Z-Axis ACME vs. Threaded Rod

[mikegrundvig]

So I've thought about this a lot and the cheapest fix might be to just have a guide w/bearing at the top of the machine on each side that the rod rides in. So the motor constrains it at the bottom and the bearing at the top. It can still "float" up and down but it will be far harder for it to wiggle in and out or left to right. Conceptually something like this:
http://www.thingiverse.com/image:131164

BTW: the printer in the picture is using ACME rod and nut.

-Mike

[orcinus]

IMHO, leaving the top unconstrained is a better idea.
Letting a bent rod whip at the top dissipates the energy/motion that would otherwise be used to destroy threads in your nut block.

Plus, it won't solve much. After a while, you'd have the middle whipping about in loosened threads of the nut block, again wasting what would be transformed into purely Z-axis motion, and once again causing uneven layer spacing.

(just my 2c, haven't actually tried constraining the top rod ends, basing this on observation - as the gantry goes near the top of its Z travel, you can notice the motion getting more and more uneven and cyclical, because the gantry itself constrains the top of the rods)

[orcinus]

Oh, and one more remark.
Mechanical solutions, especially ones concerning the Z axis, presented as an end-all solution to wobble and inaccuracy, that apply to other RepRap bots, usually don't quite apply to ORD bots, owing to the fact the game is completely different. Nearly all of them ride the Z on two smooth rods and two threaded rods. In such a construction, both of the rods serve to constrain the gantry. On ORD bots, the threaded rod only drives the gantry, it serves no constraining/supporting function (except from keeping the whole thing from falling down, of course). As such, most of the "solutions" that are meant to prevent wobble, presented on Thingiverse, don't actually do that much good in this case.

"Their" wobble is not the same as "our" wobble. Their wobble is the gantry being swung in X and Y by bent Z rods. Our wobble is the gantry moving less than the average increment in one half-turn of the Z stepper, then more than the average increment in the second half-turn of the Z stepper.

If you really are getting X and Y wobble, then something else is very very wrong (V-wheels not tight enough, or the assembly not square).

[mikegrundvig]

I guess I see this problem very differently: bent is bad, period. I'd never tolerate a bent or whipping screw on my milling machine - it's totally alien to even consider that acceptible in the machining world so I don't see why it's acceptible here. No doubt, it won't be perfect but you can certainly strive for it. This article is very useful in understanding how screws in motion behave on linear motion systems - like an ORD bot:

http://motionsystemdesign.com/linear-mo ... xity-0894/

This image shows how you can constrain a screw:


In a very sloppy way, we could say that the motor and coupling act as a fixed end. So the first design is basically how the Hadron works now. What I'm suggesting is the 3rd option. You can see this quote from the article mentioning what we are doing now - "Fixed at one end and free at the other. With one end free, the ball screw is easily bent because the free end can move like a cantilever beam."

On ORD bots, the threaded rod only drives the gantry, it serves no constraining/supporting function (except from keeping the whole thing from falling down, of course).

On the second page of that article you can see the section on stiffness calls out that there is actually a compressive force on a screw and it does deflection. Since it's cheap to improve this a bit, it seems like a good idea to do it.

-Mike

[orcinus]

Oh no, i'm not arguing that bent is good. No sirree
I'm arguing that, when you have a bent rod, in this particular case leaving it unconstrained on one end might be a better solution (until you obtain a non-bent rod) than constraining it on both ends, unless it's a very temporary solution.

The way i see it:
- constrained at the bottom = whips at the top = BAD
- constrained at the top = whips at the bottom = WORSE
- constrained at the top and bottom = whips in the middle = WORST

[Liberty4Ever]

The ORD Bot gets by with a design that other CNC machines don't because the Z axis motion when printing is only a few thousandths of an inch at a time. The concept of critical speed doesn't really apply in the ORD Bot's Z axis because there isn't enough motion to cause any whip. However, I do believe some Z axis design improvement can improve the overall design and yield improved print quality.

Despite some earlier comments, I think the problem is correctly described as wobble. Bent threaded rods or straight rods installed in the coupling at an angle are exerting a force that deflects the gantry in the XY plane as it moves in the Z direction. Rather than whip, which I think of as a dynamic condition, this wobble is a static condition, which is to say that it doesn't rely on inertia or harmonics in the Z axis motion components. At a given angle on the Z axis steppers, the Z axis rods will exert a small lateral force to the gantry in the XY plane. It doesn't need to be much force to deflect the gantry in the XY plane. The bolted together Makerslide is fairly rigid (probably more so in the X direction than Y), but the force acts over a significant distance to torque the frame, and not much force is needed to flex the frame and take up any slack in the V bearings to cause the 1-2 thousandths of an inch of motion in the XY plane that would result in lowered print quality. Remember, .002" of flex in one direction will be followed by .002" of flex in the other direction when the Z axis screw has turned 180 degrees, for a total error of .004".

I completely agree with the comment that we shouldn't tolerate any readily measurable bend in the Z axis screws. Visibly bent screws should be replaced. And as I said before, if I'm unable to quickly and reliably achieve wobble free operation, I WILL be upgrading the design as needed. I've even looked at low cost Chinese ball screws on eBay. If I needed to do that to eliminate Z axis wobble... I would.

[mikegrundvig]

I still like the idea of pinning the screw at the top with another bearing. I can see no downside to this and many up-sides. Additionally, rather than looking at ballscrews (which are massive overkill for this application) check out ACME screws from a place like Roton. For instance:
http://www.roton.com/Mating_Components. ... ly=7060750

Those are 3/8" with 16 turns per inch. That puts them very close to the M8 x 1.75 while being a bit thicker. You can certainly do cheaper than Roton, but they are considered solid entry-level ACME screws for CNC machines. Going up from there would be Nook and then Thomson but those are both significantly more expensive. ACME screws are lots more efficient than pure threaded rod though no where near as good as some of the more exotic stuff like torquespline.

In practice with 10x microstepping 16 turns per inch is a lot more turns than we can use on these machines. So the next step up would be 1/2" x 10 which is where you get into very common and standard size for desktop CNC routers and things. 1/2" screws are very durable and strong. You don't need it for the z axis IMO but you can also get great anti-backlash nuts for this size as well:
http://www.roton.com/Mating_Components. ... ly=7059284

For reference, in a traditional CNC design you'd use a lathe to create a smaller diameter to attach to the motor via coupler. Then you'd have a threaded portion that holds a nut. After that, you'd have another flat spot to hold two angular contact bearings pressed together. The nut compresses the screw against the bearings. In this design, the motor play and runout are totally ignore-able and you have a very rigid setup. At the other side, you'd have at least one skate bearing or do the two AC bearings again. This link shows lots of info about how this is done:
http://www.cnccookbook.com/CCBacklash1.htm

Last I checked, it's about 20-25 bucks to have a small machine shop finish up a leadscrew for you like I described. I don't think its necessary for these machines but it's always an option

-Mike

[mikegrundvig]

Oops, forgot to link this. DumpsterCNC has lots of inexpensive hybrid solutions too. For instance, they have threaded couplers that can adapt ACME directly to the motor shaft. They are lots more accurate than a pure smooth coupler to rod as they are machined for this purpose. 1/2x10 for instance:
http://www.dumpstercnc.com/acme_12101.html

Or you can see their other stuff here:
http://www.dumpstercnc.com/

[mattrsch]

I don't think anti-backlash measures are required or even useful on the ORD bot Z axis since it is pre-loaded by gravity. I have always liked the idea of the open bottom nut traps on the repraps that prevent z crashes from causing severe damage. If I ever retrofit my ORD to a different Z drive screw I will probably adopt something like that approach. Luckily Automation Technologies supplied me with surprisingly straight M8 rods. Once I threw some flexible couplers on the wobble was eliminated (as far as I can detect visually).

[AVRC]

The Z wobble is not actually wobble. The gantry does not shift. The uneven layers from bent rods or bad couplers are not offset, but rather different thickness layers, causing bulges every rotation.

What is happening is the motor bracket is flexing as a result of the bend or misalignment, raising and lowering the motor.
When I first put my Hadron together, I could see the motors flexing up and down with every rotation as I ran the Z axis.
In my case, the rods were straight, but the kit couplers put the threaded rods at an angle. I printed a set of the Nophead couplers and that fixed the whole problem. Occasionally, I need to adjust them, but I have proper clamping flex couplers that will go on next time it goes out of whack.

I do think ACME rod is better than regular threaded rod and I'll probably switch to that when my threaded blocks wear out, but they are not the solution to the print quality issue.