New delta printer impossible to calibrate with firmware

Hello, I just built a second delta printer, It's similar to the large kossel. Unfortunately, the calibration process doesn't seem like it can be done.

I'm using marlin and using M666 to calibrate the endstops.The issue is that the Z axis is behaving irregularly.
I calibrated the endstops and I have tweaked the values in marlin so that the center of the printer was the exact same
height as 3 positions for tower calibration ( X-95 Y-55, X95 Y-55, and X0 Y110). You would think this would indicate the printer isn't
concave nor convex but that isn't the case. When I moved the print head left and right (along the X axis in printing coordinates not the X tower)
it was concave. I then adjusted the settings to make this movement flat. However, the movement up and down (y axis in printing coordinates and along the Z tower)
was convex. not only is this movement convex, the highest position in it's arc is halfway between the center and the calibrated Z position. The degree in which the up and down
movement is convex gets exaggerated the more convex the left and right is made.

exaggerated illustration of left and right movement: [i.imgur.com]
exaggeration of the up and down movement: [i.imgur.com]

I believe the endstop calibration points are even because due to the left and right being concave, it cancels out with the Z being convex.
It's also because the convex shape along the the Z axis is centered halfway between the middle and the Z calibration point.

In marlin you can't adjust individual rod lengths so in my mind the issue must be that something is physically wrong the the printer, the Z axis specifically.
However, I have measured my printer rod lengths multiple times and everything appears identical. I don't believe this issue can be fixed with software
and I'm not sure how I can adjust the printer physically (maybe the arms are to long on the Z axis?).

Any thoughts or advice would be great, thanks in advanced.

Check your steps/mm value, and also measure your arm length between ball centres. One of them is out. I've seen this effect many times, when the wrong values are used.

Get the latest up to date Marlin firmware from GitHub

i have tried like to to tune manually my kossel
the result will be never perfect without advanced auto tune
so you should add a micro switch to make calibration

I see no reason why manual calibration using escher 3d wizard would not give near perfect calibration provided your firmware can adjust enough parameters it's just slow. However with marlin the number of parameters you can adjust, just isn't enough to get a slightly imperfect (almost all) machine to 0.00x deviation. So you can make it mechanically perfect if you want to drive yourself mad. Try Rich Cattels marlin if you aren't interested in in duet/RRF which has 9 factor autocalibration.

---

Simon Khoury

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

i have bought a smoothieboard to test auto tune..it s easier way than manuel(i have spent 2days on manuel and nothing was perfect..i used esch3D and same thing!!with smoothie you have an advanced auto calibration.i am waiting my z probe to test it )

Yes I can see that but my point is that if you look at what duet/RRF does for autocalibration it is just filling in the escher3d calculator with height values automatically so it's quicker but it isn't better. I haven't tried smoothie on a delta yet.

---

Simon Khoury

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

perhaps but sensors is more accurate than human ..mine for example is not perfect on some point of my bed!!so with auto tune it will be more accurate and i will not have any problem anymore(i hope

Good luck, sensors often are easier but not necessarily more accurate, I sometimes check the sensor by manually calibrating.

---

Simon Khoury

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

Quote
DjDemonD
However with marlin the number of parameters you can adjust, just isn't enough...

- delta height
- ABC end-stops
- delta radius
- ABC radius trim
- ABC tower angle trim
- diagonal rod
- ABC rod trim
- ABC steps/unit
- Z-offset

that's 19 parameters to adjust in Marlin! How much more do you really need?

Or give the delta auto-calibration in standard Marlin a try...

Well these must have come in with 1.1 as rc8 was the last one I used with a delta and it was not possible to adjust tower angle only per rod length trim plus the basics. The autocalibration I used then was just endstops.

---

Simon Khoury

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

Make sure your parallel rods are the exact same distance from each other from top to bottom, this is critical. Then do the following..

Quote
ohfurryone
- Make sure your end-stops are fairly accurate by measuring from the base of the printer.
- Place a piece of masking tape on one side of your bed along the X axis from center to edge.
- Lower nozzle to the bed and scribble around it with a pencil, raise and find center of circle left behind.
- Find an accurate ruler, square, fabric measuring tape ect. and mark out 20, 40, 60 and 80mm from center to edge.
- Move the nozzle along the bed on the X axis and see if it scales to 60-80mm properly, you should find it will be out by 1/2mm or so.

Change Smooth_Delta_Rod_Offset up and down to compensate. You eventually want to adjust it in .25mm intervals, until the nozzle ends up bang on at 80mm. The reason why you need multiple measurements is if one is 1/2mm wrong, it will stick out and you can re-mark it.

Once you have it close enough you can adjust the value in .1mm amounts.

Quote
DjDemonD
Good luck, sensors often are easier but not necessarily more accurate, I sometimes check the sensor by manually calibrating.

personally i have tried without sensor and no good results
if you keep an eye about my youtube video you will see that i have used esch3D website to tune my delta
but it s not perfect on all corners after spending so much time...smoothieware do it without no brain storming if you add a microswitch which cost less than 4usd(for a good one..)
so why we shouldn t use them?it s really cheap actually and will be more easy to tune
now i am agree with you we need to check some points like same rod lenght,towers are equal distance etc etc

Quote
titeuf007

Quote
DjDemonD
Good luck, sensors often are easier but not necessarily more accurate, I sometimes check the sensor by manually calibrating.

personally i have tried without sensor and no good results
if you keep an eye about my youtube video you will see that i have used esch3D website to tune my delta
but it s not perfect on all corners after spending so much time...smoothieware do it without no brain storming if you add a microswitch which cost less than 4usd(for a good one..)
so why we shouldn t use them?it s really cheap actually and will be more easy to tune
now i am agree with you we need to check some points like same rod lenght,towers are equal distance etc etc

you a smoothieware promoter lol, your banging on about it on this post and my post look end of the story there are many firmwares that do calibrations with a probe, and as you have said you need to measure but also need your machine to be square

Edited 1 time(s). Last edit at 05/29/2017 10:25AM by chris33.

---

Check my rubbish blog for my prusa i3

up and running
[3dimetech.blogspot.co.uk]