Nine prints to a good part
The cone light modifier took nine prints. This is normal and nobody says so. CAD screenshots and glamour shots of v-final make it look like parts spring fully formed from Fusion 360; the actual process is a stack of eight almost-right cones by the bin.
Here’s what each one taught, in case your part is a cousin of mine.
v1–v3: tolerance is an empirical science
The throat needed to grip a flash head by friction. I modeled the head, offset by a “reasonable” 0.2 mm, and printed a part that fit like a sticker — because printed holes shrink and printed posts swell, and the amounts depend on your printer, material, and wall temperature.
The fix was printing a tolerance ladder: one test part with five openings stepped 0.1 mm apart, printed in the same orientation as the real part. Twenty minutes of printing replaced three full-part guesses. 0.4 mm interference was the answer for my PLA and my printer. Yours will differ; that’s the point.
v4–v6: surfaces are optics
A glossy white interior sounds reflective in a good way. It isn’t — layer lines act as concentric mirrors and tunnel a hotspot straight down the axis. Test photos of a chrome spoon (harshest possible subject) showed a bright core with a dim halo.
Two changes fixed it: an S-curve flare profile instead of a straight taper, and sanding the interior to 400 grit. Matte is not a look, it’s a scattering function.
v7–v8: thermal budgets exist
A speedlight at full power dumps real heat into an enclosed cone. v7 smelled like regret after twenty flashes. Vent slots at the throat — positioned where the histogram said no light escapes — dropped the temperature to boring.
v9: stopping
v9 wasn’t perfect; it was sufficient, and the difference between those is a skill. The falloff test matched a small softbox closely enough that the remaining gap wouldn’t survive a viewer’s attention. Print nine, stop, use the thing.
The part on the printer teaches you more than the part in CAD. Budget for the reprints — they aren’t failures, they’re the tuition.