What's the expected XXL Backlash on X & Y?

Sorry to open that discussion again but I’ve read what I can find and I’m not sure whether the numbers I’ve seen apply to a regular size, an XL or an XXL and what performance I should expect.

I’m not trying to make a perfect machine, just wondering what is “normal” for an XXL and whether I’m doing anything wrong in the setup.

I have an XXL with the RepRap.nl kevlar belts (these are quite a bit less stretchy than the belts the machine shipped with, backlash went down to roughly 1/2 what it was before).

I have adjusted the V wheel tensions so that the wheel won’t quite turn by fingers on Y plates and Z backplate.

I have the HDZ installed, the left side V wheel on that was hard to install and is at the ‘loosest’ position in the eccentric nut and still is tight enough to not turn by finger.

The belts are tightened as per instructions, pulled in the length of the supplied screw, I can get a couple of fingers under the X belt with the carriage off to one side.

Using a (Mitutoyo 0.01mm resolution) dial gauge I’ve measured the backlash on the X and Y axes on the machine with no cutting loads and I’m getting approx 0.05mm ( 2 thou ) in both X and Y in lost motion when reversing direction. The actual movement distances are within < 1% of expected but on reversing somewhere between 0.04 and 0.07mm is lost depending on where on the X or Y travel the machine is, it’s worse in the middle and better at the ends. Steps of 0.025mm are clearly defined, accurate and repeatable in both directions on the gauge (on direction change the first couple are ~0.01mm until backlash is taken up).

Hand pressure side to side or back to front on the Z carriage can easily produce 0.2mm ( 8 thou ) either side of zero on the dial gauge.

When cutting, including spring passes I’m having to put about 0.15mm ( 6 thou ) of ‘backlash compensation’ into the toolpaths to get to measured dimensions (using a Mitutoyo digital caliper), the actual repeatability is closer to 0.05mm.

There seems to be around 0.02mm backlash on the HDZ and well under 0.05mm when pushed up or down so nothing to complain about there.

Having inspected the machine carefully, pulled the lock screws on all the stepper pulleys and re-inserted, on the flats, threadlocked and tight, checked for play between the Y plates and rails, Z backplate and X rail etc. The only place I can see this backlash is in the belts, pushing the Z carriage left and right the ±0.2mm it easily moves on the X rail the drive belt visibly moves with the carriage, at the carriage, but not at the two ends where it’s clamped down. Same for the Y belts.

So, is this the sort of backlash I should expect on the XXL?

If not, have I not tuned the belts up to a high enough note?

Anything else I’ve missed?


Ide say go tighter. Sharp twang instead of a dull sound. On a S3 I’ve got white steel core and repeatability is in the tenths




What spec are those steel core belts?

They are 9mm direct replacements but imo have a small window for proper tension. There are members that did quite a bit of testing on steel vs kevlar in this thread.


I can see an xxl having a tougher time with belt stretch vs a S3 but there really shouldn’t be any backlash. Just dont tighten enough to the point you break the stepper shaft lol.


Where you are is not way off.

If you haven’t calibrated for stretch, don’t forget that. There are some subtle things to know there though - grbl keeps track of the roundoff, so although every step will still be exactly one step, over a given distance there may be a fractional step required, and it will round that off (it can only make integral steps) - but keeps that rounded off portion in it’s map of machine location, so it won’t accumulate as error across the machine. The average move will be correct (within the limits of the linearity of belt stretch), but any individual move can be off a one step depending on where the roundoff is landing. That likely doesn’t matter much, as 1 step will be on the order of .025mm give or take depending on your calibration but if you’re chasing 100th’s of mm, it’s a factor.

As @Vince.Fab pointed out, you can go with more tension on the kevlar belts than the fiberglass ones, but I’d recommend a little anti-seize on the tension screws - they will gall in the clips and later snap off when you try and take them off to change belts…just saying…learn from my fail. The steel belts help a lot but do wear out faster I think - they don’t just keep uniformly (mostly) stretching like the kevlar ones, they just suddenly get a “long” spot in the middle somewhere. I run steel and like them, but it’s something to know.

There’s definitely a balance on tension, and I don’t know quite when to land it for myself - make things too tight and the tooth-tooth distance opens up too far and increases backlash at the pulley, but too loose, and the belt stretches every time anything moves. I went to “twangish” over “thudish”, but if you’re chasing perfection, there’s a balance between there that’s probably optimal.


Thanks, if I’m in the ballpark then that’s good info.

That is handy to know, I haven’t read up much on GRBL trying to avoid reading the code, hadn’t thought through how the roundoffs were handled.

I’ve been leaving the macro scale calibration until I got the backlash and wobbles sorted and making sure I machine matching dimensions on matching axes so far.

Thanks, consider that lesson learned here.

Thanks, very helpful, I’ll try tweaking the kevlar a little more and see how that goes.

Thanks for the help.