The much delayed tale of my finally realized compensating bed mounts.

The much delayed tale of my finally realized compensating bed mounts.

Even though I’m philosophically opposed to using software and extra hardware to partially compensate for a machine’s varying out-of-squareness, I still like this a lot. Just for the collision detection.

There’s two parts to this: 1) is your axes not being square, which leads to slanted printed parts. Squareness is inherent to the printer design, triangular frames are particularly poor here, while boxes like the Ultimaker or Mendel 90 are very hard to not build square. Currently no software compensates for this, even though it would be trivival to do so.
2) is the bed surface not being parallel to its (relative) movement axes, which leads to your first layer ending up unevenly squeezed down. This is what auto- tramming (or -leveling) compensates, and it can do so without introducing any additional artifact, as the firmware basically rotates the gcode into the right spot. @Whosa_whatsis will disagree with me here, but my layers actually got more consistent after i set up auto-tilt-compensation due to the constant Z-movement. Kinda like you add noise to your signal to get more precise ADC readings through oversampling.
But staying relevant to the OP: The final result is absolutely batshit crazy overkill. I like it.

@Thomas_Sanladerer It’s entirely possible to get better results by adding noise, but only if your print quality is shitty beforehand. To extend your analogy, if you’re trying to get more precise analog readings of arbitrary values, oversampling with noise helps, but if you’re trying to read analog voltages that you KNOW are supposed to be in increments that match the resolution of your ADC, adding noise only serves to introduce the possibility of one of your reading being off by one. Retaining digital fidelity is better than trying to get a better analog->digital conversion of data that has already been converted from digital to analog.

Oh, and you’re wrong that there is no software that compensates for axes out of square. The Ormerod comes with a set of calibration objects to measure the squareness of the axes so that you can input those numbers to compensate in firmware.

@Whosa_whatsis i stand corrected on the squaring correction. Didn’t know the Ormerod was already doing it.
On the layer consistency: I used to occasionally see an irregular pattern of thicker and thinner layers (not just over- or underextruded), which i now think had something to do with the z-couplers or -nuts resting in an odd spot that didn’t translate the rotation into linear movement perfectly. As they are now never sitting idle, the chance of an entire layer sitting at the wrong height is pretty much completely eliminated. Or maybe the rods turned a bit harder in a couple of spots, so that the stepper would lag behind its electrical position for a layer or two.
On the electrical side, sure, always staying on full steps is the superior solution hands down, and in a perfect world, would get flawless results. Still, a constant movement can mechanically help the z-axis to achieve greater accuracy. I have yet to see an example where the auto-compensation visibly degrades print quality.

Also check by @Josef_Prusa

I have created a page so I don’t loose it.

Very cool, thanks! I would only request that we not call it “auto bed leveling” as I find that description to be very misleading. I prefer auto platform (or bed) compensation. I know it’s an uphill battle to convince people to change the convention, but I hope this convention is new enough that we can propose something more rational and descriptive.