Need some CoreXY advice



  • Well, I'm about to rebuild Black Beauty...

    0_1561060140700_IMG_20190609_122838_2.jpg

    ...and I'm thinking about moving to a CoreXY design to get rid of that front cross member that always seems to be in my way. Most of the implementations I've seen, including @mrehorstdmd's "gold standard" implementation use belts on 2 levels and either stacked or offset pulleys. While I could go that route, it presents some challenges for me so I'd like to try a single-level arrangement. Since I have a little over 600mm between what would be the back-left and back-right idler pulley assemblies, I was thinking about doing the "twist"...

    0_1561060468261_IMG_20190620_134729.jpg

    This is just a spare-parts mock up with the near end representing the back-right and the far end representing the back-left and the belt spacing exaggerated for clarity. In the middle would be a piece of igus iGlide material on a bracket to provide belt separation.

    With that length, the twist is pretty gradual and there doesn't appear to be any undue stress as the belts interface with the pulleys. Importantly though, it makes the construction much simpler, the moving mass lower, and it keeps the forces on the 3 carriages on the same plane, low and tight to the linear bearings.

    Thoughts?



  • @gtj0 I built my CoreXY with a "twist" and I'm very pleased with how it works. I planned a belt separation arrangement, but didn't need to use it. I found that twisting the belts so that the smooth sides touched, worked without any noticeable wear or friction. I don't think it is as popular as offset belts but it works for me. Good luck with yours



  • I built a corexy sand table with belts on the same level, twisted at the crossover point. I found a couple problems with it.

    alt text

    As you can see, the belts were rapidly grinding each other into dust.

    I had the same problem at the corner pulleys where the twist in the belts was forcing them to ride on the pulley flanges, again grinding the belts to dust.

    alt text

    That span was almost 1m. Twisting the belts over a shorter distance might be even more of a problem.



  • Thanks @pawPrinter !

    @mrehorstdmd That's more dust than I would have expected. The center meet point I can solve with the iGlide but the wear at pulleys is troublesome. I wouldn't have thought that the belts would ride against the flanges as long as both ends were in the same plane.

    Hmmm.

    Edit: It's just a half twist right?



  • I have a twist in my belts, but on a stacked arrangement. So no contact. The twist itself doesn't seem to have any downsides for me over the 500mm span. I don't see any extra wear at the pulleys. By the time they reach the pulleys the twist is completely undone.



  • @phaedrux said in Need some CoreXY advice:

    I have a twist in my belts, but on a stacked arrangement. So no contact. The twist itself doesn't seem to have any downsides for me over the 500mm span. I don't see any extra wear at the pulleys. By the time they reach the pulleys the twist is completely undone.

    Yeah that's what I see in my mockup with little to no tension on the belts.



  • @mrehorstdmd Just a thought... could the dust at the pulleys be particles that chafed off at the crossover, then carried to the pulleys in the belt teeth and flung off as the belt comes around the pulley at speed?



  • I don't think so- the belts were riding hard on the pulleys. Maybe the long span is the problem and the sagging of the belts under their own weight may have caused the problems.



  • @mrehorstdmd said in Need some CoreXY advice:

    I don't think so- the belts were riding hard on the pulleys. Maybe the long span is the problem and the sagging of the belts under their own weight may have caused the problems.

    Gotcha. Maybe I'll use a forked guide in the center to make sure both belts are supported.



  • I use the same-plane configuration, with twisted belts for several months, and I have no issue. The belts cross on they back, and I have a PTFE sheet between them, maintained by a 3D-printed part.

    About dust on idlers, I had some with my previous chineese belt. As its core finally broke (it was really garbage), I bought a Gates, and I don't see anymore dust.

    text alternatif



  • @gtj0 said in Need some CoreXY advice:

    ...and I'm thinking about moving to a CoreXY design to get rid of that front cross member that always seems to be in my way.

    If I built a printer in the future, it would be a coreXY for multiple reasons including this one. However, just choosing coreXY would indeed allow omitting the beam in front, but you'd still either have the belts running across the frame opening in the front, or the motors in the front left and right depending how you orient the belt system. E3D (amongst others) have solved this problem by using one or two more idlers per belt, locating the motors in the back while keeping the open front with their toolchanger platform.

    not really answering your question, but I thought this might be interesting as well.



  • @fma Yeah I stopped buying no-name belts and got the genuine Gates belts and have been much happier.

    @sonderzug Having the motors at the front is no problem for me because they sit inside the profile of the 50mm extrusions. Interesting idea though.



  • I put the motors above the vertical frame profiles, so they don't bother me.

    text alternatif

    Note that I kept the top front profile, but it could be remove, to have full access to the printer (the frame uses 45x45 profiles I had, so is very rigid).



  • OK, I've rebuilt the kinematics but my brain is on vacation...

    For a CoreXY, when you measure and set steps per millimeter are you doing it for an axis or a drive?? The description of M92 seems to use both axis and drive interchangeably.

    Basically, should I be isolating the drives while measuring or measuring axis movement with both drives enabled?



  • @gtj0 You set steps per mm for the axes. For one thing, a single drive will give you diagonal movement. For another thing, when you do pure X or pure Y and both motors are employed, the motor movement is about 1.4 x the axis movement.

    Basically, treat it like you would a Cartesian when setting the steps per mm and let the firmware worry about the calculations.☺



  • Thanks @deckingman !



  • @gtj0 said in Need some CoreXY advice:

    Having the motors at the front is no problem for me

    This is similar to Hypercube Evolution. The front top crossmember is indeed in the way as you say but lowering it 10-15cm instead of removing completely solves that problem for me.



  • Well...

    • Stationary Y rails are straight and parallel
    • Gantry X rails are straight, parallel and perpendicular to the Y rails
    • The belts are parallel to their respective rails
    • The belts are firm but not over-tightened
    • Directing the X and Y axes to move 450mm results in 450mm of movement within what I can measure (about .25mm) I did need to adjust the X steps/mm to 160.40 rather than the calculated 160 to get that.
    • The tool carriage is solidly held by the motors. No movement in X or Y nor is there any rotation around Z.

    Yet, I still get....

    0_1564265347707_IMG_20190727_160102_HDR.jpg

    The boxes are 20mm square printed in 10mm segments in the order and direction shown. Both boxes are drawn the same. The circles are drawn from 270deg using G2 (clockwise) in the left box and G3 (counterclockwise) in the right box.

    Notice the mismatch where the lines meet in the middle of each side with the left Y line being the worst. Also notice the astigmatism in the circles is different depending on which way the circles were drawn.

    I'd immediately say "slop" but I can't figure out where. If I move each axis back and forth, I get the same amount of measured movement and the tool carriage has no movement in any direction. The only thing I can think of at this point is that I needed 4.75m of 9mm belt but I only had 4.5m of the genuine Gates so I had to use a 5m length of no-name stuff for now. Can anyone think of anything else offhand?



  • How fast was it going? Acceleration? Jerk?



  • @gtj0 Is the hot end nice and secure. If you put your finger on the nozzle, can you move it in X or Y? Could the drag of your wires and Bowen tube cause the hot end to tilt?


 

Looks like your connection to Duet3D was lost, please wait while we try to reconnect.