holy shit guys, you all are hitting it out of the park on this one, this is such an amazing innovative community i really am soo happy to be a part of this
2 Likes
i love this, gives the ability to cut larger things without it needing to be the same level because its not riding on it, absolutely fantastic
you could set up a slightly off the board skirt around the outside of the sled with the vacuum attached so it sucks up any debris before it gets under it from any direction, all the dust will naturally want to fall away and out of the cut, absolutely love this
2 Likes
To make project floor spider work the only things I would need is either a spindle to switch upside down or a longer z thread motor and linear bearing rods and support towers. I am striking out a bit on looking for the motor.
I could also lift the motor and towers up about 70 mm. All of these would work.
1 Like
So two more thoughts on Project Floor Spider. The table legs should be at a diagonal inbetween the belt anchors. A spindle with it’s bit point up into the air would probably be the easiest. solution. not much luck on longer lead screw motors.
2 Likes
Sled vacuum is cool. It just occurred to me that the sled could ride on an open work grid to avoid dust build up
2 Likes
Thanks! I am really enjoy the scope for experimentation there is. That you and the community is very engaged makes it rewarding, and the fact I can order parts individually really helps too!
@dlang - i’ll copy your points into the other thread I just started for my build (Spider Maslow, Spider Maslow, does whatever a Spider Maslow does) to avoid hijacking this thread any more 
3 Likes
Hi @bar
Has anyone looked at gordix large format desktop cnc on kickstarter. It looks, ‘inspired’ (ahem) by maslow 4 but maybe there’s interesting things in its approach to the belts.
Let’s break down what makes the M4 great. It doesn’t need a big frame to run on. It uses measurements to calibrate itself.
Cons - it’s not the most accurate.
The belt idea and self-measuring is a great idea but for great accuracy you need a cartesian frame. However they are hideously expensive and no one has a reasonably priced one that can handle 4x8s.
I’m thinking that instead of putting the accuracy in the frame that we make a rectangular frame from cheap metal, perhaps shelving or scaffold pipes. Then we put distance sensors so that the spindle will know where it is in 2D space. That means the frame isn’t so important. Belts and the sensors can tell it where it is.
Something like laser and perhaps ultrasonic together to handle big and small distances. Do it on both sides of X and perhaps Y.
It loses the advantages of the floor type setup of M4 but it gains in accuracy.
One day I’ll do it; if I ever have the time.
Grim wrote:
Let’s break down what makes the M4 great. It doesn’t need a big frame to run on. It uses measurements to calibrate itself.
Cons - it’s not the most accurate.
The belt idea and self-measuring is a great idea but for great accuracy you need a cartesian frame. However they are hideously expensive and no one has a reasonably priced one that can handle 4x8s.
the big problems with a cartesian frame are:
- they need to be rigid, any flex is error
- to be rigid they tend to be heavy (take a lot of material)
- as they are heavy they take a lot of force to move, so need big motors
- because they take a lot of material and need big motors, they are expensive,
so they are expected to also be fast to be worth the money. This drives another
round of bigger motors, and more rigidity needed.
I’m thinking that instead of putting the accuracy in the frame that we make a rectangular frame from cheap metal, perhaps shelving or scaffold pipes. Then we put distance sensors so that the spindle will know where it is in 2D space. That means the frame isn’t so important. Belts and the sensors can tell it where it is.
Something like laser and perhaps ultrasonic together to handle big and small distances. Do it on both sides of X and perhaps Y.
getting a distance sensor that can measure small amounts over a large distance
is hard. I believe that the maslow hardware is limited by the lack of precise
measurements (primarily due to belt stretch). The software imposes other limits,
but we are chipping away at those.
David Lang
2 Likes
Grim wrote:
Has anyone looked at gordix large format desktop cnc on kickstarter. It looks, ‘inspired’ (ahem) by maslow 4 but maybe there’s interesting things in its approach to the belts.
yep, there is a discussion on it.
the maslow has the belts pointed at the center of the bit so that if the sled
rotates, it doesn’t affect anything
the gordix uses two belts per corner to avoid the sled rotating
I would expect the gordix approach to run into grief at larger distances (I
don’t know how far apart they allow the anchors to be) as the belt lengths make
the separation at the less a smaller portion of the triangle
David Lang
3 Likes
Yes they need to be rigid and so more material but that does not have to be expensive. It obviously has to be metal. Where the current solutions become expensive is that they use expensive machined extruded aluminium that has to be near perfect. We are not using that to position the spindle in 2D space, so it doesn’t have to be so exact.
By using distance sensors on -x, x, -y & y, we can know where the cutting point is. We can combine this with the belts if necessary to be more exact.
I’ve seen laser sensors be mm perfect over a decent distance, so any error in the +x direction can be ignored given the data from the -x direction.
I’ve read that ultrasound sensors are cheap and have good resolution over short distances too, so perhaps a combination of laser and ultrasound would work. They could be mounted on either the frame or the carriage.
The sled/carriage (or is it called a gantry) would be on wheels and move on the x axis. Then the spindle would move on y. I can’t recall the name but I’ve read that 3D printing uses belts to move on 2D rather than motors on the x & y. Perhaps we could use this system.
as for the frame, I was thinking or either metal pipes or slotted angle steel (dexion).
That should be easier to build than the M2 frame was.
I am not sure what is being discussed with a frame here. Is it like a xyz gantry? For position feedback in a standard Maslow set up or a frame system would something like this camera tracking system work? GitHub - AsterTrack/AsterTrack: A marker-based optical tracking system for VR and MoCap · GitHub
1 Like
yes, like an xyz gantry. The idea is to decouple the positioning from the frame. The gantry just moved over the frame like it’s a rail.
When you mentioned a vision system, I had already been thinking whether I was overcomplicating it.
Put a target on top of the spindle and you can track its position. Not sure whether vision or sensors are best but it could be extremely accurate. It’s just a question of whether you use 3 positions to triangulate everything or 4 cos you put them on the anchor points.
I would go for cheap sensors but if that’s video, then so be it.
I suppose that z axis movement would stay the same but perhaps it too could be tracked this way.
Grim wrote:
yes, like an xyz gantry. The idea is to decouple the positioning from the frame. The gantry just moved over the frame like it’s a rail.
many people have suggested this over the years, as far as I know, nobody has
built one. one problem is that by pulling on the router rather than on the
gantry, you will tend to caus the gantry to ‘rack’ (rotate slightly due to
forces on one side and not the other), and this can cause the gantry to bind
When you mentioned a vision system, I had already been thinking whether I was overcomplicating it.
with video, you have to think about the resolution. If you have a 4k camera that
is looking down on a 4x8 sheet of plywood, one pixel is about .5mm (assuming no
distortion from the lens) and usually you use cameras to be within a few pixels
Put a target on top of the spindle and you can track its position. Not sure
whether vision or sensors are best but it could be extremely accurate. It’s
just a question of whether you use 3 positions to triangulate everything or 4
cos you put them on the anchor points.
if you know the anchor positions, you only need 2, but you need to have them
rotate to match the belts.
David Lang
Yeah, I’m sure that 2 would work but it would involve a lot of measuring.
Are cameras that much of a cost that we’d just use 2 when 3 or 4 wouldn’t add much to the budget?
If they are, then what other cheap sensors could we use?
Forgetting how we locate where the cutting point is, do you think that the idea of separating location from the movement would be worth it?
We’d be calibrating the motors, belts etc from observations of what they do.
you won’t know until you try, but I think that getting cameras back far enough
to be able to see what’s happening will be more work than you think.
I’ve also been spending the last month on image processing (a vehicle expense
tracking program that works by taking pictures of your dash and the pump rather
than having you type in values) and it is leaving me both impressed at some of
the images it is able to decipher, and frustrated at some others that it can’t
handle.
getting the position of the sled would require the lens distortion be mapped (if
a fisheye type lens) and several cameras so that you are not trying to see the
sled at too much of an angle.
but I’m not an expert, it’s worth trying.
David Lang