Is slicing a copy layer by layer a sane way to scan tiny objects?

It's a real technique, you haven't lost your mind. It gets used in materials labs and fossil prep work, where they grind a sample down a fraction of a millimeter, photograph the fresh surface, and repeat a few hundred times. The catch is that everythign lives or dies on the depth consistency and the alignment between slices, which is exactly where you spotted the problem. The wobble issue is solvable with registration marks. Before you start cutting, embed two or three small pins or drill tiny holes in the base well away from the actual model. Those marks show up in every photo and give your stacking software fixed points to line the slices up against. Without them, small sideways drift between shots hides itself til you view the model from the side, and then it's bands everywhere.

Depth is harder. A razor in hand will never be even enough, especally around the edges of a cut where the blade flexes. You want some kind of rigid reference plane the copy sits against, and you shave the surplus material off flush with that plane. Even then, check your first few slices against each other and measure them before committing to two hundred more passes.

Bigger worry for me: if you make the copy out of something soft, a blade will smear it. Wax and soft materials push into thier own fine crevices under a cut, which buries the very detail you're chasing. Cast your duplicate in something that cuts cleanly and holds an edge, like a rigid resin, and use a fresh sharp blade often. Blunt blades tear more than they shave.

One last thing worth racing before you build all this. Since it's a replica, nothing stops you from casting it in a matte material and going back to photogrammetry with a proper camera and a macro lens instead of the phone. Focus stacking gets you sharp detail through the whole depth of the crevices, no destruction needed, and it's a weekend of work rather than a rig you have to keep repeating for every future piece.

I actually tried something like this by accident once. I was carving a bar of soap for fun and kept snapping pics of the flat bottom as it got shorter, just to see what would happen. The stack looked okay from the top but the sides came out wavy, like the pictures were sliding around. Never thought to put little dots on there to line things up. One thing I'm curious about though, if you're cutting the copy down layer by layer, how do you get the sides of the pendant into the model? Seems like the photos only ever show the flat face you just sliced, so the shape between layers is mostly guesswork.

I did a rough version of this on a cast chess knight a while back and the slicing itself went fine, what nearly killed the whole stack was dust. Every cut left little crumbs sitting on the fresh surface and they read as fake detail once everything was stacked. A soft camera blower brush before each shot fixed it completly. Also photograph straight down with the same exposure every time, auto anything will drift on you over a couple hundred shots.

Total side thought here but I'm in a similar boat trying to scan a little wooden whistle I made, phone camera and all, so this thread is hitting close to home. Question for the OP: what are you planning to use to actually stack the slices back into a mesh? I've got a pile of cross section images from a failed project and no clue what software even accepts them as a stack. Also, if you end up casting a rigid copy anyway, would a silicone mould of the original let you pour a second backup before you start shaving the first one down? Seems like cheap insurance if slice ten goes sideways.

Update: it actually worked. Went slow with a sharp blade on a plasticine copy and the slices stacked up clean. The dust tip saved me, those crumbs whould have wrecked everything. Also clamped the whole rig to a board so the camera framing never drifted, which fixed the wavy sides problem someone mentioned. Depth consistency was the real battle but a spacer block sorted it. Still polishing the final model but the crevices came out sharp.