That’s a good point. The belt I tested is of unknown origin and was what I had lying around to prepare the test setup. I’m not sure it’s actually Kevlar. I will repeat the test with new belts of known origin.
1 Like
100%. This is why I’m keen on using actual Gates belts since they seem like they are the top tier manufacturer. Pretty sure they came up with this type of drive belt.
The 6mm and 9mm sections arrived today, so I should be able to test them soon with my janky setup.
3 Likes
Yes, I think converting belt length to cartesian is far more tricky. All we need to know is the stretch/elongation of the material and the force. The latter is what might be a little tricky because it changes based on belt angle and frame angle (among a few other things) as I mentioned in the original post and as @dlang also called out. Belt angle and frame angle are likely a good place to start trying to figure out how the force varies.
Lais wrote:
Yeah, without directly measuring belt tension, implementing any correction
model would be tricky. In theory, if you know the belt’s elastic constant, you
could correlate motor current to tension with some calibration, but I’m not
sure the constant is consistent enough between batches or over time to make
that practical. Either way, minimizing error at the source by using stiffer
belts is definitely a path that should be explored.
I also think that we can have a ‘belt stretch factor’ that can be tweaked by the
user to find tune it’
David Lang
1 Like
Absolutely. Especially since the M4 is out in the world with these steel belts, this would be great if someone wanted to upgrade to the nice Gates belt.
1 Like
It will also be interesting to see if less stretch in the belts adds more stress on the arms or other plastic components.
Dano
1 Like
Awesome work here!!!
I suggest figuring in the belt spool issue into your calculations. Remember that the motor current that we use for pull control does not directly go to line tension. You get less pounds of pull per Amp at short extension and more at long extension. That’s an easy calculation, and its something we should really be rolling into the code… There’s going to be a double whammy on this one, in that long exposed belts are stretchy, and the short belt pulls harder!
There’s also a squish factor for the belt wrapped on the drum, but wrapping at relatively constant tension will minimize this.
There’s a lot of value in rolling the drum fill radius (think layers of belt) into things so that we can specify -tension- rather than motor torque. As a really rough SWAG, the tension vs amps is likely varying 15% or so (assuming 3" empty drum and 3.5" full drum). Feels like a big factor if we are unhappy with and actively chasing 6% errors.
A really valuable test would be to set up horizontally with two anchors at useful spacing, then travel the unit between the two extremes. Data is line tension and actual position vs commanded position and tension setting. This will check a combination of the belt stretch and amps/tension/spool fill. Note that the return journey will likely look a bit different; the last motor to move (with its own drum fill diameter) will be the one that controls tension.
Another way to get at this would be to hang one vertically and take data on commanded position, actual position, belt tension (will be constant), and the tension setting required to move up.
I wish I had the time to dig into this!
FYI, doubling the belt width will cut the deflection in half.
1 Like
Rod wrote:
I suggest figuring in the belt spool issue into your calculations. Remember
that the motor current that we use for pull control does not directly go to
line tension. You get less pounds of pull per Amp at short extension and more
at long extension. That’s an easy calculation, and its something we should
really be rolling into the code… There’s going to be a double whammy on
this one, in that long exposed belts are stretchy, and the short belt pulls
harder!
it will pull harder for the same amount of current, but we don’t know how much
friction there is in the spool turning before any force is applied to the belt.
This is why I keep saying we have no way to measure tension, only length.
There’s also a squish factor for the belt wrapped on the drum, but wrapping at relatively constant tension will minimize this.
but we are not wrapping at constant tension, the tension will depend on how hard
you are pulling vs how fast you are moving, so when you are cutting with a dull
bit, you are wrapping under more tension than if you are moving with the bit out
of the wood.
There’s a lot of value in rolling the drum fill radius (think layers of belt)
into things so that we can specify -tension- rather than motor torque. As a
really rough SWAG, the tension vs amps is likely varying 15% or so (assuming
3" empty drum and 3.5" full drum). Feels like a big factor if we are unhappy
with and actively chasing 6% errors.
we are chasing length small percentage length errors, I think the tension
differences are much larger
A really valuable test would be to set up horizontally with two anchors at
useful spacing, then travel the unit between the two extremes. Data is line
tension and actual position vs commanded position and tension setting. This
will check a combination of the belt stretch and amps/tension/spool fill.
Note that the return journey will likely look a bit different; the last motor
to move (with its own drum fill diameter) will be the one that controls
tension.Another way to get at this would be to hang one vertically and take data on
commanded position, actual position, belt tension (will be constant), and the
tension setting required to move up.I wish I had the time to dig into this!
it is a very interesting rabbit hole to go down
FYI, doubling the belt width will cut the deflection in half.
yep, but they would no longer fit between the top and bottom of the router. We
would have to use a longer spindle.
David Lang
1 Like
Got around to testing the Gates belts, and they are much worse unfortunately. They are far thinner so I wonder if that’s part of the problem.
@Lais Do you have any idea what belts you tested? Those seemed to be far superior. I’d love to test the same ones with my janky setup to see how they perform with the same process.
Note: I ordered a 100mm section, and had to test slightly less than 100mm and adjusted in the spreadsheet.
Thickness comparison – a little hard to tell in the picture, but they are quite different.
1 Like
I might be the only one. But I’m having trouble understanding your data table.
I’m reading that it took 5kg for 1mm, 2.75kg for 1 mm, and 4.25 for 1 mm stretch. Is that accurate?
How do you determine the percentage numbers?
And the numbers under the force column (100, 90, 88) throw me off, what are those?
Just trying to understand your research.
Fair question – I stretched a length of belt by 1mm at a time and noted the force required to stretch it that amount. The 100, 90, and 88 were the total length of belt measured. Since the Gates ones were kinda pricy I just ordered a 100mm length and had to measure less than that due to attaching the anchors. (they actually gave me a little more than 100mm)
The percentages are how much the belt stretches per kilo of force. So 1mm stretch / 5kg force / 100mm length = 0.2% stretch per kilo.
Make sense? I should have added units/description to the length tested…
2 Likes