[farslan]

Interesting you said that, but I used G2 almost everywhere, except in the inner holes of the phone part. If you share a picture, I can happily show the Sharp3D equivalent. Maybe the G2 curvature wasn't as aggressive as it should be?

[davee5]

Huh, well apologies for the assumption though I merely have the blog images to go off. When I look at the render that shows the top ortho and the one for the bottom cable geometry the highlights end fairly abruptly. The only curve that visually looks C2/G2 to me is the acute angle blend between the tray edge and the back of the phone cradle, that has a nice acceleration of the radius in and out of the transition.

It's possible the Shapr rendering engine is not very subtle, or perhaps the G2 math is accurate in a strict sense but the output is not very differentiated from G1. It's mathematically possible for there to be a continual change in local radius, i.e. be curvature continuous, while still having local changes be sufficiently aggressive that it visually appears discontinuous at a human scale. Each CAD kernel seems to make these things in its own way, hence different industrial design studios will strongly prefer the use of certain 3D CAD programs to make their final master models (e.g. Alias). Personally I drive CREO as for ages most manufacturers overseas used pirated copies of Pro/E or CREO and thus I could send them "native" surfaces. In that program my preferred curvature continuous coefficient range was 0.52-0.57. I don't have Shapr access handy so messing around with the coefficients and finding a result that you like is outside my domain -- and perhaps you already did!

Still, all that is on the modeling side, but the best way to actually check the visual smoothness of your corners is to use analysis tools like curvature combs to check how aggressively the model is making transitions. It doesn't fundamentally matter if you use the built-in automatic tools or manually adjust b-splines in your NURBS: the smoother your combs change the smoother your corners will look. [I checked the support page for Shapr to see if it supports curvature comb analysis and saw nothing about it, so you may be out of luck on that front until future updates.] Absent that you have to just spin the model in CAD and see how smoothly the highlights roll around and hope the built-in rendering engine is doing its job well.

One last item of subjective crit in sculpting smooth models: when applying a fillet to an edge that turns a corner, such as your interior pocket, you'll have a less visually cramped and abrupt appearance if you use a fillet chord (edge radius) that's nontrivially smaller than the chord length of the turn it has to make (corner radius). Maximized fillets that come to hard corners and make a full spherical bubble, e.g. your initial models shown in gray, generally look less natural than those that allow the fillet to turn the corner. This lets the highlight work its way around in a racetrack form instead of getting "stuck" in the extremes.

Nice work dude, I wouldn't comment if it didn't seem like you're dedicated to making continual improvement and learning new tricks.

[farslan]

Thank you for the comment. I learned a lot from you and will look into these. Are there any resources I could use to learn more about these, especially sculpting smooth models?

[davee5]

Quite welcome! Sadly I don't know of any resources that are really useful when you try to put things into practice. Most of the tutorials / blogs / forum posts I've seen are not really that constructive nor sufficiently detailed to get into the truly useful practicalities. I personally learned the trade by (a) working in an industrial design studio as the token engineer & CAD jockey with "real" industrial designers pushing me to do better, and then (b) building my own kilohours of practice in aesthetically driven CAD modeling. Now as the design lead & manager on most projects it still takes me months of coaching my employees on subtleties to get things right, and even each project still requires an unreasonable amount of time tuning curves and corners. Like all professional practices this rabbit hole goes real deep. Still, you can get pretty far with brute force iteration and careful attention to detail. I think your progress thus far demonstrates that quite well.