Belts vs Screws (on a delta)

[Generic Default]

I'm in the early stages of building a very large delta printer. I already have some of the materials ($280 worth of T slot alone) and I'm not 100% about what to do for the linear motion power system. My design, as of now, can use either double GT2 timing belts on each axis OR a fast travel lead screw. I can't figure out a design for both at the same time, but I'm content with one or the other.

I know the timing belt system will work out, and it will definitely be good enough for heavy duty printing, but I'm not so certain about milling so I'm considering using the fast travel screws instead. I can think of some pros and cons right away:

Belt Drive
+No backlash
+Fast
+Almost no inertia
+Easy to install
+Micro-Step resolution minimum of about 0.007mm
-Stretches over time (creep)
-Stretches with force (not that rigid, shows up in prints)
-Can snap with excessive force
-Not rated for tolerances at all (if you know of any, tell me please)

Screw Drive
+Very rigid
+No creep
+Micro-Step resolution minimum of about 0.002mm
+Can take a ton of force
-Has some rotational inertia
-Has backlash (can be removed)
-Needs a custom made nut
-Can rust over time
-Only rated to 0.009" tolerances over a foot of length

I think I would be able to get good enough rigidity with the lead screws for light machining, I doubt that two GT2 belts would be rigid enough for it.
Anyway, since the fast travel lead screws are under 9 dollars each it makes sense to consider them. The nuts are like 30 dollars though, so I would either machine or print my own with a design that eliminates backlash. Backlash is my main concern, I want something that doesn't have any slop in it.

If anyone has worked with stuff like this before, please let me know! Until then I'll keep my design modular in case I rule one out.

[mhackney]

I've converted or built from scratch 3 CNC milling machines, 2 lathes, 3 routers all using ball screws. The ONLY advantages I give the GT2 belts compared to ball screws is "low cost" and "low complexity" and maybe "low overall weight". These small belts are not going to handle the forces for even a small milling machine (say a Dremel spindle type machine) in my opinion - unless you plan to mill only PCB or engrave.

But you seam to be asking about non-ballnut screws?

Some comments on screws...

The rotational inertia is so negligible at these sizes so to be a non-issue.

Backlash exists on the GT2 belts too, you CAN measure it. Precision ball screws with double nuts can get backlash well below .0005". The old adage "eliminating backlash costs, how much do you want to spend?" applies. Even the el-cheapo screws with a single nut on a motion system like a delta can be in the .001" range with a little care. Regular screws with anti backlash nuts can be very low. These are easy to make and can get you down to .001" backlash for lighter operations (NOT hogging SS with a 1/2" DOC).

You do not want to make your own ballnuts, that's where the magic is. You can make your own regular nuts for acme screws but I have not seen a non-ballnut screw with a fast lead, have you found something? What is "fast travel" in your mind?

".009" tolerances over a foot of length" - GT2 belts aren't nearly this tight. And good ball screws are much better than that.

On 2 occasions I attempted to use acme screws. In both cases I was not satisfied with the speeds or the backlash (one was a router, the other a mini mill) so I switched to ballscrews.

[0110-m-p]

In general what is the cost difference between a ball screw setup vs. and acme screw setup? Double? More?

[mhackney]

"eliminating backlash costs, how much do you want to spend?"

A simple single ball nut with reasonable precision ballscrews might be a factor of 4 higher cost - and this is just a SWAG and it depends on what prevision you are after and comparing. Remember, the RepRap machines use/used 1/4-20 threaded rod from the hardware store! The SeeMeCNC H-1 did too for it's Z axis. That worked fine, you could use a backlash compensating nut (printed of course), but was slower than molasses. I've not seen low TPI acme screws, but I haven't looked for them.

[Nylocke]

Open builds has 8mm ACME rod and nuts for relatively cheap. 1000mm for $30, and the nuts are around $5 I think. I know you were having concerns about the belts getting above 50 in another thread, I don't think these screws or nuts would have this problem, though I don't know what the thermal expansion will do, nor do I know your intended placement of the motion system (inside or outside the chamber)

[mhackney]

That's a pretty fast pitch so speed should be good. Backlash could be minimized with a split nut. Not a bad alternative it seems. I'd try it!

[Generic Default]

I was planning on making my own split nut for a fast travel screw, part number 99030A303 from McMaster.
http://www.mcmaster.com/#acme-precision ... ws/=vg83f7

It's a 3/8 precision acme screw that moves 1/2 inch per revolution (as opposed to about 1.5 inches for a GT2 timing belt with 20 tooth pulley). Still fast travel for a lead screw, but I could get the same thing in a range from 1/6 to 1 inch per revolution. What's temping is that I could get this for only 9 dollars per screw, which is about as much as a belt drive system.

Ballscrews are out of the price range I want to spend on this, and precision ground ones are more accurate than I would need since the rest of the components aren't accurate to 0.0002 per foot. Plus, ball nuts are huge and heavy. If I make a split nut and rigid aluminum carriage, I think I could do drilling and light milling in plastics and woods. I don't plan on making a full sized VMC mill but I want some basic subtractive ability.


Right now I've narrowed down the sources of flex and backlash in my Rostock Max. The belts, 1 inch T slots, NEMA 17s, and carriages are the main sources of flex.
So in my design I have huge 3x1.5 inch T slots (10lb per tower!), double belts, NEMA 23s, and CNC milled frame and carriage parts. The carriages are three times as wide as what comes on the Max, plus they're preloaded with two bearings in each wheel. They still weight less than the RMax V1 carriages though. I want high acceleration and smooth motion.


mhackney, since you are an expert on this kind of stuff, can you tell us of any problems that come up with split nuts? I'm planning on either printing them in nylon or making my own tap to machine them out of kevlar filled nylon or something wear resistant like that. Steel on nylon will last indefinitely I think.

[mhackney]

Expert? me?

The only significant challenge with split nuts is wear. The more pressure you put on the 2 parts the more it wears the screws and the nuts. It is not that bad in practice, especially for low force applications like a 3D printer.

[0110-m-p]

If it is temperature that you are worried about with the belts, you should check out the GT3 carbon high temp belts. Not sure the cost but might be worth investigating...

https://ww2.gates.com/europe/brochure.c ... n_id=19463

Poly Chain® GT Carbon™ High Temperature Special polyurethane compound provides superb heat resistance. It enables the belt to remain fully operational under extreme temperatures ranging up to 120°C and even up to 140°C during shorter periods

[Generic Default]

The belts, carriages, and rails are all outside of the sealed chamber, so I don't have to worry about them taking any heat. If I go with belts, I'll probably get cheap Chinese ones (glass fiber) at least to experiment with. Maybe the US made belts are better, I don't know, but they cost around 5x more. Carbon fiber or kevlar is preferable to glass, but I don't know if the price difference makes it worth the extra.


There are two ways to do split nuts I know of. One is to use two separate nuts on an axis with a spring between them. The spring pushes them against the threaded rods in opposite directions, which takes up the backlash space. They can then be locked in place with a more rigid screw to increase the preload without increasing the friction. Although pitch errors in the screw will result in spots with higher friction and spots with lower friction (or even backlash).

The other way is to make one longer nut, then cut a slit in the axial direction so it's a C shape instead of an O shape. Then clamp the two halves together with screws that pull the gap closed. The Taig mill I have uses this method, and it is very adjustable and has backlash under 0.001 if done right. They lap their lead screws though, I don't know if this method would work as well on a fast travel screw.


I just realized that my design does actually allow both a lead screw and a belt at the same time, since the motor will only be connected to one or the other, moving one will backdrive the other. Arrg why do I always give myself these dilemmas?

[Polygonhell]

The only real issue with screws is that the axis itself must be significantly stiffer than the screw, or any variance from straight in the screw tends to show itself in effector movement. It's one of the reasons that the Z-Axis design is arguably the most important for cartesian printers.

[Mac The Knife]

You also need to take into consideration what the max rpms you will need. If the shaft gets to long, it'll start whipping at high rpms.

[Generic Default]

The shaft will be supported by bearings on both ends, the non-driving end will have free axial movement to prevent binding and thermal expansion forces. The straightness of the screw is very important like Polygonhell said, but I think something in the "Precision Fast Travel Acme Lead Screw" category will be straight enough to not cause visible wobbling over a length of three feet. I know similar ones are used on most cartesian 3d printers for the Z axis now, since older repraps used cheap threaded rod and got Z ridges because of bends in the rod.

Since the travel is much larger on this screw (1/2 inch compared to 1/20 inch) any Z ridges will be much more spread out and less noticeable, although they will be translated from strictly Z to all three cartesian dimensions because of the end effector. If it shows up in prints, it won't be obvious what it is!

The Z carriage should be much stiffer than any other delta printer on the market (on par with the partdaddy). It doesn't have as many rollers, but the rollers are pre-loaded and made out of a stiffer plastic. The RPMs will be quite low, since I'm just using regular NEMA23 steppers and normal printing speeds. If the axis must move 2 inches in one second, the screw will turn four times, which is only 240 RPM.

[jdurand]

Couple of notes on screws...

http://www.use-enco.com often has Acme threaded rod and regular nuts on sale, even better when they have free shipping.

You can make pretty decent nuts for Acme thread (or even standard 60 degree thread) by taking an acetal (Delrin(tm) or Acetron(tm)) rod that can fit in a metal pipe, drilling it to I think it was the nominal diameter of the thread, then cutting it in half.

Heat a spare piece of threaded rod and clamp the plastic over it. When cool remove the shaft. If you have a mold release, coating the rod will help.

[Generic Default]

That's a good idea, I never thought of making a nut like that. Would that be considered forging? I was thinking I would have to buy an extra lead screw then grind it and cut some grooves in it to make it into a tap. First I would just try printing one in trimmer line since it's so easy though. I have a bunch of tungsten disulphide left over from my hotend experiments; now that I have someone else doing the coating, I might as well coat stuff with the leftover WS2 for a better coefficient of friction.

[Polygonhell]

FWIW one of my Cartesian printers had a very slight Z Wobble problem, it took me months to resolve it, including half a dozen different screws including 2 acme lead screws, in the end replacing the linear bearings on the axis with pairs of brass bushings fixed the issue, the bearings had very slight play in them and any even slight deviation in the screw would turn up in the effector motion.
You just have to be careful that whatever you use as a guide for the linear motion isn't affected by any slight variance in the screw.

[Generic Default]

My design has a preloaded roller bearing design similar to what's used on the Rostock Max, so there is no play at all. With that being said, nothing is infinitely stiff and screw wobble will show up in any system, just at varying levels. In big CNC machines it may only be a few nanometers of wobble caused by the rotation of the screw.

The rail on my system is a 1.5x3 T slot extrusion, which is very stiff and heavy. The oscillations will happen over 0.5 inches in a sine wave pattern. The screw will only be 0.375 in diameter, so any flex it causes in the rail will be tiny. Although I am considering using the same lead on a 0.5 inch screw; from my understanding of manufacturing, the lower length to diameter ratio of the thicker screw will increase its straightness over the same length as a narrower screw. Plus the price is the same.

[626Pilot]

Have you looked into Spectra line, or similar? It's normally used for fishing, but a lot of new deltas use it.

+ Cost per meter is tiny compared to GT2 and other timing belts
+ No backlash
+ No teeth wearing down
+ Can be found rated for 100 pounds (and more)
+ Adds less mass to drivetrain than either timing belts or leadscrews
+ This guy designed a Rostock with filament drive, and you can download his pulleys for free

I ordered a delta printer from a Kickstarter campaign, which should be getting here "soon." It uses a filament drive. I'm looking forward to seeing how well it performs. If I was going to build my own large format printer, which I probably will do some day, I would use filament drive for that as well.

[Generic Default]

I've researched it a lot, but I don't know enough about it. The components and setup would be similar to a belt drive but without pulleys. I just don't know if it is repeatable enough since it relies on winding it around a rod, and since the winding moves along the rod during rotation, it might mess up the linearity of the system.

I already bought some kevlar string for other parts of the printer, I think aramid would be much better than UHMWPE since it works over 400C and is much stiffer (The thread I have holds up to 135lbs). I would love to use it but I can't be sure of the tolerances it allows in this type of system. I'll have to test it with quantitative results sooner or later. Maybe a test rig with big digital calipers to check repeatability and accuracy.

[626Pilot]

I've researched it a lot, but I don't know enough about it. The components and setup would be similar to a belt drive but without pulleys. I just don't know if it is repeatable enough since it relies on winding it around a rod, and since the winding moves along the rod during rotation, it might mess up the linearity of the system.

Turns out Polygonhell figured out a solution: http://3dprinterhell.blogspot.com/2013/ ... drive.html

[youtube]http://www.youtube.com/watch?v=23bBKTQygBQ[/youtube]

[Generic Default]

That looks pretty cool, I'll have to make my own with kevlar thread. Plus it only needs two custom machined parts from what I can see. I'll add that to my list of drives to experiment with, now I'm up to three!

I'm assuming the bottom piece is just a simple idler that can be made with a ball bearing, right?

[626Pilot]

Idunno!