CoreXY Design Alternative

I'm a mechanical engineer and have been investigating the possibility of building a self-designed 3D printer capable of handling build volumes of 1 cubic meter (or more, depending on what the main limiting factors are). After researching motion systems, obviously CoreXY came up as a popular choice. Two the downsides of the CoreXY design regularly mentioned however is (1) the fact that the belts are complicated to get setup and (2) tensions have to be carefully and tediously balanced.

I'm by no means an expert in understanding the various tradeoffs involved in the design, so I'm assuming I'm missing something, but it seems like a much simpler design is possible. You can see what I've come up with so far here (link). The design is still far from complete, and the stuff that is "complete" is still pretty rough around the edges (e.g., I may cut the number of motors in half and drive both sides of the cross-bar with a single motor, but the mass of the cross-bars may prevent that from being an option). But the basic idea is there, which is to have the gantry connected to two cross-bars instead of just one with motors connected directly to the cross-bars with belts.

Ultimately, my question is twofold:

1. Why are there no large volume 3D printers?
2. What am I missing? Why is something similar to my design (or at least something that avoids the complexity involved with the belts) not being used already? Certainly, I'm not the first to think of it. So I must be missing a design decision that precluded its use, but I'm not sure what it is.

Look up Ultimaker printers. They use that style of crossed bar gantry. I do like it better than CoreXY. Though Ultimaker uses a cantilever bed, which won't work at your size.

I think the lack of large printers is just due to lack of demand. They're big and heavy and slow and expensive and easy enough to build yourself if you actually have a use for one. Weight increases very quickly with size. Going from 220mm cube to 1000mm cube is about 100x heavier if you scale everything up linearly. Of course you can make the frame proportionally skinnier to keep the weight down, but then you have to make the acceleration rates proportionally lower, when your print time is already going to be painfully long if you utilize much of the available volume. So most likely you'll need to go with a more elaborate bracing structure to increase stiffness to weight ratio.

When large sizes are involved, it can be better to design a printer for the particular job rather than a general purpose printer. SCARA arms are a good way to reach a large area with a small machine, though their work envelope is a circle with an inaccessible circle in the center. Not good for everything, but great for printing circular houses. I needed a printer with >1000mm Z, but normal size bed, and SCARA was just the ticket. The 270mm long arm can reach almost all of the 300x200mm bed, which is cantilever style and rides down on rails, and folds down when not in use. And the whole thing screws to the wall, so the footprint is next to nothing.

Oh! Thanks for pointing me to Ultimaker. That helps me know that I'm not completely insane.

I think the biggest struggle for me is knowing the effects of size as things are scaled up. For example, just yesterday, I discovered Rat Rig and found the design logic behind their bed design to be intriguing. But I'm still working out how the size factor changes things. In any case, it's clear that simply scaling up the design of another system won't work. At this point, I'm just creating the design based on what my gut says because that's basically all I have. In my day job, I run simulations looking at mechanical behavior, and I could at some point use something like that to look at my design to figure out where additional bracing is needed (or alternatively, where lighter parts could be used), but that's a way's out yet.

I don't think corexy is particularly complicated, and there's no "careful and tedious balancing" of belt tensions. Absolute belt tension isn't any more critical than in any other belt driven mechanism. The goal in setting the relative belt tensions isn't matching the two. The goal is to get the X axis square with the Y axis. The last time I did it, it might have taken 15 seconds. You thought balancing two belt tensions was tedious? It looks like your design uses 4 belts...

Everything about your design screams wobbly. Driving the extruder carriage around with crossbars may work fine for a smaller printer, but at 1m square I suspect that things are going to bounce around a lot, especially when the extruder carriage is near the center of the bed. If you build it beefy enough that things don't bounce around, the moving mass will force you to print slowly or use really big motors/drivers which drives up the price. Of course, unless you build appropriately, a corexy mechanism might also bounce around. Large printers of any architecture are difficult and expensive.

There aren't many large volume printers because large printers/prints are expensive/slow and 3D printing hobbyists are mostly cheap/impatient, and who has the space? Large frames require more expensive hardware to try to maintain rigidity. A bed and heater for such a large size is expensive and probably requires 220V or more. You can divide it into sections so you don't always have to heat the whole thing, but that's probably going to require some extra electronics that can talk to the controller and may not be off-the-shelf. It will be difficult to set up a large bed to be flat over its entire surface so unless you're printing in very thick layers with a very large nozzle, you'll need some automatic unflatness compensation which means independent drive of Z axis belts/leadscrews and a controller that can handle the task. The crossbars will sag more as the extruder carriage nears the center of the bed, so even if the bed is flat, you'll still need the active unflatness compensation.

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

Certainly the crossed bars would be too heavy or too flexible if you use solid steel, but thin walled tubes are much stiffer for their weight, like what's used on the MPCNC. Or you could use something like MGN7 rails mounted to carbon fiber tubes (though differential thermal expansion could potentially be a problem). Lighter still if you make foam sandwich tubes [www.youtube.com]
That would also be a good way to make the frame, since it uses very little carbon to create super stiff structure (cheap and lightweight)

Edited 1 time(s). Last edit at 11/25/2021 12:07AM by dekutree64.

Well, I'll be the first to admit that I have very little knowledge about the advantages and disadvantages of different designs apart from what I read on the internet (which can obviously be unreliable), but every place I've looked at that described the advantages and disadvantages of various different mechanisms always mentions difficulties related to belt tension in connection with the CoreXY. I originally wanted to use lead screws to drive the cross-bars directly (i.e., eliminate belts entirely), but then I realized that would run into issues with large amounts of whip. Nonetheless, using two belts that are relatively shorter in a single loop still significantly more simple than a single relatively longer belt in a more complicated belt path. I might be wrong though.

Regarding the potentially wobbly nature of the design, you may be right. The largest print volume I've seen in a commercially available 3D printer is 500mm X 500mm X 500mm. I will be taking several cues from that design for mine to maximize the rigidity as much as possible, but it is indeed difficult to know the limits of how much something can be scaled. All that to say, I greatly appreciate your feedback.

Your point about the heater for the bed is a good one. I was not expecting that to be too much of an issue because this guy's system (link) seems to work okay. Having said that, I obviously disagree with other design decisions he made, so maybe his approach to the heated bed wouldn't work very well.

Ultimately, I'm just trying to figure out where the limits are.

I'm currently using OpenBuilds' V-Slot 20x20 Linear Rail (link) in my design. Since it's aluminum, I'd think it'd be relatively light, but maybe not? My expectation is that the rails are plenty stiff enough, but the mechanism connecting them to the structure may have too much slop in it. In other words, I think the weakest link right now is the mechanism connecting parts together, not the rails themselves. Having said that, the alternatives you've suggested will be very useful if I can come up with a stiffer mechanism. Further, sources for linear rail systems other than OpenBuilds is very useful since they may allow for a more rigid connection mechanism.

There _are_ (commercial/industrial) large format printers. BigRep have their main offices walking distance from where I live, and that's just one option.

Sorry. I guess I was unclear. There's no DIY large printer designs. I suppose this is an unrealistic expectation.

20x20 t-slot for a 1m cube printer? No.....

Being a mechanical engineer, you surely understand that as cross section goes up, rigidity increases dramatically. 20x20 t-slot is barely suitable for a 200x200 mm printer (nevermind that many larger hobbyist offerings use the stuff). You probably also know that a plain tube that's 20x20 is going to be far more rigid (and probably lighter weight, depending on wall thickness) than t-slot that's 20x20. If you don't need to keep moving stuff around on the frame, and you're not using wheels-on-t-slot for linear motion, plain tubing is cheaper than t-slot, more readily available, and a lot more rigid. My 1.6 x 1 m sand table used a corexy mechanism built with 45 mm square t-slot. The belt tension was enough to cause the frame members to bow outward and I had to add a mechanical brace to try to reduce the bowing. You can see the cross member at the middle of the table in this video: [vimeo.com]

Wheels-on-t-slot can work OK, but it makes it difficult to enclose the printer and impossible to improve frame rigidity by attaching panels to the sides, top, and bottom, so you'd better use really big t-slot for a 1m cube. It is probably cheaper to use wheels on t-slot than linear guides for a 1m cube printer. A lot of hobbyist printer designs use a s**t ton of corner plates to try to boost frame rigidity without considering the flexibility of the frame members. Adding corner plates to a 1m cube 2020 frame will have minimal effect on frame rigidity. In most of those designs, if they had put the extra money into larger t-slot instead of dozens of corner plates, screws, and t-nuts, they'd have a more rigid printer that would be easier to assemble and probably lower cost.

As this would appear to be your first printer design/build, you might consider building something smaller first. Get some experience printing with different materials and solving the problems, then build bigger. An unenclosed printer can print PLA, TPU, and PETG. You might get away with small ABS prints, but not large ones. PLA is easy to print (which is why so many people print with it), but it's a junk material the softens at too low a temperature to make it useful for anything that might one day get exposed to heat. It absorbs moisture from the air (just like the filament on the spools) and becomes brittle (just like the filament on the spools). Don't even think about using PLA to print parts for your printer.

I don't know what you've been reading about corexy or where you've been reading it, but there's a lot of bad designs/information out there. I collected a bunch of examples of belt layout errors here. If you see a design in which someone uses clothespin spring tensioners on the belts, it's probably a bad design. In a properly laid out corexy mechanism, belt tension won't vary with extruder carriage position. Standing pulleys up on posts is generally a bad idea- the pulley axles should be supported at the top and bottom to keep the axles from flexing and tilting under belt tension.

This diagram holds the "secret" to a successful corexy layout:



Working belt segments labeled A-H must be positioned parallel to the linear guides. That's it. That's all you need to know. That is also true of any other belt driven linear positioning mechanism. If you feel you must add additional pulleys, you can put them anywhere as long as you don't violate the one rule about keeping working segments parallel to the guide rails, but remember, every additional pulley increases the load on the motors, which translates to reduced operating speed because torque drops as stepper motor speed increases. You can lay the belts out for single (unstacked) pulleys, but the same rule applies- keep working belt segments parallel to the guide rails.

I wrote a blog post on proper belt layout for corexy designs.

Edited 2 time(s). Last edit at 11/26/2021 12:09AM by the_digital_dentist.

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

To be clear, the 20x20 is used for the cross bars to hold the print head, but the outer frame uses 40x40. I have not done the calculation yet to look at the amount of deflection of the various different parts, but keep in mind that each 20x20 rail is only holding half the weight of the head. Ultimately, I was merely looking at the parts available on OpenBuilds and trying to figure out how I could assemble them into a conceptual model. This helped me solve a few design problems, but obviously any time the design is evaluated by someone else (especially when those people have much more experience than I do), many issues are discovered that I either never would have thought of or that I haven't had the time to look at.

If you don't want to do FEA yourself, there's a handy spreadsheet on the openbuilds forum

[openbuilds.com]

Thanks!

Quote
the_digital_dentist
This diagram holds the "secret" to a successful corexy layout:

Stepper motors and belts will be replaced with direct drive soon if all goes well with the dev, which also includes a cheap position feedback method accurate enough for 3D printing.

A direct drive setup has no belts, steppy motors, noise or backlash, is accurate to single digit microns. So no ringing from belts stretching (assuming you get the drive PID's set right), backlash from gears etc, totally silent (apart from noise from the smooth rod bearings or such like), and very fast (depending on how much current used to drive the motors).

But yes, for now it's stepper motors. belts or such like.

I think linear motors are unlikely to be used in 3D printers, at least by hobbyists, any time soon. They are just too expensive to justify the increased cost over the existing cheap solutions.

As far as using t-slot as a cross bar goes, 20mm square tubing will be more rigid. Rectangular tubing of the same size will be more rigid than any same-sized t-slot. T-slot is not rigid stuff.

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

Maybe I was wrong about linear motors... [benwang.dev]

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

... remembers me on this thread

[reprap.org]

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]