Finally got my globoid worm to mesh without interference

Mine took four tries before it stopped chewing up the rollers. I was building an hourglass worm for a small telescope tracker, and the problem looked exactly like yours at first, the throat area always grabbed the center pin no matter what I did with the profile.

What finally fixed it for me was going the other direction with the order of operations. I drew the roller positions on the wheel first, then used one of those pins as a reference to build the worm groove as a variable sweep along the curved path. I also kept the wrap short of 180 degrees because my wheel only has eleven pins and going further starved the entry angle, but I can see why more engagement would help with a denser wheel. The regeneration pain is real too. My file had something like sixty sweep segments and every small edit cost me a couple minutes of waiting. What helped was splitting the worm into three shorter bodies and only editing the middle one during tuning, then mirroring or patterning once the geometry looked right. File size dropped a good chunk when I removed the construction curves afterward.

Did you model the clearance as a single offset on the roller radius, or did you vary it along the wrap? I found the entry and exit pins needed a touch more room than the middle ones or they dragged on the flank.

My experiance with a similar setup taught me not to trust small clearances at all when the parts are coming off a consumer printer. I built a globoid drive for a slow pan mount on my deck camera, ran it in FreeCAD, and set the roller gap tight like everyone says to. The model looked perfect in the assembly view, then the first PETG print bound up on the middle rollers anyway becuase the throat came out a couple tenths fat from blobbing on the curves. Now I print the worm with a deliberate extra gap, assemble it with the brass rod rollers on my wheel, and check where it drags by putting a thin smear of marker ink on the pins and turning it by hand. Where the ink scrubs off is where material needs to come away. A few minutes with a needle file in the throat and it runs smooth. Not elegant, but it saved me from chasing perfect geometry that my printer was never going to give me anyway. On the slow regeneration thing, I stopped fighting it by keeping a second file with just the pitch surfaces and roller positions for checking interference, and only opening the fully detailed one when I am ready to slice. The detailed model on my old laptop takes long enough to regenerate that I would avoid touching it, which meant errors sat hidden for days.

Takes me back to when we'd lay these out on paper first, full size wrap drawn with a beam compass, then transfer the roller arcs to cardboard templates before ever touching the CAD. Saved me from exactly the throat interference you're describing more than once. These days I still rough the wrap geometry in 2D before committing to a sweep, old habit but it catches the pinch points early. One thing that helped me on a similar drive: print just the middle section of the worm as a short test chunk before committing to the full six weeks. A stubby section with two or three rollers engaged will tell you if the clearance is right for a fraction of the filament and time. Cheap insurance when regeneration is that slow.

I'd push back on the variable clearance idea. On my globoid drive for a camera slider I tried tapering the offset toward entry and exit like you describe, and all it did was give me a groove that wandered off the pitch surface near the ends. The rollers that were supposed to benefit ended up riding the edge of the cut. Went back to one flat offset, sized for the fattest roller I measured across, and the drag went away on its own. My guess is the entry pins only seem tight because the print squishes the first layer of the throat, not because the geometry needs more room there. Did you measure your pins before assuming the flanks were the problem?

One trick that saved me a lot of guesswork on my globoid attempt was checking the actual contact patch instead of trusting the assembly view. I smeared a thin layer of modeling clay on a couple of the rollers, rotated the worm through by hand, and the footprint showed me right away where the groove was riding off center. Looked fine on screen, was clearly tilted on the clay. Ended up shimming the worm housing maybe half a millimeter and the footprint evened out. Did you check your center distance under load at all, or just at rest? Mine shifted a bit once the mount flexed.