V2 - Linear guides and belt intermesh?

I’ve been planning a V2 build as a first-CNC project, however I’m interested in some possible modifications to the stock design and wanted to get some input.

First and foremost: I’m building towards handling a full 4x8 sheet, but want to build a more accurate/stable y-axis guide system. To this end I’m considering using an alternative roller system. I’ve considered linear rod and bearing (difficult to get rail that long), aluminum X extrusions (same), aluminum profiles with bearings, or conduit rails, such as used on the X axis. I’m strongly leaning towards conduit or iron piping, as it is relatively easy to support it using off the shelf connectors or 3D printed parts and with 608’s, you get pretty decent kinematics without relying on the table to be perfectly smooth - in fact it is possible to surface/tram the whole table once the guides are straight and parallel. Further, the axis attachment to the machine can be made quite stable

The y axis drive system is the next consideration. The longer belt lenght has some backlash to it due to the length of travel. Even with PU steel core belts, there can be a not insignificant stretch over 9-ish feet of belt. The idea here would be to use a piece of flat stock or ply cut to the right thickness to allow a second belt, contact cemented to the face to intermesh with the drive belt along the whole length, similar to modifications I’ve come across for the MPCNC. This would help limit stretch and could improve overall accuracy.

The latter is the concept I think will actually be the most difficult, likely requiring a lot of adjustment to get the pressure right over the whole length of travel. There are a variety of ways to do this, so I think it’s more of a “What is the best way” as opposed to “How” and I’m quite open to suggestions - or to be told it’s a waste of time.

Last, I very much plan on running this on a 5-driver board. Likely an SKR 1.3 with DRV chips, two axis self-squaring and possibly an auto-home for the Z (there are several options open with this board type). Definitely 24V for the better torque.

Thanks for taking the time to read!

I’ve been planning a mod, and there’s a fairly simple solution I’ve designed. 3/8 V roller bearings and put angle iron the the Y axis, and have the v wheels ride the angle iron. You can make a jig to flatten out the angle iron if you want. The KRMx01 CNC machine uses v bearings on angle aluminum after a break in period too.

It has some stretch, but not backlash. The tension on the belts stretches it a little, but thete is almost no force in the Y direction from the gantry, so there should be almost no backlash.

Perhaps this is a difference in terminology (stretch vs backlash), but would there not be a few thou or even a few hundredth of a mil worth of positioning inaccuracy over the length of the travel due to the stretch of the belt? I understand that it’s not under a high degree of load, but the un-equal stretch over the length of travel is a concern to me, which is why I’m looking at ways to limit this.

This is just my opinion, but if you are worried about “a few thou” of absolute positional accuracy on axis that is 2500mm long you might be building the wrong machine.

Backlash and stretch are very different things. Backlash is the difference in position when approaching the same location from opposite sides (due generally to slop in screw/nut or rack/pinion engagement) and does not exist in any significant amountst in belt-driven motion systems. Regardless of which direction you approach from your drive pulley will always be in the same place which will put the gantry at the same position on the belt.

Stretch (in the case of belts) is a horse of a different color. Any positional inaccuracy there depends only on axial loads applied to the cutter, is independent of the drive direction, and disappears then the load is removed. Therefore you can manage any issues with stretch by doing very light finishing passes so there is the bare minimum of cutting forces to possibly disturb the position of the axis.

Take a read through this thread if you have time. It discusses belt stretch (complete with graphs and spreadsheets), links to a thread about the double-belted MPCNC design, and contains a description of a double-belted LR2 that doesn’t require an 8’ length of aluminum angle.

1 Like

Regarding the different rail options you mentioned: My vote would be to use a 1/2 (or 3/4") EMT rail + bearings (similar to the Artisan3/PlasticMonstrosity) but with two changes:

  1. Only use one rail for the Y axis and keep the skate wheels for the opposite side. With only one rail you don't have to worry about over-constraining the axis motion and having the gantry bind if the rails are not perfectly parallel (or if your table swells from humidity).
  2. Use a 2-bearing rather than 3-bearing roller design. This allows you to mount the rail directly to the table because there is no clearance required for the lower bearings. It also leaves the top of the conduit clear to drill+screw through so you don't have to get cute with how you attach it. Finally, the weight of the gantry will be more than sufficient to resist any upward, sideways, or torquing moments the the tool would be able to create.

That’s a really good point. Although I’m not sure it isn’t overkill with even one. The ways I’ve seen to restrain that axis all seem pretty darn good to me (tracks, unistrut/indepentent rails, or routing a bull nose groove for the wheels.

Not to discourage you though. Please make this machine your own and share any measurements you make.

Yes, the “conduit+2x608” is basically an inverted version of the “skate wheel in U-channel” design so it has a lot of the same benefits. A few things that I think put conduit slightly ahead:

  1. Less deflection when side-loaded (urethane wheel may deform)
  2. Easier to 'hand fit' with shims/files/etc. to be straight/flat than unistrut.
  3. Slightly smaller risk of fouling with chips/dust.
  4. It's more MPCNC-ish
  5. Less expensive (see #4)