Fixed/Mobile choice in X,Y,Z axis and in Extruder configuration?

Hi, there,

I've been researching different 3d printer configurations lately and I saw pretty different approaches in how to choose what's fixed and what's moving in a 3d printer. Basically, 2 rules apply generally:
- the lighter the moving parts, the speedier the print can be
- the lighter the moving parts, the less inertia, resulting in the more precise printing.

I'll try to summarize these approaches, with pros and cons that I can detect in them, and then come with the question / proposal after this. If you spot additional pros-cons, please let me know.

1. Some designs (most repraps) keep the extruder fixed in XY and mobile in Z, while keeping the bed carriage mobile in XY and fixed in Z (this seems to be the most used build design choice)
Pros:
- the most moving parts are kept in the lowest position in the build, where they could induce the least amount of vibrations
- only one of the two XY motors is mobile (adding to the weight of the moving carriage)
- the extruder raises against gravity - the gravity acting as an anti-backlash force, so zero Z backlash
Cons:
- the printer's footprint needs to be at least 5-7 times bigger than the actual printable area
- the XY carriage is either too heavy or too flimsy - difficult to achieve a sturdy & lightweight combo
- any higher 3d objects are subject to serious jerking in bot X and Y directions, making almost impossible to print more "ethereal" shapes

2. Other designs try to overcome some of the above cons by re-assigning the X motion from the bed carriage to the extruder - a classic example could be the LulzBot TAZ. These move the bed carriage only in Y, and move the extruder in X and Z.
Pros:
- the carriage has no motor attached to it, resulting in a much lighter carriage. in fact, this design allows for all 3 motors to be fixed, which is a big plus, as their weight is eliminated from any moving components.
- the printer's footprint can be reduced to almost half the footprint of the previous approach (since only one axis needs to be at least 2.5-times longer than the actual printable length)
- the jolting of the model only happens on the Y axis resulting in a more stable printing model
- again, zero Z backlash
Cons:
- the extruder is no longer fixed while printing - in fact, is actually moving a lot, along the X axis. since this action happens at the top of the printer, where the frame is the weakest, this leaves room for an awful lot of potentially amplified vibrations in the X axis
- since the extruder has a motor attached to it, the X motion becomes the motion of a heavy object - see the cons of that above

3. Other designs - very few (the Chinese "MOSTFUN Sail" is an example) - choose to keep the base fixed in XY and mobile in Z. Usually, to simplify construction, they make the XY action of the extruder fixed at the top of the printer, and slowly lower the bed in the Z axis
Pros:
- the printer's footprint is the most compact - just a bit larger than the actual printing area
- since the Z moved downwards as opposed to upwards, this reduces the need for extra-strong / dual Z motors
- the part being printed stays still 99.9% of the time, which translates in the possibility to print "ethereal" (thin and tall) shapes without a hitch
Cons:
- since the bed is moving downwards, with gravity, it introduces the need for extra special anti-backlash measures in the Z axis
- the bed is usually only fixed on one side, making it potentially flimsy (like a flexible trampoline board only fixed at one end)
- the most active/moving part of the printer is moved "upstairs", where the printer's chassy is the most unstable, or where the same jolting of the XY motions produce much more inertial instability
- this is the heaviest config of a moving part - where you have both an XY carriage moving, and 2 motors in motion - so either speed, accuracy or both are very much impaired. this makes this design the most beginner-home friendly (poorer quality, less taken space in the home)

I've yet to see any incarnation of the worst case scenario ever: a completely fixed bed and a fully mobile extruder. No wonder why The only advantage remaining in that config would be the small printer footprint. Not enough.

4. The Polar-3d Delta type printers use a completely fixed bed and 3 pillars with 3 asynchronously, vertically moving carriages, achieving together both X, Y and Z positioning. In order to lighten the moving part, the extruder set is split into a mobile hot-end and a distant, fixed, extruding assembly, connected via an ideally non-extending tube (later edit: a Bowden config).
Pros:
- a very small horizontal (XY) printer footprint
- the part being printed stays still 100% of the time
- all 3 positioning motors are fixed and can be placed in the lower part of the frame
- a very lightweight moving part, resulting in potentially high speed and/or accuracy
- same backlash in all X Y an Z directions
Cons:
- a very high printer vertical footprint - several times higher than any achievable print's height
- much more moving parts are involved, with multiple universal joints, making for much more difficult to achieve precise positioning - or at least to a much more tedious/minute setup
- a lower resolution in Z, compared to other printers (thicker minimum layer heights)

Now, the questions:

1. Am I seeing correctly the pros-cons of these designs? I mean, this is where you'd point out any pros/cons I've missed out or miss-hit

2. What, in your opinion, is the best approach for best print accuracy? And what is for best speed?

3. Why haven't I seen, yet, a printer of the type 3 above (bed moving in Z downwards, extruder moving in X and Y, fixed to the top) but with the kind of split extruder you see on the Polar-3d type? This combo, to me, should bring the most pros ant the least cons.

4. Why is everybody only using direct timing belts on these babies - why not ball screws or at least a dual pulley 1:5 reduction combo - for slower but more accurate prints?

Thanks for your infinite patience on this one...
Eros

Edited 1 time(s). Last edit at 12/12/2014 03:40AM by erosnicolau.

Quote
erosnicolau
1. Am I seeing correctly the pros-cons of these designs? I mean, this is where you'd point out any pros/cons I've missed out or miss-hit

I think you mostly got them when speaking in general terms. There may be some variances between exact implementation. A few corrections would be to not say zero backlash for gravity-held z-axis, but rather minimal backlash. Printer 4, what you call a polar 3d printer, isn't a polar 3d printer. That's typically called a delta printer. This is a polar printer. There are also other non-traditional designs. Look up Nicholas Seward work at Concept Forge for some unique takes on various designs thinking sorta outside the box. Also all your designs primarily focus on FDM 3D printing. There are other types of additive 3d printing than just FDM, such as SLA, SLS, MJM, etc ...

Quote

2. What, in your opinion, is the best approach for best print accuracy? And what is for best speed?

Yes. Look at commercial printers by Stratsys et al. Those are the printers where quality and speed are most important and price isn't. What do they use? There's various implementations, but I don't think any are delta or polar style printers. A lot of quality and print speed can depend on quality of components. Many designs trade saving money for sacrificing speed or quality. For instance, generic commercial threaded rod and nuts instead of acme thread and ball screws. Cheaper NEMA 17 steppers instead of larger, faster, more powerful NEMA 23 (or other). I don't think anyone can say absolutely that one design is "the best", they just have their preferences.

Quote

3. Why haven't I seen, yet, a printer of the type 3 above (bed moving in Z downwards, extruder moving in X and Y, fixed to the top) but with the kind of split extruder you see on the Polar-3d type? This combo, to me, should bring the most pros ant the least cons.

I'm not sure what you're referring to as the split extruder. If you mean a setup where the filament is fed through a tube by a remote extruder motor and only the hot end is moving, that's referred to as a bowden tube configuration. It's not present as it's actually complicating the process and you're not gaining anything. Your total machine height is the same as you have the height each of the 3 supports of the delta configuration would need to be to reach each limit of the print bed, but you'd also have whatever your maximum height is to lower the bed as you print. You could just extend each support of the the same height and eliminate the complexities of a moving bed, additional stepper, threaded/acme rod, etc. In most cases, expanding the height of the delta is just an incremental cost with slightly more extrusion/slides, longer belts, and a bit more wire for any top mounted electronics (if any).

Quote

4. Why is everybody only using direct timing belts on these babies - why not ball screws or at least a dual pulley 1:5 reduction combo - for slower but more accurate prints?

Cost. Can you find me a meter long ball screw for less than $1.80? I use lead screws for my x-axis as they are quieter than my cheap threaded rods, but I got them free from Misumi as part of a promo so cost wasn't a factor.

And don't say everyone is using timing belts and not something different. If someone can think of something "new", it's likely that someone is, or has, already thought of that too and implemented it. Also, anything that slows movement lengthens printing time. Some prints already take a very long time, reducing it further to increase accuracy is just a trade off many don't need or want to make.

3. Why haven't I seen, yet, a printer of the type 3 above (bed moving in Z downwards, extruder moving in X and Y, fixed to the top) but with the kind of split extruder you see on the Polar-3d type? This combo, to me, should bring the most pros ant the least cons.
------------------------------------------------------------------------------

Take a look at CoreXY section. A number of people have been building them.

I have designed and currently building one using Openbeam extrusions. All 4 motors are fixed. Just finished testing all the axis motors, just waiting on a hotend to arrive before I can fully print with it.
I have build 2 other type designs, one being a Mendel style and I don't like that the bed (z axis) moves either in a X or Y direction. Once I was printing a large object which required a fast print speed, however as the object weight increased it lost its print position and I had to start again printing at a lower speed.

Only issue I see with this design is the bed needs to be held very well to keep it parallel to the printhead (no spring movement). There will be an increase of weight subject to the size of the object being printed.
So you need a good strong frame to support everything.

Also as the printed object has to be removed from the bed there is always a possibility that the bed can be damaged and become out of alignment for the gap between the print head and bed surface.
Bed needs to be moved to the lowest position. I have 4 support blocks which allows the bed to be better supported when objects are removed.

----------------------------------------------------------------------
Why is everybody only using direct timing belts on these babies - why not ball screws or at least a dual pulley 1:5 reduction combo - for slower but more accurate prints?
Reduce mass and less vibration from drive motors.


Good to see you really thinking about the different designs> This has lead people to come up with better designs for 3D printers over the years..

Quote
erosnicolau
Hi, there,

I've been researching different 3d printer configurations lately and I saw pretty different approaches in how to choose what's fixed and what's moving in a 3d printer. Basically, 2 rules apply generally:
- the lighter the moving parts, the speedier the print can be
- the lighter the moving parts, the less inertia, resulting in the more precise printing.

...
Thanks for your infinite patience on this one...
Eros

You missed one- the speedier the print, the lower the quality. There are some exceptions when comparing different machines, but on any given machine, the faster it prints, the more the print sucks. The trick is to find the architecture that provides the level of quality you want with speed (slowness) you can tolerate.

I'm not in any particular rush to have prints finish quickly. I prefer a high quality print and I'm willing to wait as long as necessary for it to finish printing. I frequently start prints before I go to bed and wake up to find them finished. I don't care if it took 6 hours or 4 hours.

Hi again, guys, and thank you for th quick replies!

First, exuse my blundering terminology - I called the Delta's "Polar 3d's" when in fact the polar ones are a completely different concept, which I didn't include for the only reason they're not as mainstream yet (but rather experimental and with extremely limited software support). Also, yes, the "split extruder" I was talking about is the Bowden tube one.

@cdru (thanks for the corrections, mate!):
- Where is backlash possible in that config, given that the movement is always in one direction (up) and always preloaded (by gravity)?
- Your argument about the importance of speed takes my mind to two things:
- the dual nozzle approach (a .3mm for details and a 0.8mm for filling) of RoboX 3D
- a (continous?) variable speed gearing. the gear change could be made either with a solenoid (for a binary kind of gearing) or with an extra stepper motor (for a CVT like a dual cone, belt-driven one). food for thought...
- You say the Bowden tube config complicates things without achieving anything. By this you mean it yields poorer quality prints? To me, reducing the mass of the moving oozer is an achievement in itself. Please detail your statement a bit, so I can better figure what you're thinking on this one

@RepRot (thanks for the encouragement!):
- The CoreXY is brilliant (because it holds both motors fixed, reducing the weight of the carriage by alot)! I was wondering when I'd see the "architect's rule" applied to 3d printers as well... (see here)
- The final position support blocks - another great idea!

@the_digital_dentist:
- You know what my dentist friend asked me first, when he heard I'm fiddling with a 3d printer? "do you also have some food-safe filament you could use for printing up mockup dentures?"
- Yes, speed and quality don't mix well. Unless you find ways to make a "2 in 1" machine (dual extruders, gearboxes...)

My quest for now remains for the lowest inertia moving configuration, and so far, from your responses and my suggestions, I get as keepers:
- good quality parts
- stronger motors
- CoreXY
- (Bowden tube extruder?)
- dual nozzle (thin and thick) extruder
- maybe a dual speed gearing

Eros

Edited 1 time(s). Last edit at 12/12/2014 04:27AM by erosnicolau.

Quote
erosnicolau
- Where is backlash possible in that config, given that the movement is always in one direction (up) and always preloaded (by gravity)?

You're making assumptions that movement is always in the up direction. At some point, you'll have to go down to start a print. So initially, there could be a bit of backlash. A growing trend is auto leveling with a z-probe. With such an implementation, the z-axis is making very small adjustments as each layer is printed, introducing the possibility of a bit of backlash. Also, gravity pulls things down, the extruder squirting out plastic wants to push it up. Now all the forces above are pretty minimal so backlash is pretty minimal, but you can't say it's eliminated. Some people also use their printer for more than just FDM. Light milling, cutting, or drawing may have different amounts of forces involved.

Quote

- Your argument about the importance of speed takes my mind to two things:
- the dual nozzle approach (a .3mm for details and a 0.8mm for filling) of RoboX 3D
- a (continous?) variable speed gearing. the gear change could be made either with a solenoid (for a binary kind of gearing) or with an extra stepper motor (for a CVT like a dual cone, belt-driven one). food for thought...

There are many ways that you can speed up a print, but they can dramatically increase cost or complexity. Why stop at 2 nozzle...get a Kracken and have 4. Or two independent dual nozzles. There are more and more designs that are coming out that implement multiple hot ends, mixing of filament, multi-color, multi-material, etc. I wouldn't say that area is in it's infancy, but it's also not fully mature and prime time yet, at least at a Reprap-affordable-for-mortals price point. But it's getting there.

Anything with variable speed is going to require more complicated mechanics. Relying on step count for positional accuracy is going to be less reliable due to lash in the mechanics. You'd probably have to use something like a rotary encoder on the final output to ensure absolute accuracy, making things more expensive. You can already do this, but it's cheaper and easier to go with steppers if you are just using fixed gearing and a single motor.

Quote

- You say the Bowden tube config complicates things without achieving anything. By this you mean it yields poorer quality prints? To me, reducing the mass of the moving oozer is an achievement in itself. Please detail your statement a bit, so I can better figure what you're thinking on this one

I kinda jumped thoughts and I wasn't very clear leading to the conclusion. I was describing what a bowden tube was, but then was saying that the type of setup you described as " (bed moving in Z downwards, extruder moving in X and Y, fixed to the top) but with the kind of split extruder you see on the Polar-3d type" wouldn't gain you anything. If I'm understanding you correctly, you're asking why you don't see a delta-style printer that has a z-bed that moves downward as the build progresses, right? If that is, then what I say still stands. It makes the process more complicated unnecessarily with more hardware requirements. You may lower the weight of the hot end assembly, but there are other ways to do that without complicating the design unnecessarily.

You can implement a bowden tube, removing a motor that has to move with the extruder, and eliminating the mass that's moving back and forth with almost any printer design. But bowden tubes aren't perfect. The filament needs to fit perfectly for optimal results. You want it to easily slide through the tube with minimal friction, but you can't have any slop that allows the filament to "snake" inside it. If it's allowed to snake, then it causes it's own "backlash" on the force pushing the filament out and for retractions. Slicing programs can account for this to some degree, but it can be a source of inaccuracy, particularly if you change materials often. And some materials, like NinjaFlex, can make it worse due to it's flexible nature.

Quote

- Yes, speed and quality don't mix well. Unless you find ways to make a "2 in 1" machine (dual extruders, gearboxes...)

Speed, quality, price. Chose two.

Quote

My quest for now remains for the lowest inertia moving configuration, and so far, from your responses and my suggestions, I get as keepers:
- good quality parts
- stronger motors
- CoreXY
- (Bowden tube extruder?)
- dual nozzle (thin and thick) extruder
- maybe a dual speed gearing

Have you built a printer of your own yet? If not, I'd suggest building a simple one to get your feet wet. It doesn't have to be a primitive one or low resolution, but dual nozzles and dual speeds adds a lot of sources of frustration.

@cdru:

- I was even wondering how in the world do all these "auto-leveling" printers take all that Z-axis fidgeting into account, lash-wise...

- About multiple hot-ends: haven't tried that yet, so having zero experience there I'm (still) wondering WHICH slicers support multi - extruder configs. It's very tempting for me too to try and build such a nice "toy" . I've seen the "0.3mm + 0.8mm" approach for speeding things up (but I'm sure the slicer needs to be smart enough for this kind of approach), and I've even seen a large scale 3d printer using a kind of vibrating-shaking head, to flatten the oozing stuff out and to get a wider deposit area. Crazy and beautiful ideas...

- About backlash: this year of my life has been dedicated to designing and building very compact and ultra-precise rotary speed reduction based on the principle of the HarmonicDrive but with less expensive manufacturing costs (about a quarter) while maintaining the same positional accuracy (~0.75 arc-minute) - so yeah, lash has been my life this year... In the case of a 3d-printer such precision is certainly over-kill, as there are many other factors involved in messing up the precision of the print (the first being the inconsistency of the filament and the fluctuations in the hot-end temperature). If you're curious about it, this is how my reduction looks in my hand, this is how it moves at 1500RPM input and this is how it fares in terms of precision (measured with a laser pointer at a 9m distance). So yep, I've been up to my elbows in machining this year and, despite having zero 3d printer building experience, I've fiddled with printer kits and I do have some machining experience.

- About the inaccuracies of a remote motor approach: I see two possible solutions for this design problem:
- the Bowden tube (which you so nicely explained to me in terms of pros and cons) and
- something I've yet to see on a 3d-printer: a throttle cable linking the remote motor to a gears-only hot-end. A throttle cable has high torsional rigidity - at least in one direction. This means that, mounted correctly, it is capable of driving very precisely the gearset of the extruder when extruding. For suckinng, you measure and take into account the lash involved by the slight torsional elasticity in the opposite rotational direction. Yes, you still have elasticity variables, like in the case of the Bowden tube approach, but at least all the variables are in a single component (the throttle cable) which has been designed for running at high speeds and a long life, not in the varying properties of each filament roll. I'm gonna get me some spare cables to test how they fare.

About having built a 3d printer - again, none so far, but I've been building other frame-based stuff, so I hope this experience comes in handy.

Quote
erosnicolau
- I was even wondering how in the world do all these "auto-leveling" printers take all that Z-axis fidgeting into account, lash-wise...

Honestly, when considering everything else with the printer the z-axis backlash is probably towards the end of the list of things to worry about. If it was CNC milling or a lathe and you had some serious forces at play then it would be more of a concern. But I think most just don't think about it or care and rather have the convenience of the auto leveling.

Quote

- something I've yet to see on a 3d-printer: a throttle cable linking the remote motor to a gears-only hot-end. A throttle cable has high torsional rigidity - at least in one direction. This means that, mounted correctly, it is capable of driving very precisely the gearset of the extruder when extruding. For suckinng, you measure and take into account the lash involved by the slight torsional elasticity in the opposite rotational direction. Yes, you still have elasticity variables, like in the case of the Bowden tube approach, but at least all the variables are in a single component (the throttle cable) which has been designed for running at high speeds and a long life, not in the varying properties of each filament roll. I'm gonna get me some spare cables to test how they fare.

I believe what you are describing is this.

LOL, and darn, nothing's ever new on this world! ) And there I was, proud of my idea...
Thanks for the link, I'm looking into that! What's your opinion on that approach, anyway?
Eros

There's a design that moves a single guide rail in X and a single guide rail in Y and uses a bowden extruder, and moves the Z axis downward during printing. A guy on Google+ named Shauki Bagdadi (https://plus.google.com/u/0/+ShaukiBagdadi) has built such a printer using cables to drive the X and Y axes and he runs the thing at 250 mm/sec. I suggest you get on Google+ 3D printing group because he and a lot of other guys are posting new ideas and designs every day.

With regard to ball screws...

I just installed a ball screw drive (from a scrapped pick and place machine, free as in beer!) in the Y axis of my RepRap type machine. Along with it I added linear guides (used, via ebay) instead of round guide rails. I am using NEMA-23 motors for all axes because when I built the machine they were available for free and because when I started on the machine it had a larger print envelope than anything else available (about 1 cuft) and I figured the increased mass would need a little more umph to keep things moving. The ball screw drive is normally quite expensive, especially if you use quality parts. Belts are a LOT cheaper. Belts are also quieter- with the ball screw drive my printer has gone from being annoying at some speeds (motor resonance, vibration, etc.) to being a real screamer because there's no flexible belt to buffer the vibrations from the motor. It's all metal on metal contact. The bed is a piece of 1/4" aluminum plate stood up off the undercarriage by 3 screws at its edges- i.e. it is a bell, and a harsh sounding one at that.

Mechanical losses (stiction/friction) are higher with the combo of linear guides and screw drive, as is moving mass, so I had to get a bigger motor (425 oz-in torque) to drive the Y-axis. That meant I needed more current so I had to build a higher voltage power supply (32V) and use an external driver. I got a DSP based stepper driver (2M542) and it has helped eliminate the most offensive resonances, but the machine is still noisier than it was with belt drive. With all the changes, the Y-axis is now a perfect finger guillotine. I'm going to have to install some safety shields to prevent casual on-lookers from putting hands where they might get severely injured.

I also upgraded the controller to a smoothieboard because it is capable of driving the steppers with 100k steps per sec (as well as being a lot more convenient to work with for many reasons) which allowed me to set the external driver for 12800 ustep/rev. At 12800 steps per rev the Y-axis speed limit is about 78mm/sec, so I may dial it back to 6400 steps per rev to let the machine get back up to 100+ mm/sec. I prefer quality prints to fast so 78mm/sec is probably fast enough..

I'm still testing and tuning the smoothieboard settings, but I have found that the X-axis (still belt drive with NEMA-23 motor) is now the limiting factor in speed. It is having trouble keeping up with the screw drive and test prints keep shifting in the X-axis. With all I've learned about steppers and driving them through this process it makes sense that it is behaving this way- the X axis stepper is a 7.2V part and I'm running it from a 12V supply. It really needs higher drive voltage to get it to perform as well as it should. I'll probably replace that motor with the old one from the Y-axis (higher torque, lower voltage). Another DSP stepper driver is on its way and I have a power supply waiting for it so that problem will go away shortly.

One thing I am seeing in the Y-axis is improved print quality. Specifically, when I had a belt driving the Y-axis, every time it started and stopped the belt/motor would act like a spring and the print bed would wobble a little, producing a wavelike pattern in the print at corners in holes and walls that were parallel to the Y axis. The ball screw/DSP driver have completely eliminated that. The same problem exists, but to a much lesser extent due to the lower moving mass, in the X axis.

Edited 2 time(s). Last edit at 12/21/2014 08:27AM by the_digital_dentist.

See the Aluminatus A One by Trinity labs (no longer produced)

[www.3ders.org]

10 start 25mm pitch leadscrews are used on the X and Y axes.

I wish someone would pick up this design and produce printers.

PBC Linear SIMO™ linear actuators may be hard to get in small quantity.