A question for the machine designers out there: A delta with towers at 90,90,180 would work about as well as the usual 120 * 3 configuration right?
I remember @Nicholas_Seward 's conversation about a wall delta way back, where he said it would be horribly wobbly. But 90 degrees seems like enough spacing to make that OK.
One part I’m not sure about is whether using precisely 90 degrees is a bad choice, since it makes the arms sort of coplanar. I could see an argument for 100 degrees, and perhaps even 80 degrees being much more stable.
Thoughts? I’m dreaming of a 5 axis platform like Oyvind Grutle’s (https://3dprint.com/77400/5-axis-3d-printer/) with this thing on top instead of a cartesian bot. Deltas will generally do better at reaching down into concave parts, and with the 90 degree configuration you could still start printing sideways very close to the build surface. https://3dprint.com/77400/5-axis-3d-printer
@Billy Thanks for the link, looks like 90 degree is pretty awful. What do you think about 80 or 100 degree spacing? Watching that video makes me think 80 degree spacing would do well, but I may be missing something.
@John_Driggers Someday… After Hyperloop’s done. I recently convinced myself that implementing straight skeletons on the surface of a mesh wouldn’t be terribly hard, and as soon as I get around to doing that Bread can do 5 axis slicing very well.
Random thought, but you can’t cross/twist the arms on each side? Making sure the twists on opposing sides are reversed, so they work against the twisting of the carraige? Can’t quite visualize the full range of movement in my head and how much it would bind.
This what you’re thinking about for a 90,90,180 delta? http://www.thingiverse.com/thing:714915 (my design)
It works just fine. No need for 80 or 100 degree angles.
I keep meaning to add the fourth arm for added stability but it’s low on my to-do list.
There are four-arm industrial deltas that build a tool-tilt function directly into the arm kinematics so you wouldn’t need a two-axis table.
I’m not sure why I never saw @Billy 's post before now. My square delta has the same kind of wiggle when the arm angles are shallow, but it’s not that bad and it doesn’t appreciably affect print quality. Regular deltas also have poor effector constraint when two arms are near horizontal, it’s just not quite as bad.
Check the print pics in the Thingiverse page I linked. I had no major problems printing in the center of the build plate at 200mm/s with that printer. (And Billy’s build looks more rigid than mine.) I simply used long arms so the angles were never worse than 25 degrees or so from horizontal within the desired working volume. True, the “opposite corner” region around max extension of the “middle” arm does have increased ringing, but it’s less than what I see in the average Mini Kossel print with plastic vertices.
If you watch the video at 50 seconds and compare with the older video linked in the text showing a 120 delta there is a quite a stability difference. Of course you’d not want to put your rods horizontal for steps sake.
100% agreed that the 90,90,180 case is less stable than 120,120,120, it just doesn’t seem to matter enough in practice to significantly reduce print quality. YMMV.
The “most stable” possible delta geometry has the three arms (and their associated kinematic constraints) perpendicular to each other, which would mean all three at ~37 degrees from horizontal at build plate center. But that makes for an unreasonably small working volume so every delta ever built uses steeper, less stable arms than the “most stable” configuration. It’s simply a compromise between perpendicularity of constraints and practicality of drivetrain design. Likewise, the more the arms deviate from the 120,120,120 arrangement, the farther away they get from mutually perpendicular, and the less stable they get. It’s just another trade-off.
The problem you will have is toolpath planning as complex pieces will require not only knowing the shape and spinning of the part, but the position of the 3 delta support arms vs the single Z axis arm. This may DRASTICALLY reduce the configuration space that avoids collisions between the printed part and arm, and severely impact your print volume. The code to solve this may be a lot more complicated as well.
@Jason_McMullan Yeah of course. I’m buried in other projects till August anyway, largely daydreaming.
@Daniel_Joyce Collision avoidance is a lot easier than you might think for printing. The shitty part of it for CAM software is usually dealing with all the stuff you put into a subtractive machine. You know what’s in a printer at all times. If you represent the shape of your layers as meshes, it’s pretty easy to prevent collisions with some intuitive rules.
You end up limiting the capabilities of the hardware a little bit because it’s a conservative approach, but for the relatively few situations where you want to work in tight almost-collision spaces, you can either throw the user a warning and let them figure out whether it’s safe, or drop down into a more computationally intensive checker.
And collisions on the hardware itself (ie, hitting your bed with an arm) are really trivial to prevent, it’s just about setting correct bounds for your working area in 5 dimensions instead of 3.
Not saying its impossible just now instead of a single arm plunging straight down along Z, and moving along X, you now have 3 arms that define the surface of a distorted tetrahedra that shifts every time the effector moves.