Best Z-axis configuration?

lifting test
[www.youtube.com]

I can't see well enough where the glass jar is with respect to the centerline defined by the two Z smooth rods. I was thinking some "micro-binding" would be happening in the linear bearings mostly when the bed would be lifted but I guess the same situation applies when the bed is lowered. Basically the motor tries to lift or lower -having a flexible clamp on the bed via the belt- and the two vertical rods act like an axis around which the bed can rotate. The rotation moment initiated by lifting the extreme right side in the picture below creates binding in the bearings that slide on the smooth rods.

Here's what I was referring to:



Nevertheless the video shows quite a resilient setup. I don't see plastic parts weighing so much. Maybe if you do some aluminum milling with another toolhead

Edited 1 time(s). Last edit at 12/08/2015 03:28AM by realthor.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

I am contemplating building a smartrapcore alu printer and am concerned about the cantilevered bed design. I will most probably build one with this feature then evaluate it, looking to move to a twin leadscrew with belts to drive both off one motor either with a prefabricated gearbox or perhaps reduction pulleys.. The technical discussion above about leadscrews and alignment etc... is interesting but is it really as much of a problem as has been made out?

There is precedent - I have an i3 with twin leadscrews and twin motors which performs pretty well and without any of the issues related to binding or banding etc... Especially now I only use layer heights that are a multiple of my full step. I am thinking of trying to drive both screws off one motor in order to prevent the motors coming to rest one step apart from one another which gradually tilts the bed, but with bed level compensation this does not overly affect print quality so its more an exercise in doing it to see how well it works, rather than to actually solve a problem.

Does anyone have a twin leadscrew - one motor corexy printer working who can elaborate?

Hello all,

I have been one of the "defenders" of the cantilevered design. Both from an experience point of view and from thinking about the mechanics.

Thing is, from a static point of view, the bed will hang a little from its own load. That is the nature of the cantilevered setup. But it wont matter because the added weight from the plastic part will be negligible so the hang will be fixed and can be compensated in bed alignment. Even it is good with the hang because it does not over-constrain the bed setup. As the bed is not moving much there are no significant dynamic effects.

Then today I got this:



I first considered it to be an extruder issue but it is not (which can also be seen as it is not present all over). It is the dynamic effects.

The dynamic effects from the wall being furthest away from the bed support (some 280 mm) and the wall being parallel to the bed supports. This allows it to start oscillating when the hotend passes in high speed - and maybe the extruder is also over-extruding. It was fixed by reducing speed and extruder output.

It could also have been fixed by having the wall perpendicular to bed supports or having the wall in the other end of the build surface.

Still, for me this makes the AndreasL ( [www.thingiverse.com] ) design even more interesting, because that design is much like the cantilevered in simplicity and static wise but it cuts the distance from suspension to furthest print point in half. I have a lot of open projects these days but it is for sure a design I have to try sometime in the future.

Edited 1 time(s). Last edit at 01/11/2016 05:04AM by LarsK.

Couldn't a cantilevered bed be "enhanced" by supporting more than one side of the bed on a single lead screw? Pulleys and cable could be used to transfer the load from one point to another, as in this diagram:

I am currently gathering materials to build a printer using this system for z; I will report back with results.

I am the OP of this thread.

After all this great information in this thread I went and designed a compound pulley system.


As you can see in the pictures my bed is supported by four rods at four corners. I added a compound pulley systems at opposing sides and connected them both by a shaft at the bottom. This shaft has 40T pulley which is run by 20T pulley at the motor. All of this gives me a resolution of 0.05mm per full step for a usual 1.8 degree per step motor. Bed does not drop by its own weight although then I put a 12V 20A power supply on top of it the bed starts to go down slowly.
Even with a motor disconnected the bed does not fall. Rather it gently lowers.
In its architecture this system mimics the corexy solution with its long belts and pulleys. I think it fits really well with X and Y axis.

Now what I do not like about it. I was planning to run it from a A4988 stepper driver set at full step mode. It works but damn stepper is loud. Everything else on my printer is almost inaudible (TMC 2100 stepper drivers for X/Y axes, extruder fans set at low voltage and passive cooling power supply). It really ruins the high tech feeling of the printer with its archaic noise.
I have also inserted a A4988 stepper driver set at 16 microsteps. This is acceptable. Not too bad really as it sounds comparable to my prusa that I am used to. That is what I left in the printer now.
I have also tried TMC2100 stepper driver set at 128 microsteps (I think) and It was completely silent. But holding torque was not enough sometimes and sometimes it did not have enough power to push the bed up.

Up next will be the printing test to see how bad single step mode really sounds. I know its bad while homing but I wonder how it is going to sound while printing (I use z-lift).

Now I am looking for your ideas on choosing the step mode on stepper driver. Am I sacrificing accuracy by going with 16 microsteps? Anyone can suggest a good .stl model to compare prints done at single step and 16 microsteps?

[imgur.com]

Edited 5 time(s). Last edit at 02/26/2016 01:58AM by Edvardas.

Interesting solution, i kinda like it. I am though a bit circumspect about having multiple belts two of which are quite long. Is tensioning going to be a challenge? Backlash might increase with the increase in length that the compound pulley requires. What are your observations from the actual using of the printer, seeing and analyzing the quality of the already printed parts?

I'd like to see the sketch or pictures from more angles.

Regarding microstepping, the rule seems to be "the minimum with which you achieve desired resolution". If you go too far you could be loosing too much torque.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

There is just one belt on each side. I took Fabrice Z-axis as inspiration so you can see how compound pulley system was implemented on his thingiverse page: [www.thingiverse.com]



Belts are a little over a meter long each. Tensioning is easy as end of belts are connected to a plate. Then this plate increases or decreases the tension by turning screws that are connected to it. Tensioning also levels the bed.

I have not done any print tests yet.

Yes, I know Fabrice's design and like it. I have had my input on his thread and he made some cantilever weight experiments at my suggestion and his design with one belt and offset linear rods pretty much handled those remarkably well. He swears by his prints too.

I imagined your compound pulley system has to be something along those lines but noticed differences and was curious about the exact ways you fix the belts and tension them, which, as you say, are different than Fabrice's, which was only the inspiration.

Good luck with testing, we're waiting to see some results.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

Quote
NathanaelXYZ
Couldn't a cantilevered bed be "enhanced" by supporting more than one side of the bed on a single lead screw? Pulleys and cable could be used to transfer the load from one point to another, as in this diagram:
[attachment 69484 compound_z_scaled.jpg]
I am currently gathering materials to build a printer using this system for z; I will report back with results.

@NathanaelXYZ: I am designing a corexy where I have your setup in the sketch, but from corner to corner and using both spectra and belt instead of the leadscrew. I've altered a bit the design because it was quite difficult to get all the plastic parts the right way and was a bit overcomplicated so now I can't really use this design but I am still trying to come up with an elegant solution . Nevertheless your sketch should be doable.

Edited 1 time(s). Last edit at 02/26/2016 05:03PM by realthor.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

The way I fixed the belts was copied from the way XY belts are fixed on Vulcanus V1 printer.

At the bottom the belts are fixed by pulling belts through 6x2mm slots. Then they are zip-tied.

At the top they are fixed the same way but to the plate. Belts can be tensioned by screwing two screws the plate is connected with to the top mount.

And now lets look at the test prints I have made.

I have used 30x30x30mm cubes with 4mm radius corners. It seems that sharp corners are not good for testing of a Z axis performance as they make X and Y axes slow down to jerk speed. This translates to a slight over-extrusion at the corners. All cubes were printed without infill with 3 perimeters.

At the left you can see cubes printed with full step motor driver. At the right stepper driver was set at 16 microsteps. Honestly i can not see a difference. At the bottom prints are 0.4mm layers height and at the top they were printed at 0.2mm
I must say that quality of the prints is really well and much better than on my Prusa i3 with m5 screws for Z-axis. 0.2mm layers height prints are extremely smooth and 0.4mm show very tidy although clear layers.


And here is an image of prints under the worst lighting possible (from above so all irregularities cast the shadows). I can not find any repeating irregularities that could be related to Z-axis being a problem.

Allright So I am actually getting better at taking pictures of these small layers. Here is a comparison against aluminium framed prusa i3 print at the same 0.4mm layer height. As you can see belt z-axis print on the left is way more even. And my prusa actually makes nice prints at 0.2mm layer height.

I am impressed. Those are very nice prints, actually both of them.
Now I would really like to see if there is any difference between your setup and Fabrice's, as he is using only one compound pulley system and offset linear rods.

Nevertheless....congrats. Would it be asking too much for a video of 1-2min while printing and go round the printer to have a better view of how things are. You might have a winner with your printer are you thinking of releasing your files?

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

I can make a video.
My printer is actually a Vulcanus V1 (it is on thingiverse and instructables) that has 1.3kg extruder carriage (three nozzles run by a single motor to print 3 of the same parts simultaneously) and now a Z-axis that has been upgraded.
I could upload the Z-axis files. It would actually fit pretty much any corexy printer made of 2020 profiles. Only requirement is at least 32mm gap (spacers could be added for a bigger gap) between the outside of bed's 2020 profile and inside of the printer's frame.

I should say that after all the bearings and pulleys have smoothed out the bed does not stay in a position anymore. It does not drop completely but If I turn off the power at Z-0 coordinate the bed drops down a few centimeters. I am going to add some 6mm thick machined aluminium plates for the bed surface and I am sure that after this the bed is going to slowly lower to the bottom every time the power is turned off. This makes me question the point of compound pulley system. This and the fact there is no clear reason to run belt z-axis at full step as 16 microsteps work just as well. I would say system from Andreas [www.thingiverse.com] should work really well after beefing up all the parts to minimize any deflection under load. They look too thin for me. Oh and I would definitely add 4 smooth rods instead of 2.

Edited 2 time(s). Last edit at 02/27/2016 02:56PM by Edvardas.

I can't remember if you have mentioned what motors you have 0.9 or 1.8. I have a thread on reprap forums where I challenged the idea that 80steps/mm should be enough for Z, as it is able to deliver 0.1mm resolution. But the conclusion of the thread, albeit not definitive, was that the increased resolution helps with smoothing out some other mechanical issues that the whole machine could have. i am thinking of 0.9deg stepper motors and simple belt dual-Z like Andreas_L and other "forks" use.

But I guess all of them have the "fall when power goes off" problem. That is just something inherent to the design.

I have seen the Smartrapcore XL300 from @smartfriendz, it has dual-Z with belt but they use 2 motors, which seems to me a waste for an axis that moves so little and seldom (unless you rely much on the auto-leveling).

Edited 1 time(s). Last edit at 02/27/2016 03:23PM by realthor.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

I have a 1.8 degree stepper. Resolution per full step is 0.05mm.
Remember that shaft at the bottom has 40T pulley that is connected to 20T pulley on a motor via endless belt. You could also use this to double the resolution on Andreas design.

Remember that reduction ratio also means that your motor needs less force to keep the bed static while not moving. This means that if your bed weights 1kg (10 N force) and you add a 2:1 reduction ratio your motor will only need 5N of force to keep the bed from dropping. Because of this it would be wiser to add a reduction ratio via belts (instead of going with 0.9 stepper) as it is going to increase the resolution and also will slow down the bed while it is going down and should get rid of a big "bang".

About XL300. I would remove the second motor and would add a long shaft between the two pulleys at the bottom. Then would put 40T pulley at this bottom shaft and connect it to 20T pulley on a motor via endless belt. This guy here has done pretty much the same thing: [forums.reprap.org]

Edited 1 time(s). Last edit at 02/27/2016 03:38PM by Edvardas.

What I don't like about long belts and even less about many belts is that they have backlash, backlash that is compounded by stacking them up.

I'd rather have one belt per axis and as short as possible. By compounding pulleys you kind of almost double the belt's length. That is why I'm thinking of a 0.9deg motor and a simple pulley system with the risk of having the bed fall. I really don't see that as a real problem, there are so many ways around it. If the current is going to stop then for sure the print is lost anyways, so having the bed fall down on its own is really a service a printer can do for you as you now have access to the failed part to remove and start over.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

I can see that you have not built your corexy yet. After you put those extremely long belts on XY axis you become quite positive about them having zero backlash. My Z-axis belts are shorter than XY ones.
And I was printing two of these cubes at the time (in order to let the plastic cool down on the other one). I was using a z-lift of 0.2mm; this means that the bed was going up and down multiple of times during the print. There can not be a better way to test for backlash not being a problem.

I would not like my bed to bang once power is turned off. I would definitely add a reduction ratio of 2 so bed goes down slowly after the power is cut off. On the other hand I have not experienced not having a reduction ratio so can not speak how bad it is.

Quote
Edvardas
After you put those extremely long belts on XY axis you become quite positive about them having zero backlash.

Yes, I'm quite puzzled about that. And kind of happy that many report it to be that way.

Quote
Edvardas
My Z-axis belts are shorter than XY ones.

Yes but if one would do a single-belt-dual-Z-axis (ala Andreas_L), it would be quite similar to the XY belt. However, the previous statement renders this moot.

Quote
Edvardas
And I was printing two of these cubes at the time (in order to let the plastic cool down on the other one). I was using a z-lift of 0.2mm; this means that the bed was going up and down multiple of times during the print. There can not be a better way to test for backlash not being a problem.

True. What about retraction? Can't that be used instead of Z-lift?

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

I have to ask this: as much as I like AndreasL's solution for dual-Z-axis configuration, I am tempted to consider fabcore's single compound pulley system and add one linear rod at Z motor side -I already have it so why not. But I can't imagine that being easy on the linear bearings as the bed is lifted from one side only, just like in a cantilever configuration, but the two middle linear rods' bearings have to follow in a forced position. (@Fabrice974) Fabrice was kind enough to record a lifting test which proved the motors have plenty power to take the load but still my physics part of the brain is in doubt (not about the system being able to lift but the damage of the forced bearings and if the platform goes up and down in a really parallel fashion and all that.

I hope my 0.9deg motors will show up soon so I might not need the extra resolution but then again I guess better resolution at less microstepping is better.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

I think it depends on the hardware you have. If you employ fully supported linear rods (or linear rails) and high quality linear bearings with a stiff bed platform then it should do fine. Digital_dentist has pointed out that while milling machines have been made in this configuration those also have heavy cast metal frames.

I use cheap smooth rods that are slightly undersized at 7.94 (mm) and cheap Chinese lm8uu bearings (2 bearings on each smooth rod). At one moment I had all 4 smooth rods in place and only one side of the compound pulley system in place. I tried to lift the bed using belt on just one side. The non-belted side started lifting only after the belt side has moved a few millimeters up. This tells me that with a cheap hardware anything close to being cantilevered would be a disaster.

You could also experiment if you choose to. Compound pulley system requires very small changes (basically just mirroring the existing parts) to add it to a second side.

I have the following setup that I am going to pursue (3 x exactly 10mm rods) and a lightweight bed with 3 point leveling (via screws). I think this is the minimal setup one can get (3 points define a plane so 3 rods, 3-point leveling). In the sketch below you see 2 bearings for the 2-rods side and only one bearing for the motor-side (this is a constraint by design, I will decrease my Z if I add a second bearing on that rod). I have sketched my transmission with both spectra and belt but it can be belt only.

I use a solution similar to AndreasL's but I don't like the motor in between pulleys, mid-road so I made it to be the driving source for on of the two end pulleys.



Now my real problem(s):
- I am still exploring possibilities for the aluminum T-shape construction (Brazing maybe, but haven't found yet those rods somewhere close and cheap - they seem pretty expensive to me). Please note that the two tubes are on different planes. It would be quite easy with 15x15 or with 20x20 T-slot and some small angle brackets if I'd find it anywhere without having to order it from Germany or other consumer-friendly country .

- I am also looking for ways to connect the linear bearings to the aluminum tubes in the most rigid way possible (will end up using 3d printed parts I am sure but I still hope for a better idea/solution).

These two connection issues are actually my biggest dilemmas right now regarding the Z-axis so if you have any ideas/links/etc I am listening .

Edited 3 time(s). Last edit at 03/08/2016 07:35AM by realthor.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

I like your 3 rods design. I have gone to 3 screw bed leveling and love it. It really makes sense. I am looking forward to hearing your results.

Are you sure you can not find any 2020 profiles in Brazil? Like not a single place with them for sale? It sounds like you have not even looked

You can actually buy linear bearings in aluminum housings. They are way more expensive (around 7 EUR each) compared to chinese bearings but cost the same as high quality lm8uu without a housing: [www.dold-mechatronik.de]

Oh and I think that you do not really need a very stiff joint between your bed structure and bearings. As your design is not cantilevered there are no forces acting on these parts.

I wish I was in Brazil lol... yesterday it snowed in Finland. I am also an expat here, I don't know the language but even so I asked all around and there doesn't seem to be a place to sell them. The cheapest would be to import from Germany but shipment is the same price with the package .

It's not that I can't afford financially, it's mentally and philosophically that I don't allow myself to go to great lengths to make something unless I find the best way with local stuff. And here's my "mind-freeze" situation because in Finland even ALU square tubing is >7Eur/m, steel tubing is >5Eur/m ...it's perversely expensive.

My design is based on tubing (ALU or Steel) but I could replace anytime the same design for alu 2020 with little modification. And I am quite frugal too, so I'd rather learn the heck out of mechanical design and engineering and force myself to innovate rather than easily spending the money and remain comfortable and complacent.

There are some aspects that make me like more tubing than extrusion for example it seems to me easier to drill tubing then extrusions with a hand drill. In my design, if you notice, I have at the base of the rods some sort of set screw centering collars because the upper clamp is the tube itself, drilled for the rod. But I can be mistaken here, as I never drilled/tapped alu extrusion. One can also insert a hex-coupler nut in the tube (face-to-face) and get a very sturdy/solid connection with a bolt that actually is solid, from end-to-end, just like that material would be a tapped block. One can't hate the ubiquity of the alu tubing too. But I see extrusions being better strength-wise and more functional too, without always drilling and cutting. But they will always be more expensive than tubing as the amount of alu that goes into each meter is greater than what goes into a square tube.

Edited 1 time(s). Last edit at 03/08/2016 08:44AM by realthor.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

I would not mount smooth rods into pre-drilled holes. You will never get them lined up that way.

2020 is very easy to drill. I use a hand drill myself. Holes are there only to put a hex key though so they do not need any accuracy.

Quote
Edvardas
I would not mount smooth rods into pre-drilled holes. You will never get them lined up that way.

2020 is very easy to drill. I use a hand drill myself. Holes are there only to put a hex key though so they do not need any accuracy.

Are you referring to the holes in the same square profile? Like top and bottom holes? I have thought about that and although there are measures to permit that kind of accuracy, there are also less preferable solutions like only rely on one hole while the other -through which the rods protrudes, is larger. For the rod's alignment I designed a fixed top end (in a 10mm hole in the tube, 2mm meat) and a free-floating lower end that is put in place with a sort of centering 3-screw collar thing. This ensures perfect verticality for one rod. The other rods can be aligned using this first one, with the help of the bed, which must slide effortlessly up and down.

I would adjust the screws to perfection then pour some epoxy in the collar to set the position in stone. After the epoxy cures, I take the screws out completely as they are no longer needed. The insert nuts are sacrificial in this case, they will remain there.

Edited 1 time(s). Last edit at 03/08/2016 10:01AM by realthor.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

Forgive me if this is a dumb idea, but would a lead screw driven scissor lift make a good Z-axis? Assuming four linear guides for stability, and the scissor under the center of the platform. It would be more weight, and possibly more work in basic construction than a belt system, but it would center the drive force under the center of the plate instead of at the edge. Would the scissor introduce imprecision relative to a vertical lead screw or belt?

Quote
ThistleDown
Forgive me if this is a dumb idea, but would a lead screw driven scissor lift make a good Z-axis? Assuming four linear guides for stability, and the scissor under the center of the platform. It would be more weight, and possibly more work in basic construction than a belt system, but it would center the drive force under the center of the plate instead of at the edge. Would the scissor introduce imprecision relative to a vertical lead screw or belt?

I guess your biggest issue would be that the scissor lifts Z motion doesn't have a linear relationship to the rotation of the motor..