For sure, especially considering the harder it ‘pulls’ (current measurement), the more it stretches!
Maybe my phrasing was off, I meant adjust the pulling force of certain arms at certain places on the frame. i.e. far left or right edges, lower the pulling force on the long belts so they don’t over power the short ones.
Not too hard at all, I actually had some code in there doing exactly that at one point, but I wasn’t able to see it improving the accuracy so I turned it off
The way it was working was just to add a little extra length to the lower belts (since this is mostly an issue in vertical orientation) when the tension was really high in the upper belts…basically if there is already a lot of tension in the upper belts from geometry, we don’t need to add to it much with the lower belts.
It looks like that code has been removed in a clean up at some point, but it should be too hard to add back. The equation for calculating the tension (from gravity in the upper two belts) is here Layout-Simulator/index.html at fae9aac7993970c854167b2db4b14203ece098e9 · MaslowCNC/Layout-Simulator · GitHub
Kyle wrote:
Maybe my phrasing was off, I meant adjust the pulling force of certain arms at
certain places on the frame. i.e. far left or right edges, lower the pulling
force on the long belts so they don’t over power the short ones.
I understand what you are saying, but I’ll say again, the maslow software
currently doesn’t really manage the pulling force, it manages the belt length
(well, mostly, there is a component that looks at the speed that the belt
extend/retracts)
It has no way to know what force there is on any belt. It can measure how much
current it’s applying to the motor, but htat doesn’t tell you what the force on
the belt is because there is an unknown amount of friction between the motor and
the belt movement.
Think of your car sitting in drive at idle. There is a constant amount of force
being coming out the back of the transmission. But you can’t know if your car
will idle up a slight incline unless you know the rolling resistance of the
tires, if the brakes are applied at all, the friction of the bearings in the
drivetrain, and how efficient the gears in the differential are
you can measure the gear efficiency, bearing losses and specify that brakes
should not be applied (and recognize that these change over time as the
lubrication in the gears and bearings ages, so these are not fixed values)
you can specify a specific tire and do extensive modeling of that tire (and/or
experimentation), but the rolling resistance of the tire is a combination of the
tire design and manufacturing, the air pressure in the tire, and the weight on
the tire.
If you don’t know the pressure in each tire, and the weight on the tire (in the
case of the maslow, this would be a combination of the angle of the frame, the
weight of the sled/cables/hoses, and the friction between the sled and the
workpiece), you cannot say if the car will move at all, let alone idle up an
incline.
If there was some way to measure the tension on the belt, it would be relatively
easy to model the belt stretch. But the maslow does not have a way to do this.
David Lang
Interesting example. Why would one belt be more slack, or possibly one belt pulling harder. Perhaps there is more friction occuring on TR causing additional force interference .
Dano
That picture is during calibration, I think the BR belt there is pulling tight and moving the whole unit right causing TR to go a little slack. TL and BR don’t have enough angle to keep the machine to the left.
Kyle wrote:
That picture is during calibration, I think the BR belt there is pulling tight
and moving the whole unit right causing TR to go a little slack. TL and BR
don’t have enough angle to keep the machine to the left.
That is possible, if the frame is pretty square, this should not be the case,
but you aren’t the only one to report it.It may help to get a video of the
calibration showing the problem
also check that your z offsets match your setup.
@bar, I’ll again call for snugging up all belts after you think they are tight
in calibration.
David Lang
1 Like
We actually do this already, the first measurement gets thrown out automatically for exactly this reason. We then take four more measurements and compare them to each-other. If there is a significant deviation between any of four measurements we throw out all four and start over
Ah yes, the force changes based on the trig angle of the belt and therefore possibly overpowering the low angle force.
1 Like
At first I thought that TL in this case was just going slack, but if you look closely BL is pulling the machine to the left causing the slack you see in TL. I think the angle and tension of TR and BR are not enough to keep the machine in place. This is during calibration BTW.
Edit: re-watching this, I know the video moves around a lot but if you pay attention to the right side of the sled it looks like the whole thing moves about 10mm to the left compared to the overhang support. It starts beyond the edge and ends inside of the edge of the wood a bit.
1 Like
Bar wrote:
We actually do this already, the first measurement gets thrown out
automatically for exactly this reason. We then take four more measurements and
compare them to each-other. If there is a significant deviation between any of
four measurements we throw out all four and start over
as I read it, taking the measurement was just reading the encoders, not trying
to run the motors. Did I miss something?
we’ve had too many people report the belts being slack during calibration for
whatever we are doing to be working
David Lang
1 Like
FWIW the belts were tight everywhere else for me, just the top long belt gets loose along the edges during a maxed out calibration. Or rather the bottom belt pulls the machine in.
1 Like
After calibration completes and you manually jog the sled to similar location does the belt go slack again or are all tight?
1 Like
All tight, but X is skewed (comes up short by 2-10mm), which is what I’m trying to get to the bottom of in this thread.
2 Likes
Any updates on all this? I’ve just started cutting with my M4.1 and trying to dial in the accuracy. I bought the machine with the intention of cutting 1200 x 2400 sheets to make furniture that slots together so i need it to be more accurate than it currently is
Yes, here’s what I’m working on now: X & Y Skew Analysis
For the X axis you’ll want to extend the belt anchors by about a foot in either direction, so the frame should be around 8 x 12. I made some little extensions out of 2x4s and 3/4 ply. The accuracy is the worst about 2-3" from the edges so either avoid that or know that it might be more skewed there. If you can orient your cuts so they are moving toward the center or origin the accuracy should improve. Also, I’ve never run a horizontal setup, but since the machine isn’t fighting gravity it should be more accurate and you should also have less sled tilting.
I haven’t run any cuts after that last post and analysis, but those tips should help quite a bit. Hopefully we’ll be able to test some software changes soon that will fix these issues. Good luck!
1 Like