DIY Z axis hiwin update

This is the place to show off your felix printer hacks and improvements.
Post Reply
dognotdog
Posts: 181
Joined: Tue Sep 04, 2012 9:22 pm

DIY Z axis hiwin update

Post by dognotdog »

So, I ordered a bunch of extra Hiwin rails from Guillaume, and now finally upgraded the Z axis of the printer to get rid of the sliding bearing. It's a two rail, three car setup, in theory giving maximum stability, but also extending the Z range due to the asymmetric construction:

Image
Image

and assembled (note the part didn't completely finish printing, yours truly was too impatient):
Image

And a test print (0.25mm layers):
Image
Image

The Good:
- A LOT stiffer than the the Igus sliding bearing, no play, and very smooth movement, it is no longer possible to "rattle" the Z carriage by hand.
- Less overshoot in Y, too, no more backlash due to the whole Z carriage moving due to hard acceleration on Y.
- maybe the 3rd car is not even necessery, as the Hiwin rails have no play, even a single car appears to be much stiffer than the original Igus contraption. I believe more flex comes from the plastic now than the bearings.

The Bad:
- Z wobble not completely eliminated, though greatly reduced, as wobble still comes from the misaligned Z axis coupling (not bent rod!)

The Ugly:
- requires 3 cars, but only 2 rails, so a useless rail remains
- needs to be revised for better access to screws
- still needs updated motor coupling, maybe acme threaded rod, too, to eliminate wobble.

Kjetilei
Posts: 334
Joined: Fri Mar 16, 2012 8:26 pm
Location: Stavanger, Norway

Post by Kjetilei »

Hi dognotdog,

COOL :)

I love it when people are exploring new solutions and taking a chance!

You might be able to buy one of the China clones cheaply (car only). Example: http://www.aliexpress.com/item/our-bran ... 26969.html

That way you get a complete set you can use yourself or sell.

Regarding Z-wobble: have you tried to use the printer without the upper ball bearing? It might reduce z-wobble, at least when the hot bed is in the upper half. You could also try to use a thinner rod (it would affect z-resolution/speed). I have done a quick proof of concept for a belt driven solution that looks promising.

You could consider to just wait and see if Guillaume's new solution is easier to purchase/install than to use a lot of time to redesign yourself with an Acme screw - but the experimentation itself might be worth it :)

Good luck and thanks for sharing!

dognotdog
Posts: 181
Joined: Tue Sep 04, 2012 9:22 pm

Post by dognotdog »

I'm unfortunately fairly sure that the cheap Hiwin clones are also quite bad. Precision in these ball bearings directly relates to stiffness, and past experience says you get what you pay for.

I'm not sure removing the upper ball bearing helps. I believe to have tried that early on.

With this design, I originally had a ball bearing at the bottom, too, but the results aren't better. I believe to see a compound issue of
- uneven thread on the rod
- misaligned motor shaft & rod (decentered, not parallel)
- unevenness on the nut

I removed the lower ball bearing today, and replaced Guillaume's solid coupling with a flexible one. However, it's quite impossible to to properly fasten the coupling to the threaded rod. Still, the results are better than with the solid coupling, with no visible Z wobble for the first 20-30mm, and then slightly increasing. I'll have to print a taller test piece to see conclusive results.

Not sure about the belt on Z, I would be worried that it doesn't have enough resolution like that to properly space the layers. I think the Z axis is more sensitive to that than X/Y.

As a final note, I was printing at 80mm/s perimiters and 150mm/s infill speed with the new setup, which is well double the max speed i could use before for good quality prints. The only thing I could see was a little more overshoot in corners, but that should be managable by reducing the acceleration factors in firmware for X/Y.

dognotdog
Posts: 181
Joined: Tue Sep 04, 2012 9:22 pm

Post by dognotdog »

Just a quick update:

I printed a new Z axis coupling of my own design, alas it's not really better. The threaded rod is still neither colinear nor parallel to the frame / motor axis. I did, however, remove the upper ball bearing, as this new coupler is solid. It may have helped, though I don't really like the idea of the stepper's bearings getting all the axial load. Are there specs available how much they're rated for? Also, I am now quite certain that Z wobble can only be reliably eliminated with a machined end on the rod that allows proper centering. I might be able to give that a try in a few days, to see if it works with the M8 rod.

However, I did a quick thin-wall test print at 0.2mm layer height, 150mm/s:
Image

I am inclined to say it looks worse in the picture than in real life, with the "large" errors to the left of the image measuring no more than 0.1mm, and they are mostly artefacts from the starting "blob" of the layer. Some Z wobble is still visible on close inspection, if you know what to look for.

Overall, the print quality @ 150mm/s looks no worse than my best prints at 50mm/s with the Felix 1.0 :)

Sündi
Posts: 16
Joined: Fri Oct 12, 2012 7:32 pm

Post by Sündi »

Wow thats great work. I nearly did the same, I only made a steel plate and bought one of this: http://www.ebay.de/itm/1-Linearfuhrung- ... 20cbe2a90a. The Stiffnes is much better than the original.
But I have still the same problem with the Z wobble. After that I made an update with an center driveshaft of the e-revo vxl.
Since then I think I have good results with Z-wobble.
Here are a few photos:
http://imageshack.us/photo/my-images/85 ... 13375.jpg/
Sorry the linking of the images don´t worked

Kjetilei
Posts: 334
Joined: Fri Mar 16, 2012 8:26 pm
Location: Stavanger, Norway

Post by Kjetilei »

dognotdog: that sure is a nice quality for 150mm/s printing!

Sündi: Nice mod. I would not have thought that the center driveshaft would work better. Have you tried with one of the flexible metal couplings available from Ebay etc.

Sündi
Posts: 16
Joined: Fri Oct 12, 2012 7:32 pm

Post by Sündi »

Yes I have tried and it was slightly better, but I removed the upper bearing and with the flexible coupling I get in trouble with the zero point of the Z axis. I alway have to adjust the zero point.

dognotdog
Posts: 181
Joined: Tue Sep 04, 2012 9:22 pm

Post by dognotdog »

Sündi, seems like a good solution also!

In the meantime, I did a print of the make torture test at 200mm/s with 0.2mm layers. Curiously, for the first time the arc printed correctly. There was noticably less curling than at speeds around 50-100mm/s. There is a bit of fine stringing, but that might have to do with me fiddling with the Z-lift values in between. And anyway it can be cleaned up with a toothbrush easily.

The only part that looks worse than with a slow print are the cylinders, but I suspect thats the combined fault of slic3r and the marlin firmware, for still not providing proper arcs in gcode.

Image

Now, 200 mm/s is also my travel speed, and I am starting to wonder if I can get it higher, printing at 250 or even 300 seems a distinct possibility, if the stepper drivers are up for it …

… but I am also waiting on yet another Z coupling variation, as well as an acme threaded rod. THE SAGA CONTINUES!

Kjetilei
Posts: 334
Joined: Fri Mar 16, 2012 8:26 pm
Location: Stavanger, Norway

Post by Kjetilei »

I'm impressed about the improvement in printable speed.

Keep on pushing the limits :)

PS: If possible it would be great if you could test how fast you can print the twisted gear lamp (http://www.thingiverse.com/thing:31722) with good quality scaled to 1.9. I test printed it a couple of days ago at 40mm/s just to be sure it would print successfully

dognotdog
Posts: 181
Joined: Tue Sep 04, 2012 9:22 pm

Post by dognotdog »

Unfortunately, I had to cancel this print half way, as I half wittedly placed it idiotically, and ran into the X axis endstop, as one can see:

Some waviness can be seen, which is some resonance in X/Y, but also slic3r did a crap job at some parts. I also mounted a TR10x2 leadscrew, and while wobble is reduced, and generally less agressive than previously, it's still there. I could mount the M8 again, as in the mean time I received a new 5->8mm coupling, but for now I tread with the trapzoidal screw.

Image
Image
Image

On another test print, no such waviness appeared, so it might be some slicing artifact:
Image

Both prints at 200mm/s

Post Reply