Auto Bed Leveling Solutions

[paradiddle65]

Hello Everyone,

I'm in the process of designing my own Reprap, and I've been pondering over whether or not implementing auto bed leveling is worth it. The idea of having an auto leveling feature is amazing, consistent print quality and bed adhesion, less preprint rituals, etc. In practice its another story. I tried auto leveling when it was implemented into Marlin a few updates ago, and it just didn't work for me. It was a manual process using paper as a spacer and frankly took much longer than just adjusting the Z-axis rods on the fly.

With this new design in the works, I thought I'd give it another shot. After some research I've come across some interesting ideas (some better and more practical than others). I wanted to ask the community if you had any experience with auto bed leveling and what your opinions are.

Are you currently using an auto bed leveling feature? Which one and what's your overall opinion of it? What's your opinion of some of the solutions out there (Commercial and Reprap)?

Thanks for your help,

Brian

[Roxy]

If you tried the 3-Point Bed Leveling that was the first version of Auto Bed Leveling. It was very useful to prove the concept but it had some weaknesses. In particular, any measurement error in the 3 points got magnified as you went out in the direction of the unmeasured corner. If you haven't tried the n x n grid based Auto Leveling, you will want to retry that. It works very well.

[paradiddle65]

If you tried the 3-Point Bed Leveling that was the first version of Auto Bed Leveling. It was very useful to prove the concept but it had some weaknesses. In particular, any measurement error in the 3 points got magnified as you went out in the direction of the unmeasured corner. If you haven't tried the n x n grid based Auto Leveling, you will want to retry that. It works very well.

Thanks for the tip! I'm guessing this method uses an endstop to detect the print bed?

[Roxy]

Thanks for the tip! I'm guessing this method uses an endstop to detect the print bed?

The typical implementation uses a servo to kick down a probe leg with a switch mounted on it. With that down, measurements are taken. But people are going in all kinds of new directions. Some have sensors built into the nozzle so they can tell when that touches. Others have wired up the metal nozzle and if they have a metal bed can detect when it touches. Etc. Etc.

[maar1201]

Hi,

Is possible to change Marlin to do manual bed leveling? I don't want to use the servo.

My ideia is to acouple to x-carriage a switch and using G29 measure many points on the bed. This give-me the equation coefficients and the matrix.

I need help to find a way to store this information on configuration.h and use it in normal printing doing axis Z compensate the lite diferences. Can someone tell me what code on marlin_main.cpp i have to change?

Thanks

[Roxy]

The Enhanced G29 code will print out a topography map of your bed (normalized to the mean height). That is very helpful for manually leveling the bed. But other than that, you could connect up the probe with no way to retract or engage it... And just let the G29 do its thing. (and print out the desired information).

[maar1201]

Thanks for the reply.

So, i have the topography map of bed collected using fixed switch near extrude nozzle which i know the offset.

Now, i want to change Marlin modules (i don't know where to start) in order to use the topographic map created by The Enhanced G29 previous stored.

Thanks for your help,
Mario

[Roxy]

Marlin doesn't store the topographical map. What it does is it uses those data points to generate a 'Bed Level Correction Matrix'. And when it needs to map a coordinate from the GCode to the 'corrected' coordinate system, it just does a matrix multiply of the coordinate and the correction matrix.

What I think you want to do is expand the M500, M501 and M502 commands to save, restore and load the Bed Level Correction Matrix from EEPROM.

[maar1201]

Yes, expanding the Gcodes M500, M501 and M502 commands and save, restore and load the Bed Level Correction Matrix from EEPROM are a solution to me, but how to do this?

Can you give me some clues on Marlin code to find the function where as You said: "it just does a matrix multiply of the coordinate and the correction matrix."

Thanks

[Roxy]

Yes, expanding the Gcodes M500, M501 and M502 commands and save, restore and load the Bed Level Correction Matrix from EEPROM are a solution to me, but how to do this?

In configurationstore.cpp there are 3 functions: void Config_StoreSettings(), void Config_PrintSettings(), void Config_RetrieveSettings()

In the store function... I would change:

EEPROM_WRITE_VAR(i,zprobe_zoffset);

to

EEPROM_WRITE_VAR(i,zprobe_zoffset);

EEPROM_WRITE_VAR(i,plan_bed_level_matrix[0][0]);
EEPROM_WRITE_VAR(i,plan_bed_level_matrix[0][1]);
EEPROM_WRITE_VAR(i,plan_bed_level_matrix[0][2]);

EEPROM_WRITE_VAR(i, plan_bed_level_matrix[1][0] );
EEPROM_WRITE_VAR(i, plan_bed_level_matrix[1][1] );
EEPROM_WRITE_VAR(i, plan_bed_level_matrix[1][2]);

EEPROM_WRITE_VAR(i, plan_bed_level_matrix[2][0] );
EEPROM_WRITE_VAR(i, plan_bed_level_matrix[2][1] );
EEPROM_WRITE_VAR(i, plan_bed_level_matrix[2][2] );

Corresponding changes would need to be made to the Config and Retrieve functions.

Can you give me some clues on Marlin code to find the function where as You said: "it just does a matrix multiply of the coordinate and the correction matrix."

Thanks

The matrix name with the correction matrix is called: plan_bed_level_matrix

You can find references to it in Marlin_main.cpp, Planner.cpp and Planner.h The Planner code is where Marlin 'Plans' its moves. That is where the coordinates are actually 'corrected'. It does this by doing: apply_rotation_xyz(plan_bed_level_matrix, x, y, z);

This function is defined in vector_3.cpp It looks like this:


void vector_3::apply_rotation(matrix_3x3 matrix)
{
float resultX = x * matrix.matrix[3*0+0] + y * matrix.matrix[3*1+0] + z * matrix.matrix[3*2+0];
float resultY = x * matrix.matrix[3*0+1] + y * matrix.matrix[3*1+1] + z * matrix.matrix[3*2+1];
float resultZ = x * matrix.matrix[3*0+2] + y * matrix.matrix[3*1+2] + z * matrix.matrix[3*2+2];

x = resultX;
y = resultY;
z = resultZ;
}

That would be a [1x3] matrix multiplied by a [3x3] matrix. [x,y,z] * [bed level correction matrix]