Improved Z-stage

[jonnybischof 60]

Hi community

For my "Ultimaker Black Edition" I want to make an improved z-stage which must be more rugged and stable, less prone to constant re-levelling and easier to level in general. Also, I don't want to lose too much build space (none if possible) and I want to keep at least the front of the printer open for accessibility.

So, that's quite a task because there's many requirements to be met.

Also, I'm not a trained mechanical engineer, so I appreciate any help from you guys!

I figured that the main problem with the current design is, that the whole z-stage rests only on one side (the back) without any support on the front. This works well as long as there is no increase in weight. But with pretty much any HBP, the weight IS increased so more stability would be appreciated.

So, imho it is necessary to add support for the z-stage in the front of the stage. Meaning either just two more straight rods + linear bearings, or even a whole second leadscrew.

Now of course it's easier to just put in two more rods with bearings, but are these bearings made to carry that kind of force? I don't think so - but I lack the experience to judge on that...

To give some numbers:

I want to design the z-stage for a maximum total weight of 5 kg. This gives headroom for most builds. When using a really heavy HBP like qu-bd's Basalt plate which weighs 2kg, there is not too much headroom at all.

So, if there are only two straight rods on the front, and the z-stage connects to them with linear ball bearings (same as for the back of the machine), then the z-stage's weight will put a "lever force" on the bearings instead of a proper "straight downwards force" because the front lacks a support point:

 

Now, 1st: Is that a problem for the bearings?

Most ball bearings are clearly made for only one particular direction of force. (usually the axis). Twisting, shearing, pulling is usually bad for the bearing. What about linear bearings?

And 2nd: Would that setup run without friction?

-> My guess is that this introduces friction on the linear bearings which may lead to the bearing hanging / slipping instead of running smoothly:

There are reports of "hanging / slipping" z-stages (no movement even though the leadscrew turns, later it suddenly jumps to the intended position). This must absolutely be avoided!

Comments on that part?

I think it would be more stable to not only add two rods to the front, but also a second leadscrew.

Now, this makes the build MUCH more complicated because not only do the leadscrews need to be synchronised (meaning coupled), but the front leadscrew would ideally sit right in the middle part of the front, where there is supposed to be a window to reach into the printer and take your prints out.

That would be a serious "SIGH" for the looks of the machine going down the ditch.

So I want to avoid that if possible. But if it's necessary...

I think it would be too much trouble to switch the rods and leadscrews from the "front - rear" to the "left - right" sides.

Besides, I probably want to add a crossflow-fan which blocks an entire side and also complicates things on the other side a bit (need a rounded wall to guide the air).

----------------------------------------------------------------------------------------------

I think rather than breaking my head over the last solution, it would be better to think about something entirely different.

For example: Why not put the z-stage on 4 rods (each corner, held with linear bearings) and move it up / down with timing belts:

One belt in every corner, both front belts as well as both rear belts connected by another rod in the bottom, and both rods coupled to the Z-motor.

The four timing belts would be close and parallel to the supporting rods, so the platform's weight force is straight downwards - clean without any friction.

This eliminates the leadscrew entirely by replacing it with a different system. It also means that there needs to be some kind of gearing for the motor in order to get a good positioning resolution (as well as lower the necessary torque the motor needs to deliver).

Imho this system would be cleaner and easier to achieve than adding a second leadscrew. This is mostly because getting a second leadscrew which is exactly the same as the original one might get tricky, whether as getting some belts, pulleys and straight rods is easy and required anyways.

Also, if you think about it, the actual design of the z-stage would be pretty simple. It doesn't even need to be extremely sturdy because it's supported in all four corners, which is where the weight of the bed connects anyways.

Or, how about that:

Maybe, if that can be constructed sturdily enough, there is no need for belts nor additional leadscrews, and the whole mechanism is just underneath the build plate instead of around it.

 

[drayson 75]

Currently I´m working on stiffening the Z-table by replacing the two inner arms (which are holding the stage where the little black brackets for leveling are placed into) by 6mm aluminium arms.

Nevertheless, your approach is also very interesting!

 

[burki 1]

I was thinking in a similar direction like you, Jonny. But i was considering having the current setup from the back (straight ros and threaded rod) on each of the sides (omitting the one on the back), and coupling the two units, for example with a timing belt. i guess only to have the straight rods and bearings will be counter-effective. Since the bed is still only supported on the back, the front is still free to move. This will imho lead to canting in the front bearings.

If you had a rather stiff design, where two sides (or back and front) are controllled, you should be able to get a very rigid print bed.

On the other hand, you might get decent enough results if the arms were a bit higher, so the bearings get longer, resulting in a better angle to get the forces on the brint bed into the bearings.

 

[xeno 108]

You might want to have a look at the UM2, the platform looks much more sturdy then the UM1,

I solved it by extending the length of the support on the rails, and make the canting almost non existent,

and add weight to the platform, and is around 5KG

 

[jonnybischof 60]

That's an interesting point. The UM1 z-stage has some headroom upwards for increasing the length of the bearings (meaning: adding another bearing on top of the existing one). That could be done without losing any build height.

Another thing to prevent canting (thanks for the word ) could be to provide a counterweight on the opposite side of the rods. I've seen that on other machines. But there is little space on the UM1 to do something like that. And it increases the friction on the leadscrew because the whole z-stage would be even heavier.

So...

Do you guys think it would suffice to just double up the linear bearings on the rear rods? That would be great of course, KISS

I'd also design a sturdier platform arm with 3-point levelling, but that's detail for now.

/edit:

That picture shows your custom z-stage, not the UM2, right?

Did you replace the rod + linear bearing with a rail system? It kinda looks like it...

 

[jhertzberg 19]

If you decide to add a leadscrew, then why not add two? Place the new ones in the front corners. Tie them all together with a timing belt. Add a release mechanism so that you can turn them separately, and you could do your bed leveling right at the leadscrew.

 

[gr5 2,224]

The photo of the scissor lift is interesting. Note though that if you rotate 1mm worth of movement on the screw horizontally, the vertical movement is only equal to 1mm when things are at 45 degree angles. The formula is something like:

(Z position)^2 = (arm radius)^2 - (servo position)^2

So Marlin might need this formula included in a few places. Or you could fudge it with a cura plug in that changes all the Z heights.

 

[jonnybischof 60]

The photo of the scissor lift is interesting. Note though that if you rotate 1mm worth of movement on the screw horizontally, the vertical movement is only equal to 1mm when things are at 45 degree angles. The formula is something like:

(Z position)^2 = (arm radius)^2 - (servo position)^2

So Marlin might need this formula included in a few places. Or you could fudge it with a cura plug in that changes all the Z heights.

 

What? That's not linear? Didn't give it much thought I'm afraid

But this kills the design idea imho... If the Arduino can't compute the formula exactly, then it introduces a new and unnecessary error to the movement... No go :(

 

[jonnybischof 60]

If you decide to add a leadscrew, then why not add two? Place the new ones in the front corners. Tie them all together with a timing belt. Add a release mechanism so that you can turn them separately, and you could do your bed leveling right at the leadscrew.

 

Interesting proposition...

That would make the levelling process a lot easier, and it would eliminate the straight rods + linear bearings altogether (see below why). You could even go really crazy and put in two more steppers instead of connecting the leadscrews. Then you could completely do levelling in software

Use an inductive sensor and the software can automatically level the platform as a standard procedure on startup...

Some thoughts:

-> You can't leave the straight rods and do levelling with the leadscrews, because then you would create tension between the leadscrews and the rods.

Okay, it's possible to work around that by using ball-joints on the leadscrew-mounts. But that's going to be complicated.

But maybe still worth it? Have to think about that...

But this introduces another problem: The z-stage would probably rattle in the x-y plane because the nuts connecting to the leadscrews are (afaik) not as tight as the linear bearings...

*is there smoke coming out of my head already?*

How about leaving the two straight rods in the back where they are. They will fix the platform x-y wise.

Then, connect the back leadscrew with a balljoint.

Add two leadscrews in the front corners, connected with balljoints.

Every leadscrew gets it's own stepper motor. That means I have to make a new electronics board (which will be expensive, but otherwise easy provided there's some free I/Os on the Arduino)

Add a positioning sensor to the printhead that can measure distances with 0.05mm accuracy. Shouldn't be too difficult, again provided there are more free I/Os on the Arduino.

Then there's going to be some tinkering with the firmware (that'll hurt...).

But the result would be a completely self-levelling and extremely sturdy z-stage like the world has never seen before

/edit:

!!! :-P :-P :-P

 

[jhertzberg 19]

Well that is a fantasy z-stage! As long as we are spending buckets of cash. - To get around the issue of tension between the straight rods and the ball screws, how about a mechanism that releases the linear bearings to float in their housing (like the stock z nut capture) on the x and y axis during leveling, then clamps them in place from above and below to hold the new location. A couple of solenoids would do nicely if you want to completely automate it.

 

[xeno 108]

That picture shows your custom z-stage, not the UM2, right?

Did you replace the rod + linear bearing with a rail system? It kinda looks like it...

 

I changed to the rail system because I built a taller UM+, and the rods are too flexible in that height, en the rails were the best KISS solution

I also dabbled with the idea to have more support for the platform:

 

[3dcase 37]

Hi,

nice brainstorming here

I know what I would like to change on my machine, as we all know (or will find out) the difficulty in modification is, you have to come up with something drastically different or take a design that's fits in more easy.

I bought my printer as a hobby, mostly for the Printing, no time nor money for building or buying another one to chop it.

But from a 30 year of experience of building, maintaining and designing 'real' world machine's , I have some thoughts.

And trough some of them in.....

Stay away as far as you can of a scissor lift, not only that it's not linear (as mentioned above), but the most unstable option there is. You definitely would need some kind of axial bearings or rollers on the side to keep from moving around.

I would not use more than 2 guides to have as less "schranken" (dutch, perhaps in german it makes some sense as well) as possible. it's the effect of all bearings, guides and Hydr,pneum. cilinders. (Slip/stick effect) the friction or load on one is always different than on the other one. movement in the 2 never the same distance nor time!

Moving your printhead around by hand in the centre works in the middle easier than pulling on only one slider block .

Using 4 would only double your trouble.

It would really be effective, and all Z-Problems solved, when you use 4 (perhaps 3) Z-crews. As stable as it can get, lifting and guidance in one. the only downside is you need 4 pulleys and a timing belt to synchronize them.

A less drastic solution and the most easiest one is, put the Z-motor on top. why? Because the construction in the UM1 is a big balancing act. Pulling is always a lot easier than pushing. most people walk around with their briefcase handle on top, instead of pushing the briefcase up . so pulling the z-stage will straight out all play and movement in bearings and threated block, not my idea but Mr newton's.

You need a bearing in the bottom. but it could be a printed housing with only one ca.8mm bearingball (marble).

Almost forgot one thing, looking at your UM1 now, shows you, threated block and bearings are on one horizontal line. Moving up the threated block as high as possible (30mm?) gives you a triangle, and even with the motor in the standard position, you reduce that balancing problem, because mister newton takes care of the play and movement in the bearings and threated block.

Like to read some comments . But it is only I.M.H.O....

 

[chopmeister 0]

Nice writeup 3dcase!

I agree on the scissor-lift. At one time I thought it was a good idea and then spent hours reading through everything I could find on those. Even did some basic 2d simulations in autocad to see how it would work, only to conclude it wouldn't. I really like the motor-on-top idea, though!

What say you about a belt driven system? I have a design somewhere (just reinstalled my PC and I can't seem to find anything) which uses two linear rods (placed on diagonally opposite corners) and two belts hooked on the other two corners. Both the belts are driven by the same motor via a worm gear, so that it stays level even when powered down (it would be really funny otherwise ) That whole system is a fraction of the price of good leadscrews I think. You get loads of torque, very little backlash, if any, and no wobble. At least that's how I see it, with my very limited engineering knowledge. If anybody is interested I'll whip up a new sketch.

 

[owen 19]

I bought 2 more leadscrews and nuts from UM to put in the front corners and was going to attach a stepper motor on each and wire the 3 steppers in series (= no extra drivers or belts etc.) (also = you don't want one stepper to get out of sync with the others). There isn't much room at the front to do it however.

I also made this to lessen the back lash in the nut.

The nut is held tight in the yellow piece and it uses 4mm shafts and bushings. It has no play Z and takes out all Z wobble but I think I have a bit of play elsewhere which still gives me levelling problems.

It is too big for the front corners though so I would need a more compact version for there.

 

[jonnybischof 60]

Thanks for all the input!

So you can buy leadscrews from UM? That's good to know, thanks!

Ok, scissor lift is dead then

Also, adding unsupported rods in the front is a no-go. Thanks for clarifying!

About putting the motor on top: I honestly don't get why that would make any difference Supporting the leadscrew on the opposite side of the motor is one thing to be fixed, but what good is it to reverse it?

I'd say for now there are three different, promising approaches that came up. I'll try to sort them by simplicity, simplest modification first.

1. Simply increase the length of the linear bearings, and make the z-stage's arm more rigid.

+ Very simple to design and put to reality. No changes to the firmware or the function of the machine in general.

-/+ I'm not sure yet whether it would fully solve the rigidity issues.

2. Use belts instead of leadscrews, as chopmeister (this time I get the name right!!!) suggested.

+ Even distribution of forces: minimal odd tension

-/+ Changes the z-stage behavior. But this should be manageable without problems

-/+ I dare to question the stability: Isn't belt stretching an issue?

3. Use 3 leadscrews, make a complex coupling system, and use linear rods for x-y stability. Synchronize the leadscrews (either with 3 steppers wired together, or by using timing belts)

+ Unquestionably the most stable solution z wise, while...

- ...a bit more difficult to get a rattle free x-y plane

-/+ I don't know what a UM leadscrew costs, but I guess it's somewhere around 20 Euros. Plus there is more hardware needed. So this will get rather expensive.

4. Go totally crazy and use 3 leadscrews, control them independently and use them to level the build platform in software.

+ Same stability as (3)

+ Very easy levelling process, can be driven up to the point where the user doesn't have to care about levelling at all.

- Needs a complete redesign of the electronics and major firmware modification

- EXPENSIVE - rough guess 100 Euros for the hardware, 200 Euros for the new electronics. Plus many hours of programming.

My thoughts so far:

I will give #1 a shot first, just because it's so simple that it would really be stupid not to try and compare. Maybe it's totally good enough, and that's what I want to have in the end.

I'm not sure about #2. It sounds promising but it's rather hard for me to imagine it in detail.. The total lack of sturdy leadscrews is somewhat disturbing

Also, it might be difficult to realize that project without making too many irreversible modifications to the frame - so if it fails, you have a somewhat wasted or "stained" UM...

Maybe if there was detailed 3D sketches, this would become more attractive?

# 3 and 4 are a bigger project. I think I want to wait for results of #1 before I make this more than brainstorming. Also, I shouldn't over-estimate my abilities...

But if #1 fails, I'd love to do a community project in that direction! It's probably too much for one person alone, but if only a few of you amazing guys collaborate, this should be peanuts

# 4 can be done as an (optional!) upgrade to #3, so the "EXPENSIVE" part can really be made a completely optional upgrade to an otherwise perfectly functioning #3.

 

[chopmeister 0]

2. Use belts instead of leadscrews, as chopmeister (this time I get the name right!!!) suggested.

+ Even distribution of forces: minimal odd tension

-/+ Changes the z-stage behavior. But this should be manageable without problems

-/+ I dare to question the stability: Isn't belt stretching an issue?

 

The behavior is only different in the sense you have to calculate a new value for steps/mm.

As for belt stretching, nobody ever said you have to use 6mm belts. Wouldn't be much of a problem to use 2x9mm GT2 belts, and you have ample space to add adjustable belt tensioning. Plus the weight of the platform will work for you, like with the inverted Z motor setup, but you're putting pulleys on top instead of the motor.

EDIT: I forgot to add, this is my hypothetical setup meant for my own 3d printer design, I have no clue how to shove that all in an Ultimaker, just throwing ideas out there.

 

[Daid 306]

We tried Z belts for an experiment (early UM2 stuff). Main issue was resolutions and that when you lose power the bed drops like a brick.

 

[chopmeister 0]

We tried Z belts for an experiment (early UM2 stuff). Main issue was resolutions and that when you lose power the bed drops like a brick.

 

Yeah, not nice at all. That's why my design is worm gear activated, so it always stays put. Also, with a 20:1 worm gear, and 1/16 microstepping you get half a micron per step resolution. You can then use larger stepping to get more speed and still stay within a few microns per step.

 

[drayson 75]

Hi guys,

as I´m nearly the whole time sitting in front of my UM1 thinking about possible mods (and looking to my post´s there are a lot of ideas but unfortunately less time to finalize) I recently started with exchanging the acrylic platform by a alu bed with a simple 3-point leveling similar to UM2 leveling.

As a next step I will exchange (only) the two Z-arms by solid aluminium parts, re-using the other wooden stuff between those parts and the z-screw.

I don´t want to make the whole Z-stage out of alu as I think, a bit of flexibility in x/y is needed from a mechanical point of view (bearings - on free, one fixed).

I´m currently in designing the two parts in SolidWorks, with respect to the "producability" in our companies trainee workshop. (yes, abuse of manpower :-) )

It will take a few days, but I will post the files soon, in case somebody is interested in too...

 

[poshman 0]

From my point of view the most important part is minimizing the vibrations, and as I see it they stem from:

1) The bed is not totally fixed in z-direction. Lifting it will cause a "loose" feeling.

2) The weight of the bed is to low. Generally making sturdy and exact machines requires weight and vibration absorbing materials (like cast iron e.g.)

3) The table is only fixed in the back, which leaves the front to be able to create upp and down movement.

And moreover, the bed kan also swing in x-y direction, caused by rapid head movements.

So, IMHO I think the following will yield enough sturdyness to get a sufficient accuracy of the machine.

a) Add a singel support rod in the front of the machine using a roller bearing like in the back. See to that the rod can be adjusted in the y-direction so that the design can be "tensioned" slightly. This will minimize the posibility of disjoint movement of front and back.

b) Increase the body weight of the machine itself so that the body picks upp vibrations.

c) Increase the weight of the bed to lower its resonance frequency. The lead screw setup can handle much more weight than a heavy heat bed.

d) include a second lead screw nut to take out the possible backlash.

A question: Does anyone that has modified the z-axis note a big difference in print quality??

Printing at high speed I recon that it could leave vibrationmarks in the x-direction, but besides from that I doubt that you see it - or am I wrong??

I think the

 

[jonnybischof 60]

Some news about my improved z-stage!

I have made a lot of progress since January, and the new design is almost ready. The latest update made me change lots of things concerning the z-stage (Z-motor is now on top).

I need to switch from the standard 12mm leadscrew to a thicker 14mm version (because there are no axial bearings for the 12mm screw available - long story).

Now, I'm not sure if I understand the "pitch" correctly.

If I buy a 12mm trapezoidal leadscrew from Misumi, it comes with a 2 mm pitch. The 14mm version has 3 mm pitch.

Could it be that the UM original leadscrew has a 2.5mm pitch? That would be the distance between two thread profiles. Not sure if that's the right way to determine the pitch..

Would using a leadscrew with 3mm pitch become a problem with accuracy? I know I'll have to recalibrate, but that's not too difficult.

 

[anon4321 16]

Jonny, I believe the original UM1 lead screw is 3mm pitch. If you check the UM1 marlin firmware, the steps per unit of the Z axis is 3/200/8 where 200 is the stepper steps per rotation and 8 is the microstepping setting. This leaves 3 as the pitch.

I believe pitch is distance between threads. Or in other words, a nut would move X mm on a screw with a pitch of X mm of the screw is turned one full revolution.

 

[anon4321 16]

Oh, and one more thing.... gimme,gimme,gimme,gimme.... (for those not from the US, gimme is slang for "give me" implying that I want the design really bad).

 

[jonnybischof 60]

Don't worry, I will publish the full design once it works (part of the UM black edition). Just don't forget that it will demand for high machining capabilities (such as lasercutting 3mm steel). Maybe it's possible to replace the lasercut steel parts with something easier to process, but the solution is really sweet

A little teaser:

 

 

These three parts already feel stronger than the whole UM original assembly. And they are completely missing the "strenght-giving" center pieces. They're also about the same weight as the whole original platform...

Using 8mm Trespa HPL as the base platform, the steel arms wouldn't even be necessary in the first place (I guess). This is definetly going to be strong enough

 

Sadly, I got the holes a bit too tight so I can't fit the linear bearings in there. Next version will have an ultra-tight fit

If the original leadscrew is 3mm pitch, shouldn't that be measurable on the actual part? Could it be that while it is a metric diameter part, that it is still an imperial 1/10" pitch? 2.54mm is just about what I measured...

I guess I'll just try out the 14mm leadscrew with 3mm pitch. I don't have many alternatives anyways...

 

[anon4321 16]

That looks sweet!

I noticed yesterday that the original Z stage is really wobbly not only front to back which you might expect from the cantilever design but even at the linear bearings it can be "twisted" such that the left side drops and the right side rises or vice versa.

Do you even need a tight fit if the bearings are flanged as they appear to be based on the design? I would just use a rotary tool with a sanding drum to open the large holes up or drill out the small holes. In the end, won't the screw through the flanges clamp them in place regardless of the amount of "slop" you open up in the holes?

Sorry, can't help you with the leadscrew pitch as I can only supply what they firmware uses. I'm pretty confident that at least from a firmware perspective, the pitch is 3mm from thread to thread.