[Analias]

I'm in the design stage for my own 48"x48" CNC machine. I would like to use my CNC machine for working wood, plastic, aluminum, and do PCB milling. Due to limited space, I have chosen to use a moving gantry for all three major axises.

Looking over some of the designs I have seen on CNCzone, I notice some use belt driven X and Y axises. A belt driven solution, on first glance, appears to be considerably cheaper than a lead screws or ballscrew solution - as much as a third of the cost of the others.

What are the pro and cons to using belt driven? Do they stretch while in use, or deform over time? How does accuracy compare to threaded solutions?

With a belt solution, what considerations should one use for selecting components?

With a belt it looks like it's easier to implement a dual drive system for the X axis which carries the full load. Any comments about the good and bad of this idea?

[ger21]

A belt driven solution, on first glance, appears to be considerably cheaper than a lead screws or ballscrew solution - as much as a third of the cost of the others.

I don't think they're cheaper at all. Good belts and pulleys are not cheap, and you need to have some type of gear reduction, which usually means more belts and pulleys. When you add it up it's more expensive than good acme, and more complicated to build. My opinion of course. Belts are very appealing, but every time I start looking into a belt driven design, I always end up going back to acme.

[Hirudin]

I've been looking at belts for a while too (I'm also a novice)...

I've also noticed that they're more expensive than ACME's, but if they'll work I'm willing to splurge. I'm looking the hardest at Gates Powergrip GT2 belts and pullies.

I've looked a lot, but I can't find many people using belts with their CNC router (here's one). Is there something inherently wrong/inferior with belts? For instance, do they stretch? Are belts really only suitable for low-torque applications (like torch/laser machines)?

Another thing that I haven't been able to quantify is the amount of backlash that can be expected. Most of the literature I've read says that the backlash is "negligible". Does "negligible" mean zero (or maybe <0.001")?

[LeeWay]

I think belts become a viable or more appealing option when you start getting over a 48" length on the axis. This is when I see them used most often. Smaller screws can have some whipping after that length and the belts negate that concern. Stretching is a concern as well, but I have seen some nice solutions on that too.

I usually use ball screws now on my projects, but I am working on a rack and pinion on a plasma table. I am starting to use some nice belts from SDPI for steppers and spindle drives and such. These are a lot nicer than my first try with some other type and size belts.

I am a little more open to belt use now than I was a year ago. Nice material, but as others have said, it does mount up on cost and complication. That said, I don't think the belt installation need be all that accurately installed. Not like screws anyway.

[kawazuki]

Flame me if you like cos this isn't my area of expertise or experience but it occurs to me to try something we used to use in our drawing office before CAD - wire motion transfer as used on the parallel motion drawing boards. Its basically a system of two wire cables either side of the sliding element and wrapped several times round a capstan mounted top and bottom on a common shaft with a spiral track (cast or milled into the outer face) which the wire runs in. Because of the multiple wraps there is no slippage or 'backlash' and cable tension is via a small turnbuckle between the bearings.
I don't know if it'd work or not but its got to be a damn sight simpler than cogs, belts etc.?

[ger21]

I think belts become a viable or more appealing option when you start getting over a 48" length on the axis.

My thinking is the opposite. Although belt stretch can be negligible with a short belt, it becomes more of a concern on longer axis.

[LeeWay]

That WAS my way of thinking too, Gerry until I saw the thread about the best belt drive ever or something to that effect.

Effectively it was two pieces of belt. One adhered to a rail and the other held down to engage this belt fully except where the upper belt mad it's loop around the idlers and the timing gear. The only stretch then is in that short loop and negligible. That thing was incredibly fast and it has opened my eyes to other possibilities. The belt engages itself along the full rail length and about 65% around the timing gear. Much better accuracy than on a gear and rack. No whipping like on a screw.

[ger21]

Lee, that belt drive was competely different to what I'm talking about. Sure, with that system, any stretch is eliminated. Very sweet.

[Ed_R]

I think I'm gonna go looking for that thread.

[ger21]

I think I'm gonna go looking for that thread.

http://www.cnczone.com/forums/showthread.php?t=59570

[r.r.mccoy]

look here for some belts.

you have to dig to find part numbers but the prices are good

Also call before you order to check availability. They don't update their database to often. Good prices on pulleys as well.
http://kscdirect.com/item/GAT%2B7787...M-LLUKFDA%250A

[Mike Everman]

I just noticed this thread. Our site hasn't been updated yet, but here is the current state of info available on the net. We're not to the point of sales to the DIY'er yet, but I'm happy to help with a project.

See the following links for some light reading and pics:

www.bell-everman.com/ServoBelt_Techsheet_web.pdf
www.bell-everman.com/servobelt%20x-sect.jpg
www.bell-everman.com/ServoBelt%20Q%20and%20A.doc
http://www.bell-everman.com/SERVOBEL...NEMA_34%20.jpg
http://www.bell-everman.com/SB232-90...33-180-520.jpg

And a video:
www.bell-everman.com/ServoBelt.wmv

In the cross-section view, all idler bearings are widely available. There's a limit to the back-bend radius (that's why the larger bearings on the closest idler set), and steel belts are required, not Kevlar. The other thing is that you need to use T-xx series belting or HTD because they mesh with themselves. We also use 3M VHD double stick tape to put the lower belt down.

[r.r.mccoy]

Hi Mike,

Why steel belts?

[Mike Everman]

Kevlar does not like the plus and minus bending. It fails.

[r.r.mccoy]

Thank you very much.

I was not aware of this. I thought it would handle such bending better than steel.

I was thinking of using Kevlar thread in another application. Now I need to rethink this.

[Mike Everman]

Maybe not. I believe that the issue is that the cords are molded in to the rubber and end up taking alternating compression and tension cycles. I don't have a reference, but think you'd be fine if the approach has the fibers able to find and live on the neutral axis if they are going to be flexed.

[HankMcSpank]

Ok, newbie here....first post!

So I got a couple of old DEC LA75 Dotmatrix printers - (for the stepper motors) & have just received my controller board....

http://www.tinyurl.com/648wrp (BTW This seller's postal service to the UK - is woeful. It took him 4 days to post it & 3weeks to arrive. Very slow to respond to emails, his parcel tracking number didn't work, etc - be wary!)

I'm now thinking in earnest what rails to move the X & Y tables on.

The Dec printers I took the stepper motors out of, have a very nice rail (for the print head) - driven by a belt drive. I had been thinking about adapting these for x, t *and* Z axis (& after all the motors will mount perfectly onto these rails!)

Are we saying here in this thread that I should save the heartache, bin 'em & use threaded rods?

I aim just to make very small PCBs (just as a hobby), so I'm not talking a huge amount of XY movement.