An orthographic 3D camera measures in the same pixels a sprite does: a model one unit wide is one pixel wide. That is what lets a 3D piece be told to stand on a tile instead of being nudged into place by eye.
Y points up in 3D and down in 2D, because every model is built that way and mirroring it would turn each one inside out. The bridge is one subtraction. Arrow keys scroll the board, and the piece stays on its tile.