Inconsistently Stretched Belts?

Short story - my cuts don’t match my CAD. I’ll spell out what’s gotten me to this point.

I’ve had my X-Carve for a while and just love it. Done a bunch of signs. Everything came out great.

I upgraded to the new wide maker slide. Which is awesome. Made more signs with no problems.

Then I decided to get fancy and see how accurate my cuts were. I found that my CAD dimensions didn’t match my cuts. But I figured, no big deal. You just need to calibrate your X and Y axis.

I did this by inserting a V bit to act as a pointer into the spindle. Although I did the same for the Y axis, I’ll only describe the X-Axis here for time sake.

I positioned the bit over the 10 cm mark of my measuring tape. And then using UGS, moved the spindle 650 mm to the right. I don’t remember the result, but let’s say it was 660. I did the math, and reset my $100 to the appropriate step size. Ran this same process a couple of times until I got a consistent answer for my steps/mm. Then did the same for the Y.

$100 = 34.019 steps/mm
$101 = 37.229 steps/mm

Got very repeatable results.

So I cut a 1" diameter circle, and then a 1 x 2" rectangle. The media was placed in the lower left part of the table. I was off.

Circle was 0.885 in the X, and 0.931 in the Y. The rectangle was 1.733 in the X, and 0.937 in the Y.

Let me be clear. I’m not one of those guys who thinks he can buy a hobby machine and get 0.001" tolerances. But I’d like my cuts to be within 0.05" inches.

I figured I just needed to recalibrate. It had been a day or two since I had done the calibration. But my calibration came out dead on. Starting a 10 cm, issue a move X 650 mm, ended up at 660mm.

So I decided to move across the x-axis in 10 cm steps. Each 10 cm movement gave a different number of millimeters moved.

Moving 300 mm actually moved 350ish.

So here is my question. Is it possible to stretch segments of the belt so that your teeth pitch is not constant across the belt?

It’s pretty clear to me that since my steps per mm are 34.019 and 37.299 versus the 40 that an unstretched belt claims, I may have over-tightened my belts.

A few other points that folks may wonder about.

Yes, the machine is square.

V-Wheels are properly adjusted.

Yes, I have dialed in the POTs according to “the overload” method. Can’t remember where I read it. But it works.

Cut are repeatable. If I position my media in the exact spot every time, the parts I cut out are exactly the same size every time. If I position my media in another part of the table, then the cuts will be different from the other cuts. (If that makes sense.)

Easel, or Fusion 360 through UGS makes no difference. Results are the same.

Some folks have reported that they have issues where each progressive cut shifts a bit to the (pick a) direction. To me that indicates skipped steps. As far as I can tell, I’m not skipping any steps. Unless I’m skipping the same step whether the spindle is running around a path clockwise or counter-wise. Maybe I’m not saying that right either. Anyway, my vertical walls are perfectly smooth. No “ghosting.”

So what do you think? Am I missing something really simple here? Or have I pulled sections of my belts past their yield point? Z-axis - of course - is dead on. So I will be upgrading to ACME lead screws for the x and y in the future. But before I can do that, I’ll want to be able to accurately cut my new aluminum plates. :wink:

My next step is to remove the belts and compare that to a spare belt I have to see if I can spot areas that are excessively stretched. But I don’t want to do that until I hear some responses. That’s obviously a bit of disruptive work if there is an obliviously simple solution. That and I just can’t shake the idea that it would be impossible for me to stretch the same sections of the Y axis belts the exact same way. But maybe I did.

Looking forward to the response.

I don’t know of any way to unevenly stretch the belt. With so many teeth engaged on the pulley, even a little inconsistency should be pretty much corrected. You should be able to inspect your belt pretty thoroughly without removing it, I’d look for any obvious damage. Second thing I’d check is the pulley screws, it’s VERY common for them to loosen up and the pulley to slip just a little bit. If there’s a flat on the shaft, make sure one screw is centered in it. If not, you may consider making one with a file, saw, or dremel.

If your machine is calibrated under no-load, then any change while under load is an indication of lost steps. It may be belts (tight/loose), tight V-wheels, loose pulley screws, cutting too deep/fast, or any combination (plus more).

I can explain the circles pretty easily. Consider starting at 12 o’clock and moving clockwise. At first, your X stepper will get full power and Y gets none. Then as it moves around toward 1:30, the X power decreases while the Y power increases. At 3 o’clock, the Y power is at full power while the X power is at zero. If you have enough friction somewhere, then when your stepper is at a low enough power it may not be able to move and you may lose steps.

Having said that, the rectangle is a good place to start, and you should probably use something like MDF so you don’t have to worry about the effects of grain. I would recommend starting by loosening your V-wheels up until they’re too loose, and re-cutting the rectangle and circle. If there’s no change, then your V-wheels were good and you can set them back. If there was a big change, then your V-wheels were too tight. If you’re willing to bet money that your V-wheels are perfectly tensioned you can skip this step, but I’d suggest doing it just to be able to show that they’re good.

After checking your V-wheels, try loosening up the belts a little bit and re-cutting it, then measure. If it gets worse, tighten them up. Continue until you notice no change or until you are getting the size you need.

Let us know what you find!

After re-reading, I thought maybe I could be a little more descriptive.

With my “pointer” at 100 mm, I issue a move X 600 mm command. Pointer moves to 610 mm just like it should. I issue a move -X 600 mm command, point moves back to 100 mm just like it should.

Now back at 100 mm, I start issuing move X 10 mm commands. The first number is the actual reading. The number in parens () is what it “should” be reading.

108.5 (110.0)
118.0 (120.0)
117.0 (130.0)

Basically, every time I told it to move 10 mm to the right, it would move 8.5. So by the time it “should have” been to 600 mm, it was only to 525 mm.

I did this same thing for steps of 20, 50 and 100 mm.

The interesting thing is that all my numbers were very consistent. For 10 mm step, I actually got 8.5 mm.

For 20 mm step, I got an average of 16.98 mm.

For 50 mm step, I got an average of 42.56 mm.

And for 100 mm step, I got an average of 84.92 mm.

This seems to indicate that my belt isn’t stretched in some sections more than others.

If I break this down to actual mm traveled per expected mm of travel, I get the following ratios:

X 10 mm = 0.850 mm traveled / mm expected
X 20 mm = 0.849 mm traveled / mm expected
X 50 mm = 0.849 mm traveled / mm expected
X 100 mm = 0.849 mm traveled / mm expected
X 300 mm = 0.850 mm traveled / mm expected
X 600 mm = 1 mm traveled / mm expected

So what the heck!

Hey! Thanks for the reply @RobertA_Rieke!

Your calm suggestions are just what I needed. I’m going too far into the math weeds. I tend to over complicate things. I was getting such decent results from my signs, it would never have dawned on me to loosen anything up. I’ll try that tomorrow. But for now, after measuring all those stupid steps, it is time for an adult beverage.

1 Like

I wrote this earlier in an other thread:

The single belt on the X-axis is very much prone to localized stretching.
I was making a small series of throw levers for rifle scopes out of
POM-C and I was milling 4.7 mm holes in them to accommodate steel
threaded inserts and screws for mounting them. The piece (that was
milled from a sheet of POM-C) I was milling was mounted in a machine
vice. I milled a hole for the threaded insert, turned the piece and
milled a recess for the screw head. Very soon the the holes turned oval
in shape. They were as bad as 4.0x4.7mm.

Belts are a bad idea, at least when there is only one of them per axis.

Edit: added a photo to illustrate what a throw lever is:

Pick a place in the X-Y space and mill a hundred holes in that same place. There.

Gotcha, good point. For runs like that where there’s several hundred operations in the same point on each belt, it’s probably a good idea to (a) offset jobs in different areas to prevent localized wear and (b) keep extra belt on hand and replace on a regular basis.

I bet it would be possible to just install a ridiculously long piece of belt in such a way that you can feed it through without replacing the whole thing. That way you could just cut off the worn section (say, the first 12") and pull another 12" through…