Originally Posted by
Patrick Chase
To do this by clamping you have to ensure that the two flat steel plates (I use Starrett's precision-ground O1) will apply pressure exactly parallel to each other and to the sawplate. You can do that with a good machinist's vise, and you can also accomplish it by floating one of the plates so that it registers to the sawplate and thereby to the other plate in a common mechanic's vise. The "floating plate" approach is preferable for saws with tapered plates even if you do have a parallel vise, because then you want the clamp plate to follow the taper.
You then find a sheet of paper, the thickness of which is roughly equal to (or slightly larger than) the desired final pre-side set. Fold the paper over the teeth such that it covers both sides, insert that package into your vise, and clamp away. It's remarkably accurate, and is based on the observation that the tooth points apply enough local pressure to pierce/compress the paper and therefore register to the clamping plates, while the sawplates apply much less pressure and register to the ~full thickness of the paper.
The clamping plates don't need to be harder than the saw plate, because you're just bending weakly suuported cantilevered beams (the teeth) as opposed to actually indenting material as in a hardness test.
EDIT: The big advantage of the clamping approach is that it doesn't add another facet to the tooth. Stoning does, since it creates a side-facet ~parallel to the plate at the full set distance. While avoding such side-facet is generally held to be a plus on Western saws, I don't know about Japanese ones.