Print speed limits?

I would like to understand what currently limits most printers to printing under 150mm/s or less before quality goes to pot. I realize that the mass and stiffness of the moving parts play a big role but I can design those parts to accommodate high speeds/accelerations. I don't know enough about the control side to know where the limits are there.

I ran a few searches and didn't find much on this. I'm wondering, what are the practical print speed limits in terms of mm/s move speed for the following components. I realize I'm oversimplifying. Feel free to point me in the right direction:

1. Stepper motor. How fast can a typical NEMA 17 move a carriage using a 10mm dia pulley?. If the stepper is limited to 300 rpm then that puts the linear speed at around 150 mm/s. Can I run the steppers any faster without risking missed steps?
2. Typical extruder system. Assuming a 0.35mm nozzle and 0.25 layer you would lay down around 35 cubic millimeters of plastic per second. How fast can current extruders feed filament before they jam or slip?
3. Controller card. Not sure how to characterize it but is processing or data transfer rate a limiting factor once you get up to let's say 400 mm/s print speeds? Are some cards/firmware combos much faster than others?

Assumptions:
Stepper/belt drive for XY axes
Open source control card/firmware

Any links to some good discussions on the topic of designing a faster printer?

Thanks for any help!

AFAIK, the limitations are the extruder itself. Here's the thread I started to ask about the problem.

[forums.reprap.org]

I've spent the last 9 months designing a printer; one of the design goals was high speed operation without a loss of precision.

I think in general, the limiting factors for many printers is poor design choices. Primarily, moving mass in the form of the bed platform and stepper motors is high. In some cases there are also issues with frame stiffness and the layout of the axes, how things are constrained, etc.

In the case of a well-designed printer, which I think mine is, the limiting factor appears to be control over the extruded filament. Not necessarily the peak speed at which you can extrude, but the ability to precisely start and stop the flow, and modulate it around corners (the flow must be slowed as the printer decelerates to go round a corner, for example). In my case, this is somewhat exacerbated by using a bowden instead of direct drive, but the problem is still present for direct drive extruders.

In theory it is possible to compensate for these issues with firmware. This is what Sailfish does with JKN-Advance and other features. I've asked for those features to be ported to Marlin or some other RepRap firmware, but no one seems interested in undertaking this. (Marlin has something called ADVANCE but my experimentation with this did not yield good results. I am not a good enough coder to dig into the code and figure out what's going on.)

EDIT

There are some secondary limitations concerning the stepper motors and how high you can set the jerk and acceleration before you see artifacts in the print. There's a bit of disagreement in the source of those artifacts. To some degree higher driver current helps until you hit the stepper clipping limit....

Edited 1 time(s). Last edit at 10/23/2013 10:30AM by crispy1.

Crispy1 what sort of design did you use? Would you care to share the juicy details? I'd like to know what you've come up with. I'm aiming for the same goals but have only got a few weeks of research under my belt so tips would be greatly appreciated.

@crispy1

I've also been designing my own, but am no where near starting to build.

As far as I can see the moving bed on most repraps is a major issue to speed. Not just its mass, but that objects simply break off the bed if you move it too quickly.
The taller the print the worse this gets.

Aside from a deltabots, I've looked at the Makerbot / Solidoodle and also the Ultimaker, to attempt to work out which has least mass to move.

Looking at Ultimaker prints, there are often loads of wisps of filament, as the extruder can't stop the flow quick enough when moving from one part of the object to another.
So I'm not sure that a bowden cable unit is ultimately going to give the highest speed, without some loss of quality or need to post process the print.

Ultimaker's geometry is interesting in that the X and Y motors are fixed, but the trade off for this is possibly the flexing of the X and Y bars go from the X and Y carriages to the print head assembly.

H bots also look interesting as X and Y motors are also fixed, but the trade off is a very long belt as well as twisting / slack in the Y bearings, causing minor errors in print head position.

However as I'm not a mechanical engineer, I'm not sure how you'd model the dynamics of each of these types of system to determine the best configuration.
I presume this requires some sort of physics simulation. where the parameters like stretch in GT2 belt, and flexing of 8mm smooth bar and looseness in bearings could be modelled.

Assuming that the extruder motor needs to be at the print head, one other thing is reduction of weight of the stepper motor.
I'm not sure if there are light weight stepper motors that use rare earth magnets, but this would be one way to reduce the weight of the stepper, because as far as I can see a lot of the weight of the stepper is the permanent magnets, and they don't seem to be designed with weight as a consideration

Regarding extruder weight, the amount of motor power required to extrude is really only a few watts. Most NEMA 17 motors used on extruders have way more power than needed, but the gearing and motor drives are mismatched for the task. You can get steppers to run well over 1000 RPM, but you need a pretty high voltage driver. Another option would be a small DC motor with encoder and a servo control system. These options are both expensive, but that comes with the territory with any high-performance solution.

edit: Also see http://forums.reprap.org/read.php?1,263680 for a discussion on closed-loop stepper control

Edited 1 time(s). Last edit at 11/14/2013 01:17PM by LoboCNC.

That is only true when you ignore retraction where you need much higher speed and the same torque. This extruder used a tiny stepper but it was too slow at reversing to get blob free prints.

---

[www.hydraraptor.blogspot.com]

Quote
nophead
That is only true when you ignore retraction where you need much higher speed and the same torque. This extruder used a tiny stepper but it was too slow at reversing to get blob free prints.

How close were you to having suitable retraction speeds? I'd guess the little motor you were using might only deliver about 1w of power. My back-of-the-envelope calculation is that you'd need more like 3-4w.

Quote

Looking at Ultimaker prints, there are often loads of wisps of filament, as the extruder can't stop the flow quick enough when moving from one part of the object to another.
So I'm not sure that a bowden cable unit is ultimately going to give the highest speed, without some loss of quality or need to post process the print.

Given my (very positive) experience with bowden extruders, I think retraction issues like you describe are specific to their implementation. They are direct-driving 3mm filament, which gives a very low steps per cubic mm and low torque with which to retract the filament. That said, I have seen plenty of prints from ulitmakers that didn't have stringing. So it's partially a tuning/maintenance issue.

Quote

H bots also look interesting as X and Y motors are also fixed, but the trade off is a very long belt as well as twisting / slack in the Y bearings, causing minor errors in print head position.

You should look at this website: [www.corexy.com]

@crispy

I have looked at CoreXY, and have contacted a local laser cutting company about getting the components cut, but I'm still not sure it's any better and a conventional XY head type, ie where one motor is still on a carriage, because the Y gantry looks like a substantial bit of metal, albeit aluminium

I'm not sure how much lighter this is than just using plastic carriages and a stepper on one of the carriages.


Howerever is still an option for me, as I already have a lot of the parts I'd need e.g a 4 length of gt2 belt and a load of 10 mm linear bearings

Quote

but I'm still not sure it's any better and a conventional XY head type, ie where one motor is still on a carriage, because the Y gantry looks like a substantial bit of metal, albeit aluminium

Maybe I don't understand what you mean here. But I don't see a reason the Y carriage (I call it the X crossbar assembly) has to be any heavier or larger than the one in a traditional axis layout (eg with a flying X motor).

Hi Crispy1

I've no experience of H bots, but all the designs, pictures and videos I've seen, made it looks like the components needed to be quite rigid because of the twisting forces.

Perhaps this only applies to the frame.

Would it be possible to have a H bot with plastic carriages and 8mm smooth rods?

Quote
rogerclark
Hi Crispy1

I've no experience of H bots, but all the designs, pictures and videos I've seen, made it looks like the components needed to be quite rigid because of the twisting forces.

Perhaps this only applies to the frame.

Would it be possible to have a H bot with plastic carriages and 8mm smooth rods?

yes, i've built one from M12 threaded rod and it runs on 8mm smooth rods, prints quite well but i'm having a steel frame welded up for it

---

-=( blog )=- -=( thingiverse )=- -=( 3Dindustries )=- -=( Aluhotend - mostly metal hotend)=--=( Facebook )=-

Quote

I've no experience of H bots, but all the designs, pictures and videos I've seen, made it looks like the components needed to be quite rigid because of the twisting forces.

Perhaps this only applies to the frame.

Would it be possible to have a H bot with plastic carriages and 8mm smooth rods?

CoreXY does not apply any wracking (twisting) force to the crossbar during moves, so the frame and bearing assemblies do not have to be any more rigid than any other printer. This is the crucial advantage of CoreXY over hbot.

Sorry @Crispy1,

I was under the misapprehension that CoreXY was a HBot design.

Any idea why this geometry isn't widely used ?

Or is it used and I've just not come across printers that use this. I've researched various printers both commercial and non-commercial, and aside from DeltaBots, most XY moving head printers seem to be quite traditional, i.e have a motor on the Y (or X carriage). With the exception of the UltiMaker, which has both X and Y motors fixed, and a cross arrangement of bearings at the print head assembly.

I'm currently trying to design a reprap which has a moving XY head, and uses 2 x Z axis lead screws (to overcome the need for very strong bed supports on the side that would be attached to a single Z axis lead screw and guide rods, i.e because this appears hard to build in a reprap).

I have seen one other posting (quite some time ago) where someone else was building a XY moving head with twin Z lead screws, but was using 1 motor not 2, I was intending to use 2 Z motors like conventional repraps ( can't find the link at the moment, but is in the forum somewhere.

I find it strange that most repraps are not going over to moving XY head design, as this seems much better than moving the bed in Y (or X for that matter).
I have a MendelMax 1.5, and the moving bed has always been a problem, as if I increase the print speed, I end up shaking the model's off the bed, unless they've got a large surface area on the bed.
And even if they do stay stuck to the bed, I'm sure the vibration of the movement starts to affect the quality of prints that are tall and have thin sides etc.

I was intending to build the printer from aluminium L sections, as my local DIY store sells a lot of different sizes, e.g. 25x25x3mm, 30x30x3mm, 40x40x3mm, 50x25x3mm, for around $10 per M for the largest size (40x40x3mm)

So I suppose I could build a CoreXY using 50x25x3mm, as the 50mm side of the L is large enough to mount a normal nema 14 motor (which is around 42mm square)

I did get price for laser cut 3mm aluminium for CoreXY and it came out around $120, which didn't seem that expensive. But looking at CoreXY, it looks like the example on their site is made from 5mm aluminium, and I think that it would need to be water cut, so I'll need to find a local company that does water instead of laser cutting if it needed to be that thick, as the company that quoted me on laser cutting, said their max thickness on small holes eg. for the motor mounts is 3mm (not sure why this was)

Anyway, if you can shed any light on CoreXY vs other designs, I'd be grateful

Thanks

Roger

Quote

Any idea why this geometry isn't widely used ?

Some possibilities...

Because until very recently good mechanical design was not a real limiting factor in 3D printer performance.
Because it requires a more complex and larger frame than moving bed designs - you need a full box frame vs 2 plates (Mendel90) or a plate and some threaded rods (i3).
Because (for the most part) people are very reluctant to change what they know works ("mendel style seems to work and everyone is using it so it must be the best thing out there").
Because the benefits of CoreXY are not clearly explained anywhere and most people cannot or do not work it out for themselves.
Because it requires a more complex drive system layout vs your typical open-ended belt run in a loop.
Because the advantages of CoreXY are not aligned with what most people want to accomplish when they build or design a 3D printer.

Regarding your z axis questions...

All things considered I found designing a good z axis was by far more difficult that designing the CoreXY drive. I experimented with a cantilevered bed design (like a Makerbot or Ultimaker) and found it to be unsatisfactory. It is difficult to achieve the required stiffness and resistance to oscillation as you scale the machine up. If you ever get the chance to touch a Makerbot or Ultimaker in person, you can wobble the bed all over the place if you grab the front and gently push on it. 2 leadscrews is not a perfect solution, but it's better than cantilever. I won't say exactly what I am doing for Z because I want people to work it out for themselves

I used aluminum t-slot extrusion to build my machines. It's relatively cheap, easy to work with, and very easy to modify the design on the fly. For instance, I can adjust the mounting point of stuff by just sliding the bracket along the extrusion. Getting plates laser cut doesn't give you this design flexibility, and it's probably more expensive for a 1-off machine. L-beams are going to be harder to work with, primarily because you're going to be drilling lots of holes in them which gets old after a while.

while the seperation of x and y makes the frame much simpler, you can achieve a much more efficient build area in comparison to machine size with a corexy machine, even in comparison to a delta which has all of the axis tied together its overall build volume is much less than its machine volume.

Thanks crispy1

Quote
crispy1
Getting plates laser cut doesn't give you this design flexibility, and it's probably more expensive for a 1-off machine.

Isn't the CoreXY on the website made from water cut aluminium ( [corexy.com] ) ??

Quote
crispy1
L-beams are going to be harder to work with, primarily because you're going to be drilling lots of holes in them which gets old after a while

I was going to make a jig out of mild steel for the corner holes, and also build a wooden jig for overall alignment. But I agree its not as easy as I beam. But I beam is $25 per linear metre, which is between 2 and 3 times the price of L section.
However I guess for prototypes I beam is much better.

I was hoping to just get the main assembly of CoreXY (http://www.thingiverse.com/thing:22005/#files) laser or water cut, then build the rest of the printer out of L extrusion.

Re: Z assembly

I've not had hands on Makerbot, Ultimaker / Solidoodle etc etc, but reading the various blogs there are obviously issues cantilever Z assemblies, but my main reason for adopting dual Z leadscrews was just that it looked difficult to build even a basic cantilever system. Dual lead screw design looks like its far more producible without a foundry ;-)


Thanks

Roger

@crispy1

BTW. Do you know if it would be possible to build the CoreXY frame (plate) from MDF ?
i.e like early Mendel 90's ( [hydraraptor.blogspot.com.au] )

Quote
crispy1
I've asked for those features to be ported to Marlin or some other RepRap firmware, but no one seems interested in undertaking this.

crispy1, would you be interested in opening a bounty for this issue? There's more info on bounties in this thread.