Class-A surface is really the single span one. But the whole idea behind Class-A is the whole process, problem solving to reach as cleanly as possible (and in timely manner) to that type of surface.
There are many aspects that determine that level of precision: CV structure, Hull flow, corner conditions, ogees (overlooked flat spots), inflexions (if not desired), overall proportionality with regards to curvature, close to G3 transitions (a.k.a. pseudo 3G) to save some constr. degree (math), and of course ā the redesign facilitation (a.k.a. ālast minute changesā) ā and how easy those redesigns could be maintained.
Class-A design is a whole process. A single span surface (in both directions) on itself does not mean Class-A, but it is a part of it.
ŠŠ½Š¾Š³Š¾ ŠæŠ¾Š·Š“ŃŠ°Š²Šø Š¾Ń Š¼ŠµŠ½!
Iām aware what the Class-A term stands for, this is why I mentioned āand other techniquesā. I prefer to create clean models close to whatās referred to Class-A, but from my experience 99% of the time itās much slower compared to the basic non-Class-A modeling. This is why, for commercial projects I often use the latter to save time.
For example, this shape only takes 1 second to create a single network surface with complex shape (assuming that the curves are already existing). Yes, itās rough, but gets the job done in a timely manner. However, it would take many minutes to build the same shape in Class-A surface quality with G2 continuity between the gazillion individual patches. Network surface (tutorial) - YouTube
Another example with āwrongā surfacing that saves time is this model of a plug for manufacturing. Every Class-A guru would laugh seeing the amateurish technique that I used, but in fact it would take hours to achieve Class-A surface quality on this particular geometry, because nearly every trimmed surface would require shrinking and rebuilding to make it simpler and get rid of the majority of trimmed edges. Then, every pair of adjacent surface edges would require matching to G2. The existing long blend surfaces used as G2 blend fillets would need to be split into multiple short single-span surfaces that also require matching to G2. That process alone makes the work multiple times longer than the ābrute forceā I used on this particular model with a single multi-span blend surface. Rear bumper mounting panel plug (part 1 of 2) - YouTube
I think it is allways about a trade off between surface quality and modeling time and depends heavily on the product in question.
There is a good reason that strict class A surfacing is so prevalent in automotive design and much less in other areas: not every product is a giant smooth, shiney shell of organic surfaces that is sold mainly on its distinctive emotional aspects.
Some NURBS modelers are just obsessed with doing everything with single-span surfaces. Multi-span surfaces are not worse in quality. For example, Iām pretty sure that the blend shown here would be optically smoother if it was made with a single multi-span surface rather than using multiple shorter pieces.
I have seen Alias modelers struggling for hours to achieve perfect G2 continuity on tens of short surfaces that were the result of splitting a long, perfectly smooth multi-span blend surface into a number of Bezier surfaces with Degree 5. They do it for the sole reason that they just want everything to be single-span with 6 control points in either direction, even though the perfect quality of the multi-span blend canāt be improved further.
They donāt do it to reduce the amount of control points either. In many cases a surface with 7 control points is just enough to produce perfect quality, however, Class-A gurus will feel really bad if they see a multi-span surface on their monitor consisting more than 6 control points in either direction, so they immediately split the said surface into two single-span surfaces, thus increasing the amount of control points from 42 to 72 and the time needed to adjust the curvature continuity across the surfaces.
@Rhino_Bulgaria, @theoutside
To be fair to the āClass A crowdā: the level of obsession one needs to model this way often can only be maintained at 100% and leads to being pedantic about āthe rulesā
Very fair point, but remember the ārulesā for class A came about a long time ago (and for good reasons at the time) but have not necessarily updated to reflect the current state of modern modeling tools capabilities.
in my opinion this is the most important drawback.
test yourself the matchesrf command in this example and observe the curvature graph. match multi-spans.3dm (51.9 KB)
Here is a video I recorded on my youtube channel.
you are right
it is true that a multispan patch is quickly modified, especially if it is time-limited work and it requires being productive. multi span patches are a good compromise between quality and time.
but they also have some other significant disadvantages, more or less serious
for example,
a multi-span surface
must have many spans so that it can give the desired shape. compared to what can be obtained with divided surfaces.
this is why class A modelers divide surfaces
it depends on the degree of the surface and the multiplicity of nodes (itās a vast subject, I canāt summarize it in a few lines).
but I give you a concrete example of a surface that I divided in two, I improved the continuity between the adjacsent edges
through division.
I donāt think anyone will be able to put a multi-span surface with less than 3 spans with good continuity as in the example.
it is a challenge that I propose to you. try it yourself to see. divided surface .3dm (534.6 KB)
another thing also related to what I said.
itās that we donāt have access to the internal points hidden and captured by the nodes for the multi-span surfaces, we canāt play on their distance and their orientation.
one is forced to accept the configuration of the internal continuity of the surface as it is.
you will say but why that!
since we can also obtain continuity g2 not only by moving the third row of points (P3)
but also with other techniques which gives access to continuity g2
end bulge of the points corresponding to the tangency alone. sliding them along their tangency vector in both directions
by the centric rotation of the tangency vector (vertical rotation at the normal of the surface.
by the isoparametric direction of the rows of points to directly affect the continuity of the lateral edges of the surface.
but all that depends on the skill of the modeller.
the division of surfaces is not a random thing.
besides, we must know well where we must divide our surface,
otherwise the division will have no meaning and no use.
this is what happens with some car designers who do class A, they donāt know where they have to divide their surfaces and even worse they donāt know why they have to divide the surface. they simply do not know the usefulness or the reason behind this practice.
Thatās exactly what you mentioned bobi. you are mistaken with these people around you who pretend to do class A.
I am not someone who has worked in the automotive sector.
and I have no professional experience.
I am passionate and practicing modeling in rhino and most of my experiences do not go beyond 3d printing for personal and sometimes professional purposes with my circle of acquaintances.
but I like the idea of āācreating something in this sector or working for someone or in a company that is my goal in the future. I draw a lot of cars I have a lot of designs on paper and on my tablet
I have been doing research for a long time to learn more about the practice of class A modeling. on the net or by rubbing shoulders with people who have worked in this field. I had the opportunity to see class A models that were destined for production (itās achievable entirely with rhino alone)
it was difficult to find resources about this subject it looks like a nuclear field for the atomic bomb.
my tutorial of the audi A7 model is not class A
I already mentioned it in my first message a year ago.
in celebration itās just the workflow that i aim showed
everyone can improve their work to their taste.
after this tutorial I would offer class A tutorials.
Hi Fares, Iām all for using clean surfaces where itās possible, but as already mentioned by several people above, myself included, itās all about balance between time and quality. I have seen Alias designers split a singe surface with 7 control points in one direction and 6 control points in the opposite direction into two surfaces, simply because they want every surface to include no more than 6 control points in either direction. Then they spend a lot of time adjusting the split surfaces. Whether itās worth the extra time is a subject to personal preference.
As for your example, if it was mine project I would make sure that the large surfaces are not split. I noticed that your vertical surfaces are split at the bottom end, so I used the RefitTrim tool to convert them to untrimmed surfaces that closely resemble the original split versions. Then I applied Match surface to all major surfaces there. Next, I untrimmed the flat base and trimmed it again with a newly created blend curve (the green one in the file below) between two lines (the red ones) that I slightly extended in comparison with your original design intent, because that looked more natural to me. I deleted the two horizontal blend surfaces at the bottom and created new ones whose side edges flow naturally along the edges of the vertical surfaces above them. Since the vertical blend surface is no longer split, that let me use a single surface in the corner that replaced your two small surfaces there. The continuity I got is not ideal, bust is acceptable for general purpose. That was a nice challenge and thanks to the RefitTrim tool it was easy to preserve the main shape of the vertical trimmed surfaces.
indeed the challenge is to use the input geometry as it is.
this is to show the capacity of the split patches
to handle such a situation.
in my example the adjoining surfaces have the possibility of being raked with the _refitrim command, but this is not the case all the time. Often we are faced with situations where you cannot use the _refitrim command.
thatās why I proposed this challenge.
you cheated bobi.
I would not call it cheating , but rather using appropriate modeling techniques. Itās a common practice in Class-A surface modeling to follow two main rules that I also did in this example:
Refit the majority of trimmed surfaces (wherever possible) so that they will no longer be trimmed, hence they will offer as clean edges as possible. That will eventually let the adjacent surfaces to be simpler, too;
Align the direction of the edges, in order to achieve a more natural flow of the reflections.
I get your idea and in some complex scenarios itās really like what you have mentioned, but the best Class-A designers always refit the primary surfaces and keep the original ones in a hidden layer. That forces them to spend a lot more time into adjusting the individual Bezier patches in comparison to the quick but sometimes lower quality use of multi-span surfaces.
By the way, you will notice that your small rectangular surface on the top edge canāt be aligned to the adjacent surfaces with the necessary precision, but (here is the funny thing) if you use Match surface with the āRefine matchā option to add extra spans, you can actually achieve far better flow across the surfaces in that particular area. This is one of those examples where a multi-span surface can offer better quality while also being super fast to apply. However, all pedantic Class-A gurus will consider this āa bad modelingā, because they canāt stand any surface with more than 36 control points (6x6), so they would split the said surface into 4 smaller ones and then will struggle for a long time to achieve G2 continuity across all of them.
P.S. Do you mind if I post these two videos in the Rhinoceros CAD 3D group on Facebook?
@Rhino_Bulgaria you missed what I wanted to explain and show about divided surfaces, I hope you didnāt do it on purpose.
I did not share the 3d model to say. what is the best approach to model this. or how to modify the model to find the best solution. . that was not my goal.
(the goal of the challenge was clear).
create your surfaces on the input geometries. and see how many spans in the surface you can get away with, against a split surface. it makes it clear how useful splitting is, and why split patches are preferred in certain situations.
that was the purpose of my response to your question regarding the division of surfaces.
hope i was clearā¦
if you see that the approach you are following is the right one for you, thatās great. everyone is free to choose their method. all roads lead to Rome.
but, you cannot say that the divisions of the surfaces are useless.
on top of that I showed in the video the major handicap of multi span surfaces. particularly with the _matchsrf command, I think that no one is really interested in class A will be able to work with surfaces that do not obey to give g2 continuity easily.
as I said before everyone is free to choose their
model.
you mention the links of the discussion in this forum and the origin of the basic 3d model by my name. share it wherever you want I have no problem with that
Hi @fares.boulamaali , looks like there is some misunderstanding. As I mentioned above, I would not use such input geometry in the first hand. Both, the direction of the bottom horizontal blends and the general shape of the surfaces above them (with no trimmed edges wherever possible) would be different on my model, which I what I did with this particular example. On the 3d model that I showed in one of the videos only the top surface remains the same. I converted the vertical primary surfaces and re-matched them again, then I re-trimmed the flat surface at the bottom, then I built a new vertical corner blend and two horizontal blends that this time follow the same direction as the primary surfaces above, and finally, I added the lower corner blend.
If you mean that I only have to make a new lower corner blend (where the two red surfaces are located) while using your other surfaces with trimmer edges and control point flow that does not match to each other, then yes, based on that more complex geometry itās obvious that a single multi-span surface will have more control points than the amount of control points of two split surfaces.
But as I mentioned already and showed in the other video where I compare both approaches, using both, untrimmed input geometry and matching the flow of control points across the adjacent surfaces is the main goal in those situations.
Also, you will notice that a few posts above the conversation started when I mentioned that I have seen Class-A Alias designers feeling bad if they could achieve perfect G2 continuity with a surface having 7 control points in one direction, while their goal was to use surfaces with no more than 6 control points. That forces them to split the said surface and spend a lot of extra time to match the control points of the newly created two single-span surfaces with degree 5, ultimately achieving virtually the same optical quality as the single surface that they replaced but having 72 control points in total (6x6 + 6x6) instead of 42 (7x6). Of course that there are exceptions like yours where the primary surfaces on the opposite end are trimmed, then a couple of single-span surfaces would add a lesser total bumper of control points. But that also leads to not-so perfect continuity, because the flow of the control points is not optimal due to the split edges. This is why most Class-A modelers prefer to convert their 4-sided split surfaces into untrimmed ones via tool similar to Rhinoās own āRefitTrimā.
P.S. Your last video above shows a different approach. You first convert one of the surfaces into one with an unnecessary extra amount of control points, then you apply āMatch surfaceā. On the other hand, I talked about using a simplified geometry that subsequently receives extra control points as a result of the āMatch surfaceā.
Hi bobi
I know what you showed in your video is the shortest way to succeed smoothly, without facing you in the wrong direction against the NURBS.
NURBS are complex and delicate mathematical geometries.
absolutely it is not recommended to go against the wind while modeling, it is mathematics behind the geometries.
if we rebel against these mathematical rules we end up losing for sure.
In my example I broke slightly and on purpose the general rule of laying out the patch if conditions permit, by creating edges that donāt flow exactly in the direction of the isoparametric direction of neighboring surfaces.
this is on purpose to show the power of splitting patches in similar situations.
I know this patching technique that you did in your example, I already showed it in a previous video a few months ago, itās between 20:40 and 24:00 in the video at the top for the hole has six edges.
I knew that this class A subject is an endless debate, even we will talk all our lives we will not be able to put an end to the question who is right and who is not right.
I have no doubt about your bobi modeling skills, you are one of the best, your experience in the automotive field is gold. I am one of your admirers.
but me i like this unique patching workflow, i adapted it in my modeling, even if itās just for 3d printing .
i have a lot of fun while modeling.
and I will not let go of the subject of class A
I will continue to deepen my knowledge on it.
her rules are very real.
I donāt think they were put at random, there are valid reasons behind it, (I could be wrong).
it is true that the CAD packages have evolved since their beginning as it was mentioned by kyle.
the subD is part of the evolution but the route of the path has not changed.
people have changed their way of modeling.
but single span surfaces always remain sublime and beautiful surfaces
Donāt get me wrong, Iām all for using single-patch surfaces, but in my opinion they are useful mostly where the project is really important and the time frame and cost are not set tight. I often break the Class-A rules in my workflow, because time pressure is what makes me choose shortcuts wherever possible. I admire your dedication and the time you can afford to devote to optimize your surfaces. Basically 2/3 of my work could be considered ālow qualityā by the Class-A gurus , because Iām a heavy user of multi-span surfaces as they let me save time.
@Rhino_Bulgaria@theoutside
it seems to me that you are partly right as far as the division is concerned.
I was able to put on a multi span surface of two under surfaces while training tonight on the challenge I shared the day before yesterday.
the transition between the edges is just a tiny bit less G2 quality compared to the two patches I put on before. I did not try to improve the first example to make a real comparison. divided surface 2 .3dm (238.8 KB)
but it gives me the question if it really deserves to divide the surfaces to gain more quality, especially if it is small connecting surfaces.
(for large surfaces I have no doubt).
I heard that one of the reasons may be the simplification of geometries into bezier surfaces to minimize bugs and facilitate import and export between CAD packages
which is a repetitive process throughout the design of a product.
Iām not sure if this supposition is real or not.
. I spent the evening yesterday downloading 3d geometries from other packags to find out if I might encounter any problemsā¦
I downloaded an igs file from catia
I found bugs with rational or weighted surfacesā¦ they were weird and half transparent.
but no problem so far with mutis spans surfaces.
if someone who is used to swinging geometries between CAD packages
can give us a clarification about this subject. that would be a big plus
You did a nice job on this one. If you take a closer look at my example posted above, the whole problem comes from those split edges that force Rhino to add extra control points to be able to meet the file tolerances. Projects that are not than important donāt really need a tangency of less than 0,1 degrees since this is a pretty much enough precision for basic CNC-milling or 3d printing where the manual finish of the real-life object is hand finished anyway. The CAD designers at the OEM car manufacturers devote a lot more time into converting their primary surfaces into 4-sides untrimmed surfaces with tools similar to Rhinoās RefitTrim, because that lets them use the bare minimum of control points per surface. Iām pretty sure that you already know those rules.
This is why, for commercial projects I often use the latter to save time.
, because Iām a heavy user of multi-span surfaces as they let me save time.