For some reason there is no mention in any of the guides that I saw about calibrating the diagnal of the axis. For example your x, y and z can be perfectly calibrated but you can still be off in that your x and y are not completely perpendicular. You may not even realize it because all that happens is your prints are slightly off diagonally. I only noticed this after I designed a specific print to be used for a model.
Very possibly that this issue is specific to moving Y bed designs. Anyhow printing a cross and measuring the corners to be 90 degrees is important. To fix the problem you need to move the X or y axis so that it is perpendicular.
Actually a large L would be better because you would have a longer path to ensure that the paths are or are closer to 90 degrees.
The cross is not as accurate, that’s just true.
Or a great big square in which you can measure corner to corner, then with simple math applied to the differences in the two measurements, you can figure out how much to adjust one axis to make it sqr. It’s similar to how I used to square my milling machine w/ a 10’ x 20’ bed (yes those are ‘feet’ measurements!)
@jerryflyguy sounds like an awesome plane building machine or something but you can measure corner to corner with an L as well. That’s a right triangle.
XY gantry printers CAN have this problem, but it’s usually very obvious and easily fixed.
I thought about this before and maybe there could be a way in firmware to adjust Matrix for skew of the gantry, with some calibration inputs.
Was that researched before in Marlin for example?
@Ryan_Carlyle unless the problem is caused by the frame being skewed.
The bukobot build includes instructions for using a machinist’s square on the X/Y joint.
Another quick way to do it on Y-platform machines (if you’re sure that your platform edges are perpendicular to the Y rails) is to look at the machine from directly above and line up the back edge of the platform with the X rails. Doing it right can get it square to less than 0.1 degrees. Of course, the machine needs to have the necessary adjustment points to fix it.
This is why I moved away from the Wallace design, which not only didn’t have adjustments for this, but actually had heritable error. If the machine making the parts wasn’t square, it could print parts for another machine that would have the same error (though at least in the case of X/Y, careful orientation of the prints could cancel it out).
A machinist square is the fastest way to square up your diagonal. It will also show you if one side is not the same as the others, as in bent, long or short. Careful measuring from diagonal to diagonal is the best way.
@Mark_Rehorst Perhaps my biggest take away from the Wallace design was that if you use printed parts, you need to be able to adjust them to get proper alignment. This way, it is possible to get high-quality prints from low-quality parts.
I am basically a novice at 3D printing, only about a year under my belt. But this is what I have found so far.
Calibrating a 3D printer is a bit different than say calibrating my mill. Once my mill is calibrated it will stay in the ten thousands of an inch. A 3D printer oozes and mashes and changes with small environmental changes. And then if you try to remove it before it cools it can all go out the window!
So calibration seems to be a fleeting thing! LOL
So I get it close as I can then out comes the drill bits, files, belt grinder, torch, soldering iron or what ever is takes to get the part to fit right.
@Lob0426 That means your printer isn’t very good 
Or your filament isn’t very good, or both. You can easily 3d print to +/-0.1mm (0.004") repeatability on the same print done a month apart. Just takes good hardware and settings. I know that’s not as precise as a decent mill, but that’s more a function of the construction style and speed of motion than anything. Try milling at >200 ipm on a $400 mill and let me know how precise it is…
People tend to have fairly low standards for desktop 3D printing, but spend a few grand on a modern commercial-grade machine (not Stratasys, I mean like TAM or Fusion 3) with pre-made slicer profiles and they’re VERY precise.
Yep; Machinist square! and an accurate way to measure your diagonals.
For your printer it will be a matter of checking the frame for square and true. Then checking that the extruder is running true, plumb and center. Bed running centered and true to its axis. Bed level checked against the extruder. In that order!
Then time to run a print. Use the square again to see if the X printed line and Y printed line are being printed straight for the entire axis. Then print a circle and measure at least eight points and see if it is out of round. If lines are not straight and circles not round then you try to find out why. Many causes, like binding during travel, improper drive settings on the steppers, steps per mm set right. There is lots of time involved in making sure everything is built to design, after the building is complete. That is part of the challenge in building a kit or building from scratch.
Then it is tome to Calibrate it with one of the Calibration prints. Measuring the object then calculating the changes needed in steps per mm to get it as close as you can to the true sizes on all axes.
your values will differ from these but those are the calculations needed.
Current M92 value = 80.2 (value in firmware)
Desired movement = 70mm
Actual movement = 70.87mm (measured)
Here’s a formula you can use to solve for the New M92 value
New M92 value = Desired movement / Actual movement * Current M92 value
70/70.87 = .987724002 (New M92 value should be 98.77% of the Current M92 value)
80.2*.987724002 = 79.22 = New M92 value
The comments were about the printed printer parts (Whosa whatsis comment)! You get them to print as close as possible, by calibrating the printer you print them with, then you have to fix any out of round or out of square problems.
I have built a Printrbot Simple Makers kit, then upgraded it to a Play, so I have been through this myself. The play upgrade had printed parts So I know how that can go too! I have carried out a bed extensions on the maker and the play so had to reset it all again a couple of times.
+Ryan Carlyle My Printrbot Metal is very close, but +.004" means a loose fit and -.004" means dig out the drill bits. Or hammer the rod in and hope the print does not crack! I see around +.02mm (.0078") at 80mm/s. Slow it down and the accuracy gets better.
Most of the time this is close enough. We have to have realistic expectations as to what accuracy we can expect from any given printer design. And you can cut very accurately on a $400 mill but you are not going to cut at 200ipm. Spindle quality can be a problem. Backlash is the main problem as you go to cheaper mills. Backlash can be negated by proper operation habits.
My Sherline desktop CNC mill cuts a 2" circle to .0025" out of round time after time, If I do my part! It has backlash compensation.
I am surprised that there is no setting, that I have found, for backlash compensation in 3D printers. Z does not need it as it only moves in one direction during print. E the extruder does need it to control the flow of material. Belts do reduce backlash but inertia can have the same effect as backlash, especially if the belt is not tensioned properly or print speeds are excessive. One problem that the new Simple V2 is going to have less of. The TinyG controller it uses has controlled jerk rather than constant acceleration. It slows before changes in direction.
I know another long post!
that process is called squaring.
@Lob0426 there have been some of these types of calibration options, but they’re generally not used correctly and lead to more problems.
With your comment about getting out the drill bits, have you heard about polyholes? Remember that everything has to go through a format that doesn’t preserve curves. A little math at the design phase can make a lot of difference.
Then I am doing the math wrong! LOL.
Or asking for too much from my printer!
It is very close. But sometimes you need a better fit. I have the tools so I dig them out and make it right. If it is too far out then I make adjustments to the .stl.