Expand and read the bold parts for the tl;dr. I kinda buried the lead, and if you don’t expand, you will get the wrong idea from just the first paragraph…
Although they are not necessary (and don’t even work) when your Z screws are not over-constrained, I really like the simplicity of this style of isolator. The concept is that the nut is trapped and the guides on the sides fit around the smooth rod (some versions instead fit inside a T-slot or similar) to prevent the nut from turning so that it will move up and down the rod. The carriage for the X axis (usually the ends of the X axis) then sit on top of this part, and allow it to slide around to absorb any wobble in the center of the rod so that it does not apply a force perpendicular to the Z axis. A cool side-effect is that if the lowering axis hits an obstruction (like the extruder colliding with the platform or a print left on it), the axis will stop moving down while these isolators continue to move, keeping the two nuts in-step so that once the issue is solved, you don’t have to re-square the Z axis. I’ve been wanting to implement something using this concept for a long time…
Unfortunately, they don’t work.
Consider what happens with a very bowed Z screw constrained at both ends. In the middle of the screw, the nut will be visibly moving in circles around the rod’s axis of rotation. This circular motion has two components: a motion along the X axis that moves it closer to and further from the smooth rod, and a motion along the Y axis that makes alternatingly move in front of and behind the axis of rotation (this axis of rotation is parallel to the smooth rod that is guiding its Z motion). This latter motion is the problem. When the nut moves in a direction that is not coplanar with the Z smooth rod and the Z screw’s axis of rotation, the whole Z screw isolator part must rotate to keep the nut in its trap and the smooth rod between the guides. This rotation translates into movement up and down the screw, so by isolating the wobble of the Z screw in this way, you are creating an oscillating error factor in the Z position, aka Z-ribbing. http://www.thingiverse.com/thing:45454
Speaking as the designer of the original version of that part (thing 20147), I must partially disagree. They do work in so far as they isolate the X axis from Z screw wobble. And they have saved me from a Z axis crash into the platform. But they do introduce the new problem you describe. I think the induced Z error will be very difficult to detect though.
I haven’t run the numbers yet… Let’s see: With M8 screws, the error is about 0.0035 mm per degree of nut rotation. I guestimate I’m getting something on the order of 3° total nut rotation from wobble (certainly not as much as 10°), so about 0.01mm error per rotation max. That’s 0.01mm error per 1.25mm of Z travel. Compared to the Z wobble I used to get, that’s pretty good!
Of course, any new printer or Mendel X carriage design should inherently decouple all axes from Z screw wobble without introducing any other kind of error. This kind of component should never have been necessary in the first place.
Thanks for running the numbers! You forgot one thing, though. You have to consider layer height. The standard layer height that any decently calibrated machine should be able to achieve these days is .1mm. At this layer height, that .01mm error is a 10% error, which is huge. This uneven compression of the plastic varying by 10% every 12.5 layers will cause some layers to be severely over-compressed while those directly between them are correspondingly under-compressed.
Z-ribbing, as always, is much more noticeable for high resolution printing (whereas Z wobble is merely a bit more defined), and I think if you try printing at high resolution, you will notice this effect.
That’s a 0.01mm error per Z screw rotation, so less than 0.001mm per 0.1mm layer. A 1% error, not 10%. Twice that, really. The error averages to zero for every rotation, so the 0.1mm Z error swings negative for half a rotation then positive for every second half rotation (or vice versa).
So let’s say you have a layer height of .1mm and a thread width of .5mm with 3 perimeters laid down from inside to outside. The extra spread of plastic will mostly be on the outside because the partially-cooled inner loop will act as a barrier to the spreading plastic, so you’ll see half of that 2% error on the inside (hidden) and 2.5x that error on the outside, so the perceived error will be 5% of .5mm, or .025mm, so the magnitude of the oscillation will be 2% of its period. From my experience, this error will be visible with an opaque plastic (especially white), but probably not with a translucent one.
As fun as all this math is (that’s not sarcasm, I really do find this kind of math enjoyable), it’s kinda beside the point. A proper solution to one type of error should not introduce another type, even if it is smaller. Not to mention the fact that you can solve the problem by removing parts rather than adding them so that the Z screw is no longer over-constrained.
OK, then. So, I need to update my firmware so I can tune my heater PID, so I can unmask this error which I probably won’t be able to see in my natural ABS! But I do have a project coming up in black that will require a good finish, so I’ll have to plan for this.
If the coupler is flexible, and the other end of the rod is free, then your threaded rod isn’t overconstrained, and you shouldn’t need an anti-wobble device at all. If you need anti-wobble in one direction, why not both? If your nut carrier could be made free in both directions without inducing rotation of the nut, then you could constrain the free end of the threaded rod without Z wobble.
None of that should be necessary, unless you need to constrain the free end. If say, your motors are at the bottom, and the nut can travel close enough to the coupler that the rod is in danger of falling over or whipping during rotation.
Fwiw, I tried these isolators while I was troubleshooting my wobble/ribbing issues and in my case they definitely improved things, but not as much as when I later swapped out the M8 threaded rods for smaller (in diameter) ones.
not tl:rd, just confused.