From what I understand the three plate method is when you are going from 0 to flat as the errors are averaged out.

Doing the "covered with dye and rubbed together to find the high points, which are then scraped off" thing is only if you already have a flat reference surface as you wouldn't have a way to know if the thing you're trying to make flat is really flat.

The real question is how do you get the first flat reference surface when all you have are a few somewhat flat things?

▲

KaiserPro 6 days ago | parent [-]

> The real question is how do you get the first flat reference surface when all you have are a few somewhat flat things?

my understanding is the threeplate method allows you to build the reference plate in the first place.

▲

bluGill 6 days ago | parent [-]

Right. You start with any two plates and make them flat with respect to each other. One will be convex and one concave of course, then you take one of those and your third plate and make those two flat with each other, the switch out again using the third plate and the one not swaped out. Keep repeating until all three are flat with respect with each other - only true flat will have all three flat and the repititions keep getting closer.

of course if you have a known flat surface you can save effort by making the new plate flate to the known flat.

▲

KaiserPro 5 days ago | parent [-]

> of course if you have a known flat surface you can save effort by making the new plate flate to the known flat.

All the effort!

one of the good things about granite flat references, is that they last for ages, so you can get them reasonably cheaply second hand, if you can find a second hand machine shop specialist.

	▲	bluGill 5 days ago \| parent [-]
		Even if your second hand plate is no longer flat enough it is still close enough so there is very little effort needed to bring it back to flat. (machine shops often check them on a regular basis and bring them back to flat as needed)