How much wobble is acceptable on a DIY macro focusing rail?

Some drift is normal on a stacked shot, but the frames themselves should be dead still between captures. Ghosting on a beetle's edges ussually means the camera or subject actually moved relative to each other, or the step interval was too coarse and the alignment software had to stretch frames to match. Those two problems look similar in the final stack but have diferent fixes.

Easy test first: home the rig, take a frame, run the stepper back and forth, return to the exact same position, take another frame. Compare them pixel to pixel. If they differ, you've got backlash or slop in the drivetrain. Lead screws are notorious for this, especially if the nut has any play or the carriage lifts slightly when reversing direction. If you always approach each step from the same direction, that eliminates most of it.

The wooden board is a real suspect too. Wood moves with humidity and it damps vibration slowly. Tightening bolts helps but the board itself can flex. And yes, having the camera hanging off the end of the rail puts a twisting load on the whole assembly, which gets worse as the carriage travels. The classic fix is exactly what you guessed: lock the camera down solid to the bench or a seperate base, put the rig under the subject, and move the subject instead. The beetle weighs nothing, so the moving mass is tiny and the loads stay low. It's a rebuild, but it's the setup that stops this class of problem. On step size, don't guess. Look up the depth of field for your actual magnification and aperture, there are free calculators online where you plug in magnification and f-stop. Then step at roughly a quarter to a third of that depth of field so every frame overlaps heavily. At high magnification with a tube and telephoto that depth is tiny, which means lots of frames, but overlapping frames are what let the stacking software align cleanly instead of ghosting. Too-big steps with heavy overlap requirement is a common cause of exactly the edge ghosting you're describing.

I'd also check that the shutter isn't shaking the rig. Use a delay or electronic first curtain if your body allows it, and let the rig settle a second after each step before firing.

This whole thread is making me nervous because I built somethig similar out of a drawer slide and a bolt I turn by hand, and I photograph little mushrooms. I noticed my shots got sharper once I started waiting a few seconds after moving before I press the button, like the whole thing wobbles for a bit. I also put a heavy book on the board to weigh it down and that helped somehow. Do you think the settling time matters as much as how big the steps are?

Agree with moving the subject instead of the camera, that's what I did on my rig and the ghosting went away. I put the camera on a stiff column off the rail entirely, so the only thing that moves is a little clamp holding the specimen. Also found my flexible coupler between motor and screw was adding slop, swapped it for a rigid one and the back and forth test finally matched pixel for pixel.

I'm way newer to this than you guys but I ran into something similar on a little time-lapse project. Mine was sliding on a smooth shelf and I assumed the motion itself was the issue, but it turned out the cable to the camera was tugging slightly every time things moved. Unplugging it and using a short cord with slack made the wiggle go away almost entirely. Maybe check if any wires or the tether are pulling on your camera when the carriage shifts? Also a dumb question from me, does the ESP32 hold the stepper energized between frames or does it power down? I heard motors can relax a bit if you let them go loose, not sure if that applies here or if I made that up.

Worth checking your step timing too. On my rig I run the ESP32 to step, pause 3 full seconds, then trigger, and I learned that the hard way after chasing 'drift' that was actually the screw still settling. Also, if your stepper is doing microsteps, some of them don't have the torque to hold under load, so a full-step move then settle works better. What driver are you using? Some of the cheap ones run hot enough to warm the screw and everything attached to it, wich can creep things around during a long stack.