Automatic bed leveling issue—Bad rods?

Hello everyone,

I am not 100% sure that this is the right sun forum, since it's possibly mechanical or possibly some sort of settings issue on my end. Apologise, if this is the wrong place.

I am having very uneven and difficult results with auto bed leveling on my printer. I'm new to Marlin, in general, so I'm not sure exactly what I need to do to correct for it.

I've included some screen shots showing what my G29 and M48 returns typically look like, as well as a shot of how extrusion looks, after setting the mesh. I do not issue a G28 after the G29.

If I roll my rods, the results do change, and I know that no rods are ever perfect anyway. However, given that there is a pretty sharp variance between some readings, I'm assuming something is up. If I manually level the corners using paper, the center regions of the bed are far too close, which suggests that it's not just something wonky with Marlin.

Am I missing anything here, or should I just wait for new rods?

(Marlin 1.1.3 RC7)

I'm using Marlin 1.1.3 RC7 on a FlSun Cube printer, with an inductive probe on the Z axis. Mesh leveling is bilinear with 5*5 points.

(Please ignore the round scars from past prints, just the square is of interest here.)

Pictures can be seen here:
[goo.gl]

I'd hazard a guess its the rods deflecting, if you have the central "hill" in the bed, this is caused by the rods bending in the centre.

No real way around this without changing some things, but using mesh compensation (presuming you can taper it off at a set z-height so you don't get wonky parts all the way up) is a valid way around this issue. I've seen it on my old I3 and my corexy, until I switched to linear rails, now its gone. Using really big rods would probably help but the extra cost and mass would not make this a better solution than rails. Cheap chinese rails work fine I have some 9 of them all working perfectly, but search the forum for instructions on how to make them run smoothly, or spend some money on Japanes/German ones.

Substantially lightening your print head (consider flexible cable driven extruders, bowden systems - in that order, replacing heavy sensors such as industrial induction proximity sensors with something lighter, moving to tiny fans - all this helps).

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

I appreciate the help, DJ. Is there a reason that my bilinear mesh would be so far off? As you can see in the first picture, it's clearly not anywhere close to correct.

It has a MK8/V5 clone hotend with a 30mm clip on (V6 style) cooler and a fairly light inductive probe. It can be lighter, for sure, but it's currently Tyler much more nimble than the MK10 on my Wanhao Duplicator I3.

I'd consider manually measuring the 5x5 grid points see if they match your sensors output or set marlin to do manual mesh then you can jog up and down at each point and set it using paper.

If that matches your auto mesh values then your sensor is accurate it's your printer that's sagging. A mk8 etc... Is all well and good but I'm taking about seriously lightweight print heads, my deltas effector weight is 185g total, your extruder motor probably weighs 395g on its own. This sounds extreme and it is, but 8mm rods deflect,under their own weight let alone with 600+g print head resting on the middle.

Try a deflection calculator you need the length of the rod and the mass you want to support on it. You'll be surprised how much a typical 500mm 8mm rod deflects. They really aren't a good choice for either linear motion (lm8uu bearings are basically junk) or for rigidity, but they're cheap.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

I figured out the issue. My probe X and Y offsets were incorrect in firmware. I did a more accurate measurement set, updated the firmware, and am now setting correct perimeters.

I feel like a bit of a goober, but at least I'm printing now!

Well thats good news, I bet there's still a hill in the middle of the bed though?

To be fair x and y probe offsets are inherently confusing, is it the offset of the nozzle from the probe or the probe from the nozzle?

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

I ended up drawing a dot on my bed and putting the nozzle over it. I wrote down the coordinates and then moved the probe over the dot, and subtracted the coordinates. I feel like my Z height isn't spot on, but it's close enough to work.

Whatever was causing the high (or low, depending on you're frame of reference) spot is still present, but I'll not worry too much about it until the new rods arrive.

The carriage weight is pretty much fixed, until I pull the trigger on a V6 hotend. Even then, I'm not sure I'll save that much. I will admit that I'm eyeballing the Nimble as a way to bypass the Bowden setup all together. It's far more challenging to dial in than my direct set-up.

Cable driven extruder is definitely preferable to bowden.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Would you explain why?

Yes. Filament does not take to being pushed through a tube well.

Let's imagine a perfect case. Your bowden tube has just 0.01mm larger ID than your filament, which is perfectly manufactured to 0.01mm tolerance, not oval in places perfectly round and completely even. Your tube is made from a material with zero friction, not low friction like PTFE, zero. Your filament is inelastic.

In this example bowden works as well as direct.

We all know though that filament diameter varies by 0.03mm if it's good and more if it not. As such bowden tubes for 1.75mm are 2mm ID, this allows the filament to spiral around inside the tube, which is not zero friction, as such if it's long enough the friction increases. Filament is elastic, maybe less for pla than tpu but elastic nonetheless. So a certain amount of pressure at the extruder has to build up before the filament begins to extrude at the hotend. You can see the less accurate results if you place a direct extruding machine next to a bowden machine and print the same object, assuming all other variables are constant. Can you get away with 1mm retraction on a 70cm bowden tube? No you need 5-6 mm due to the combination of elasticity in the filament and apparent elasticity caused by the loose fit of the filament in the tube. This takes more time and unless tuned very well results in less accurate extrusion.

Now as a compromise bowden was a reasonable one, for deltas (and corexys/others) where a lightweight printhead was desirable/necessary bowden was preferable l to putting an extruder with motor, on the effector/head. Even highly geared lightweight motors have failed to achieve acceptable extrusion, take e3d titan and 20mm pancake nema 17 stepper probably the lightest combination that works well, it's still 280g.

There are enhancements for bowden, pressure advance helps if you tune it right, making the tube as short as possible makes a big difference, situate the extruder better to achieve this or even suspend it on elastic above the head (flying extruder) making the tube just 20cm instead of 60-70cm. Use 1.8mm ID tube (as long as you use expensive, precisely made filament). These are all ways of making an imperfect system less imperfect.

If you take a flexible cable extruder you can have any motor you want driving it as it's stationary, it only weighs 28-45g on the effector/head. It behaves like a direct extruder on the head but with 10% of the mass meaning high speed/high acceleration is much more Achievable. But if you print slow and are happy to do this, just put an extruder and motor on the head, the only compromise is you can't fling the head about as quickly.

TL/DR: Bowden tubes used to push filament, are an imperfect solution, for many reasons, used to achieve lightweight print heads. With appropriate finesse they can work almost as well as direct extruders. Cable driven extruders achieve direct extrusion without having the mass of the motor on the print head with higher cost and slightly slower retractions the drawbacks.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

There are many people using this printer therefore if sagging was an issue it would be more common complaint. Did you check flatness of your bed?

PS.: I would just get new rods and drylin bearings. This kit is known for using low quality linear rods.

Edited 1 time(s). Last edit at 08/29/2017 09:53PM by newbob.