Houston we have a problem, belts

Updates and News
1

Well to my wonderful crew mates, let’s not get me walking the plank over this one.

a pirate walking the plank holding some machinists calipers
a pirate walking the plank holding some machinists calipers1024×1024 192 KB

Best I can tell we have a belt issue. Our 2mm pitch belt does not seem to be 2mm pitch.

I was asked about this during the LR4 beta and blew it off, We have been asked about this many times and blew it off. We have a fixed gear system and calibration should not really be needed. That is only true if all the components are as advertised. I have been using this same belt supplier for a few years now. I am not sure if this is a new problem or has always been there. JJ really pushed to convince me to take another look and he was correct in the issue and I am very thankful for the persistence.

A 1200mm gcode move results in a 1198.25mm actual move. I was careful to only move in one direction to eliminate as much backlash/slop as possible, I tried 8th stepping and 16th stepping. I used 3 tape measures, and calipers. In the end to verify we just measured the belt with some calipers to find it is indeed not perfect, not even very close in my opinion. What we have on our hands is a case of high precision and not high accuracy.

1024×684 188 KB

What this means, is the steps per mm need to be adjusted on most, if not all builds. Something I and most others warned everyone against.

On my LR4 I found 50 steps per mm needed to be adjusted to 50.073 on both X and Y on the Jackpot, I have not verified it on the SKR Pro yet but those would be 100 to 100.146. But I think for now it is best to see how everyone’s numbers play out.

If you want to test your build I suggest doing it just like the square calibration but you only need two points at the extremes of your Y travel. I suggest only moving on one direction so home Y, move Y + 5-10mm, pop a hole, move out to your Y max (or close to it the further, the better), pop a second hole and measure as accurately as you possibly can. I tape the tape measure at one end and gently check with the other, I made several tests all exactly the same as far as I could tell. From there you can get the difference in a percentage by (commanded distance)/(actual distance).
In my case 1200/1198.25=1.001460
To get your corrected steps multiply the current steps per millimeter by that number.
In my case 50*1.001460=50.0730
Round to 3 decimal places
50.073

Once we get some more confirmations, we can see the spread and how and if I should correct from here.

Apologies for the carelessness and assumptions of the parts perfection. Luckily, this is a pretty easy fix and only costs a little time.

9 Likes
2

Is this a fairly recent issue, or does it apply to long standing builds as well?

3

@vicious1 Thank you for taking the time to work on this with me, and I am happy we learned something from it! As I told you previously, you owe no apology. I thought the same as you did this whole time.

I had the same issue on my LR3 as well. And now that I have made the adjustments on my LR4, a 1400mm strut I cut earlier this evening came out dead on. I am beyond happy with the accuracy!

3 Likes
4

It could be recent, it could be since the new belts, it could be from a long while back. I have been through a handful of belt vendors.

I actually have a new sample sitting here that I will be testing very carefully.

edit I think the new belt does look spot on, just comparing it to the current belt. I will need to test it more tomorrow.

2 Likes
5

Crap, should I/we beta testers have been more vocal/observant about this during Beta/RCs? Sorry for not providing more data clearly at the time https://forum.v1e.com/t/update-or-new-version/40835/4732?u=azab2c

1 Like
6

I guess there’s a reason why Creality uses Gates belts :slight_smile:

Doesn’t seem like a major, honestly, to me at least. As I said in another thread, I’ve found that most of my tape measures are incorrect vs a class-1 reference steel rule and a most of them by more than your error here.

For clarity, that’s 0.15% oversize. That’s enough that I probably wouldn’t even bother correcting it.

1 Like
7

Well crap. .75mm short on a 1200mm move.
That’s .06% error.
Frankly not terrible.
[ edit- 1.75mm error. I apparently can’t read today…]

I’ve replied a few times on and off the forums that estep calibration isn’t needed, so I’ll need to check in with a few folks that I may have given bad advice to.

Does new belt fix it? If so I’ll be ordering replacement belt rather than taking any machines and tweaking the esteps.

This affects the belt for the MP3DPs as well?

2 Likes
8

No, I heard it, just didn’t think it was possible. Seemed like a waste of time to check…oh how wrong I was.

That was my second argument. That is why I tested three then went to 200mm caliper tests. The movements are most certainly more than a tenth of a percent off.

1.75…pretty bad

I only have a 10M sample, I have shipped probably more than 60km of the current stuff.

Yup, I wrote it off to material shrinkage since it is so small over 100mm. I am very interested to know if this helps with the ripple issue though.

1 Like
9

No. I am very sorry for not listening to both of you. I am definitely an asshat for that one. I was convinced it could not possibly be an issue. I will surely be more critical before writing things off.

I am pretty confident I told you both your measurements were wrong and we had fixed gear ratios…not possible to be wrong. I was.

The assumption was any error would lead to poor gear meshing and we would see belt movement issues. So no movement issues no distance issues…bad assumption.

I am interested in what the manufacturer says.

1 Like
10

I thought the same thing, so it’s not just you.
And apparently my reading comprehension is going south on me.

So with the small error the meshing isn’t an issue.

I wonder, though, if you can HEAR the difference between the two belts…

Chill a bit and give my friend Ryan a bit of a break. You’re not the only one that looked past the reports.

Is there any kind of not horrible to do lot sampling test you can do that would check the various incoming material lots for the issue?

1 Like
11

LOL, if they are a bit short on the measurement, will a slightly overtensioned belt give a better result?

1 Like
12

Hmm, interesting. Since my cuts came out the way I wanted them to, I am not sure whether they are actually too short or whether I actually want to know they are. :sweat_smile:

4 Likes
13

:cold_sweat:so my 2 lr4 kits will need steps revision in firmware?

Cutting big circles (or drawing them) is the best way for tuning steps. Any different than perfect and the circle is not a circle.

14

So, i have often wanted to say that I feel if a 3d printer needs adjusted, then why no the cnc, it is just a larger version. There are so many variations in processes anymore, who is to say there is not a tiny variation of the pulleys, idlers, hell even a printed process.

To me this is not a big deal. Compensation is easy, measure, calculate, fix, move on.

We all can be wrong, it is how we handle it when we are, correct?

Maybe instead after the Crown we carve out a cube and calculate and fix :slight_smile:

happy thanksgiving everyone!

You know can something similar (no way as advanced) as the calilantern also help us to Auto Square as well as set our steps???

15

Thanks :+1:, I really just feel a bit lazy on that or grumpy old man. Being so set and in my ways that I didn’t even consider it, when the test takes about 3 minutes total.

As for testing I am hoping ine of my nicer rulers will quickly show the issue. I am actually hoping my free digikey ruler shows it. Simple test. Well that and change vendors, unless by some miracle my current vendor listens, believe me and can adjust.

It is the same situation a printer should not need XY or Z step adjusted, it is a fixed gear system as well.
Extruders always need adjustments, even between different filament because the ratio changes depending on how hard a filament is and how far the teeth bite.

Skew adjustment is needed on both and is already recommended by me, we just call it squaring and leveling.

The error is just to small for a small cut to measure. The longer the distance the better, and poking a hole in tape or a dot with a pen is far better for the sake of accuracy.

Same as above. That is an easy test to see if you are perfect but very difficult to actually measure.

Marking two points is ideal and easier that cutting anything.

Those bring up a wonderful point. 99% of the time the issue is not going to be non-perfect belts. What I mean by that is I really do not want to see a bunch of step calibration suggestions for people cutting a 3 inch square and chasing zeros. This is a simple test and fix. If someone is getting a square with 100.539mm sides the answer is generally “bad cam”. Feeds and speeds are very difficult, step calibration is simple when it is set it does not change.

4 Likes
16

That’s like, 6000 full sheet machines. I’d love to see more builds, where are all those people hiding? :sweat_smile:

2 Likes
17

FB or not online at all :smile:

1 Like
18

The biggest fear I have w.r.t. step calibration is measurement. I would not mind at all having every belt be 1.998mm pitch and making that the standard. I just don’t trust myself to measure to 0.1% and I don’t want users to calibrate incorrectly and then have trouble with accuracy (and looking at CAM or bits or hold downs when the issue is the tape measure).

Best solutions (in order IMHO):

  • Getting the belts sourced that are to spec. Maybe occasionally sampling belts for spec. Maybe 2-5 samples per shipment? If that is easy enough to do. This won’t solve anything for self sourced kits.
  • Having a community average/knowledge/calibration for the belt spec. If we all agreed on 50.073, then that solves the issue to me. But not all belts are the same. Maybe a table with different manufacturers? That seems like it would be out of date immediately.
  • Having a known “yellow brick road” calibration step. No 100mm square cuts or letting people come up with their own calibration gcodes. We would have to prescribe an exact procedure with a reasily available and usable measurement device.
1 Like
19

For what it is worth… Here is another data point for you.

I just finished a full sheet LR4, but I used belts purchased from V1 about one year ago in December 2023.

I just measured movement of 1100 X, and 2100 Y and it looked spot on to me.

1 Like
20

I had two orders yesterday with 20M of belt each. Some people buy a lot of belt. I kinda wonder if they are mixing up meters and feet sometimes, and not seeing that it comes with belt already.

Yeah, On the drive over I think this is not actually horrible thing. Having this calibration step before the squaring step is a pretty good training step. Mark two points and measure them, change and save the config, save and reboot. Then squaring is a just a bit more involved.

I am just about to see if I can easily test some belts with the cheapie ruler, and hopefully it does not require anything more fancy.

Absolutely. Anything over 1000mm is pretty obvious (~1.5mm error), anything under (100mm at 0.15mm ) I am assuming we could have some issues with.
I am also confident this is so small that none of us really paid attention to it, knowing that millimeter accuracy over a full table is statistaccly noise in the tape measure.

My biggest fear, kinda like the tables, is everyone telling people to calibrate their XY steps anytime the 20mm hole is not exactly 20.000000mm. To be very clear, at anything less than 100mm this has almost no effect in anything other than metal, and in metal your feeds and speeds need to be freaking perfect to even see this.

3 Likes