Strange non-linear movement of the bed (Y axis)

Hi everyone,

I've built a longer version of the RepRapPro's MonoMendel in the last view months (bed size 218x318) with a Duet v0.8.5 board, GT2 belts and an E3D v6 hotend. The machine has worked quite well so far, printing sets of smaller objects until recently. A few days ago, I printed a large circular object, about 180mm in diameter and couldn't help noticing that the Y dimension was a bit off. The ruler told me that the object was 180mm in X direction, but 188mm in Y direction.

So I attached a ruler to the bed, a small nail as a pointer to the hotend and started measuring to get my steps/mm setting right.
I succeeded in finding an M92 value that creates a perfect move from Y=0 to Y=300 when sending G1 Y300 after G1 Y0.

Strange thing is, when I use values in between, the hotend doesn't stop where I expect it to stop:
- G1 Y0 stops exactly at 0mm
- G1 Y0 followed by G1 Y100 stops at 97mm
- G1 Y0 followed by G1 Y200 stops at 201mm
- G1 Y0 followed by G1 Y300 stops exactly at 300mm
What's happening here?

I've already tried a few things, but without any success:
1) I replaced the Y stepper motor with a new one -> problem remains.
2) I swapped the axes on the Duet's connectors (along with the endstop connectors and adjustments of the config.g, bed.g and home*.g files), because the movement of the X axis is dead on -> problem stays with the original axis.
3) I've updated the Duet's firmware from DC42 fork version 1.14a to 1.17c -> problem remains.

The only possible reasons for this non-linear movement I can think of are
1) A mechanical issue, but I can't see how. The belt moves parallel to where it's fixed to the bed and the belt doesn't seem to be damaged or loose. The pulley has a firm grip on it.
2) The firmware doesn't command the correct number of steps when given a G1 code. Is it possible that I accidently triggered some maths for a delta printer which messes with my cartesian machine?

I hope someone's got an idea of what's going on here, because I'm running out of ideas.
I've attached two pictures of my calibration attempts and the config.g file.
Any hints would be very much appreciated.

Thanks in advance,
Tim

Are the belts parallel to the guide rails? Post a few pics of the underside of the bed so we can see the Y axis mechanism.

Edited 1 time(s). Last edit at 01/20/2017 06:22PM by the_digital_dentist.

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

Here are two pictures of the Y axis mechanism...

Did you try the same ruler test, starting from 300?
Do you have any binding with the linear bearings? Maybe a little binding compresses the frame a bit until the tension gets to strong and the frame relaxes. Can you clamp down the Y- frame to a flat plate?

The pictures don't show whether the Y axis belt is parallel to the guide rails. If the belt slopes upward from the end of axis pulleys to the attachment point on the carriage, you will have distorted motion because the angle and therefore the length of the belt between the attachment point and the pulleys will change. The attachment point on the carriage has to be at exactly the same height as both of the end of axis pulleys.

The top drawing shows the wrong way to do it and the bottom shows the right way. Ignore the XY designator in the drawing. Z is up, Y is horizontal. I have omitted the guide rails, but they should be parallel to the bed plate (the top rectangle in each drawing). If you leveled the bed then the bed plate is parallel to the guide rails (as parallel as it can be with end supported rails).

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

Check the number of teeth on the Y Axis Drive pulley Bet it's not the same as the X and therefore your Steps/mm for the Y will be off.

Doug

Thanks for your suggestions so far. I really appreciate it.

I'm including some images that should show that the belts are as close to parallel to the rails as I could plan it. I've actually gone to great lengths to make the machine as exact as I could manage (as you all do, I suppose).

I'm going to double check that the rails are aligned so that they don't disturb the bed's movement in any way. When I move it by hand (against the resistance of the powered down stepper), it doesn't feel like there is anything wrong. But you never know.
And I will try to tighten the belt a bit further.

I will let you know if there's any change after that.

So, here's a short update:

Today I got out the digital calipers and fine tuned the position/distance of the guide rails to make sure that the bed could travel without binding or bending the frame. I also tightened the belt a bit further.
The results of the next 0..100, 0..200, 0..300 tests with the ruler produced exactly the same outcome as before.

Getting somewhat desperate, I decided to find out if the Duet really didn't play tricks on me.
I put a microphone near the stepper motor and recorded the sounds it produced while I performed my tests one more time at very low speed (F100).
Looking at the waveform I could easily calculate the number of samples (= time) that the stepper motor had been active on each move.



The samples for the 0..100 and 0..200 tests deviated only about 0.05% from the expected values that I got from the 0..300 test (1/3 and 2/3 of that time respectively), which of course doesn't explain the 3% error I get with the 0..100 move.
That means that the Duet is no longer a suspect and that the issue's got to be a mechanical one.

So my next step will be the replacement of the belt with a new one. Maybe a long shot, but you never know...

Check you pulley is firm on the stepper motor

That fact that its hidden makes me think this issue is in there... (it's just asking for it to fail without being noticed)

If you have stepper with a flat on the shaft and the pulley grub screw is loose it can catch on the edge of the flat, so some moves are correct, but others (particularly but not limited to change in direction) will be out.

There was a thread about a delta which had weird positioning problems too ( leaning prints )
Turned out, it was the steel-reinforced belts.

You wrote, you've dialed in the Y-steps to match 300mm, would be interesting to compare this value with the theoretical one.
I wouldn't be surprised seeing a bigger error with tighter belts, if it is the frame acting as a compressed spring.
The y-stepper is hovering above ground and the brackets that extend the threaded rods look suspect to me.
This..and the diagonal rod that only works well in pull direction, but gets pushed ( bent? ) when the belt got tightened.

Edited 2 time(s). Last edit at 01/23/2017 03:19AM by o_lampe.

o_lampe, you were right. It was the belt after all.

I replaced it today with a new one and now every move stops exactly where it's supposed to.
Even the steps/mm value corresponds to the theoretical one (80).

Thanks everyone for your help!
Can't wait to see how the next prints turn out...

How could it be a belt? Was there anything obviously different about the original belt?

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

Well, there was no visible damage to the belt. But when I examined it after the replacement turned out to be working nicely, I found that I could stretch the belt in the area that had been near the pulley several millimeters by hand. It seems as if the reinforcing fibers inside the belt have broken there. It's just stretching on a length of about 20cm, the rest of the belt is fine.

Ah, so either it was poor quality belt (wouldn't surprise me at all) or maybe it got severely over tensioned (much less likely) at some point. It's good to know that sort of failure is possible.

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

Oh, good find. I hadn't even considered that possibility.

Greetings! I’ve registered just to say THANK YOU! I nearly have blown my brain to solve the problem. For the ones that can possibly get similar headache I will describe my issue too.
I had a problem, that was in non liniar behavior of x-axis on self made cube printer, based on Anet A6 motherboard. On the X-coordinates from 80 to 110 I had an increment of 1 mm per each 10 mm, measured it with caliper and mooved the extruder with printer menu. I’ve switched stepper motors, checked connections, bearings, shafts, gears, all seemed to be fine except this tricky increment. Within ukrainian internet segment there was no infirmation but prusa’s quick guides and useless recommendations like “have you checked power plug?” Nearly the same situation was in english i-net. At last I’ve found this thread, bought the new belt and what do you know? Everything fixed like magic! The issue was in belt with metall cord, the part that was on motor gear when the extruder was in 80-110 coordinate was rusty, the cord came out of plastic belt and rusted through, so after it started to pass the gear the belt began to strach and gave this freacking increment! Now I bought the belt with fiber cord, we’ll see how long it will last. Thank you once again!

Edited 1 time(s). Last edit at 12/29/2018 12:39PM by Arezus.

Steel core belts need large bending radius or they will fail this way.

I made a poor belt clamp design for a steel core belt and recently had a similar failure- the clamp just held the ends of the belt in a straight path.

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]