even a smooth curve, you have a curve with the narrowest
portion at the edge centers that goes out towards the anchors that intercepts a
curve in the opposite direction ancored on the corners opposite each other.
David Lang
arjenschoneveld wrote:
good point! would it not be possible to shrink the side dead zones too?
not easily, the other two hold the linear bearings, so they can’t just be cut
off. it may be possible to squeeze a degree or two out of them, but not much
I don’t have my machine yet, so I am deducing sizes by looking in the assembly instructions and importing the beltguard stl into my CAD.
and I’m working from the top/bottom .step doc until I go out and get more
measurements tomorrow. It actually says the dead zones are 41 degrees, but
that’s assuming the motors exactly match the diameter of the notches in the
arms, they are close, but I think the motors are just a little smaller
for a 8x10 frame, the grid should not be greater than 800x1600
This is good info for @bar to check. If he can throw out calibration points
outside that range and see if it improves his results, it may help stabilize
things.
David Lang
I didn’t forget, but I am assuming a vertical orientation (I should have said) and am assuming that the weight of the power cord and vacuum hose will try to orient it right side up and thus not allow it to rotate and improve the situation. It may be forced to rotate some, but there will be force on the belt and the belt will be deflected some as a result.
If true, then I wonder if the firmware could be changed so that when the Maslow is in say the upper left corner region, the upper left belt is treated only as a support and its encoder reading is ignored. The location of the Maslow would be computed from the other belts which should not be deflected.
It might even be that two belts have to be ignored and the position calculated from the other two. Assumptions about rotation or non-rotation would have to be made.
WFD wrote:
If true, then I wonder if the firmware could be changed so that when the Maslow is in say the upper left corner region, the upper left belt is treated only as a support and its encoder reading is ignored. The location of the Maslow would be computed from the other belts which should not be deflected.
It might even be that two belts have to be ignored and the position calculated from the other two. Assumptions about rotation or non-rotation would have to be made.
I think it’s a bit worse than that. When you hit the limit, the affected belts
are shorter than the maslow thinks they are, so there is more tension on all the
belts, so they stretch more…
and when you hit the limit, it’s not that one specific belt is inaccurate, it’s
that you don’t know the sled rotation (it will depend on the tension in all the
belts, drag from the vaccum hose, etc) so you don’t know which belt or belts are
wrong.
David Lang
WFD wrote:
I didn’t forget, but I am assuming a vertical orientation (I should have said)
and am assuming that the weight of the power cord and vacuum hose will try to
orient it right side up and thus not allow it to rotate and improve the
situation. It may be forced to rotate some, but there will be force on the
belt and the belt will be deflected some as a result.
those forces are pretty smal in comparison to hitting the limits.
David Lang
I looked at some linear bearings today, and they were 15mm outside diameter, so they should fit in to a 20mm tube. I guess that would be enough, I think the existing upright is 30mm?
Is that stepfile on Github?
Arjen
arjenschoneveld wrote:
I looked at some linear bearings today, and they were 15mm outside diameter, so they should fit in to a 20mm tube. I guess that would be enough, I think the existing upright is 30mm?
Is that stepfile on Github?
David Lang
I tried to find the exact spots there they made contact so here’s the corner:
And here is the top center:
This is on a 8’x10’ which is the minimum that is workable.
I think this is a valid concern, but I don’t think it makes our top ten list of things to worry about. I’m absolutely positive that it’s not causing the calibration issue that we are seeing because the calibration process never runs that close to the edges of the sheet.
The length of the belts is actually 14.5 feet / spool so a little more headroom there.
Overall I think that this is a valid concern, but I don’t want it to distract us from the real issues which are preventing folks from cutting things. Let’s get to the point where we see some evidence that this is really an issue in practice before we stress about it too much.
1 Like
it is looking better than my drawings 
Thanks for the pictures
1 Like
it looks like we could potentially get down to 27-28 degrees around the lead screw (would require a slight spacer below the bottom arm and the screws getting trimmed off (or euqivalent) and the top arm of the anchor being trimmed
and stock looks like 39 degrees (note that draft angle may tweak this a degree or so in either direction)
with the benefit of 20-20 hindsight, making the arms a bigger diameter and putting them outside the router is looking extremely attractive 
You mean, making the spools larger so the router, z-axis, and vertical supports all fit inside? I think that would be a great idea.
Also, I think that means the spools and arms would not / might not go up and down with z-axis movement, so that source of error can go away.