In the end I had to completely unscrew the PCB mount and remove around half of the cables to be able to get it to completely clear the router. Not a very difficult task, around 2 minutes or so.
Still have to do a full powered on test.
1 Like
It hit me (when I woke up in the middle of the night) that the motors for the belts donât have to be mounted to the arms, so many hours later, I have a draft CAD model (this shows a single arm so you can see the stuff in the middle)
this mounts 4 worm motors to the sled, each driving a shaft that has a gear that turns the spool from the inside, and an âidler gearâ for the other 3 spools (so the spool is supported on 4 gears) This not only eliminates any chance of the belts getting into the gears, but also should shrink the thickness of the arms
It then has 3 lead screws driven with a belt from a single stepper for the Z axis. This allows for spindles >80mm diameter with an overall size smaller than the current version
I havenât yet modeled all the plasic bits (the thing that rides on the lead screws and holds the spindle, the sled, the dust collection, and the vertical stuff from the sled to support the bearings and stepper at the top, extra material on the arms so you have a place to put screws)
I thought about putting the stepper on the bottom, but having it at the top leaves a nice hole for dust collection
since the motor will spin the entire arm when it pulls tight, you have to rely on the encoder telling you the belt isnât moving rather than measuring current (and you have to watch out that you donât damage your encoder connectors/wires by wrapping them around the sled)
I figure that the electronics, spindle power, and power brick can all be mounted to the sled between the main motors
2 Likes
the more I look at it and think about it, the more useful it seems to be to be able to run the spindle off the top rather than having to unclamp it, so I may just go with 3 steppers instead
Why not make the spools/arms the diameter of the sled and make the z axis independent of the spool/arm cluster. Your cad drawing may show that , but not quite sure
I do have the Z independent of the arms
spool size is a balance between larger spools hold more belt, but take more
material (more expensive) require larger systems to manufacture (more expensive)
As I currently have it sitting in CAD, it can hold a 85mm spindle but arms can be printed on a 213x230mm print bed
I did a bit more work, this may show things better
not counting the stepper sticking out the top, itâs just over 6 in tall (and I think Iâm using much thicker bearings than the stock version, once I measure them Iâll adjust and it will shrink a bit)
Iâve got everything parameterized, so I can set different sizes pretty easily (things get a bit messy when the drawings overlap so itâs not quite seamless)
since I use stock motors and then have a 15mm tall coupler to the shaft to drive the arms, there is a lot of wasted space at the bottom, switching to motors that have long shafts on them would save a lot of space there.
If the motors are fixed then when the assembly applies tension to the belt, the torque reaction will cause the arm assembly to rotate out of true and the arm will not point at the anchor, but off to one side a little, putting a bend in the belt and changing itâs effective length.
In addition, as the Maslow moves about the work surface, and the arms move to align with the anchor points (if it is working as you intend) then as the angle of the arm changes, the belt will wrap and unwrap from the spool, again changing the belts effective length.
I think the only way to avoid this is to do as Bar has done with the design and have the gearmotors anchored to the arm assembly which is free to move. This closes the loop and eliminates the torque reaction.
I like your older suggestion of enlarging the ID of the arms and fitting the entire z-assembly inside, so the belt motors do not interfere with any supports.
1 Like
While we are talking about belt driving gearmotors, it is important to note that one of the reasons we have so much trouble ensuring proper tension on the belt is that we use a gearmotor that is nearly self-locking. âSelf-locking meansâ that internal friction is high enough that you cannot backdrive the motor by pulling on the belt. This is convenient in that if you remove the power the Maslow does not fall down to the ground (if using vertical orientation) but it makes it hard to ensure proper tension on all belts, especially during calibration when you donât know where the anchors are.
One solution would be the use of a direct drive motor with no geartrain, and a brake that engages automatically when power is cut. In this kind of setup, the motor current is exactly proportional to the belt tension because friction is minimal. On a vertical setup during calibration you just set the tension current on the lower two belts (set a constant current and ignore the encoders) and steer the Maslow with the upper two motors and encoders. This will perfectly prevent loose belts and the resulting problems, as well as allowing you to use the full calibration matrix right from the get go.
This kind of direct drive setup is used on industrial robots and works great. Unfortunately it is horribly expensive.
A compromise might be some kind of belt tension measurement device, perhaps using strain guages and flexion beams. This is done quite inexpensively inside virtually every digital kitchen and bathroom scale. Using this kind of device along with gearmotors like we already have, the lower two belts would be maintained at constant tension using a control loop in software, and the upper two gearmotors would drive the sled around while referring to their encoders.
Once calibrated, the Maslow would operate as it does now, but because you could monitor the actual belt tension, you could potentially implement a smoother movement control algorithm, and detect error conditions quickly. You might even be able to compensate for errors caused by the forces on the router bit as it cuts, or even the effects of frame flex of belt stretch.
If the motors are fixed then when the assembly applies tension to the belt,
the torque reaction will cause the arm assembly to rotate out of true and the
arm will not point at the anchor, but off to one side a little, putting a bend
in the belt and changing itâs effective length.
good point, with nothing at all stopping the rotation, it will just rotate until
it tears off the encoder wires.
but if there is a stop, then it will rotate until it hits the stop.
the question is if the remaining torque will move the arm enough to matter
compared to the weight of the motor on the arm (shrug)
I wonder how much it would help if opposite belts spin in opposite directions
(as the belts are tight, torque will cancel out, as we move, one will pull
harder than the other is releasing). with the spool design I have, you could
just flip them over
In addition, as the Maslow moves about the work surface, and the arms move to
align with the anchor points (if it is working as you intend) then as the
angle of the arm changes, the belt will wrap and unwrap from the spool, again
changing the belts effective length.
True, but since the length of the belt is controlled by the encoder at the end
of the arm, not the motor position, it will automatically adjust for this (I did
think of this)
I think the only way to avoid this is to do as Bar has done with the design and have the gearmotors anchored to the arm assembly which is free to move. This closes the loop and eliminates the torque reaction.
I had forgotten about the torque issue, but that was something we worried about
with the old maslow as well and didnât end up being a problem in practice.
I like your older suggestion of enlarging the ID of the arms and fitting the entire z-assembly inside, so the belt motors do not interfere with any supports.
this shows one way that can be done in any case
David Lang
Also as far as torque goes, I expect that the torque from the belt motors is
much less than the torque from the router, especially with a good size bit.
David Lang
WFD wrote:
If the motors are fixed then when the assembly applies tension to the belt,
the torque reaction will cause the arm assembly to rotate out of true and the
arm will not point at the anchor, but off to one side a little, putting a bend
in the belt and changing itâs effective length.
there is a potential problem, but itâs not the torque applied by the motor to the spool and sled
itâs that without the motor and spool both acting on the arm, the straight line will be to where the belt is tangent to the spool, not to the center of the bit. If the spools all go in the same direction, it will just unspool every belt.
since all the forces on the belts have to cancel out, I wonder if flipping belt directions on opposite sides would end up working?
(being able to move the drive to the other side of the spool from the feed is such an attractive thing. It lets the arm be thinner, and eliminates the possibility a loose belt getting tangled in the gears)
I wonder if itâs possible/reasonable to steer the arms? if you know the anchor locations and belt lengths you also know the angles needed, and you only need to rotate the arms through about 90 degrees (and one could be fixed with the entire sled rotating as it moves)
Curious of thoughts of having the spools be set radially, one in each quadrant.
- Smaller, simpler spools
- should require a similar or event potentially smaller amount of material, as well as a smaller mold
- no more potential of creating a linkage between arms when screws come loose
- Motor/gears could act on the inside edge of the spool, removing chance of interacting with belt
- keeps belt drawing a straight vector from anchor to bit
- puts all belts at same z offset
- reduces complexity of core assembly and overall assembly process
- should all but remove the issue of limited angles for the arms
- would potentially be able to be modded onto existing Maslow units with the only changes being removing the existing arms and modifying the sled to add mounting holes for the new ones
- should require minimal software changes
- sled orientation is maintained and doesnât tilt anymore
- the sled shouldnât tilt anymore from the belts because the angle of attack will be lower and the force is being applied further from center
- encoders are further from router and cutting
Pair this with a wider opening in the center and the Maslowâs inherent versatility shoots through the roof.
to make the math work, you really want to have the belts pointed directly at the
center of the bit. If you donât do that, then you become extremely sensitive to
the angle of the sled (so if you have a vacuum hose that limits the sled from
rotating, exactly how much force does it apply?, how much does friction on the
workpiece slow the sled rotation, things like that) The sled will rotate, if
only due to the torque of the router.
so you really want the belts to anchor to the sled in a way that allows the sled
to rotate underneith you without the bit moving.
If you put the spool out on the arm, then in the vertical mode you have more
weight on the arm that can distort them from hanging vertically. The spools on
the arm will also limit how close the arms can get to each other.
your sketch matches one of the early ideas of the âeverything on the sledâ
designs.
look at the linkage designs from the old maslow and you can see some other ways
of keeping the belt/chain pointed at the center of the bit
David Lang
Iâm realizing now that the vector through center isnât maintained unless the belts are at 45 degree angles in my previous drawing. In light of that and a few other things you said, I would change the orientation of the linear rod supports so they donât stick out so far and create routes for the spools to travel around the center within their quadrants, which would only impose limits where the spools would come into contact with each other. It may require a larger sled, or attaching things to an existing sled. May need to mirror the groove that the spool travels along above it as well to keep it from tilting when in tension.
replacing belts would also be much easier. Almost all of the servicing would be, tbh.
itâs also woth pointing out that there is nothing requiring the spool to be
parallel to the sled with the belt perpendicular to the workpiece. you could
rotate the spool 90 degrees so it looks like a snail
David Lang
I would hazard that the ideal orientation would be the one that doesnât propagate a twist in the belt and matches the orientation of the belt at the anchor.
with the original maslow, we found that if you find the center of gravity along
the Z axis and put the force there, with the belt/chain parallel to the
workpiece, you end up with very little tilting force on the sled.
now, that CG will vary depending on the length of the bit, how deep you are
cutting, etc. but getting it close works very well.
David Lang
Carson Barry wrote:
I would hazard that the ideal orientation would be the one that doesnât propagate a twist in the belt and matches the orientation of the belt at the anchor.
We control the anchor as well as the sled, the belt can go into the anchor in
any orentation
David Lang
I realize that, and also that it would not be as easy to have the post for the anchor to be oriented differently. Trivial for some, but would require more complex woodworking for others. That has me leaning towards not changing that being better, even more so when I consider the motors rotating and potentially being more problematic than the spool with how far they stick out.
vertical post for the anchor, but then a slot that the belt goes through and
loops back on itself (with either a clip or wire ties to anchor it in place).
David Lang
I am kicking myself a bit for not remembering ring-head screw bolts.