Hi Kevin,

> Yes, I've seen that functionality in modo (It doesn't
> exist in 3DS Max as far as I know).

Possibly with Polyboost? I think another option is Silo which I believe has some retopo tools in it and is really cheap and accessible as a companion program.

But yeah this adds another kind of tool into the mix with another skill needed to be learned, that part is not so good but on the other hand that kind of tool can also be useful in pure poly modeling as well.

- Michael

Hi Kevin,

> If Moi had variable radius fillets maybe you could come closer
> but I doubt that would look right either. Same goes for
> the cockpit.

Variable radius fillets will definitely get in here at some point, it just has not quite bubbled up to the top of the list yet...

Another approach which can work for this is to trim the parts back so that there is a gap between them, and then use Blend to make a smooth juncture (G2 option in blend). That is a lot more of a "natural" and seamless type of juncture between pieces. You can kind of get a similar effect as variable radius fillet by varying the width of the gap that is trimmed between the pieces.

But that is getting into much more of the fancier and advanced part of the NURBS toolset, kind of away from the easier and quicker parts.

Probably the best result for the kind of highly controlled blend that you are describing would be trimming a gap and then doing a 2-rail sweep with some custom profiles to shape the blend, with an option to force the sweep to be smooth to the edges it is sweeping along. MoI does not currently have this option, but it is another piece that I want to add in eventually... Again it is part of a kind of advanced toolset so it just has not been a focus for the earlier part of MoI's toolset.

- Michael

Since this is a learning thread, here are some points.

The detail to the creation is "KEY". If you examine the curves of the model that didnt work you'll notice these sharp "CORNERS" that dont exist on the REAL object surface. It may appear to be a sharp spine in a picture, but if you put your hand on the surface of the F9F, you'll not feel this sharp edge.









The precise nature of NURBS calls for extremly accurately placed points and curves. If this is not done accuratly, then the results will be according.

Kinked curves produce kinked surfaces! this model has designed kinks all over it.

If we look at the software capabilities for Complete accurate depiction, yet Rhino falls short then what are we doing?

It seems there is an opposite conundrum here in that, NURBS would do the REAL object and Poly can easily replicate it to look somewhat real. Not the other way around.

BurrMan

>Kinked curves produce kinked surfaces! this model has designed kinks all over it.
Please reread Michael Gibsons post 20 and 27. It explains why my surfaces have warped and twisted all over the place and have not given me the results I wanted or expected.

> Probably the best result for the kind of highly controlled blend that you are describing would be trimming
> a gap and then doing a 2-rail sweep with some custom profiles to >shape the blend, with an option to
> force the sweep to be smooth to the edges it is sweeping along. MoI does not currently have this option,
> but it is another piece that I >want to add in eventually... Again it is part of a kind of advanced toolset
> so it just has not been a focus for the earlier part of MoI's toolset.

This is what I have done in my model except the swept surface doesn't have the ability to maintain a G1 or G2 tangency blend with its adjacent surface (limitation of MoI and last time I used Rhino it didn't have this feature either). Even if it did I'm not sure if my surface will still buckle and twist due to the cross section transitions I need the nurbs to make.

Not to worry though, I'll keep on eye on MoI's future development and like I said will still be using MoI for the detail components of my aircraft where booleans are far easier to do than polymodelling them.

Hi Burr,

> It seems there is an opposite conundrum here in that,
> NURBS would do the REAL object and Poly can easily
> replicate it to look somewhat real. Not the other way around.

It's pretty much a non starter to use polygons/subd for the "master definition" of an engineering model, because they are unable to represent conic section shapes like circles, spheres, cylinders, etc... You can get closer and closer to a precise sphere shape by using a lot of points, but it is inefficient to use a lot of points and still does not get a perfectly defined sphere. This is really bad for mechanical parts which tend to have lots of things like circular holes in them.

So that's the fundamental reason why NURBS is used for the "real" engineering tasks - it is a kind of unified object definition that can have exact circles, or also wiggly surfaces not just only one or the other.


Polygons in their simplicity though do give an easy ability to tweak and edit them, you are more free to grab a point somewhere in the middle of your shape and pull it up a little bit, or widen a piece a little bit by yanking points around.

The NURBS toolset does not tend to work like that and instead has more of a "construction" approach with building surfaces from drawn curves, booleans, cuts, fillets.

If you want to work more of in a "sculpting" type way of kind of nudging and tweaking your surface, then polygons work more for that.

If you want to work more of in a "drawing" type of way where your pieces are more defined by profile curves and you want to have things like fillets where pieces join, then NURBS works more for that kind of thing.


A lot of times organic modeling for stuff like faces and characters fits more with the "sculpting" method and mechanical and industrial type shapes fit more with the "drawing" and profile driven method.

An airplane fuselage and a car body are examples of man made objects that have a much more organic shaping to them, that can sort of put them on the "sculpting" side of the fence moreso as far as what is easiest to mess with them.

No matter which one is easiest for an individual person trying to whip out a particular model for rendering, for actual engineering "master plan" data the NURBS method is always the one used because it has that ability to have stuff like exact circles.


People build models for all kinds of different reasons though, if you're rendering the object and not constructing it, it is probably not a problem to you that your sphere is 0.05 units deviated from a true sphere because it is hard to see that in the rendering.

So the right tool for the job depends on a bunch of different stuff - what your goal with the result will be, etc...

- Michael

Hi Kevin,

> This is what I have done in my model

No, not quite - what I was talking about there would be using the kind of "larger independent component" parts like I was showing earlier. You were mentioning that a fillet would not do the right thing that you wanted where the pieces connected - you would use a sweep in there instead of the fillet, kind of like this:

The 2 larger components booleaned together:



Then since a fillet does not give you the right shape, you would trim away some parts to make a gap:



Then the sweep would fill in that gap making a kind of custom fillet - the sides of the gap would be the rails for the sweep and you would put in some cross-sections to control the shaping of the blend to get more control over it.

That would be the way to solve your problem of a standard fillet not being what you wanted in between those 2 pieces. But it will be a while before that would work with continuity in MoI.

- Michael

Thanks Michael

I understood what your meant even without the images or extra post and thankyou for you help and advice. This is pretty much how I have gone about creating the fin and cockpit in my model. I've booleaned off the areas of the aircraft where the fillets begin and used curves to define the ends but the transitions of the curves are just too tough for nurbs it seems and results in a lot of surface joints that look bad on the meshed model.

I still leave it open to Burrman, Brian or anybody who wants to have a go to show me how this aircraft should be done with nurbs. I would be most interested to see their effort and how they went about it. All the reference photos they need, blueprints and MoI setup is contained in the first post of this thread. My only requirement is that they closely capture the look of this aircraft as it easy to distort the shape of the aircraft to suit the software.

The challenge is there and there is no time requirement

Hi Kevin,

> This is pretty much how I have gone about creating the
> fin and cockpit in my model.

Is that in a version that you did not post any screenshots of?

The reason why I ask is the last one that I saw was this:




That one is definitely not done in the same way, again the difference is an edge-to-edge type construction there rather than building separate parts that intersect one another - that is that punch all the way through each other.

At any rate, the actual method that I was describing is not possible with MoI yet, I'm talking about a way that would be possible in the future with some enhancements to the sweep command to ensure continuity. It would be what would be needed to use instead of the fillet for joining the "solid" components together since you don't want a standard fillet there.

- Michael

Michael

Yes that is correct, this is the last image I posted and its current stage of construction.

I've created a series of swept surfaces to try to recreate the complex shape of this area (Unsuccessfully).

Your last post also needs the user to create a swept surface to create the subtle curvature around the cockpit.

For anybody who wants to have go modelling this aircraft and has difficulty understanding what shapes need to be created I attach a polymodel version of the fuselage. Perhaps these images make it more clear of the complex curved surfaces involved around the cockpit and fuselage/fin blend. As you can see it is not a one radius fillet but a series of complex curved surfaces that all seamlessly blend together.

Seeing this was my very first, ever, plane model I was pretty pleased with my 75 year olds new MoI skills.

Had fun anyway.

Brian

Hi Kevin, I wish I had enough time available to work on all the small details that you're really interested in. But unfortunately it is just not going to be feasible for me to spend much time on this.

But here is generally what I would probably go for with a NURBS approach:





Model is attached as PantherNURBS_3dm.zip .

I fully understand that this does not have many of the details that you need, but it may give you some ideas of what could be possible.

But I don't think though that there is really any good reason for you to switch off of your polygon modeling workflow for this task. At the level of detail that you want to achieve, I don't really think that you are going to gain any benefits by trying to do it with the NURBS toolset, there are a lot of detailed things and problem areas that you would have to sort of navigate through. If you didn't care so much about the specific details you could possibly get a speed boost for this but with the target that you have in mind I think you'll be looking at decreased speed for this particular type of task. And that would be after a ramp up period to get accustomed to a different toolset. It does not leverage the "sweet spot" of NURBS that would deliver a speed increase for things like heavy booleans.

At any rate, you can see the results here of some of the more component based structuring that I was referring to earlier, and can at least show that you can get good surface quality with the right kind of strategy.

- Michael

>I wish I had enough time available to work on all the small details that you're really interested in.

Michael, thanks for taking the time to do this. Much appreciated. You've given me some more food for thought with this post. When I have some more time I'll try build on your idea to see if I can develop it further.

>I fully understand that this does not have many of the details that you need, but it may give you some ideas of what could be possible.
Thats fine, it shows the principle which is the main thing.

>I don't really think that you are going to gain any benefits by trying to do it with the NURBS toolset,
>there are a lot of detailed things and problem areas that you would have to sort of navigate through.
Fully understand that and I will continue to do some parts of the aircraft with nurbs to speed up the workflow. For example on my polymodel of the F9F it might be better to do the wings and tailplane with nurbs where I can do some quick booleans to cut out ailerons, elevators, trim tabs and rudders etc as this is always a tedious task on high poly model. At the moment it is good to have both a nurbs modeller in MoI and polymodeller. At least with both I will always have the right tool for the job at hand.


You've made some great posts here about the best way to go about creating models with nurbs or the philosphy behind it. Maybe those posts should be compiled into a new sticky thread for reference.

Cheers

>Seeing this was my very first, ever, plane model I was pretty pleased with my 75 year olds new MoI skills.
Looks pretty good Brian. Certainly interesting camoflage.
It reminds me of the Salvadorian type camouflage http://hsfeatures.com/features04/fg1ddwa_2.htm

Hi kevjon,

I can understand your frustration on this. Myself, I am the complete opposite to yourself, as I come from a background of using nurbs packages. I did move on to trying poly programs and found them very limiting to what I was used to. It did take quite a while before I was comfortable with poly programs before I added them to my pipeline (for home hobby), but even now still prefer nurbs.

I do agree that MoI could be seen as limited in the toolset, but that is down to possible comparison with other packages, (and there is also a need to consider the price on those other packages), but what is within MoI works quite well, various functions that are missing, such as the variable fillet you mentioned, can quite easily be worked around using such as the blend, which, although at times it cannot be used well directly with split surfaces, it is just a need to blend projected to surface construction lines and sweep.

I will try and find time tomorrow to build in MoI part of the model you mention.

Regards,

- Steve

Hi Steve, thanks for your post.

>I am the complete opposite to yourself, as I come from a background of using nurbs packages.

My background is Autocad and Inventor which I use all day, everyday at work for past 15 and 5yrs. Both those programs would be totally unsuitable for a project like this or to model up a car. Both are great for architectural and mechanical type work though.
As I have a interest in aviation and creating digital aviation art I decided to search for the best tool for the job. At first I tried Rhino V2 and V3 but found getting a a good clean mesh from that program problematical. The problems are mostly bad mesh at the joints between the surfaces which creates shading errors when rendering the model. I hoped V4 would address this area but they have done little to no work in this area and may tackle it in V5. I then decided to learn poly modelling and found that much more suited to creating a nice smooth seamless mesh that renders beautifully from all angles. It also has the added benefit of being able to continually tweak the model until I am satisfied it looks pretty close to the original. This of course is all done by eye using photo reference as I do not have the original construction drawings of the aircraft, just some blueprints aimed at scale modellers to correct their models.

I will be most interested to see your attempt at the model and I appreciate you taking the time to have a go.

Hi Kevin, I thought a few details on how I arrived at that model and a few other tips and background might be useful.

If you want to end up with a shape that has a kind of long and evenly smooth type form it is really hard to get that by building things only in small individual panels that touch edge to edge. Even when you have continuity control for matching the edges it is still not a good way to do it, because the continuity control will tend to bend around the area just near the ends of your patch, which produces kind of lumps which are technically smooth where they touch edge to edge but does not have the kind of totally even distribution that you want in a case like this. For it to work your initial curve network has to have a very even curvature distribution in it and it is hard to create that.

There are several industries that do construct their smooth things in a panel by panel basis, however it is a very time consuming way to do it. They have some fancy software to help a bit, but really the role of the fancy software tends to be fairly exaggerated for this particular stuff, the results are more about a lot of expertise in setting up the framework and also in nudging surface control points around, and probably the biggest ingredient is a lot of time and effort spent on the problem. It is just not a very practical method for your kind of situation.

So for a project like this I would really avoid that kind of a strategy.

Instead to get smooth pieces it works a lot easier to build surfaces that are larger and also try to be more sparing in the number and complexity of curves used.


So I started by drawing a profile following the edge of the fuselage in the top view.

Then I took a look at the front view, and I noticed that this cross-section pointing to in red here:



seemed to be an exact circle.

So exact circle made me think of revolve, one of the nice things about revolve is that it will always make a very predictable and evenly shaped smooth form. So it is nice to use it when possible.

So the revolve builds a large smooth section of the fuselage all at once:




That's a good example of the kind of componentization that I was referring to earlier.

Forget about some of the other pieces like the canopy for a minute, just focus on getting a particular element of the model right. Later on the canopy will get built separately and then blended in to the main shape.

It probably would have been a good idea to rotate the profile 90 degrees down so that the seam was on the bottom, that would just make a more symmetric topology for later on when putting in wings.

Then looking from the side, it seemed like the front part was good, but along the back part there is kind of a rising section. So I turned on surface control points and pulled some of the up to stretch out that part a bit, and to start some of the transition into the tail:






Then I traced a profile of the canopy from the front and side:



Note how the side profile is extended beyond where the line ends stop on the blueprint, this kind of extension is a big ingredient for building components so that they punch through each other and can be intersected:




First I tried doing a sweep, using side profile as the rail:




But that kind of gave a sort of flat-sided shape (may be just I did not pick a good profile to do...), I noticed in the photos that the canopy has a lot more rounded bubble type shape to the sides:




Instead of messing with the profile much, I decided to try a rail revolve, which is a variant of revolve that can also take a shaping curve. So in this case the revolve profile is the long side view piece, the front profile is the rail, and the axis is snapped to the endpoints of the long side view piece, that generates this individual piece:



And together with the main body:



Then to fuse them together, select both of them, run Edit/Trim, at the "Push cutting objects or Done for mutual trim" prompt, press done or right click to say that you want one to cut the other, switch to Mode: Keep and pick on the 2 pieces you want to keep and then Right-click and the inside pieces will removed. Then join them together, and I used Fillet to round off the edge, but switching the Shape: option from the standard Shape: Circular to Shape: G2 Blend - that will put in a more organic connecting piece that is not an exact circular shape, it tends to make for a more seamless and less distinct boundary to the blend.

I hope this may be useful to you or others, I mean you can see the results give you smooth surfaces without the kind of warping or bunching that you were seeing with the other kind of strategy.

Unfortunately I probably won't be able to do any more extensive posts on this, got to get some work done on the software too! :)

- Michael

Thanks Michael, some more great tips and ideas posted there.

With the canopy surface you've generated I could not cut the windshield to the right shape. I think you solution in the canopy thread was a far better although that is a trial and error exercise with the windshield side profile. However I do understand the principle your getting at and I think you've come to realise how difficult some of the shapes that make up a aircraft are. Same goes for cars.

I actually built the fuselage of my model as a revolve as that was the simplest and most accurate way to do it. It also highlights the errors in the blueprints used as the side profile when revolved did not match the top and bottom profile of the aircraft but from the front view it shows the aircraft as a perfect circle.

I like the point pulling idea to give me a leg up in the tail area and reduce the seams in the model. The only problem is that a user can't point pull after a surface is trimmed so he can't tweak the shape till it looks right compared to photo reference. Nor can the user define how many points should be put on the surface to aid them in the point pulling exercise. As we have discussed point pulling is probably best handled with a subd poly modeller which is designed for that purpose.

I don't want to build the model panel by panel (even if a could) as it would take to long and would be difficult to keep it all straight and aligned. I'm interested in reducing my time to create each model and creating a nice smooth mesh that renders perfectly from all angles. That is my one and only goal when it comes to a project like this.

Hi Kevin, yes I still don't recommend that you use this process in place of your existing one, like I mentioned previously the level of adjustment and detail that you want to achieve would not be any easier and most likely more time consuming with this method rather than your current one.

This does not change any of that.

However, I wanted to show you what the suggested approach would be if you (or someone else) did want to build this using NURBS. You were previously saying it was not possible to build such things with NURBS because of bunching and warping issues. I wanted to show that there is a method that can be used that does not create warping and twisting, but you do have to use a different strategy than your original one.

I mean you can clearly see that the overall form that I created there does not suffer from bunching or warping issues.


> With the canopy surface you've generated I could not cut the
> windshield to the right shape.

Yes - please understand that I did not take a whole lot of time to do that shape, if it is not the correct shape you would need to edit the curves and tweak things. That is not a whole lot different than your polygon modeling approach - doing things in polygons does not automatically create them at the perfect shape immediately, you have to do a lot of tweaking to manipulate the shapes. The same principle applies here as well.

So even though you could not generate the windshield to the right shape with the specific canopy I have created here, I hope you do not think that it is not possible to create a suitable shape with further edits, because that is certainly not the case. I wish I had time to demonstrate and prove to you that it is possible to create every single one of these little details to a very exact specification, but I'm sorry I just do not have the time available to do that.

I would probably have to spend quite a bit more time studying a lot of details of the shape to become more familiar with it before I could try to create something to meet the level of detail that you are targeting.


You're certainly right that it is a challenging shape to work on!


> The only problem is that a user can't point pull after a surface
> is trimmed so he can't tweak the shape till it looks right
> compared to photo reference.

No, that's not quite accurate - you can still turn on points and manipulate them on a trimmed surface, but not on a joined assembly of surfaces, you have to use Edit/Separate on a joined together cluster of surfaces to break them into individual pieces and then they can be point edited.

But there is a general problem of after trimming if you yank things around you can pull trim edges apart from one another.

Again though the general strategy is to form the shape of each component more correctly and fully before you do any trimming. You would not want to trim and combine them until you liked their shapes.


> Nor can the user define how many points should be put on
> the surface to aid them in the point pulling exercise.

That's not quite accurate either, for example if you use Rail revolve that will create an output surface that has the exact same control point structure as the input profile and rail curves combined, so you can insert points as you want on the input curves to control the point structure of the final result.


> As we have discussed point pulling is probably best handled
> with a subd poly modeller which is designed for that purpose.

I agree - definitely it is a focus of that toolset and once again if that is your main focus for how you want to operate then I would recommend going with that toolset instead.

Again, my last post was not an attempt to convince you to use NURBS instead of your current workflow, but rather as an explanation on how one would go about this task if you did want to use NURBS for it.


> I don't want to build the model panel by panel (even if a could)

The reason why I mentioned panel by panel, is your previous post here:
http://moi3d.com/forum/index.php?webtag=MOI&msg=2036.31
showed that you were using a panel-by-panel type approach when you were trying to build this in MoI, and you were complaining about buckling and seams between pieces. That strategy is not going to be fruitful with NURBS, you have to use a different strategy of making components that intersect like I was showing in my last post. It is a totally different way to approach the construction than the way you were doing it...


Again, sorry if my last post was not clear, it was not meant to convince you to abandon your current method of working in polygons, but just to show you what type of an an approach you would need to adopt to construct such a thing with NURBS without the kind of buckling, warping, creasing problems that you had encountered.

It is not an easy method, it kind of requires use of many of the detailed and advanced parts of the NURBS modeling toolset. It may require you to do things like subtle edits and visualizations of how pieces intersect before you get a good result. I do not really think it would save you time in your particular case of needing such very minute details and tweaking, but for other people's purposes who are not so particular about replicating very minute details it could be useful and just generally shows off the approach that you would use for creating more seamless skin type models using NURBS.

- Michael

Hi Michael

To some extent when modelling anything you have use a panel by panel approach. Why am I doing that ?

The answer lies here in your own statement
>It may require you to do things like subtle edits and visualizations of how pieces intersect before you get a good result.
It is pretty darn hard (impossible) to visualise what shape the canopy needs to be so that when the windscreen is booleaned from it that shape matches the look of the real thing or plans. Where it is far more logical to create the windscreen shape, get it to look right by tweaking curves and then create the canopy shape from there. Which as we have seen doesn't work. And yes this is also how I would do it with polys except this approach does work.

The other reason is that the plan view & side view shape of the cockpit opening and windscreen are known entities from the plans so I am trying to model up to those know entities. What are not know entities are all the subtle blends that lead to those items which can only be done by eyeballing the shapes from photos. The plans are not much help when it comes to the shape of all those blends.

>I mean you can clearly see that the overall form that I created there does not suffer from bunching or warping issues.
Yes, I can see that the way you have tackled the transition of the fin from the fuselage is a good approach much like polymodelling.
But trying to get the shaped pulled to match the airfoil shape of my fin would be almost impossible to do on a first attempt. Once I've trimmed it so I can see what cross section shape I have got, I can't change it ? I would have to undo, trim again, nope still not right, undo, tweak, trim, undo, tweak trim undo etc. Bit of nightmare for someone trying to model this thing accurately. Sure if you're modelling fantasy stuff, who cares what shape it is.

As I stated in Jasons Clarks thead, nurbs almost gives you the shape you want but without the ability to tweak those surfaces to get those surfaces looking right compared to your reference photos, the user is in for a lot of trial and error and a very frustrating time. But as you've said if it were possible to do this, all your joints would start springing apart.

So this all comes back to the wrong tool for the job at hand. Best to use nurbs for what it does best which is boolean operations of more mechanical type parts rather than the complex curvatures and blends associated with trying to accurately portray a real car or aircraft.

Michael
I feel we are both going around in circles. Your telling me how nurbs works and how I should model. If you feel your not getting through to me your wrong, I fully understand exactly what your saying. For the reasons I mentioned above (please read them carefully), I'm sorry that approach is beyond the average modeller or even the above average modeller who wants to model a car or aircaft accurately which is probably the reason there are so few Nurbs cars or aircraft that are accurate looking representations of the real thing. All this work being done with polys. I honestly wish this was not the case as I like the precision and speed of working with nurbs but unfortunately for projects like this trying to get nurbs to bend to the complex curvature of this aircraft is just too difficult and time consuming for the average modeller.

Aside from solving all the modelling problems, getting your nurbs model to good clean mesh ready for uvmapping and rendering is another whole set of problems.