Shapeoko 3XXL goes good until it doesn't :(

Hey Everyone,

I got a (second hand) Shapeoko 3XXL a bit more than a year ago and didn’t do much with it, so I took some time this holidays to get into it and try a bunch of stuff and I’m having some super weird (not to mention frustrating) problems with it.

Basically, it goes ok on quick simple stuff, but if I get it to do something more complicated that needs more time, it goes ok for about an hour then just freaks out and starts cutting randomly into the stock:

In the video you can see how it did the first part of the pattern fine but then just went off the rails (figuratively speaking). At the end, I show the wood I used on my first attempt and you can see how it lost control in a different random way (again after about an hour).

Is this something anyones seen before? Any suggestions how to make it behave properly?

Here’s the model I’m trying to make:

When the cut doesn’t match the preview for the file it is caused by an electro-mechanical issue.

Check that all the wiring is in good condition and all connectors secure, and that all wiring leading into and away from the connectors will not shift as the machine moves.

Verify belt tension and pulley set screws

Ensure that nothing will mechanically interfere during the cut (hoses and cords are likely culprits).

2 Likes

Thanks Will,

I will check everything with a fine tooth comb today, it’ll be a relief to feel like there’s something I can try :slight_smile:

I didn’t build this though - I bought it second hand - can you recommend any links to more information about what I should be looking for?

Make sure your set screws on the motor pulleys are centered on the flat of the motor shaft and tight.

It was common to have issue with these and many of us replaced them with hex head bolts to make them easier to maintain.

My theory is the heat from the motor warms up the pulley and it slips.

Just out of curiosity, are problems like this things that go away after you fix a bunch of stuff up or is it more of a ratio thing where I should expect a certain failure rate with job runs?

If you learn to treat your machine properly, keep it well maintained & learn the limits of feeds/speeds for different materials, it can work just as expected every time. But getting to that point takes a bit of effort, practice & knowing what to look out for & what to check.

It helps to develop a routine that allows you to get repeatable results, so then if something does go wrong you’ve got a baseline from which to troubleshoot. These machines only do as they are told by the G-code - when things go wrong it’s usually due to operator setup/design error or the operator not noticing that something is not quite functioning properly with the machine.

And always, ALWAYS, test or verifty a complex/time-sinking-design with scrap material the first time around before using & wasting high-cost material. Until you’ve got experience & trust the machine to do what you want it to, don’t expect it to not screw up.

3 Likes

The set screws for me were a blue loctite and never had an issue again. But the hex heads allowed me to mark them and check them during belt changes.

It’s a machine. It’s going to need maint.

V wheels will need adjustment and replacement.

Belts will break.

For me I was pushing it very hard throwing a 18 lbs spindle at 180ipm cutting through baltic birch for 40 plus plus hours a week.

Steel belts got replaced monthly. The rest of the maint got done at the same time.

Polyester belts will last a lot longer. They just have more deflection especially under high loads.

I really wish I could still get the work I used to do with that machine now. My s5 can do it so much faster and better that I could make a lot of money.

1 Like

Aight, I put new nuts in with some (purple) loctite and left it for 24 hours to cure:


Then ran my gcode again on another polystyrene box lid and got this far:

So there’s none of the weird random X/Y stuff (at least on this test) but something’s still screwing up on the Z axis. Interestingly, the failure has the same pattern as my first test run in pine just in the opposite direction (and not at the same point in the carve fwiw):

So… my 3XXL is not the belt driven Z axis one:

Does this mean I need to pull that whole scary looking bracket apart and look for more bolts to replace?

I suspect that this problem is an X-axis belt tension or pulley set screw issue.

The Z-Plus is pretty straight-forward, maintenance points are:

lubrication:

https://carbide3d.com/hub/docs/maintenance/

the anti-backlash nut

I would recommend lowering your trim router in the mount so as to match the instructions.

Then try cutting a test file such as:

1 Like

Welp, tied some loose leads down in case they were getting pulled around and ran another test.

Wife rings me from shed: “I think it’s broken”.



Here’s the test file I was running:


Not seeing anything in there about going beserk and cutting random holes everywhere…

It’s starting to look like there isn’t going to be anything I can do to sort this out, might have to sell it for parts and buy a new one (one day when I get enough money again).

Can’t guarantee it’ll be a Shapeoko at this point.

Does the CM screen indicate that the job is running normally, even when it isn’t? Does it disconnect from the machine? If you replace the stock and start the job over, does it immediately mess up, or does it work for an hour again? So far, all of the responses have assumed a mechanical issue, which in my mind, would persist immediately upon starting the job again. If it runs a while and then misbehaves, I’d suspect an electrical noise issue confusing the controller somehow. When I first started using my S3XXL, it would just randomly freeze and disconnect from the computer. I traced it to static discharge in the dust collection hose. Once I properly grounded it, the problem went away. This isn’t the same as what you’re experiencing. But the fact that it runs a while before misbehaving has me wondering if it’s still a similar root cause.

CM just keeps chugging along, it doesn’t appear to be aware that something has gone wrong. I’ve been manually powering down everything to stop it ramming the router into the spoilboard.

I used to get disconnects last christmas holidays when I tried it but I “solved” that by getting a big powerboard from the local hardware store with the largest surge protection rating they had and it hasn’t disconnected since.

I don’t know anything about tracing static discharges sorry, but all my instincts are saying it’s something to do with the controller board given how the first stages of failure are to climb up or dig down in a staircase pattern in a very structured way (one step per line).

I did just email support at carbide 3D, it’d be super awesome if there was a way they could help but I did buy the machine second hand so there might not be much they can do for me :frowning:

Okay. If the controller doesn’t disconnect, it seems unlikely to be static discharge. Based on what I see in the photos, I’d take another look at the Z axis. Either something is mechanically loose, causing the motion to be different from what is commanded, or there is a loose electrical connection to the motor, causing it to miss steps. It might possibly be a bad stepper motor, I suppose. But usually they either work or they don’t. I wonder if you set up an “air cut” with a bunch of repetitive Z moves if you could isolate the problem? Best of luck. Hopefully C3D will help, too.

Are you by chance crashing your z axis with too tall of retract height?

2 Likes

Great thought Ryan. If the retract is too high, the Z axis runs out of travel and loses steps. Each Z axis extension after that goes incrementally lower, which would look just like what you’re seeing. The later versions of CM warn of this, BTW, when you start the job.

Hmmm… I don’t think it’s that, though that is an excellent suggestion.

(Some of the failed cuts climb up not down and I’m using the latest version of CM).

Your suggestion has made me realise another thing I haven’t tested though. I’ve been using Blender 3D with the FabEx CNC addon, I should really try something else (i.e. fusion or freecad) in case there’s a bug in the addon which is adding something dodgy into the gcode. I don’t think it’s that because the problems (mostly) start at different points in the job but I should still test it.

Also, I just wanted to say, I’m super impressed by how friendly and helpful the community here is :slight_smile: