CoreXY or Cartesian which is better ?

[fcwilt]

@maxgyver said in CoreXY or Cartesian which is better ?:

Cons:

• Longer belts always add "springiness" to the system

• Higher number of belt idlers, that each add a little "wobble" to the system, since even good idlers are often far from running perfectly true

• More idlers mean more parts and also higher wear on the belt since it is bend back and forth around multiple radii.

• The necessary interpolation of X and Y Moves drastically limit the top speed of diagonal moves

• Can take a lot of fiddling until all motors move in the right way

Have you any data to back up these "cons"? The last two are particularly suspect.

I certainly have not found these "cons" to make any noticeable difference between my printers.

[MaxGyver]

@fcwilt said in CoreXY or Cartesian which is better ?:

Have you any data to back up these "cons"? The last two are particularly suspect.
I certainly have not found these "cons" to make any noticeable difference between my printers.

No, these points are only based on my personal experience and should be seen as such.

Right now I am designing a copy of my markforged-style printer but with a Cartesian belt setup, I am happy to share some data when it is finished.

[peter247]

@maxgyver

So a prusa type cartesian printer is faster than a Ender 6 core XY , EH ?
I think you are the only one who thinks a cartesian is faster than core xy in general.
The question is wrong to start with ? what is cartesian printer ?
In the hobby market a prusa and ender 5 are both cartesian ?

[MaxGyver]

@peter247

I have experience with neither of both. My cartesian printer is a gantry-style, not a Prusa style "bed slinger".

@peter247 said in CoreXY or Cartesian which is better ?:

The question is wrong to start with ?

Absolutely, the correct answer to the question is: It depends on what you are looking for in a printer...
Print speed, print quality, reliability, cost etc. These are all factors that have to be considered before building or buying a printer.

[peter247]

@maxgyver said in CoreXY or Cartesian which is better ?:

Absolutely, the correct answer to the question is: It depends on what you are looking for in a printer...
Print speed, print quality, reliability, cost etc. These are all factors that have to be considered before building or buying a printer.

I couldn't agree more and the place to start is size.

The best printer is 2m x 2m x 2m , that will print everything ( Joke )
It is a case of the bigger the print area the more weight , the slower the printer. but if the printer is too small that is more of a limiting factor , so in the end you finish up with 2 printers.
one printer with does all the normally small items fast and one printer for the larger items slow.

[deckingman]

@peter247 said in CoreXY or Cartesian which is better ?:

It is a case of the bigger the print area the more weight , the slower the printer..........................

It never ceases to amaze me how this myth propagates. It generally goes along the lines that higher mass needs more force to accelerate it. Therefore, for a given motor torque, a lower mass can be accelerated faster. But that completely ignores the fact that what really limits print speed is not how fast the hot end can travel though space, but how fast one can melt and extrude filament. It is a constraint that cannot simply be ignored. Much like a Ferrari might be capable of much higher speeds and accelerations that say a Fiat 500 but on the school run, where both vehicles are constrained by the same heavy rush hour traffic and speed limits, the Ferrari won't do it any quicker than the Fiat, regardless of what it might be capable of without those restraints.

For info, by feeding 5 filaments into a 5 melts chambers using 5 extruders and a mixing hot end with a mixing ratio of 20:20:20:20:20, I was able to achieve a melt rate in the order of 45mm^3/sec which allowed me to print at 300mm/sec using a 0.5mm nozzle and 0.3mm layer height as documented here in October 2018 https://somei3deas.wordpress.com/2018/10/14/real-3d-printing-at-high-speeds-and-even-higher-melt-rates-with-a-large-nozzle/

What may surprise you is that a 5 input Diamond hot end, slung between two parallel rails has (in my case) a moving mass in the order of 2Kgs in the Y direction, and that above tests were accomplished with Nema 17s running at 1.8 amps and using CoreXY kinematics. To this day, I still use 350mm/sec non print move speed (even though I rarely print at much more than 100mmsec).

[peter247]

@deckingman said in CoreXY or Cartesian which is better ?:

It never ceases to amaze me how this myth propagates. It generally goes along the lines that higher mass needs more force to accelerate it. Therefore, for a given motor torque, a lower mass can be accelerated faster.

I don't think that , it is more to do with the mass deaccelerating and newton law on the conservation of energy.
The more the mass the more ghosting you have to deal with at a fixed speed , Using your example if you have a 2 ton 4 wheeled drive V a fiat 500 on a head on crash the one with the most mass wins .

[deckingman]

@peter247 Sorry mate. You said quote " ...........the more weight , the slower the printer". That is a sweeping statement and the one which I was refuting. So you can't now turn around and say (quote) "I don't think that".

Also, how on earth do you arrive at this statement - quote "The more the mass the more ghosting you have to deal with at a fixed speed" ? Please elaborate the logic behind that statement. Given that the resonant frequency of any object is inversely proportional to it's mass, I'd say the opposite is true.

[giostark]

@deckingman
All your exchange is interesting , especially for noob like me. Hope you will continue without start fighting
Please be patient if I dare ask trivial questions...
"The more the mass the more ghosting you have to deal with at a fixed speed" . This phrase doesn't mean that if an heavy head is thrown the belts will suffer of more elongation despite a small mass head? (generating more "wobbling"). The resonance frequency is applied to static object or in movement or in what way? I would understand.

[peter247]

@deckingman Sorry !!! , My logic is you would think that lighter is better due to conservation of momentum , or why do people go to the trouble of getting super light extruders ?
Yes my bigger is slower is like the which is best statement, a little too over generalised

[deckingman]

@peter247 said in CoreXY or Cartesian which is better ?:

@deckingman Sorry !!! , My logic is you would think that lighter is better due to conservation of momentum , or why do people go to the trouble of getting super light extruders ?
Yes my bigger is slower is like the which is best statement, a little too over generalised

Exactly so! Why do people concentrate on reducing mass far beyond the point where mass itself is not a constraint or limiting factor?

If, as I have demonstrated and as simple mathematics will confirm, it is not only possible but relatively easy to build a printer with a moving mass of over 2kgs, which will exceed the speeds and accelerations that filament can be extruded, why do people spend so much time and effort on reducing that mass? The answer can only be that they either ignore all other factors and constraints and look at mass in isolation, or that they simply repeat what they have heard/read.

Why do manufactures make and promote ever lighter hot ends? Because people will buy them thinking that they will be able to print faster (because lighter must be better), then spend the rest of their lives printing at 60mm/sec because that's the melt rate limit of the super lightweight hot end that they bought in preference to a one with a higher melt rate but which might be a few gms heavier. It's simply absurd.

I repeat, one cannot make use of the extra acceleration or speed that the reduced mass will theoretically allow one to achieve, because it's impossible to accelerate the melt and extrusion rate of the filament to match it.

Also, a higher mass has a lower resonant frequency and is less susceptible to being affected by external sources of vibration. I don't get any artefacts that can be attributed to ringing or ghosting - it just doesn't happen like it does with super lightweight, and possibly less rigidly mounted hot ends.

Even with my 2Kgs plus, the limiting factor on how fast I can print is the melt rate of the hot end which I can get up to 45mm^3 / sec. If I could increase that melt rate further, I could print even faster than 300mm/sec without reducing that 2Kgs of mass.

Or to put it another way, if I halved my 2Kgs of moving mass, I still could not print any faster than the 300mm/sec that I have demonstrated because, as I have repeatedly said, with FDM printing the moving mass is not the limiting factor - the melt and extrusion rate of the filament is what limits the speed.

[deckingman]

@giostark said in CoreXY or Cartesian which is better ?:

@deckingman
....................
Please be patient if I dare ask trivial questions...
"The more the mass the more ghosting you have to deal with at a fixed speed" ...........................

I have no idea what that means either - that was a statement made by @peter247 and I wish he would elaborate on how he comes to that conclusion.

[dc42]

Warning: long post with lots of maths, some of which may be wrong.

As I see it, the main problem with higher moving mass, assuming sufficient motor torque to accelerate it at the desired rate, is the reduction of the resonant frequency of the system. The resonant frequency of a spring+mass system is given by:

f = (1/(2*pi)) * sqrt(k/m)

where (for example) k is the spring constant in Newtons per metre, and m is the mass being moved in kg. So higher mass reduces the resonant frequency. Low resonant frequencies normally cause more issue than higher ones in 3D printers.

The main components of the spring constant k are the belts and the motors. It's easiest if we take the reciprocal of k to give a "stretchiness" measure (in metres per Newton) instead. Let's define this stretchiness factor as S = 1/k. Then:

f = 1/(2*pi * sqrt (S * m))

Reducing S counteracts the effect of increasing m.

We can break S down into two main components:

s_total = s_belt + s_motor

s_belt is the stretchiness of the belt subsystem. Mostly this is caused by the belt, but also includes any leaning of the shafts bearing the pulleys that varies with belt load.

s_motor is the intrinsic elasticity of the motor. This is dependent on the holding torque of the motor at rated current, the angle per full step, and the actual current used.

[Skip the next section if you are not interested in the maths.]

The torque of a stepper motor with a microstepping driver is approximately:

T = (holding_torque/sqrt(2) * (actualPeakCurrent/ratedCurrent) * sin(2 * pi * lag_angle/(4 * full_step_radians))

This assumes that the holding torque is quoted for both phases energised at the rated current. At low lag angles this simplifies to:

T = (holding_torque/sqrt(2) * (actualPeakCurrent/ratedCurrent) * 2 * pi * lag_angle/(4 * full_step_radians))

Divide by the pulley radius R to get the force:

F = (holding_torque/sqrt(2) * (actualPeakCurrent/ratedCurrent) * 2 * pi * lag_angle/(4 * R * full_step_radians))

The distance moved during elastic motor movement is:

D = lag_angle * R

The stretchiness is just dD/dF. Therefore, at low lag angles:

s_motor = 4 * R^2 * full_step_radians * sqrt(2) * (ratedCurrent/actualPeakCurrent) / (holding_torque * 2 * pi )

The distance moved in metres per full step is R * full_step_radians. To put it another way:

R = 1/(full_steps_per_metre * full_step_radians)

Substituting for R:

s_motor = 4 * sqrt(2) * (ratedCurrent/actualPeakCurrent) / (full_step_radians * full_steps_per_metre^2 * holding_torque * 2 * pi)

Substituting full_steps_per_metre = 1000 * (steps_per_mm/microstepping):

s_motor = 4 * sqrt(2) * (ratedCurrent/actualPeakCurrent) / (full_step_radians * (steps_per_mm/microstepping)^2 * 1000000 * holding_torque * 2 * pi)

Converting the full step angle to degrees:

s_motor = 4 * sqrt(2) * (ratedCurrent/actualPeakCurrent) / (full_step_degrees * (pi/180) * (steps_per_mm/microstepping)^2 * 1000000 * holding_torque * 2 * pi)

Simplifying the constants:

[this is the result, so you can stop skipping]

s_motor = 0.0000516 * (ratedCurrent/actualPeakCurrent) / (full_step_degrees * (steps_per_mm/microstepping)^2 * holding_torque)

Note that holding torque here is in N m whereas for Nema 17 motors the datasheets usually specify N cm.

Example: 17HS19-2004S1 motor (holding torque 59Ncm @ 2.0A) run at 1.6A with a 20 tooth GT2 pulley (80 microsteps/mm @ x16 microstepping):

s_motor = 0.0000516 * (1.6/2.0) / (1.8 * (80/16)^2 * 0.59) = 1.55e-6 m/N

This means that with a 1kg print head and an infinitely stiff belt, the resonant frequency is:

f = (1/(2*pi)) / sqrt(1.55e-6) = 148Hz

In practice the belt will add to the elasticity, so the resonant frequency will be lower.

[Luke'sLaboratory]

@deckingman

Yeah, definitely in this camp. Most printers if built right will outrun plastic meltrates unless you're in the supervolcano/magnum+/your crazy mixing hotend land. The thoughtfulness put into the motion (stiffness of all the parts moving, Eliminate slop in belting system, etc) matters more than the individual gantry motion. The only exception is acceleration - for smaller printers with smaller prints, this is definitely king, with the CROXY gantry style taking the "ultimate" cake. 4 Motors, 2 per axis moving a very light crossed gantry is the way to go if you have to have the absolute zaniest acceleration. Scales well into larger machines in addition.

I have a 400x400 bedslinger with a 5kg bed assembly that I run at 3000mm/s2 and print at volumetric limits of the skeeter magnum that I have installed on there (30-35mm3/s, ABS). For parts printed on a small .4nozzle with .2mm layer heights, sure, I run out of available speed at 300mm/s when I hit EMF limits, and I guess, yeah, reducing weight would get me a slightly higher top speed, before I'd have to look at changing my electronics architecture (to go to 36 or 48v) to get significantly past that point anyways.

My corexy builds (400x400 or 622x622) use 2040 for gantries, and again, don't run out of chooch before my motors hit an electronic wall at 350mm/s. This more than handily keeps up with Volcanos and Magnums, and will probably only struggle with the two magnum+'s I have in the mail, but only with nozzle sizes .6 and below at lower layer heights.

That being said - All of the motion components in both printer designs are in Double-shear and use quality components, and in addition, are using 9mm belts instead of the common 6mm belts. I suspect that changes to 9mm from 6mm would drastically assist quite a few larger corexy designs.

[deckingman]

@luke-slaboratory For info, I'm still using 6mm belts on my CoreXYUVAB. The XY gantry carries the heavy 6 input mixing hot end which has a moving mass of around 2gs. But the UV gantry carries the 6 Bondtech BMGs and that tips the scales at over 3Kgs. Admittedly, I had to use Nema 23s to get the required torque to accelerate that mass to the desired speed, but I'm still using 6mm belts with it.

My only concern about using 9mm belts is that they require significantly more force to tension them correctly, and I wonder if the relatively tiny bearings found inside the average stepper motor are up to the job. If I ever go the 9mm belt route, I'd likely mount the motors remotely driving an intermediate shaft with bigger bearings.

[Luke'sLaboratory]

@deckingman

Yeah - 6mm belts can definitely drive those masses!

My experience for the past year has been that as long as the pulley is mounted as close as possible to the motor, no bad effects have been noticed on my largest machine that I haven't upgraded to double shear yet. On the smaller version I made of it, I made a double shear support that floats with the motor casing to keep it from bending and causing issues. So Far, So good, and it spins plenty fast and tensions as high as you'd like it.

The belt dust is from when I had a light interference with the belt from the edge, which I fixed with a deburring tool. Now we're at low clearance