I know Rhino is a swiss army knife and secretly wielded by lots of auto/vehicle designers, but that it ultimately falls short of tools like ICEM / CATIA which are specialized tools for building Class A surfaces. And I get that Class A surfaces are crucial for cars in particular, because the look and integrity of the car is all about the curvature/shape/profile, and how reflections/light hit it, etc.
Makes sense.
But… isn’t that also true for all architecture, particular any non-rectilinear architecture? Look around the world and you see countless dynamic, swooping, curved, futuristic, smooth surface buildings, facades, overhangs, and structures.
And think about the vast majority of industrial design… take Airpods, or hair dryers, or literally any object known for multiple curved surfaces.
And as I understand it, many of these structures are not only conceptually designed in Rhino, but also fully produced in Rhino. I hear repeatedly how common Rhino is in arch firms / industrial design firms.
If that’s the case, why aren’t Class A surfaces such a big hot stink for industrial design / architecture? Why do auto designers spend years and years of specialized Class A surface training with these extremely esoteric/expensive specialized tools (ICEM, CATIA), but it doesn’t seem to be the case for architecture or other types of industrial design? Certainly lots of objects behind cars require “perfect” surfaces… no?
I sense another heaty discussion between industry experts.
Class-A do exist in other fields, but I guess it’s rather impractical to achieve such “perfection” of the exterior surface or not necessary when you put the triangle of money-time-quality.
The explanation I heard AGES, AGES ago was that car tooling is made using the highest-speed CNCs possible where a continuity difference you couldn’t even see on the model could result in a change in the acceleration of the tool and leave a visible mark.
Of course cars are very very “shiny” while most every other product is duller, or has a finishing step after molding or whatever.
The other big answer is that “Class A” is a meaningless marketing buzzword. It’s basic definition could be applied to ANY CAD tool, anything is technically capable of this stuff, what it’s usually applied to is “Whatever the Alias guys are trying to sell.”
Typically most car body panels are being made via series of giant steel stamping dies that are huge and extremely expensive to develop. A huge hydraulic press applies massive forces to two or more pieces of stamping dies to turn the flat sheet metal into a desired shape. The cost of creating each of these dies is so high that it would be a disaster for the company to use a 3d model whose surface continuity is compromised. Since the car body panels are so big, it’s very difficult and slow to manually fix any imperfection in the continuity of the stamping die.
On the other hand, manufacturers of small electrical devices such like hair-driers make them via steel or aluminum moulds that are initially CNC-milled to a slightly rough state with visible milling markings, and then manually polished by workers with a number of different sandpaper grit. During that process, any imperfection leaved by the CNC-malling that was caused by the 3d model’s lack of perfect continuity could be fixed to a certain degree, because the matrix is small. Lastly, a polishing paste is used to give the matrix a final touch of smoothness.
Also, keep in mind that many plastic devices have a matte finish or a texture (created by using either sand blasting or chemical etching), so their 3d model don’t really need to be 100% perfect.
PS: On a side note, the Sony Dual shock 4 controller is a good example of a badly designed shape. It has plenty of G1 fillets that look especially bad upon looking at them towards a direct light source. There is a visible lack of continuity in several areas. To this day I still can’t believe that “Sony” approved such an imperfect 3d model that was intended to be used for the manufacturing of a plastic with semi-gloss finish (not to mention that later they also released a few modifications with glossy paint). A G2 blend would make it look sooo much better…
According to me the answer is: the development costs and the time.
Why to spend one year by designing shell for e.g. hand-mixer, consumer electronics or fittings or buttons…
In fact, Alias and ICEM could speed up the design process, because their advanced surfacing capabilities save huge amount of time. Not to mention the superior surface quality they provide. However, most private industrial designers and small design studios can’t afford (or don’t want) to rent those expensive programs annually.
In the case of the Sony Dual shock 4 game controller, making smoother blend surfaces would take no more than 1 day maximum. Then, the product itself will look better forever. Sadly, their CAD guy opted for the quicker, automatic G1 fillet.
The same goes for the wooden handle from @Stratosfear 's post above. Spending couple of hours more would make that product look considerably better and of higher quality.
I don’t think it’s just time, but also tools available (and the money they are willing to pay for them).
I doubt the Sony controller or that kettle was made with Alias/ICEM, because then getting single span G2 or G3 would have been easy.
I bet it was made using Solidworks/NX/Rhino/Inventor, where single span G2 and G3 isn’t as easy (Catia is somewhere in between… if they had a Catia license, I agree that it’s just down to time).
(And it’s not that the other tools can’t do single span G2/G3, it’s just that fixing, or even detecting, all the problems that may appear is that much easier in Alias/ICEM because there’s more tools available.)
I wouldn’t bet on Airpods not being Class-A, because Apple does care (and does use Alias). However, at least the 1st gen iPad was multi-span which isn’t pure Class-A…
Class A is not only about continuity and reflection but also about precision and refactoring until perfection. Since tolerance sum up its really important create data as precise as possible, so that only the deviation of the production process is relevant. Further its about teamwork. If someone begins to work lousy other people may get a problem. And last but not least. Light and smooth models are much simpler to modify and processed further. A low controlpoint count makes a non-parametric model almost semi-parametric, because any cp you move has a noticable effect on the shape. Direct CP modification is what class A experts do alot.
I guess many of these requirements are not as relevant for other professions or they do not know enough about it. Its difficult to explain someone who did no class A what the benefits are. Also many claim to do class A, but don’t actually do. Misusing it as a marketing buzzword.
But… isn’t that also true for all architecture, particular any non-rectilinear architecture? Look around the world and you see countless dynamic, swooping, curved, futuristic, smooth surface buildings, facades, overhangs, and structures.
In architecture for the most part it would be a wasted effort. Class A takes time and skill, architecture is a large tolerance industry and any curvature becomes panalized anyway so you will have seam breaks and rationalizations. Conditions where there are no seams typically involve manual labor on site like a sheetrock finish which would deviate from the model anyway based on the workers craftsmanship. Architecture digital documentation is always an abstraction and rarely what exactly gets built on site. Unlike a car, which entire parts of the body can be made and assembled mechanically, exactly to the model, requiring little to no human finish.
And think about the vast majority of industrial design… take Airpods, or hair dryers, or literally any object known for multiple curved surfaces.
In product design it really matters how much the company cares. Apple cares a lot. The PlayStation controller example here could obviously be better but it is probably not their primary concern as the controller isn’t the visual focus of the product so it is good enough for them, the screen is the focus point.
Cars are largely about their outside appearance and because it is a product that is continuously in various lighting environments, bad continuity is easy to spot.
Here the problem is in fact not the fillets, but he flat face on the side. When a G1 fillet ends up in a flat face it makes the flat face appear concave. + That part of the object is the thickes and moste prone to shrink, further adding to the concave shape. So it should have been modelled convex.
All designers would benefit form understanding what makes good blends and how it affects the visual appearance. One thing is light flow along dull faces like in the handle example, but a whole different ballgame is understanding the flow of reflections along an organic object.
Light might travel nicely over a G1 blend, but reflection will look horrible as it changes speed. So for reflections to travel nicely over surfaces (glossy bonnet of a car etc) those surfaces needs to have the same radius where they meet.
That and keep the surfaces to a minimum and the complexity of the surface to a minimum and you have the basics of Class-A
There are several ways to make an Y-branch smoother than G1, and in the past I used a technique with creating 6 “Sweep 2 rails” surfaces that I had to rebuild with “Rebuild surface” to be able to manually adjust the control points to my liking, then I applied “Match surface” with the “Preserve isocurve direction” option. Something similar, but with better surface quality done in Alias, is shown in the following video:
Some CAD programs are also capable of doing similar multi-surface patch automatically, though it comes with some imperfections where the 6 patches meet together.
Automatic multi-surface patch is also possible for Rhino with the XirusCAD plug-in. I have never had a chance to try it, but it seems to use the Catmull-Clark subdivision type that’s maybe similar to SubD in Rhino 7.
Thank you for that detailed response. I’m actually curious if you could pinpoint/mark the specific areas of that PS4 controller that are so egregious. I definitely agree the controller is gross (even if it were perfectly surfaced, I hate the overall design) but the only things that stand out to me are those ugly pinched sections at the bottom of the controller, where the handles meet the body.
I’m curious if there are other more specific bad G1 areas in other areas you could point out so I can identify these flaws at a technical level better in the future.
Is this dicussion about actual product surface quality, or about a methodology for surface design?
“Class-A” originally meant the perceptual quality of readily visible surfaces of a product, for example the exterior surfaces of a car. The surface quality is frequently assesed visually by the smoothness and flow of reflections, highlights and shadows but can also be assessed tactilely (by touch). Also contributing to the perceptual surface quality is the size and uniformaty of gaps and offsets/steps. Class-A surfaces predate the use of computers in design. In some instances achieving Class-A quality is highly dependent on the skill of the craftsman involved with making the tooling an/or object. Assessing Class-A quality can be context and observer dependent, similar to many if not all aesthetic assessments…
More recently “Class-A” has been used to describe methods of computational design which are based on use of single span Bezier patches or equivalent with suitable matching between patches. More complex surfaces are usually modeled by use of higher degree patches with more control points and increased number of control points. Not infrequently the statement is made that the use of multi-span patches is not allowed by “Class-A”. These methods are commonly used in the design of products which are intended to have “Class-A” surfaces. It should be noted that use of these methods does not guarantee the resulting surfaces will be “Class-A”, nor does the computational design of “Class-A” surfaces require the use of these methods.
