Class-A Modeling: Deep Dive with Alias V2022

BARRY KIMBALL: Hello and welcome to Class A Modeling Deep Dive, Alias 2022. My name is Barry Kimball, and I'm the Autodesk Alias technical product manager. For this session, we'll be discussing Class A modeling and how you can achieve high-quality services in Alias. So let's get started.

What is Class A surfacing? So "Class A surfacing" is a term many people hear in the industry, whether you're in automotive or you're in product design or you're just making simple parts. Everyone talks about Class A surfacing and what it means to them or to their company.

So in general, "Class A surfacing" is a term used to describe a level of quality or a set of techniques to produce a final object. So firms around the world all have their own take on what "Class A" means. And the job of a Class A modeler is to ensure that a design is feasible and flawless. The task of a Class A modeler includes but are not limited to analyzing surface models, incorporating the engineering criteria, and most importantly, protecting the original design intent.

The Class A modeler is the last person to touch the data prior to it going into production or manufacturing. Oftentimes, the Class A modeler has a lot of overtime, has a lot of pressure, and is involved in having a lot of people work together on a single project to come to an end result. Class A modelers are typically very detail-oriented. They're very good at thinking in three dimensions. And they usually typically take a lot of pride in their work because there's a lot of time invested.

In creating the best CAD model, there must always be a bar that you need to achieve or a level that you need to achieve. In the area of freeform surfacing, the highest level is found in the automotive industry, typically. We've got a lot of product design firms out there now that are also wanting to achieve this very high-end quality. And that quality is perceived by customers in the tactile feel or the look of their designs.

And people are becoming very picky, if you will. There's a lot of choices for them out there, and they really want to have high confidence in their purchase. And as they see things that look better and are more refined, they see that as a measure of quality, and they want to buy that type of product.

So a lot of times Class A surface people-- let's just go back one slide really quick-- Class A surfacing is not necessarily just about the surfaces themselves. I could make a box and make all G3 continuous or perfectly curvature continuous surfaces and say that that's a Class A surface. In my view, you'd be incorrect because there's no criteria. Class A surfacing means that you have a design, and that design also meets all of the criteria required to make that design manufacturable. So just because you have really smooth surfaces doesn't mean it's a Class A surface.

Class A surfacing means that you've got a design, and you've also incorporated all the information to be sure that that design is manufacturable and meets all the criteria necessary throughout an entire company with respect to being able to paint it, to manufacture it, to ship it-- all of these things have to be taken into account.

So what are some aspects of Class A surfacing? One aspect would be it's an advanced modeling level and requires years of experience. You can't just say, oh, I'm a Class A surface person and worked in the industry for six months. It's not going to work that way. It requires a lot of experience. Also, Class A, it's commonly agreed upon between engineering and styling or design. Not one side gets to decide, not engineering decides, and not design decides. It's a commonly agreed baseline on what we're going to try to achieve.

Some of the basics of Class A surfacing would be referred to as lean and light modeling. And what does that mean? Lean, the fewest surfaces possible to describe an object; light, the most simplest surfaces out of that set to define the object. So if a Degree 2 surface is OK, use a Degree 2 surface. Don't use a Degree 6 surface where a Degree 2 surface is good.

Also, special training and care can be needed to optimally and efficiently create Class A data. It's a constant learning experience because no two designs are alike. So each time you are into a new design, that means you have to learn new things, some new techniques, and always be expanding on your expertise. And in the end, Class A surfacing is mostly or all about the beauty and the harmony of the control vertices that make up the surfaces that describe the part.

So let's get in and do a little deep dive here and talk about some aspects of Class A surfacing. So here, we have a body side of a car. Now, we're just going to have a free talk here about this model. So this is a body side of a larger vehicle and you can-- I'll turn the model off here, so you can get a look at what this-- visualize what this looks like.

So we've got a pretty long wheelbase and some pretty interesting body side shapes here, and you can adjust the perspective. Now, I like to use viewing for Class A modeling. It's extremely important. In this case, we're looking at it in Perspective, and I like to have my camera set with the Lens Properties. 44 is the default. I like it at about 32. This gives an appropriate screen style or screen space amount of perspective. It's not overly perspective, if that's a word we could use. So here's the basic shape.

Now, we could say this is a Class A surface. Some people would say this is a Class A surface. It looks pretty smooth. You've got some nice highlights through here. Everything looks pretty good, and you know what, in reality, a design. So a designer, or a person who's not a Class A surface person, may look at this and say, yup, that's sweet. It looks awesome. But the Class A surface is a little more deep than that. And that's where we start to talk about the flow of control vertices, the number of surfaces used to describe the model, and things like that.

So right away, as a surfacing person, I can look at this and say, well, right away, I see that this surface right here is a NURBS surface. So a NURBS surface-- everyone knows this term, "non-uniform rational B-spline--" well, not everyone, but most would know this term. So in this case, we have a multi-span surface. So in this case, it's Degree 5 with four spans, and those four spans mean more control vertices. The more control vertices, the more opportunity for undulations. It's not the lightest weight as possible. So it's harder to analyze the actual shape because there's more surfaces.

If I took four playing cards and laid them out on a table, and they were supposed to be very square and not angled to one another, very square and very orderly, if I did that with four cards, it'd be very quick and easy to see if they were organized really, really well. If I took an entire deck of 52 cards and laid it out on a table, you'd really have to search around and look and see if you could find the cards that are out of place. So that same concept applies for control vertices, where these control vertices need to be arranged very nicely, and the fewer that there are, the easier it is to analyze.

So I would look at this surface first and question, why is this surface more complicated than the two input surfaces? And by "more complicated," I would say or show you that this surface is Bezier, and this surface is Bezier. So two Bezier surfaces, when they get blended towards one another-- in this case, it's requiring a multi-span NURBS, so I'd want to take a look at that.

The next thing I would start to look at is in this area here. This button does random color shading, so now we can see the surface structure, how many surfaces they used around in this model. So here, I see a common thing. This looks like two surfaces that are perfectly continuous. And then over here, we have two surfaces that are not continuous. So we're making a transition from a blended surface on its way around this corner to no blend. So it's faded out. And I see this a lot, where we have one surface and then, in the end, are incorporated two smaller surfaces. And we're going to go in and take a look at each of these areas once we get them flushed out here.

So in this case, we have a blend here, Degree 5 by 8, again, highly complicated, and it actually builds all the way to the end here. But then the user chose to put two small patches to blend that out and be able to work across this dual boundary with two surfaces versus a single surface matching to two.

We can also do a little foreshortening here. So if we go to the left side view, this button allows us to shrink this here, and I can see this is going up and then down. It looks pretty good, but I'd want to investigate how all of this happens, and more importantly, I'd want to investigate the layout of these vertices that make this up. Because even right here, if I look, here, we have a surface, and then this surface tends to bunch up to one side, and I can see some irregularities in these control vertices.

So when I look at a theoretical intersection-- so a theoretical intersection is where the two slabs hit before you put your blend in-- I'm looking at this theoretical intersection right here, and if we look down this theoretical intersection, I can see some undulation in these control vertices. Again, viewing is very important. And when I'm looking at control vertices, I never have Perspective viewing on it. I want to see if things are parallel or if they're converging at a nice rate.

With Perspective on, that's fairly tough to see, and this may illustrate that fairly clearly if I switch right now to Orthographic mode. And now, we can really see the difference here and the undulation of these CVs. This should be a nice, smooth shape of a theoretical, and that would result in a smooth blend.

A couple of other areas we might look to consider, I'm just going to-- this button. So here is setting Non-Proportional Scale. This is toggling it off. So now that we have it, we can toggle from on or off right here. Again, viewing is key to Class A surfacing. So I'm going to look at these surfaces up here. And this is a common thing. We've got a flat line across here that comes to a blend. Let's look at these. So these surfaces are not built to a common intersection. That is typically a no-no in Class A surfacing. We always want to control the theoretical intersection, and in this case, there is no theoretical intersection.

And when you don't have a theoretical intersection, you can start to run into issues with a surface that isn't just cramped-- so we have a theoretical, and then we want a blend in between. When we don't have a theoretical, now, this blend in between, if these are in these positions, we can get an S shape in between. So to judge that, I would then look at this surface, and I might look at this surface from over here. Again, viewing is your friend.

So I'm going to go over here, and I can already see this S shape right here. And if we do some non-proportional scaling of this view and we look down this view, you can see here how this surface goes down, steps up, and then is continuous. That's a direct result of not having a controlled theoretical intersection. Again, good design model we're looking at here. Shapes seem to look OK. But when you get in and you analyze the data technically, so really technically, this wouldn't pass as a Class A surface.

And another way to check this would be to turn on some curvature combs. And I'm just going to come in here and scale this down a little bit. And a blend, if this were a true-- a nice blend, this would go-- this would start flatter, increase in curvature, and then come back down. In this case, you can see right here, there's a flat spot along this edge. And that flat spot is a direct relationship to this structure of control vertices. So we'll take a look at how we might be able to clean a couple of these things up now. And we'll just delete our locators, and let's get back to traditional viewing.

So first things first, I'd like to take a look at this data. So I'm going to turn on some shading here, and I like the shady sky for this. And I just want to look in Perspective here. I just want to look at the shape of this. And what I notice right away, if I turn this model off, what I notice right away is this is all negative. And when you look at these things on oblique angles-- so when I mean "oblique," like right along the surface. So we're not looking at the side of the car. We've rotated the car, so now we're looking along this surface.

And when I look along this surface, I see this constant negative. Typically, that's not good. It blends itself, or you visualize it as hollow, so a low spot. And I've looked at this model in VR. And if you haven't used VR before to analyze surface data, I highly recommend it. You can see, it's just like standing at a clay. This looks peculiar to me, especially when I'm looking-- so if we had this with Symmetry turned on, and I'm standing on one side of the car over here, but I'm looking at this side, that negative, so much of that negative is peculiar to me. So I would like to go in and take a look at how we might be able to fix that.

And on some of these, we'll do some live demos on how to do it. On others, we'll just look at what I've created here. We only have just under an hour to chat about this.

So here, we have a surface that I've built, and this is a Blend surface, so it's just a Freeform Blend. But what we've done here is taken these two surfaces, and we've added crown to them. So when I say "crown," if I turn off these vertices and we can compare these two, so I'll turn both on at the same time, and we look down this surface. So this is the original down here. And you can see that a lot of this negative right here, I have taken away. So I've pulled these surfaces up.

So just to give you a quick idea here, I'll copy and paste these, and I'm just going to do a new layer and assign them and then untrim them, so we can see their theoretical intersection. So now, if we look at this theoretical intersection, this theoretical intersection is positively crowned. And what that's going to mean is the fillet is going to build nicely on that.

So if we do the same thing-- Layer, New-- and I'm going to turn the original on, grab the same surfaces, and copy and paste them, and assign it, and we'll do our same untrim. And we look down this line, it's got crown, some crown. But if we do a comparison here, not as much crown. So here's the revised. Here's the original. So adding that little bit of crown-- and you can see the tangent lines right here-- just tends to add volume to that surface, and it doesn't look so hollow. Now, this is a choice for design, not for engineering.

So Class A surface person can also-- when you're designing things, it's a very fast-paced environment. You might get two days to build this model. You may get a day to do some blending. Class A surfacing guy is going to look at this, and he's going to really deep dive into this data and have a look at how well it looks. This is one of the things that he'll look at, and we may then go back and forth with design. In the places that I've worked, the Class A modelers are in the studio, working with the designers. So we may go back and mill this on one side of the car and check it. Design might say, nope, it's fine. Let's just stick with what we have.

Now, the other thing, though, that we can also see here is that I've created a nice blend now. So previously, we had this. So the difference is in the number of vertices. So if you look at the original-- remember, we talked about the placement of these vertices. If we look along this edge, and now I switch, and we go back and forth. If you notice, I believe this one is nice and uniform. And I believe this one is a little more haphazard. It moves from left to right, so not as perfectly placed and not Bezier. Now, we can come in and do a continuity check on this and say, hey, is this curvature continuous? Yes, in fact. It's curvature continuous. That passes a check.

Another check I would like to look at, though-- because this is our wheel flare down here. This is our wheel flare. So we need a blend to put right here. So I would like to check and see, tangent-wise, how this looks. And here's another flaw. In this data, this surface is not continuous with this surface. There's a 0.15 millimeter gap. That would not be allowed in a Class A surface model. Everything would have to meet at its theoretical intersection.

Now, what I wanted to see was this angle between these two surfaces, these and their friends. Now, when we're out of positional tolerance, we cannot check angle. But there is a way to do it. If you go in and pick the locator, and we go to the Information Window, and we set the positional tolerance to be something higher than its value-- so I'm going to set this to 0.16-- then it will show us the angle.

And I want to get the angle here the same-- or see if the angle is the same across these. So I'm going to move this locator. I'm going to add in some of these. I want the height of these to be the same. If you wanted them perfect, you could grab this one and say Info Window. And what is the scale? So the scale is 8.6. I'm going to make that scale 8. Grab this one, make it scale 8. And grab this one, and make it scale 8. So now we have consistent scale. And if you look, we have a larger angle here. So we've got an angle that might be this big. And then, as we move to this area, that angle starts to get shallower.

So it's going-- that angle is doing this as we're moving around this wheel opening. What that's going to cause is the highlight or the-- yes, the highlight on the fillet that's built across there, that fillet it is going to get flatter and tighter as it goes around this opening. That's something you would really want to control in your model. So if we go to the modified version and we take a peek at this, and I'm just going to show you a little trick here. Check tangency right now. It's on Auto Scale. So it's based on the size and the scale of the surfaces. So I'm just going to say I want it to be 8. So now I can check this one, his little friend here, and this guy.

Now, we got a little angle in here. It's less, but let's check this out the same way we did the other one. Let's set this gap to 0.3-- no, 0.7. Let's get its scale. I bumped the scale with my mouse when I was doing that. And let's just increase this. So we're similar to the angle. So we might want to go in and tweak that a little bit. Because you can see here, we still fall a little short on the angle, so we're still doing this a little bit with this surface. Just something to note-- it doesn't mean you have to go back and fix it. And we'll just delete those locators. So that was an area we wanted to go in and investigate.

Another area we want to go in and investigate really was this area. And we wanted to look at these theoreticals, so by these theoreticals, I mean these. We wanted to check the difference between those theoreticals and these modified theoreticals that I built.

So if we go to a straight side view here and we zoom in, there's our theoretical intersection line. If we turn on the original and we do a Pick Objects, that's the original. So if we now do another non-proportional scale, and then we make this flat on the screen. I'm just using the right mouse button, and I am in Azimuth and Elevation viewing, so Object-Based Tumbling versus Camera-Based Tumbling. And we look at this.

So we talked about this theoretical wanting to be nice and smooth. So we can see here that this theoretical-- I have the original highlighted. So the original goes up, down, and up. If you look in the background, you see how that's smoothed out. It's controlled. And if we look at it from this direction-- again, viewing is critical in these cases-- if we look at it in this direction, we have a nice, smooth-- in the dark, we have a nice, smooth, controlled theoretical, where this big angle right here at the end-- so by "angle," I mean the attitude between this vertice, this one and this one-- that has to be mimicked on the other side.

So since this surface has a ton of acceleration right here at the end, and it's theoretical-- so its theoretical is going like this-- the next surface has no choice but to continue that. And that's why you see the next surface goes this high because it's having to compensate, so a little bit more control in this theoretical intersection. Again, we're talking about very minute changes in the CV structure and, distance-wise, very slight changes. But that's what Class A surfacing is, perfection in these surfaces. So let's turn these off really quick.

And then we can look at the blend in between. So in this case, we have the blend here, Degree 5 with four spans, so we get a lot of vertices. And if we look at the modified version-- so here's the newer one. This is a Bezier surface, very controlled, very lightweight. And then we move along, and we move to this next surface. We have this one, which is the original, a lot of vertices. And then we have the new one, here, controlled, very few vertices.

The next thing that I'd just like to show-- and I think I'm going to be able to do this without totally rebuilding it. If you remember a few minutes ago, when we started, I spoke about this surface and then requiring two surfaces to do the transition. I want to show you a little trick in the Align function that will allow you to work around having to do this and actually do it in a little bit better way-- or what I feel is a better way.

So this was built, and if we Query Edit it, we have metadata. So I can see right here this was built by the Surface Fillet command with Chord at G2, so on and so forth. There's all the settings. And I'm just doing a Query Edit, which is found under Object Edit. Query Edit and right mouse button will bring up all of the metadata. So by metadata, if you're not familiar with that, that is all the settings that were used in the Surface Fillet command. So this is giving us access to what every one of these settings was.

And I believe if we grab this surface-- so if we grab this surface right here and I go into Surfaces, Surface Fillet, it says right here, you had a surface picked. It has all this metadata. Do you want me to fill out the Surface Fillet tool? Do you want me to set all those settings? And if I say Yes, now, I have all those settings set exactly how it was built.

So I'm going to try this really quick and see if it works. If it doesn't, we'll just revert here. But I'm going to grab these two surfaces. I'm going to copy them and paste them, put them on a new layer. And actually, I want these two also because that's what I want to show. So I'm going to copy and paste those and put them on the layer and then turn this off. Oop. So I must have missed a couple of surfaces here. So let's get this one and this one and paste them over here.

So here's our basic setup. And we'll do an Untrim on those, so we can get back here. And let's just go ahead and see what the Surface Fillet command does. It's an interesting situation for a command like the Surface Fillet tool, because if we look at this, we have an angle between these two surfaces. And that angle-- I might have turned off my non-proportional scaling-- that angle goes away at the end. So how does the Surface Fillet tool know where to position that fillet section when the input surfaces are already perfectly continuous? So that's a challenge for the Surface tool.

So Multi-Surface Fillet, again, it's got all the settings from before. So I am going to just ask it to put a fillet in here, and we'll see what it comes up with. And it comes up with something kind of crazy. So I'm just going to do Default on this end and see what we get. So it only builds this far. It only builds that far.

So what I did in the original is built it, and it's just a slightly different manner. But if I take this surface right now and in the V direction, in the V direction, I go to 5. And then I'm going to do an evaluate tangency over here. We've got tangency. Evaluate curvature. We could do curvature over here, and we've got curvature. Now, that's because I had already fixed this surface.

But what I wanted to show you all is that-- let's just say this was off. So I'm going to grab this vertice and just tweak it so that it's off. So now, we are not continuous. And we'll see if we can get this to happen. So the Align command-- and I'm going to do a G0 Edge Alignment-- and I'm going to align this fillet surface to here and then here. So now, it's aligning one surface to two edges. And all I'm going to do is turn on Partial.

So now, since I've done Partial, it's keeping the original locations. And as you can see, it's fixed the continuity problem that existed before. So in Alias, you can align one surface to two surfaces. And that's something a lot of people aren't aware of, and you can do it in Project and in Edge Alignment.

Now, we could come in here and say the Continuity Check. And we could say let's check for curvature, and we are curvature continuous. We might have to tweak a little vertice right here. So we might have to say, you know what, this broke my curvature in this direction a little bit. I'm going to delete my construction history. I'm going to come in here to my Move CV command, so this is just Transform CV. Space bar brings up the controls.

And I want to move-- since it's curvature, I'm going to try to move this vertice, and I'm just going to use my arrow keys and just bump it. So we've got way too much movement, so I'm going to up the sensitivity. This will make it move a lot smaller. And I'm just going to watch that. So you see, right here, I'm getting into why I needed a Degree 6 surface before. So I'll grab this, maybe go down here. I just moved CVs just to touch, and now we've achieved the continuity there. So in this point, Align command, you can use to align one surface to two surfaces.

And if we go in-- and we'll go in at the end here and just check out some of these highlights when we're finished. I'd like to show one more area. And we checked this surface up here. If you recall, we talked about the fact that there's a theoretical intersection that wasn't defined. So I'm going to take this surface and untrim it. So we'll untrim this surface. And let's just pick this curve on surface now and delete it. Remember, we have this surface, and we saw this undulation in control vertices right here, and we want to take care of that.

So let's go to Object Edit, Align, and I'm going to align by projection because I'm aligning to the middle of a surface. So I'm going to align this edge to here, and I'm going to do it in Y. And the reason I'm choosing to do it in Y is, I guess in this left view over here, I think that this arrangement of vertices is nice in that direct view. So I'm only allowing the Align function to move straight into that view. I'm not allowing it to go all over in three dimensions. So I'll always have a good structure in this view, and I'm going to ask for curvature. So I now look at this, and I see that I have curvature if we turn on these curvature combs again.

And I'm just going to bump up the scale a touch and increase the sample so we get a little better view in the right. We want this to be nice on both ends. This end should look like this end. It should be decelerating because this surface is typically flatter than the fillet that we're moving to. So how can we manage that? So I'm going to grab this surface now and turn some vertices on. So I'm just going to tweak a couple of these CVs. Normal-- it doesn't really matter. You could go Normal. You could go in Y only. But I just want to show you the principle here as I move these. So I'm just going to move these.

As I move these, we have history in a list. So I have a surface, the big slab that I'm moving right now. I've got another surface aligned to it, so back here in the middle. And now, I'm watching that surface move around. And what I'm trying to do is get those CVs to be nice and smooth and, in reality, get these curvature combs to be nice and smooth. So I'm just going to grab a few of these and begin to tweak them. I might grab these and move them down.

And as you can see here, as I do that, I get a more balanced curvature comb. Now, I say "balanced." It doesn't have an undulation in it. It doesn't have a flat spot. The reason why we've got curvature bias to the right-hand side in this view is because there's no theoretical intersection controlling this. It's user-defined. So the best we can do here, if we want to maintain this design, the best we can do here is to fix the highlights in the blend.

We could go and start tearing up this whole fender, but that may-- design may say, you know what, we like the shape of that fender. We're OK with the fact that it's a little unconventional and that the blend has more bias up to this side. We're OK with that. This is design versus perfect surfacing.

Perfect surfacing is ideal. Design can sometimes influence the perfection of the surfaces because they want a certain shape. In the end, we're not selling Class A surfaces. In the end, we're selling a design to people that they like and want to buy. They're not really aware of what the surface's construction is. They just know that they like something when they see it. So we perfect these the best we can in the time allotted and then move on.

So before we wrap up here, let's just take a look at some of the highlight differences between the before and after. So if we look at the before here-- and I'm just going to do a Delete-- oop, not Delete Model, no. I'm going to do a Delete Locators. And let's just take a quick look here.

So a really nice tool for looking at highlights is this tool. This is called the Iso Angle tool, and I've had other videos that you can watch back in older AU classes where I explain exactly what this tool does. But basically, it's analyzing how highlights could potentially flow based on a constant angle, not your view.

So if we look here in the back, this just-- some things draw my attention here. This arc, which is not repeated on the other side, this little maybe S shape and this highlight, it looks good, but it's not 100% controlled, I would say. So if we switch to the newer model, we have this one, and I just believe it's a bit more controlled. If you look, we still have a lot of the same characteristics. So here, I'm just going before and after. Slight changes, we probably moved the surface five millimeter here or something. So we still have this highlight. We didn't really change that that much. But I believe what we've done is control this in a little bit better manner.

So if we look down here, we've added some volume to this area. And I think that the highlights flow straight up the core a little bit better. If you look here, these taper off. And if we look here, they go straight. We could turn down some of the blur here, so we can really see these edges. We want a little bit of blur, but we can look at this here now, and we go to this one, and you can see these tend to taper off to the right, and these tend to fan pretty straight in the middle. So that's one area that I think we've tuned up a little bit.

So if we look here, here's another set of highlights versus these. We can just turn the model off and just take a look, but slightly different, only slightly different. We're Class A surfacing. We don't want to redesign the car. They already had a design that they liked. We're perfecting the surfaces. Now, we could probably go in here and add some more reflection lines and show some more differences.

But let's just look at it, in general. There's probably something that we could have done up in this area. I see some little flaws and some little wiggles. We probably could have done something up in that area. And then if we go here, here's the original, and there's the modified.

And if you look here, that's highlight, big sag. And the sag moves up this way, and it goes away. Now, big sag, and I think the sag more moves up through the model. Now, this over here, this is just because we didn't finish and fix this side. We modified this surface and never modified this surface. So that would be something, again, where you might want to take the Align command. And we're just going to do an alignment between these two, and we're doing a projection, and we're doing it in Y, and Alt-O, and delete my locators. And now, if we do a Show Model again, now we've cleaned that up. So now, we've gone like that.

When you do this back and forth, you can see they're different. I see this as an arc, a flat, and an arc. Maybe the flat keeps going, and it just-- it doesn't look like it knows what it wants to do, to me. And when we look at this one, I think it's pretty specific-- arc, curved, arc, arc, curved, arc-- and it just tends to flow back and forth across the surface.

So those are some detail aspects of Class A surfacing using Alias and specifically this version. It's 2022. We'll go back to our PowerPoint and move down. I appreciate your attendance and your participation. Thank you.