Hey! This thing is still a Work in Progress. Files, instructions, and other stuff might change!

# Advanced Delta Printer Calibration (tower position + individual diagonal rod)

by dolpin Mar 28, 2015

### Thing Apps Enabled

Amazing tool! Thanks a lot for that and kudos :) Was essential in helping me equalize the dimensions for perfectly square squares while also achieving almost mind-blowing (at least for my mind ) 10 micron precision of my delta.

Dear sir,
Thanks so much for this guideline. My home made delta is thanks to your clear instructions where I hoped the machine could be at.
The overall accuracy is within 0.1 mm field, the height of course is more precise 0.04 mm field.
The process capability is much higher than expected. (Cpk >1)
My build is D=300 and Height 434. Push rods 329mm. Magnets d=8x16mm. Balls are 14mm. Dual floating extruder, dual E3D nozle Controller = 32bit Lerdge S board + TFT screen.
Calibration done with nGEN material from Colofabb. 230 dgr and bed temp at 85dgr

Hello everyone,

I would like to know where you get all those data so accurate form the print?
how do i get that from the print? what tool do you use to do the measurement?
So far on the paper i can see 120 degrees on all axes and still my cube is
x, +2mm (fucking gap it's huge)
Y, +1mm
Z, 0 (correct) - i use BLtouch i think thats why...

I've printed out the small calibration object and while the boxes are 60mm they are only 4mm high. What could be causing this?

On the "Small Calibration" they are supposed to be 4mm high. I checked my layer count over layer height (0.1mm x37 + 0.3mm initial = 4.0mm and 38 layers).
While this is a great calibration model, the instructions are fairly generic and unclear...
For example, it says to change sin and cos angles in the firmware to match what you've calculated to... mine are listed as follows:

define SIN_60 0.8660254037844386
define COS_60 0.5

How exactly am I supposed to change sin(60) to sin(61) if those are variables to begin with? Am I supposed to calculate them out again? And since they are variables, how do I change just one tower? Manually place them in at each instance it calls for it on the one tower?

Just an update for anyone with that issue I had above
Calculate them by hand (as degrees NOT radian which is usually what the calculators default to, for google you can write out "cos(60 degrees)" and it'll output what you need) and replace the variable with that for the specified tower, there is only one instance of cos and sin for each tower.

Is there an alternative calibration for Marlin 1.1?

Dolpin, This thing and your instructions are outstanding for delta owners, Thank You!

How did you create this to be align to the 3 towers

Latest Marlin features a complete delta auto-calibration. Did you try it yet? It corrects for tower angels and rod length

I printed this out, and the angles are right on for me. But it's unclear to me from reading the instructions WHAT to measure for the size/dimensions for each tower with my caliper.

(6. measure the print sizes in each tower direction using a sliding caliper (60mm).

What am I measuring? The distance from both outside round parts of the "nubs", or the distance between the flat inside par of them? Either way I fall short of 60mm by 4-5mm, I just want to make sure I'm taking the correct measurements.

Thanks for this!

Hey people! I've been calibrating my printer with ESCHER3D calculator and I got rid of dome/bowl effect almost perfectly but with this test I figured out that there is 0.5 degree error in X and Y axis. What will happen if I now will follow the instructions in this tutorial, won't it mess my earlier calibration? Won't the bowl/dome effect appear again?

Doing this should really only further tune-in your current calibration. But always back-up your current settings just in case.

Very big THANK YOU! It is a great job and really useful!

i dont know actually

This is an awesome way to calibrate a delta. I am using smoothieware on a smoothieboard so it was a little bit more work to change the rod length for each tower but worked in the end.

Sonicspeed17 sent you a message, just incase I wanted to ask here as well. How were you able to change each rod length in Smoothieware? I have not been able to find that information anywhere and I need abit of tweaking on one tower.

avez vous la réponse à votre question ? je chercher à faire la correction dans smoothieware moi aussi.

merci

This is great. My X and Y towers were each of 0.5 degrees so I declared my own sine and cosine terms and now my prints are dead on. Thank you!

Fantastic! I had no idea how to get my printer to stop making everything a half millimetre too large until I found this guide. I had to iterate 4 times to get final values, but now I'm within 2 hundredth of a mm of the expected size. That's more than good enough for me.

I'm still working out how to tweak dome/bowl, but for now the objects I'm printing are small and it's not having much effect on them.

but i did just find
in configuration.h

# define DELTA_DIAGONAL_ROD_2 pow(DELTA_DIAGONAL_ROD,2)

The SQ function and pow(x,2) actually give the same result.

I based my procedure on an older version of Marlin, which is constantly evolving and changing.

For the later versions use a define: DELTA_DIAGONAL_ROD_TRIM_TOWER_1 / 2 / 3 to adjust each rod. These defines are usd in the marlin_main.cpp code.

Note that my procedure is not exact science but just an approximation. Nowadays better procedures are to be found on the internet, like: http://escher3d.com/pages/wizards/wizarddelta.php

i got delta printer
using marlin 1.6.1 version
i have calibrated angles successfully and now trying to calibrate the diagonal rods per instructions and formula.
i can not locate
float delta_diagonal_rod_2 = sq(delta_diagonal_rod); in marlin_main.cpp
or may be you could direct me in other way please.
only my Y tower prints larger dimension by 0.5mm more. thank you.
great job by the way!!!

I can't figure out why but this method doesn't work for me... I follow the instructions carefully but I don't have good results with dimensions. Here's what I do:

I Print the calibration object.
Angles are ok, so I check the dimensions:

I get:
x = 59.16
y = 59.05
z = 59.57

so I adjust my diagonal rod with the most accurate axis (z):
old_diagonal_rod = 196
new_diagonal_rod = sqroot((196^2) * (59.57 / 60) = 195.2964038

I adjust the measures with the gap of z axis, so:

I use marlin 1.1rc3 so I define in configuraition.h:

define DELTA_DIAGONAL_ROD = 195.2964038

define DELTA_DIAGONAL_ROD_TRIM_TOWER_1 (DELTA_DIAGONAL_ROD59.59/60 - DELTA_DIAGONAL_ROD)
define DELTA_DIAGONAL_ROD_TRIM_TOWER_2 (DELTA_DIAGONAL_ROD
59.48/60 - DELTA_DIAGONAL_ROD)
define DELTA_DIAGONAL_ROD_TRIM_TOWER_3 (DELTA_DIAGONAL_ROD*60/60 - DELTA_DIAGONAL_ROD)

I print a new calibration object expecting to obtain values really near 60 for all axes, instead I get:

x = 59.83
y = 59.59
z = 60.27

what am I doing wrong?

I've been having the same problems for a long time now trying to get my delta calibrated. I've gone as far as replacing my vertexes with milled aluminum ones instead of printed ones to eliminate frame issues.

For my printer, my measurements came out to:

x = 60.00
y = 60.18
z = 60.10
(multiple prints just to be sure)

My issue is that there is something like 3 different calculations going on in the instructions and comments on how to adjust the delta diagonal rod to calibrate all three towers. The original instructions were:
sqrt( (diagonal rod^2) * (print measurement/60) )
Using my Y tower as an example, that would calculate the Y diagonal rod at 300.45 - an adjustment of +0.45.

Then there i the further instructions using Marlin 1.1 RC2 to define the adjustments using
diagonal rod * (print measurement/60) - diagonal rod
Which would calculate the diagonal rod adjustment as +0.9, or a new diagonal rod of 300.9.

The newest formula from dolphin (thanks so much for your work getting this all assembled and your continuous help btw!) is
sqrt( print measurement^2 + diagonal rod^2 - 60^2 )
Which gives me 300.04, an adjustment of +0.04.

Now all three of these can't be correct. But here's the thing - none of them are work the way I expect. The best one, and the one that makes the most sense, is the last one (0.04). When I use that formula for my Y and Z towers, and enter in the adjustments for them in Marlin as DELTA_DIAGONAL_ROD_TRIM_TOWER_2 (0.04) and 3 (0.01) here's how the print measures:

x = 59.88
y = 60.01
z = 59.88
(again, multiple prints)

So yeah, y adjusts correctly, but somehow x drops by 0.12 without being adjusted, and z drops by 0.22, overshooting the goal by 0.12. If I try to then adjust x up, y increases and is no longer at 60. I've tried fresh firmware, switching motors, stiffer frame components, different electronics - at one point tried Repetier firmware instead - and I can't get all three towers to align. Either there is something wrong with my printer at an electronics/hardware/frame level that I can't find (prints flat, towers are straight, parallel to each other, as close to 90º from the print bed, belts are tight, motor drivers are adjusted), or Marlin isn't adjusting things in the way that I expect it to.

I'm willing to try just about anything at this point to get this printer calibrated. If anybody has something that might cause my delta to do this, let me know.

The more you came near the correct size, the worst the formula works.
Don't ask me why... I'm not a mathematician.
But be aware the fact that changing a tower value, it will reflect over other towers, because of the delta math design.
As an example, I've corrected x-2, y-1.8, z-1.4 to correct an x deviation of .56, and changed the radius by 3mm to compensate build plate convexity.

Indeed the corrections all influence eachother because of the delta algorithm. I tried to work it out but noticed that i just can't get it right. When calibrating the diagonal rod length for the object size, not only the deviation and original rod length should be taken into account, but indeed also the delat radius. For calibrating the individual rod length, it not only depends on the length of one rod, but al three influence eachother.

Ideally a spreadsheet would be great, were you insert the current defined rod and radius and the measurements from the calibration obect and will result in the new individual rod lengths. I tried to resolve this, but i am not able to.

I declare my calibration object to show the errors and deviations and give some advice on how this could be adjusted using a long iterative calibration process.

Last week it took me 6 calibration prints to get all dimensions in the directions of the towers within 0.1mm

The biggest one for me is the adjustment of the tower angle. This made my printer (btw mostly cnc milled) suddenly workable

I made a spreadsheet to do this, if anyone is interested I can share it.

dolpin - in reply to Gianky_

no clue, your calculation checks out.

To resolve this, first only focus on the delta_diagonal_rod size, because as you notice the z value is also not correct after the correction, and this value is only corrected by the delta_diagonal_rod value.

Gianky_ - in reply to dolpin

I have just rebuilt my diagonal rods cause the stock ones were of different sizes.
I built them using a jig and they are pretty close to the size they should.... 196mm.

I don't know where to start now to calibrate my delta...

Maybe the clue to calibrate my printer is in the step: "calibrate the print bed on each tower position by adjusting the z-height for each carriage".

I thought this means that the spider must be at the same distance from the three axes... Is this correct? Or should I put the three axis carriages at the same height without looking at the spider position? (Ok in a perfect printer l should obtain both the things simultaneously, in my printer it doesn't happen).

dolpin - in reply to dolpin

I am just recalculating the way the diagonal_rod values are corrected. Originally I used a different calculation, but I changed this to the calculations as given by others. I think this is indeed still not correct.

dolpin - in reply to dolpin

I really have to look over the whole instructions again to make it super correct. But I need to free up some time.

A few things that can be off on a delta printer

frame:

• print bed not level and square to the towers
• towers not vertical (leaning in/out/sideways)
• towers not positioned exactly at 120 degrees apart
• steps/mm for the carriages

platform:

• diagonal rod size (average and individual)

With my calibration object I try to correct

• the diagonal rod size, which alters the object size in general and in the directions of the towers individual
• the tower angular position, which results in printed object which are not for instance square in the x/y direction

What I would like to do:

• update the instructions for the latest marlin firmware
• check all the used calibration calculations
• create an overview off all possible errors in a delta printer, the effect of the error and how to correct
• add a calibration for individual delta_radius, this will eliminate the first layer thickness issues between the towers!

tips/help always welcome

I added a new picture to the object showing another way to calculate the diagonal rod size.

dolpin - in reply to dolpin

Your diagonal rod size should then be 195.87 according my new calculations (see pictures)

Gianky_ - in reply to dolpin

Ok I've seen your calculation. can't wait to try with the new diagonal rod value...
Thankyou!!!

I've tried with the new value but still no luck... I suspect that I have my print bed not level and square to the towers...

dolpin - in reply to Gianky_

ah ok, yes you may first check the frame/towers itself

Hey Just trying to calibrate... Not 100% sure i understand the way marlin rc3 calculates the Trim Towers value

The angles i have gotten are:
XZ : 120
YZ : 119
XY : 121

Does that mean the following is correct?
#define DELTA_DIAGONAL_ROD_TRIM_TOWER_1 (DELTA_DIAGONAL_ROD * 60/60 - DELTA_DIAGONAL_ROD)
#define DELTA_DIAGONAL_ROD_TRIM_TOWER_2 (DELTA_DIAGONAL_ROD * 59/60 - DELTA_DIAGONAL_ROD)
#define DELTA_DIAGONAL_ROD_TRIM_TOWER_3 (DELTA_DIAGONAL_ROD * 60/60 - DELTA_DIAGONAL_ROD)

i tried to re calculate each individual rod but the arduino ide is saying that they are not defined! please help

this is the code

// Diagonal rod squared

# define DELTA_DIAGONAL_ROD_2 pow(DELTA_DIAGONAL_ROD,2)

float delta_diagonal_rod_2_x= sq(delta_diagonal_rod(60.26/60));
float delta_diagonal_rod_2_z= sq(delta_diagonal_rod
(60.22/60));
float delta_diagonal_rod_2_y= sq(delta_diagonal_rod);

Hi,

I noticed you used different coding then as per instruction, which version of Marlin do you use?

Be aware that The C language is case sensitive, and the variables created can only be used in the same file when globally defined or in the same function when locally defined inside the function.

Hello bro!
I have developed also a calibration method based on the measurement of a printed object, and I would like to have some feedback, to check mistakes.

It worked wonderfully for me, I just applied some trigonometry and particularized the calculus for specific horizontal radious.

http://www.thingiverse.com/thing:1212012

Impressive work !

I am just printing it to calibrate my new GEEETECH G2S PRO. As my printer has two extruders (one at {-13,0} the other at {13,0} ) i was wondering if one print would be sufficient or if I had to print with each of the extruders ?

Hi,

Mechanically it does not matter what the position is on the platform, which is suspended between the rods. However the firmware will translate the platform position after setting the offset position for the extruders in the firmware. I am unsure what the effect of translating the extruder position in the firmware will be on the calibration.

To rule out the extruder offset translation you could set the extruder offset temporarily to 0,0 for the active printing extruder. The calibration object will just be printed slighly off center, but it is still usable as normal

Maybe someone can help me out with this. I'm following the directions for calibrating print size and adjusting the delta_diagonal_rod values. My Kossel mini has a small print bed currently and can't fit the full size print, so I scaled the print down to 75% - meaning the distance between the blocks should be 45mm.

After printing, my measurements were:
X: 44.95
Y: 44.95
Z: 44.82

Previous delta_diagonal_rod was 215.4. Calculated new diagonal_rod as 215.28. Made the changes in marlin_main to add delta_diagonal_rod_2_x, y, and z. X and Y were left as sq(delta_diagonal_rod). Z was entered as sq(delta_diagonal_rod*(44.87/45)). calculate_delta function was updated to use the new numbers.

After updating the firmware and re-printing, my measurements are now:
X: 45.00
Y: 44.99
Z: 44.88-44.90

Z is still a full 0.1mm off. I was under the assumption that after making the adjustment, they should all be much closer than that. Is there something i'm missing here?

Might be that your towers are not square to each other and the printbed. Maybe your z tower is slightly tilting inward/outward, or the printbed is not square to towers. Otherwise I have no real clue, perhaps you can re-calibrate again using the re-print

Thanks for the suggestions. I realized late last night that I had another thing working against me - my default delta_diagonal_rod was stored in EEPROM throwing off my calculations. I fixed that and ran a few more tests but still didn't get to where I wanted to be. I wasn't sure what version of Marlin I was running (it's been a while) so i updated to the newest RC and made some progress. Still not there yet and I'm running a few more tests, but as a warning to everyone who ends up here: make sure when you make changes to your firmware that your EEPROM isn't overriding those changes!

It's fantastic ! My printer thanks to your design prints round cylinder ! Today I am going to drink a beer to your health !

Works great. Also works in Marlin 1.1 RC2. When i get home i'll post the parameters that i adjusted. I have an error margin on all towers of +-0,03 mm measured on this test object.

EDIT:
in marlin_main.cpp find float delta_diagonal_rod_2_tower_1 = sq(delta_diagonal_rod + delta_diagonal_rod_trim_tower_1);

There you can see that the 3 rods are already configurated. Tower 1 = X, 2 =Y, 3=Z.

in my case, i edited it like this:
float delta_diagonal_rod_2_tower_1 = sq(delta_diagonal_rod(60.3/60) + delta_diagonal_rod_trim_tower_1);
float delta_diagonal_rod_2_tower_2 = sq(delta_diagonal_rod
(60.05/60) + delta_diagonal_rod_trim_tower_2);
float delta_diagonal_rod_2_tower_3 = sq(delta_diagonal_rod + delta_diagonal_rod_trim_tower_3);

Works a charm.

If I correctly understand from the Marlin code, in recent version (1.1.0-RC2) you can just define in Configuration.h:
#define DELTA_DIAGONAL_ROD_TRIM_TOWER_1 (DELTA_DIAGONAL_ROD*60.3/60 - DELTA_DIAGONAL_ROD)
#define DELTA_DIAGONAL_ROD_TRIM_TOWER_2 (DELTA_DIAGONAL_ROD*60.05/60 - DELTA_DIAGONAL_ROD)
#define DELTA_DIAGONAL_ROD_TRIM_TOWER_3 (DELTA_DIAGONAL_ROD*60/60 - DELTA_DIAGONAL_ROD) // Means zero :)

And no other firmware modification is needed anymore.

Also you can use "M665 A### B### C###" to modify these values on the fly.

Hello :)

I can't find theses define in the configuration.h file.
Does this mean I have to add them and not modify existing value ?

Thanks

Yes. These defines do not exist in configuration.h file by default. But they are used in the code. They are defined elsewhere as "0" if there is no definition in configuration.h file.
So, just define them in configuration.h file and these values be used.

dolpin - in reply to M4vrick

Hi,

These defines are found/to be added in the marlin_main.cpp file. Just read the instructions carefully :)

dolpin - in reply to GrAndAG

Interesting issue. When I replace the calculate_delta function with the code you provided, the machine refuses to move. It'll home, but that's all it will do.

I already have the code below in place of the function. As I say however, when replacing it with yours, it doesn't work!

delta[X_AXIS] = sqrt(DELTA_DIAGONAL_ROD_2

• sq(delta_tower1_x-cartesian[X_AXIS])
• sq(delta_tower1_y-cartesian[Y_AXIS])
) + cartesian[Z_AXIS];
delta[Y_AXIS] = sqrt(DELTA_DIAGONAL_ROD_2
• sq(delta_tower2_x-cartesian[X_AXIS])
• sq(delta_tower2_y-cartesian[Y_AXIS])
) + cartesian[Z_AXIS];
delta[Z_AXIS] = sqrt(DELTA_DIAGONAL_ROD_2
• sq(delta_tower3_x-cartesian[X_AXIS])
• sq(delta_tower3_y-cartesian[Y_AXIS])
) + cartesian[Z_AXIS];

The code you use is not the code from the instruction. The code is formatted wrongly by the webpage but I do notice that there i only one ( character after the sqrt function and the instruction shows two. Use only marlin v1.0.2 and follow the instructions carefully. There might be an error in my instructions as well but the code you posted is equal to the code in the instruction

When I run this through slic3r it tries to fill in the gap between X and Z. Just one layer. Rotating it fixes, but obviously this wont work.

dolpin - in reply to kprom

no clue why that is, i checked the design and did not notice that a segment of the outer circle was made out of segments instead of a circle segment. I solved this and uploaded the stl. All works fine for me with slic3r, the model is watertight and flat on the surface.

The skp design file is included, which you may use for findgin the issue for you

Hi friend.

Really great work. But as @kprom, when I run this through slic3r it tries to fill in the gap between XZ and XY. Just one layer #10.
https://imgur.com/a/yliAK

And your skp design dont help me, because i dont know how to edit sketchup in Linux.

I corrected this problem and uploaded the corrected STL: https://www.thingiverse.com/thing:3240091/
I just oppened the stl on soliworks and re-exported to stl. When I opened the solidworks identified a problem (by an error pop-up), but it allowed me to export it anyways. I think it corrected the error by itself.

Remix - Advanced Delta Printer Calibration

Works with Repetier as well and is really helpful! This is the solution for round circles! Thanks a lot!

Hey exactly do you mean by "Works with Repetier as well and is really helpful" ?

What works? What did you change? I use Repetier because of a diamond head.

Great, thank for the feedback

I tried this with great result. My circle are actually cirlces now.

There is one issue though. After doing this, my hot end get really low between the towers which is in front of the tower that I adjusted the diagonal rod length of.

I couldn't print anything big since it would crash into the bed.

Did you have this issue?

Good questions, i do not have this issue, but i only needed small corrections. I do need to work on the exact science about these corrections related to all possible errors in the printer geometry.

Did you do an accurate check on the angles and sizes after correction using the calibration object? Values should be within at least 0.5 degrees and 0.5mm otherwise you will not get an flat first layer as well.

Is your printbed flat? Also ensure that your printbed is square angled with the towers, you can use a big square angle to check.

I need to work out which kind of geometry errors there, how these are corrected using the calibration object and what result this gives. For example one rod pair is 5mm longer then the rest and corrected in the firmware. How does this effect the endstop position, height above the printbed etc ...

If you have eeprom enabled you can do all the calibration without recompiling or changing the code.

For the angles just change the Alpha A/B/C values like you would change the angle in Marlin formula.

For the diagonal rod length correction change the Corr. diagonal A/B/C values in EEPROM. Here you do not enter a factor like the instuctions computes but the difference between the target diagonal length and the one you have set. Just think that the value you enter gets added to the nominal diagonal length.

If all changes are in one direction, reduce them by the mean value and add that to the nominal value.

Thanks
I printed the tool and my tower angles are spot on so no corrections required. For X, y and Z size I had some variance.
The target is 60mm and My rod length is 213 currently.

X=60.39 mm
Y=59.49 mm
Z=59.23 mm

So for correction on Repetier firmware all I have to is enter

Corr. diagonal A = 213 (60.39/60) - 213 = 1.3845
Corr. diagonal B = 213
(59.49/60) - 213 = - 1.8105
Corr. diagonal C = 213 * (59.23/60) - 213 = - 2.7335

In the EEPROM settings and not modify DELTA_DIAGONAL_ROD ?

Exactly. Corrections get added for each axis to diagonal rod and since they differ it is not possible anyway.

I have a Rostock max v2, in repetier host I dont see a "Corr. diagonal a,b,c", but I see Alpa and "Delta Radius a,b,c". Is that the correct one?

If you are using Repetier-Firmware you should see them directly after Delta Radius params in eeprom. For other firmware I can not say what the names are, but radius is not diagonal.

Here's a screenshot of my eeprom. After Delta Radius param it shows Z prob. Why am I missing corr. diagonal?
http://i.imgur.com/9x8QGFK.png

I'm measuring 55-56mm on all sides so I definitely need to make these adjustments. Any advice would be greatly appreciated!

Ok, I see they are in deed missing in your case. Is this the original repetier-firmware or from a vendor where vendor has removed the output for these 3 values assuming he has no error? If I check this on my firmware it is between "delta radius C" and "z probe height".

Yes this is the link to the firmware directly from Seemecnc's website:
https://github.com/seemecnc/Repetier-091-ROSTOCKMAX

What are my options? Things print a little too small, I still get about 55-56mm for each tower. I tried shortening the "Diagonal Rod Length" but still got the same results.

Thanks for your help and replies!

Simple solution is to add the 3 lines for eeprom output to your firmware. See eeprom.cpp in you should find the first 3 lines with the next 3 lines missing:

I think this gives you the options back.
Harder method is to upgrade to current firmware which includes setting all parameter also you can find them in your configuration.h.

Hello people
I printed caldeltaV2.stl
my value is:
X=60.67 mm
Y=60.73 mm
Z=60.42 mm
Diag Rod is 310
For correction on Repetier firmware set this
Corr. diagonal A = 310 (60.67/60) - 310 = 3.461
Corr. diagonal B = 310 (60.73/60) - 310 = 3.771
Corr. diagonal C = 310 (60.42/60) - 310 = 2.17

My angle measured is:
XZ : 121
YZ : 119
XY : 120
in repetie firmware I put:

ALPHA A (210): 209
ALPHA B (330): 329
ALPHA C (90): 90

I think works fine, I tested now.