A sine bar works because its two rollers sit at a precisely known centre distance, and tilting it by a slip gauge stack of height h gives the angle through \( \sin\theta = h/L \). For this to be accurate, the rollers need to be identical in size, round, and their axes must be exactly parallel to each other and to the top working surface, and the top surface itself must be flat. All of these directly affect how correctly the angle gets set. Whether the top surface is parallel to the bottom surface of the bar, however, plays no part in this calculation, the angle only depends on the rollers and the top surface, not on the bar's underside. So parallelism between the top and bottom surfaces is not something the accuracy of a sine bar actually depends on.