Why are Class-A surfaces such a big deal for cars, but not industrial design / architecture?

from clay to press? im sure that’s the next question!

when i see what was achieved in the 40’s i design and machining a P51 together which was reworked in the middle of the war for better production efficiency… makes you wonder… how easy digital manufacturing can be…

Ten years ago the studio workflow at my empolyer (a very large automaker) involved cycles of refining clay models, scanning the model, creation of a math surface, refinement/modification of the math surface based on designer input and to meet legal, packaging, manufacturing and other criteria, refinement of the clay models, etc. The “designers” did very little or no 3D modeling, either in clay or digitally. Each studio had a team of modelers who worked on the clay models and the digital models.

Once the surface was released by the studio it went to a separate group who refined the surfacing using computational methods, verified compliance with criteria, and eventually released math surface data for the design of dies and other tools. In the early 1980’s the surfaces which were released by this group could differ significantly from the surface released by the studio. By the 2000’s the studio released surface was expected to conform to criteria with no major changes needed.


The most shocking mistake in the design of the Dual shock 4 controller is the area where a 4-sided patch (marked with #2 in the picture below) was created between the two main G1 fillets. It has some weird concave shape in the middle that’s super noticeable under a direct light. The versions that are coloured with a shiny coat expose the issue even further.

From a pure physical stantpoint, ergonomics are compromised by the large radius below the G1 fillet that I marked with #3. That radius is considerably larger than the cross section of the middle finger that goes in that area while the controller is held normally. Many people complained about the aforementioned inconvenience, myself including. I had a PS4 6 years ago and that intrusive shortcoming in the ergonomics department was the primary reason for me to sell it quickly after the purchase.
A G2 blend would make that area both, more comfortable and better looking.


Well, at the risk of getting mud thrown at me, here is a different view angle.

Why put 100% design effort in if you are selling almost as many products with 80% effort.
80% of the end clients of most mass produced items like a playstation are price oriented and wont notice the extra 20% design & surfacing effort. Neither would they pay for it.
Yes, there are exceptions, like Apple, but in pure numbers a huge amount of people more buy something else and are not willing to pay for a very advanced surface finish and price in many products.

Size. in architecture and shipbuilding & yacht design the products are very big, so the tolerances go up a lot too. So comparably small differences in surface continuity don’t show as much.

A creative person involved in design tends to look much more critical on such things as surface continuity than any other person does. Remember, the clients out there are usually not trained to spot these differences at all. In the end the client does not find so many things intrusive as a designer (and perhaps perfectionist) might do.

So it boils down to 80% effort sells almost as good as 100% and you rarely get paid for the difference.
Which in turn results in many smaller offices working in less expensive software which easily allows to achieve the 80% but perhaps not so easy the 100%.


As I mentioned above, me, several of my friends and many other people around the World were forced to get rid of their PlayStation 4 because of the bad ergonomics of the Dual shock 4 controller. The very large G1 radius at the bottom that I pointed in my image above is very intrusive during playing with the controller. That particular area of the controller could have been done much, much comfortable. Especially knowing that it was expected to reach the hands of millions of customers (PS4 just reached nearly 110 000 000 units sold last month, and more than 150 000 000 DS4 controllers). It’s one of those consumer products that are continuously used for hours each day, hence the ergonomics must be of the highest priority during the design process.

On the other hand, the Xbox One controller don’t suffer from such an intrusive area for the fingers, because it uses a very smooth G2 blend with considerably smaller “radius” that was designed with actual ergonomics in mind. :slight_smile:


This is correct, but still many production model of many manufacturers are almost completly made of single span surfaces. Because a single span model is more likely to be of good quality as a multispan model. So it is true that you don’t need single span surface to achieve class A and vice versa, but its unlikely someone is skilled enough to achieve the same quality with multispan surfaces. Many manufacturer will also reject most multispan parts because they do not understand why someone uses multispan if the same output could be single span. Rational? Not always…

I think the main reason is, that a multipan surface is really hard to make them truly smooth. So even if the unsmoothness is not visible, it might affect the curvature of other surfaces depending on this surface. If you apply a curvature graph to a multispan surface you will have a real hard time make it look right. Between two single span surface you instead have impact on the matching in between them, whereas a multispan internal continuity is fixed to its math.

This is what happens when different opinions get bashed all the time in this forum by a certain number of “Regulars”


In shipbuilding we’re working with class-a surfaces constantly, the problem especially for large steel ships is the deformations that happen after welding. We do care about deviations and the tolerances are not that big but sometimes the only solution to make these big structures (structural blocks) to come together is to use the “Big Hammer”.

As for the PS controller, I think it’s not about the effort there or the price. Why on Earth would one care about how smooth a controller is. You don’t look at it you hold it and you are immersed in the game. By all means the best console experience for me was with the controllers of Nintendo and SEGA Mega Drive 2. You can see how edgy they were. I didn’t care how smooth they are or how they reflect the light :man_facepalming: I cared that I had to buy another one because I bashed the buttons of this one :stuck_out_tongue_winking_eye: .

People give too much credit to how products look.
A smooth surface of a car won’t necessarily make it look better, let alone safer. (Talking about consumer cars not the ones dedicated to break speed records)

1 Like

Regarding the welding distortions and similar. Well, thats what makes the class A effort less worth it.
The shipbuilding industry has the budget for those software packages and uses them for a lot of different reasons. I guess the class A is more of a side effect for them.
I spent many years in yacht design, I understand the need for fair lines, still class A is in many cases overkill though.

If you ask me, unless class-a is necessary for performance, energy efficiency or safety, it’s an overkill.


If someone has worked with class A, such a person does always creates quality models by nature because otherwise it feels like doing garbage. I think the whole term is misleading, the german word for this is “Strak” and is originiated in shipbuilding industry. A designer doing class A would not consider him doing “Strak”, whereas a person doing this kind of work always produces class A. Sounds more like extra nice modelling, instead of doing something meaningful… :wink:

1 Like

Oh don’t get me wrong, I can’t release an non-smooth ship’s hull I feel sickened if there are edges visible in the 3d model.

Imagine how I feel seeing what the welding shrinkage does to my smooth form :face_vomiting: :smiley:

With yachts especially composites it’s easier to maintain smoothness as it is mostly done with epoxy and fibre. (unfortunately I’ve never worked at a company designing yachts, only passenger vessels and commercial cargo ships, and workboats)

1 Like

I can soo relate to this. Not neccessarily regarding pure class A surfacing (which I don’t think I am really capable of) but regarding the “proper” way to model…
I can’t count the times I had to correct a model before sending it out, even if it was only a slight detail nobody (but me) would ever notice. :laughing:


Simple and more insidious explanation for some of the ‘WTFs’ when examining a plastic part.

Often, capable designers will deliver surfaces that are ‘perfect’ for the intended manufacturing process (class-whatever) yet, downstream, ‘something’ happens. And it happens…

Example: a mechanical engineer or tooling engineer decides your part needs 3 deg draft, not the 2 deg you delivered, (maybe because the VP of Marketing said in some conf call you weren’t on - “can we make the texture stronger”) - or a tool engineer thinks your part needs to taper from thicker at one end to thinner at the other by a mm or two for mold flow. Or here’s a great one: they delete all your ‘fancy’ fillets for ‘some’ reason and add their own. (suspect in the Sony case, unless that was a 100% ‘engineer’ project???)

Either way, ‘someone’ whips-out their fancy MCAD and steps all over your ‘perfect’ model, and you don’t find out until you see T1 shots, and you’re like - WTF - and no one seems to understand or care, or when they do, usually it’s - “oh well, the tool is cut, it’s not so bad, NEXT…”

It happens…

On the other hand, I’ve worked on projects where the engineers are fantastic (culture actually) and you’re in the loop on everything, with opportunity to go back into the model and give them what they want, but ‘your’ way.


That’s why most expensive projects don’t use a straight line workflow process, it’s done doing iterations until the departments agree.

Rhino has a single command to fix that problem: ConvertToBeziers. A multispan surface instantly becomes a set of single span Bezier surfaces with exactly the same shape. :grinning:



I see this as butchering the surface not a fix. If you have to edit this afterwards…well, good luck.

1 Like

Not only Rhino, and I know this trick as well :slight_smile: Its not that I’ve done this. If you ask me it would also okay to say, take as few cps as possible, if there are very few spans, so what… But I think its about the people fitting in patches with 10+ spans, and then you can even tell that it used to be multispan before because nobody would match 10 beziers in a row if there is no strong curvature change involved. The only way you see something valid like this is on very difficult corner blend situations with multiple directions involved. And even there you reduce the surface count to an absolute minimum.

Actually the other way around is quite handy, and I don’t know if Rhino can do this. Take 10 surface patches and create a surface with as few cps out of it. Kind of a merge surfaces command with reapproximation behind. I use this often to clean up converted multispan NURBS

Do you think this was more a failure at a foam modelling kind of stage, rather than further downstream when an idea is being surfaced?

No, they all just couldn’t stand the fillets! :rofl:

(Sorry, had to.)


Which software do you use to do this? ICEM?
I wish to see this in Rhino as well.