Designing a reliable and accurate Z axis.

Alright, not 20 microns then 40 or 50!
The smallest nozzle I have is 0.25mm.

So what about the Z axis? Which leadscrew should I buy?

On reprap.org I read about somebody printing at 5 micron layers. On my printer, the best apparent quality is at 70 microns. 50 is still acceptable, 20 is a mess.

I find bridging very tricky at thin layer height.

Edited 1 time(s). Last edit at 06/04/2015 12:51PM by cristian.

Two linear bearings on each shaft (one above , one below)

The delran nut us interesting --- this nut can be "loaded" or adjusted with bolt, to minimize backlash
I WOULD use an anti-Backlash nut!

Use a motor with seperate leadscrew so you can change coupler and leadsrew
and replace the motor when you fry it.
You will be frustrated with couplers and how they mount to motor / leadscrew

I would use a "leadscrew" NOT just a threaded rod --- the acme thread much better than screw thread

I don't know of a mechanism that is going to get you to 20 micron positioning

There are 50TPI screws

[www.youtube.com]

I think the glass on my bed is NOT flat to 20 micron???

?? Why such a tight / precise layer size ??

Do you have a powerful inspection microscope to verify size?

Quote
FalloutBe
Thanks again cozmicray, for all your help (;

I'm going for the cantilevered platform, 4 bearings, 2 smooth rods and a single lead screw.

What do you think will be best?
[www.ebay.com] (delrin nut)
or
[www.ebay.com] (motor attached, regular nut)
or
[www.ebay.com] (motor attached, regular nut, coupler)
?

The delrin nut states to have much better resolution, and it is cheaper.
The regular lead screws all seem to have an 8mm pitch, which I think is too large.

Anti backlash nuts or regular nuts?
The sets on ebay don't have anti backlash nuts :/ Do you think it is necessary?
Gravity should keep it down, but my platform is moving downwards along the print, so it might stick.

EDIT: I want at least 20micron layers!

Quote
cozmicray
I think the glass on my bed is NOT flat to 20 micron???

Of course the first layer should still be thick enough to stick to an imperfect bed, but from the second layer on one may use arbitrary thin layers.

Couplers can indeed be tricky to mount. A lot of times, the screw ends up rotating off-center.
That's why I wanted to buy motor and lead screw in one piece. But hmm.. I'll use a coupler (: In the end, I will be glad to have it in two pieces.

why such a tight layer size?
When the layer size is for example 20(+-1) microns, that means the layer thickness will vary by 5%
I can indeed not see with the naked eye if a layer has the correct thickness, but the smoothness of the surface, with 5% inaccuracy is bad!
It really looks ugly then, and it won't be worth getting to such thin layers.

If I can however create a perfect Z axis, let's say 20(+-0.1) microns, you almost wouldn't notice that the part was 3d printed!
I once printed a part at 50 microns, on my old prusa I3 and I was so amazed about how pretty it looked.
(see attached image, the infill has a layer thickness of 0.4mm, AAA battery for size reference)
That was just threaded M5 rods btw, so what I'm asking should be perfectly possible

Oh, and an M5 rod has 0.8mm pitch. the ACME lead screw has 8mm pitch. Are you sure the ACME screw is better because it has a finer finish?
I tried to look for Metric lead screws that have a smaller pitch but most on eBay are 8mm :/

EDIT: Will a screw with imperial sizes be bad for me, if it does not match Z steps/distance? This is the only reason why I don't want the imperial rods.
That derlin screw is an M8 rod btw, it is not a lead screw.


EDIT 2: You know what, screw all this! I'm buying a ballscrew
It is the smoothest way to eliminate backlash and it actually doesn't matter if I buy a 20$ lead screw, or a 40$ ballscrew.
The smallest type seems to be a 1204. I will just get one of these, which is end machined. One end has a 10mm long, 8mm diameter finish, which will fit a standard 5 to 8 coupler.
The other end I will just cut off, because I don't need it. What about this! Good idea?
[www.ebay.com]

Edited 6 time(s). Last edit at 06/04/2015 02:43PM by FalloutBe.

How did you print the outside at 50um and the support material at 400 um?

Quote
the_digital_dentist
How did you print the outside at 50um and the support material at 400 um?

Looks like the infill is printed at thicker layer height, slic3r allows this in the "combine infill every N layers", I guess other slicers do as well.

Quote
the_digital_dentist
How did you print the outside at 50um and the support material at 400 um?

In Slic3r, there is a setting for the infill: "Print infill every # layers", which I set to 8.

Quote
FalloutBe
If I can however create a perfect Z axis, let's say 20(+-0.1) microns, you almost wouldn't notice that the part was 3d printed!

If you think this depends only on Z, you are going towards a huge disappointment. You will need a similar precision on X and Y (deflection, bearing quality, ...) and, even worse, on the extrusion, with all the problems that gearing or a Bowden may give, not to mention the variance in the diameter of the filament, the warping of the piece, uneven cooling, etc. that become all relevant at micron scale. Many printers pretend to print at 20 microns layer height, but nobody knows their real variance I think.

If somebody has a microscope powerful enough, it would be very interesting to see what is going on at that scale. In my opinion, at some scale layers must disappear completely because of surface tension.

Hmm well yes, the Z precision of my XY axis will probably be worse than the Z axis.
But right now I have a terrible Z axis, and I just want to get it as good as I possibly can.
I don't care if my XY axis becomes the worst factor in play. At least I will be happy with my Z axis then.

I attached a picture of what I have in mind right now. What do you think?
It is a 1204 ballscrew

Quote
FalloutBe
Hmm well yes, the Z precision of my XY axis will probably be worse than the Z axis.
But right now I have a terrible Z axis, and I just want to get it as good as I possibly can.
I don't care if my XY axis becomes the worst factor in play. At least I will be happy with my Z axis then.

I attached a picture of what I have in mind right now. What do you think?
It is a 1204 ballscrew

Same here, I want the best Z I can afford. I saw more than one printer with the same idea of cantilevered bed, but I don't trust it. You will probably need very robust axes to minimize deflection. I cannot find a suitable formula to give an estimation of deflection for that setup, so you will probably have to look at the rod diameter and weight of the bed in similar models.

Using fully supported guide rails or linear guides mounted on a suitably stiff base will help prevent bed deflection due to flexing rails. You'll still have to deal with the cantilever itself.

If you want your work to look good
Have you thought about using ABS
and using a vapor bath to smooth your parts?

You are NOT going to get smooth even 20 micron layers
You won't even get smooth even layers at 70 microns
and printing so many layers per inch perfectly (X, Y , Z position perfect, extruder perfect) Good luck
Just disappointment after 60 hours of printing

Metric vs imperial screw your gonna set XXXX steps for a set distance
Slicing in metric and printing in imperial may have a problem

Leadscrews can be had in different pitches (youtube ref in previous post -- showed 50TPI screw???)

Look into photo resin printer!!!!
[www.youtube.com]

This below falls into un-obtainium
why such a tight layer size?
When the layer size is for example 20(+-1) microns, that means the layer thickness will vary by 5%
I can indeed not see with the naked eye if a layer has the correct thickness, but the smoothness of the surface, with 5% inaccuracy is bad!
It really looks ugly then, and it won't be worth getting to such thin layers.

You are NOT going to get smooth even 20 micron layers
You won't even get smooth even layers at 70 microns
But in my previous post I attached an image, 50 microns and it indeed was smooth! I don't understand why you say it is impossible, if I could do this on a cheap Prusa I3

And I did often smooth my parts with acetone. This makes them look great, but it doesn't change the fact that I want a better Z axis.

Quote
dc42

Quote
FalloutBe
"The one that doesn't move"
Haha, indeed (;

But Something has to move in the Z direction.

This is the big advantage of delta printers: the bed and the print don't move. The head moves in all 3 dimensions; but it is much lighter than the bed, assuming you use a Bowden extruder, and therefore easier to move.

Quote
FalloutBe
I'm quite stuck now with the idea of "really wanting the bed to be my Z axis" because I am really happy with how well the XY assembly performs.

There are at least the following ways of moving the bed in the Z direction:

1. Single leadscrew or threaded rod, and linear bearings. May be the best solution for small beds.

2. Multiple (typically 3 or 4) leadscrews driven from a single stepper motor via belts.

3. Multiple stepper motors driving leadscrews to support the bed, driven in series or in parallel. Problem is sync the motors when you turn the printer on.

4. Three stepper motors driving leadscrews, driven independently by the electronics. The firmware can use a Z probe to probe at least 3 points during Z homing, and adjust the 3 motors to get the bed level. I think this may be the best approach for really large beds. But it needs another 2 stepper motor drivers.

In the case of three threaded rods and one motor, do you think that an autobedleveling use with inductive sensor can solve the problem of synchronization?

Edited 1 time(s). Last edit at 06/10/2015 06:07AM by mdcompositi.

Quote
mdcompositi
In the case of three threaded rods and one motor, do you think that an autobedleveling use with inductive sensor can solve the problem of synchronization?

With three threaded rods driven from one motor by a belt, once you have set up the system you shouldn't get any synchronization problems.

btw there are other reliable height sensors besides inductive, for example the differential IR ones I list here [forums.reprap.org] are designed to work well on glass.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
dc42

Quote
mdcompositi
In the case of three threaded rods and one motor, do you think that an autobedleveling use with inductive sensor can solve the problem of synchronization?

With three threaded rods driven from one motor by a belt, once you have set up the system you shouldn't get any synchronization problems.

btw there are other reliable height sensors besides inductive, for example the differential IR ones I list here [forums.reprap.org] are designed to work well on glass.

I have 250x300 bed, when I calibrate it with the four springs in the corners, I start a print and all work well!
the next print, or when z axis move simply, need a new little calibration!
belt is tensioned ad all work good!!
what do you think about?

Quote
mdcompositi
In the case of three threaded rods and one motor, do you think that an autobedleveling use with inductive sensor can solve the problem of synchronization?

Auto-tramming does absolutely nothing for syncing the screws. All it does is level the bed so the first layer will stick. It relies on three assumptions: that the screws will remain in sync after auto leveling is done, and that the axes of the printer are orthogonal, and that the bed is a flat plane. Losing sync is a disaster and guarantees your print will fail. Using threaded rods instead of lead screws is asking for problems, as is using four screws instead of three. Threaded rods are not straight. One motor will bind and slip. When that happens with four screws, the whole mechanism will bind. With two or three screws, the print may continue for a while (with a tilted bed) or the bed may bind and stop moving. You'll probably have to manually adjust the screws before you can get the thing to home the bed again.

The other problem- nonorthogonal axes- may let your print finish but it will be poor quality, so it too, is a disaster (for printing anything but Yoda heads and tugboats) because it wastes time and material.

Using four screws to level the bed or lift it in the Z axis almost guarantees the bed won't be flat.

Edited 1 time(s). Last edit at 06/10/2015 07:47AM by the_digital_dentist.

Quote
the_digital_dentist
The other problem- nonorthogonal axes- may let your print finish but it will be poor quality, so it too, is a disaster (for printing anything but Yoda heads and tugboats) because it wastes time and material.

Some firmwares can do orthogonal axis compensation. But it would be unwise to rely on it to compensate for other than small errors.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
mdcompositi

I have 250x300 bed, when I calibrate it with the four springs in the corners, I start a print and all work well!
the next print, or when z axis move simply, need a new little calibration!
belt is tensioned ad all work good!!
what do you think about?

Are you saying you have belt synchronized screws lifting the bed? How are the screws anchored- can they move up or down in the end bearings? Is there only one end bearing or are there bearings at both ends of the screws? Are they threaded rods or lead screws?

Using four screws to level a bed or to drive the Z axis is asking for trouble. Three points define a plane. Adding a fourth does nothing but increase the risk of the system failing.

If you use 4 screws to level the bed and assume the bed is perfectly rigid, when you adjust one screw, the diagonally opposite corner of the bed must move in the opposite direction. Each time you adjust a screw, you are tilting the bed around a diagonal which tilts it in both the X and Y planes. Reality is that the bed plate is not perfectly rigid. There is a spring under the diagonally opposite corner that is trying to push that corner up. If you reduce tension on one spring to raise that corner, the opposite corner spring must compress, but since the bed is not perfectly rigid, the plate flexes instead. Now you don't have a flat surface. How are you going to print on that?

Using 4 screws is also a problem for raising and lowering the bed. Instead of a spring at each corner, you have a screw thread that prevents the opposite corner from moving up or down when you adjust one screw. That guarantees that the undercarriage plate has to flex, which immediately increases the load on the other screws. The motors are going to slip and the whole thing will lock up.

Even if you belt-synchronize the lifting screws, four of them is a bad idea. The fourth screw does nothing useful but increases the load on the motor and adds to expense. It also makes set-up and maintaining the alignment of the machine that much more difficult.

Edited 1 time(s). Last edit at 06/10/2015 08:18AM by the_digital_dentist.

Orthogonal axis compensation (which should more accurately be called nonorthogonal axis compensation) doesn't keep the bed lifting screws synchronized. It compensates for poor axis alignment and makes the same assumptions that autotramming does- that the bed plate is flat and that it will remain at a constant XZ and YZ angles throughout the print (and assumes that the angles between X, Y, and Z axes are all fixed and don't change, ever).

Nothing can compensate for the bed tilting during a print. If the bed tilts, the print will fail. The screws have to remain synchronized or the bed will tilt. The best way to ensure that is to use the minimum number required to lift the axis and level the bed (you can argue about how many is the minimum required all day, but 4 is definitely too many), use proper lead screws instead of threaded rods, use bearings at both ends of the screws, and guarantee that they always remain in sync by connecting them with a belt.

In machines that have the X axis lifted by screws, the axis alignment depends on the screws remaining synchronized. If the screws lose sync you have to realign the machine. Orthogonal axis compensation doesn't help because it assumes that the axes remain in the same alignment they had when you printed the alignment test jig. If the X axis has tilted because the screws lost sync, you have to start over again.

Edited 3 time(s). Last edit at 06/10/2015 08:24AM by the_digital_dentist.

Quote
the_digital_dentist

Quote
mdcompositi
In the case of three threaded rods and one motor, do you think that an autobedleveling use with inductive sensor can solve the problem of synchronization?

Auto-tramming does absolutely nothing for syncing the screws. All it does is level the bed so the first layer will stick. It relies on three assumptions: that the screws will remain in sync after auto leveling is done, and that the axes of the printer are orthogonal, and that the bed is a flat plane. Losing sync is a disaster and guarantees your print will fail. Using threaded rods instead of lead screws is asking for problems, as is using four screws instead of three. Threaded rods are not straight. One motor will bind and slip. When that happens with four screws, the whole mechanism will bind. With two or three screws, the print may continue for a while (with a tilted bed) or the bed may bind and stop moving. You'll probably have to manually adjust the screws before you can get the thing to home the bed again.

The other problem- nonorthogonal axes- may let your print finish but it will be poor quality, so it too, is a disaster (for printing anything but Yoda heads and tugboats) because it wastes time and material.

Using four screws to level the bed or lift it in the Z axis almost guarantees the bed won't be flat.

as I said before, I have three screws and not four
on one of these three I connected the motor
the prints are of good quality and everything works quite well.
But it happens that every time I start a new printing plate must recalibrate slightly.
What I understand is whether the configuration that I (I know that the lead screw is the best solution) can overcome this problem since the start of each print is always a new calibration of the plate!

I would look at mechanical factors such as play in any of the bearings, flex in the X and Y guide rails and play in the extruder carriage and overall frame flexure. You shouldn't have to relevel/zero the bed with every print.

When you recalibrate the bed, are you adjusting the zero position or are you releveling it or both?

when I recalibrate the bed I adjust a little the screws on bed, not the zero!
seems that there is something that cause a very little lost
how do you say you make me understand that my problem is not related to the configuration of my z axis and sync but due to other!!
what I would like to understand now is that mine is not a problem in the synchronization of the three threaded rod!!

Maybe your problem has something to do with thermal expansion?
Do you always wait until the bed is fully heated, before you start recalibrating?

Not that it would change much in the Z height of the bed, mostly XY would scale,
but maybe the printbed flexes as it heats because of uneven stresses in the glass/pcb?

Edited 1 time(s). Last edit at 06/10/2015 05:29PM by FalloutBe.

no, I haven't heated bed!!
for the moment I print only PLA!

Hey,

Like cozmicray suggested, I decided to use an SFU 1204 ball screw to move the bed up and down.
Cantilevered bed, held in place by two M8 smooth rods. 2 LM8UU on each side.

The ball screw is cut off to length, and the side which has an 8mm diameter is connected to the stepper motor, 5x8 flexible coupling.

This works fantastically! I have printed parts down to 50microns and the edges look perfectly flat.



Oh, I had very ugly edges in the beginning, but it was due to my bed temperature which was varying by about 5degrees.
Thermal expansion caused the platform to go up and down a tiny bit, but it was very visible in the print!
Changed heating mode to PID and it is perfect now (;