CoreXY for bed instead of extruder(s), for small printer?

For a printer with a small printing area (for example 100x100mm) and more than one extruder, in your opinion would it make sense to have a coreXY to move the small bed and have the (direct) extruders lifted only vertically?

Hi,

Well, in the CoreXY desing, the printhead don't move along the Z axis.
So that XY mobile part is very stiff and dead flat.
Only the bed is moving verticaly.

You can apply a coreXY to any printer.
But keep in mind the overall dimensions of a CoreXY printer is pretty larger than the bed,
because you need some extra room for the motors and the belt/pulley system, and also for the bed lift.
If you choosed a 100x100mm print area, maybe you care about overall volume of the machine.
So, a compact design like the SmartRap can be smaller on a desktop.

Actualy, if you want to discover 3D printing or if a 80x80x80mm volume is enough for your projects,
you should consider : [www.hobbyking.com]

++JM

Edited 1 time(s). Last edit at 08/31/2015 03:40AM by J-Max.

Indeed the size of the printer would be considerably bigger than the printing area. On the other hand, since I wanted to build a multi-extruder printer without Bowden, the weight of two extruders would be more than that of a small bed. So I suppose the coreXY system would be lighter, faster and cheaper if the bed was moved horizontally and the extruders vertically, instead of the opposite. But maybe there are other disadvantages I am not seeing.

Conversely, the Mini FabriKator is a very compact and cute machine, but I would not rely on a plastic (acrylic?) structure despite the small size.

Hi again,

Look at the investment.
That mini printer works fine with acrylic, for its dimensions.
I mean much better than a Prusa I3 acrylic,
with its building volume more than 4 times larger !
But to have a first contact with 3D printing,
with a plug and play machine around 150€, it's not a bad idea.
You can keep it as a secondary printer or sell it easily.

I mean you can't design a proper printer if you did not experience 3D printing for a while.

Geared or not, bowden and direct drive systems have pro and cons.
None are perfect. I experimented both, and I prefer bowdens,
because there's less influence of the drive system and filament traction on the printhead,
and the moving parts are lightweight, and that allows faster acceleration.
Once a printer is well set up, you will be looking for speed.

Whatever the printer's architecture, 3D printing means travels along 3 axis.
The most important is to move components along the best possible axis.
If you move the bed back and forth, it will needs to be lightweight... or slow.
You want your heating bed to be dead flat, and to diffuse heat evenly.
The best way to get that is to use at last a 6mm aluminium plate.
It aint necessary to be fast. If the bed moves along the Z axis only,
no matter its weight or its speed.
Compares to a good bed, a printhead is light, expecialy with a bowden extruder.
You want your printhead to stay alligned and move quickly, at last along one axis.
No matter if you use that same purpose along 1 axis or 2.
The lighter the printhead will be, the fast direction changes you will get.

Compares to a carthesian design (like the Prusa I3),
the CoreXY have a lot of advantages.
That's what makes CoreXY exciting, and that's why experienced makers
ends with a CoreXY printer after experiencing carthesian ones.

According to its name, the CoreXY allows the fastest moves on the X and Y axis.
The momentary slow steps Z axis moves (or no moves at all) belongs to the build plate.
That's all you want for a 3D printer, a laser or a plotter.
Interdependent XY axis are simpler to square up,
and the 4 ends push-pull gives the printhead more stability.
Only one motor for each axis versus 2 for the Prusa's Z alike.
No screws missynchronized, no alignment problems, no Z wobble...
If anyone objects or knows CoreXY disadvantages, I'll be happy to hear about it.

Build a carthesian printer if you like, but to me, it will be difficult
to find a better system than the CoreXY for a 3D printer.

++JM

You both make some good points. I would like to put in that we here in the forum sometimes might be a little too biased towards fighting weight.

You can make any heavy build move fast, you just need to build it with proper parts. Obviously if you compare two builds made of the same parts then it will impact performance negatively if one gets a heavier hotend (...) - Though nobody has yet made any real good investigation into how much it affects.

For sure, these days the 3d printer seems to have more in common and be moving towards the consumer ink-jet printers and their low cost injection molded plastic parts then with the modern industrial robots and/or CNC milling stations - Those can move blazing fast and still be really heavy.

The_digital_dentist is a strong promoter of this philosophy - Building heavy but stable and rigid - he is very active on other parts of this forum.

Concerning the topic - Moving the bed - I think it is an interesting idea and worth investigating. I think maybe it is more attractive for a CoreXY CNC mill. If you just want more extruders then I think you are better off sticking to the hotend moving CoreXY and just upping the overall quality of the build.

Occasionally some dreamer comes along who wants to make an all-in-one 3d machine (multi color 3d Printer, CNC, Laser, plotter, coffee maker). If such a machine should be made, I think coreXY with the bed moving would be the way to go.

Quote
J-Max
Look at the investment.
That mini printer works fine with acrylic, for its dimensions.
I mean much better than a Prusa I3 acrylic,
with its building volume more than 4 times larger !
But to have a first contact with 3D printing,
with a plug and play machine around 150€, it's not a bad idea.
You can keep it as a secondary printer or sell it easily.

I mean you can't design a proper printer if you did not experience 3D printing for a while.

I have been experiencing 3D printing for a year now. I have never designed and realized a full printer, but I have modified my smartrap enough to make it unrecognisable, to solve the most important problems it had (Z wobble, impossibility to print with flexible filaments, orthogonality, very weak structure, ...). Not all the problems are completely solved, but it "works" "sometimes". Now I would like a printer that just works.

Quote
J-Max
Geared or not, bowden and direct drive systems have pro and cons.
None are perfect. I experimented both, and I prefer bowdens,
because there's less influence of the drive system and filament traction on the printhead,
and the moving parts are lightweight, and that allows faster acceleration.
Once a printer is well set up, you will be looking for speed.

My smartrap has a Bowden that works relatively well, I have never tried a direct extruder so I would like to give it a try.

Quote
J-Max
it will be difficult to find a better system than the CoreXY for a 3D printer.

One possible weak point of the coreXY is the length of the belts, that are elastic to some extent. How this affects printing quality (in relation to the weight of the carriage, bearings friction, ...) still remains to be measured, to my knowledge.

Quote
LarsK
Concerning the topic - Moving the bed - I think it is an interesting idea and worth investigating. I think maybe it is more attractive for a CoreXY CNC mill. If you just want more extruders then I think you are better off sticking to the hotend moving CoreXY and just upping the overall quality of the build.

I will probably give it a try if it turns out that the bed is considerably lighter than all the extruders.

Quote
LarsK
Occasionally some dreamer comes along who wants to make an all-in-one 3d machine (multi color 3d Printer, CNC, Laser, plotter, coffee maker). If such a machine should be made, I think coreXY with the bed moving would be the way to go.

I would be happy with a printer that can print small parts in 3 colors, I can resist without coffee for now.

He, he !

I hate coffee anyways !
The simpler, the better : one machine, one function.

LarsK, I don't believe on a CoreXY CNC mill
mostly because the interest of CoreXY is speed,
and a CNC mill don't realy need fast moves, even for wood machining.
Through, a belt system on that type of machines
should be very strong to hold the machining forces.
And you should need huge motors too, or a big reduction.
The best drive system for a CNC mill is definitely the screw.

christian, the Corexy belt system is a push-pull loop transmission.
What you could lose at a end will be compensaded at the other end.
Movements are assumed simulteaneously by 2 belts too.

There's max extension values for each belt, that's true.
But, that does not means a belt is so flexible.
That value is max extension before the belt brakes.
That means you should overtake the admissible load for the belt.
For a reprap, what are we talking about ? 0.2Nm ?
The load on a 3D printer is definitely negligeable, even for a GT2 belt.
There's different belt quality, but a low quality armoured belt already stands a lot.
To be confident you can even upgrade to a fiberglass armoured belt too, its no big deal.
But it ain't necessarary so.

++JM

Edited 1 time(s). Last edit at 09/01/2015 04:41AM by J-Max.

Quote
J-Max
christian, the Corexy belt system is a push-pull loop transmission.
What you could lose at a end will be compensaded at the other end.
Movements are assumed simulteaneously by 2 belts too.

There's max extension values for each belt, that's true.
But, that does not means a belt is so flexible.
That value is max extension before the belt brakes.
That means you should overtake the admissible load for the belt.
For a reprap, what are we talking about ? 0.2Nm ?
The load on a 3D printer is definitely negligeable, even for a GT2 belt.
There's different belt quality, but a low quality armoured belt already stands a lot.
To be confident you can even upgrade to a fiberglass armoured belt too, its no big deal.
But it ain't necessarary so.

I am not talking about max extension. I am talking about the fact that we need precise positioning in the order of magnitude of 0.1mm, and a belt in a coreXY system can be as long as 1m or more. So what is the force needed to cause a misplacement of 0.1mm in a coreXY system with such belts? Given such a force and the weight of the carriage (although one should also consider bearings static and dynamic friction, in particular for cheap bearings) one may obtain the maximal allowed printing acceleration over X and Y (and the maximal allowed friction for bearings). But we need some data about belts elasticity to estimate that.
While a coreXY system would of course allow you to go very fast, I would not be surprised to see that the maximal acceleration allowed to obtain such a precise positioning is quite low.

Of course we know from experience that even cheap belts "work" somehow, but I don't think that posing such questions hurts, and I wish there were more precise measurements made and reported online.

Hi again.

Under 1.3Nm you won't have any issue vith a GT2 belt.
Seems like you need manufacturer's data to make up your mind.
So just ask your provider or a manufacturer.

++JM

Quote
J-Max
Seems like you need manufacturer's data to make up your mind.
So just ask your provider or a manufacturer.

Definitely, provided that one buys from a serious manifacturer. It should also be possible to some extent to measure the reliability of the coreXY system in almost static conditions for small distances, such as one or a few motor steps. It would also be interesting to see how this changes as a function of belt tension.

I will do some tests when I get the dial indicator.

Cristian, have you got anywhere with this idea? Design seems similar to 3drag or K8200 printers except that corexy layout allows all the motors be connected to the frame (on mentioned printers Y motor is moving with a bed).
I am currently building a normal corexy and am considering swapping it upside down later.

Not yet started the build, I purchased a cheap linear guide, my first dial indicator, a scale (the third one...) and I am getting ready to take all the measurements to decide what hardware should be used to get the precision I want (small printer ==> high precision). I will first make some measurements to choose between spectra line and timing belts (and in the case of belts, which ones). Then I will see if cheap Chinese guides are up to the job, or if I really need to get some pricey stuff.

The 3drag is the most similar design, yes, although I discarded already smooth rods and linear bearings because they seem too heavy (and sellers don't give enough info on clearances, generally). I also will try to avoid as much as possible plastic parts, if not at the beginning at least in the long term.

If you try something similar, please share here your experience.

How many extruders Cristian?

I could see it working. Take something like the X3D mechanism modify the Bowden Carrier to hold the bed. Make the axis lengths long enough that all the extruders can reach the entire bed area. Mount the Z axis above it in the middle like a CNC router where the head drops down from the top, and your done...

I would like to start with 2 extruders. Maybe that will be enough already, maybe not, depending also on the effort needed for their calibration, how well multi-"color" printing will work (although I am not really interested in "colors" as much as in mixing different materials), etc.

The whole idea was that if the bed is light enough one may get all the advantages of a (multiextruder) corexy system without need for a Bowden setup.

Technically I agree that there is nothing new to invent, it is just a remix of existing setups. The difficult part for me is that I want to use metal as much as possible for the structure, but of course I will start easy with printed parts for the most difficult pieces.

I will post updates on my pre-build and build progress as soon as I make some.

I must say that we both have a pretty similar mindset on what we want to make.
I am also eyeballing linear rails. I think it would be smart to use linear rail for an X-axis as it is a moving one and should be kept as stiff and light as possible. Are you from Europe? This is a cheap system although there are no specifications on preload or pretty much anything: [www.dold-mechatronik.de]
For Y and Z (which are fixed to the frame) axis I would probably go with fully supported linear rods and bearings that have adjustable preload: [www.dold-mechatronik.de] (I am very familiar with a play between cheap linear bearings and rods but hope that his setup should take care of that)
I really like this configuration of Z-axis: [www.flickr.com] It is great that motor is fitted on a center as there is no need to mess with two motors or belts to connect multiple lead screws.

Yes I live in Lille, France. I know www.dold-mechatronik.de as a partner of motedis.fr, they both look interesting but I get stuck on dold-mechatronik because of the language (google translate does not help much because you cannot really click and see details of the objects or your cart...).

The "cheapest" preloaded guides that I found are Misumi's, which are of course not cheap at all but may be worth every euro. Chinese stuff gets expensive too when looking for preload. On the other hand I started wondering about how the friction given by preload may affect the hysteresis of the belts in the corexy system: it is probably negligible but I will be able to tell more as soon as I measure line and belts elongation properties.

I also concluded that for the X axis a single linear guide should be used, while for Y the choice is more free. Bearings with adjustable preload are probably a very smart idea, if you try them please let me know how they work.

In this setup, Z is for sure the less critical axis. I was also considering some central lifting mechanism to avoid belts or double Z motors, so we definitely have a very similar design in mind. I don't know yet what would be best, either some very simple mechanism like the one that I have on my current printer already (but that would not work with preloaded guides or bearings, which may be bad), or a classic lead screw.
I am fully open to discussion.

Hi,

Un nordiste... Bon, on te garde quand même... ^^

Christian, don't be affraid about the belts. Belts are not elastic stuff.
Stretch values of 0,4% max are in extreme issues, just before beaking.
A GT2 6mm quality belt allows 120Nm !
You won't get the quarter of that on your printer, even with preload or friction.

I recommand polymer bushings instead of linear bearings.
The preload is adjustable, and they're quite noiseless.
Look at Igus.fr for alternatives.
They makes a double rail, a very smart solution for the Xrail.
You should consider the V-slot technology too.
That guides are quiet and very sturdy too.
It's possible to mill some Motedis profiles by yourself with a simple router.
Both are not very expensive guides.

++JM

Hey guys, I'm not an engineer or anything so I'm hoping you can clear something up for me as I'm following along... Why are belt tensions being expressed as a torque value in Nm? Wouldn't a belt under tension strictly be a one dimensional force that should just be in N? Sorry if this is a dumb question.

Thanks.

Quote
J-Max
Un nordiste... Bon, on te garde quand même... ^^

Les Chti sont sympa, oui, mais moi, je suis pas Chti, je suis Italien !

Quote
J-Max
Christian, don't be affraid about the belts. Belts are not elastic stuff.
Stretch values of 0,4% max are in extreme issues, just before beaking.

Then I guess the belts of my printer are quite low quality: I can easily stretch them about 1.5% and I am far from breaking them.
I have no idea of their origin (they came in a kit). Where did you get info about 0.4% max elongation?

Quote
J-Max
A GT2 6mm quality belt allows 120Nm !
You won't get the quarter of that on your printer, even with preload or friction.

I don't know how to interpret N*m, I would expect newton only in this context, since belt dimensions are given.
And I don't know how to explain my concerns further: the problem is not the maximal elongation of the belts, but the displacement caused by perturbations along the belts, which consist in accelerated and decelerated loads, friction, etc. This article is an interesting read, although lacking a lot of info about the setup they used.

Quote
J-Max
You should consider the V-slot technology too.
That guides are quiet and very sturdy too.

Aluminium profiles are usually not guaranteed to be really straight. I know that "in practice" they are straight "enough" for most purposes, but maybe not for the precision I hope to achieve (or maybe yes, who knows?).

Hi

Quote
Ryan_M
Hey guys, I'm not an engineer or anything so I'm hoping you can clear something up for me as I'm following along... Why are belt tensions being expressed as a torque value in Nm? Wouldn't a belt under tension strictly be a one dimensional force that should just be in N? Sorry if this is a dumb question.

I realy don't know. Maybe it's because of motion and speed while N is for a still mass ?

Chrstian, I got that values from manufacturer's specs.
FDM technology allows some kind of precision,
but you can't reach extreme precision too.
V-slot and belt drives fits the precision FDM can provide.
Christian, what precision would you expect ?

++JM

Quote
J-Max
Christian, what precision would you expect ?

For my needs, accuracy is probably not as important as precision, reason why aluminium profiles may work for me as well. On the horizontal plane, I would like to print reliably at 0.25 mm layer width, while vertically I would like to reach 0.05 mm layer height, which is the main limiting factor. I guess I am looking for precision in the 10 micron scale then?

Hi,

Christian, you can't reach a 10µ precision with any FDM printer IMHO.
The best you can expect is +-150µ. You'd rather look at SLA or DLP printers.

A 10µ precision printer is obviously expensive.
Only C0 preloaded ballscrews allow that kind of transmission precision.
A single 300mm C0 ballscrew is more expensive than most of the reprap kits.
You must be very precise when building and squaring the machine too !

Look at first quality manufacturer's printers,
and if you find a 10µ capable machine, look at the retail price
That's what we're talking about here.

++JM

Quote
J-Max
Christian, you can't reach a 10µ precision with any FDM printer IMHO.
The best you can expect is +-150µ. You'd rather look at SLA or DLP printers.

Okay, now I am starting to understand what you expect from a fdm printer and why belts are fine in your opinion.
I think that in some cases repeatability can be better than 150µ, otherwise nobody may really get good perimeters at 200µ layer height. But in general I agree that most printers cannot achieve much better than that (I suspect that my current printer is worse, actually).

Quote
J-Max
A single 300mm C0 ballscrew is more expensive than most of the reprap kits.

If it worked well, the price may even be fine with me (and 120mm ballscrew would be enough for the small printer I am trying to build). But wouldn't it be too slow then?

You can often get better resolution than 150um in the Z direction from an FDM printer, but not usually in the X and Y directions, because the extrusion width is typically about 400um.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
dc42
You can often get better resolution than 150um in the Z direction from an FDM printer, but not usually in the X and Y directions, because the extrusion width is typically about 400um.

The problem for me is not just the resolution. It is true that details smaller than 400µ cannot be printed with a standard nozzle, but if the positioning system allowed a precision of 10µ along X and Y, then a vertical wall printed at 50µ layer height would really look good, even if the layer width is 400µ. Conversely, with a precision ( ~= repeatability) of 150µ, a vertical wall printed at 50µ looks horrible independently of the chosen layer width I think.

Quote
cristian

Quote
J-Max
A single 300mm C0 ballscrew is more expensive than most of the reprap kits.

If it worked well, the price may even be fine with me

After seeing the price for a preloaded C3 ball screw from Misumi I am no longer so sure that it is fine...

Hi,

Well, I gess there's missunderstanding between resolution and precision.
Resolution is what the printer theoricaly can give.
Precision is what the part is compared to it's virtual 3D model.
You can't reach your model's dimensions within 10µ with a FDM printer.

You need a damn good SLA or Powder printer to get close to that !

It's not depending of the mecanics, but of the fused filament technology itself.
There's an aleatory factor with fused filament.
You need to squeege the filament on the previous layer to make a good bond.
So, if you have a 250µ nozzle maybe your filament path will be +-275µ
depending of various factors, like speed, vibrations, heat variation,
fused filament expansion, and many little random parameters.

On a SLA printer, the path is created by a laser, which is thinner than any FDM nozzle.
If you use a quality laser module properly focused, then you get much better results.
Theres some light diffusion in the resin, but it smooths the parts edge.

There's an example of a good FDM print :


And now an example of a good SLA print :


Just a question Christian, what is that printer for ?

++JM

Quote
cristian

Quote
dc42
You can often get better resolution than 150um in the Z direction from an FDM printer, but not usually in the X and Y directions, because the extrusion width is typically about 400um.

The problem for me is not just the resolution. It is true that details smaller than 400µ cannot be printed with a standard nozzle, but if the positioning system allowed a precision of 10µ along X and Y, then a vertical wall printed at 50µ layer height would really look good, even if the layer width is 400µ. Conversely, with a precision ( ~= repeatability) of 150µ, a vertical wall printed at 50µ looks horrible independently of the chosen layer width I think.

My delta printer has a resolution of better than 25um in the Z direction (i.e. I can move it in 25um Z steps and detect that it moves - I haven't tried smaller steps), which means it should be around 12um in XY. At present there is some backlash in some of the Traxxas joints that spoils the XY resolution in some prints, but I should be able to overcome that.

I have no experience of CoreXY designs, but I am guessing that stretch in the long belts may be a limiting factor. If it isn't, or it can be overcome, then I think you should be able to achieve better than 50um resolution (but not accuracy) in the XY plane. You may need to use 0.9deg/step motors to get higher incremental torque, and small enough pulleys to get a high steps/mm factor.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
J-Max
You can't reach your model's dimensions within 10µ with a FDM printer.

I am becoming aware of that...

Quote
J-Max
It's not depending of the mecanics, but of the fused filament technology itself.
There's an aleatory factor with fused filament.
You need to squeege the filament on the previous layer to make a good bond.
So, if you have a 250µ nozzle maybe your filament path will be +-275µ
depending of various factors, like speed, vibrations, heat variation,
fused filament expansion, and many little random parameters.

If we are talking about the worst case, I totally agree (still wondering where you get those numbers ).
However the "average case" can't be so bad, otherwise nobody could obtain decent surface look at layer height lower than 0.5mm.
In my opinion, several printers+slicers can obtain an overall precision of 100µ or better if you do not consider the most critical points or conditions (sharp corners, retraction points, etc.).

In any case, the worse the precision of the positioning system, the worse will be the overall precision including filament extrusion. So having a precision of 10µ for mechanics, if achievable, would be already a fantastic starting point and I would be very happy with that.

Powder printing is interesting, but it seems to me less feasible at the moment (not considering industrial products of course).
SLA gives wonderful results but it is out of my budget, and I don't know how delicate are the printed objects and their surface.

Quote
J-Max
Just a question Christian, what is that printer for ?

Well, officially for small plastic jewelry. In reality, it is my personal crusade to get the best precision out of homemade (or home-assembled) FDM technology given a reasonable budget (which is not 300€, but not even 2000€).