Mstn+Luxology vs 3dMax: Preferred AEC rendering app?

Good story  dwy.seah,

I think MicroStation could indeed be improved on some points, but if you don´t have to model, materialize and render a dinosaur, or make advanced animations, why should you already buy an expensive 3D studio Max licence next to MicroStation? Does the extra costs of the program weigh up against the few extra functions you maybe use? 

I think that for architectural or engineering visualization MicroStation is an excellent 3D application. The quality of the end product is in my opinion, already comparable with advanced 3D applications like 3D studio Max. The biggest influence is the knowledge of the user. Look for example in the Gallery to images of Kurt S.

Regards Louis.

Louis,

" why should you already buy an expensive 3D studio Max licence next to MicroStation? "

I can only speak from the architectural side, and I would say that most firms do NOT use Mstn, even those who are Mstn shops, as their viz application. The most popular package is 3dMax.

I have always thought this to be a problem, as there was always the translation barrier and the coordination headache, because last minute changes tended to be done in Max and the rest of the CAD info would fall out of sync, among other problems.... not very BIM-my.

Regards

Dominic

If everything was imported as geometry than it's no wonder there are problems. Hopefully, this instancing / LOD stuff will be solved soon.

Anyways, does anyone have some DWG's with materials etc for testing? SS3 apparently will include more support for rendering info. Please forward to Don.

The problem is though even using with shared cells approach for such geometry, these objects are so complex that the scene is going to get quite large and quickly if using many different complex models due to MS's requirement that at least 1 cell definition needs to be in the file. With a Vray proxy you take a max model, apply the textures and mapping then export a vrmesh file. Though you can see it in the view ports and copy the mesh many times and it does not exist in the scene as the geometry is only imported from the external mesh at render time only.

Yes..... I hear that MR or V-Ray  proxies in Max can be stored as external meshes with materials that are only called at render time by the renderer. A lot of game apps use the 'deferred rendering', and multi representation type replacement. In the GIS sector, this example DILAS is actually associated with Bentley, but uses a specialised terrain viewing app G-Vista.

Modo already has instances (sort of like shared cells, but without the pesky drop and redefine to modify restriction) and replicas (based on particles/points that are stand-in's for render deferred objects). Replicas are lightweight, but are not meant to be manipulated/transformed. The interesting thing is that replicas are recursive and can include groups of elements.

In Max, it's possible to use bitmaps based on aerial photographs to distribute or paint particles/proxies for trees etc. Draped rasters on contour meshes are a bit old hat and don't give near enough fidelity close up. Thematic maps showing vegetation location are now available, and there are even algorithms to separate out trees from buildings in point cloud scans.

It's probably more a QV or Mstn data issue and not a rendering/LX one? Mstn already has 'fast curves/cells' view attributes setting, which can be limited to individual reference attachments. Maybe Mstn needs some more adaptive and LOD-aware display styles.

1. View distance / occlusion aware: Near / visible cells are more refined, far off cells dynamically degrade to anonymous bounding box or point cloud (bounding box vertices or centre points if far off enough) representation. Voxel-style sorting?

2. Maybe there should be different levels of degradation/detail: Bounding box (with max size, that generates separate boxes for outlier elements), planes for groups of elements on a plane like 2d drawing refs, lines (or cylinders/cones) for linear elements, points/particles (via Pointools?) or plain circles/globes for discrete stuff. Bounding boxes alone is pretty limited and not very informative.

3. Hierarchy / nesting aware: Maybe proxies can be restricted to nesting depth or pre-defined cells like 'bolts'. I guess this is what proxies are already in Modo/Max.

4. Option to leverage GPU-based texture billboard handling. RPC's on cruciform planes for trees, baked normal maps for building facades?

5. Instances vs near-Instances: Just wondering if it is quicker to treat everything as a reduced number of optimised proxy 'prototypes', and only differentiate/refine the objects in the foreground or snapped/selected? Quick and dirty sketchy display style for speed?

6. Adaptive Nurbs vs Mesh vs vertex cloud representation? IntegrityWare SubD-Nurb switcheroo.. good B splines?

7. Dynamic Shrink wrapping ? Internal or mostly enveloped elements are automatically culled, unless exposed to view.

8. Can Bentley Map's XFM be used to guide things? Mapping features already simplify or remove themselves based on zoom / camera position..?

9. ....?

In V-Ray 2.0 you can specify a low-poly object to represent proxy object.

VERY useful.

"The problem is though even using with shared cells approach for such geometry, these objects are so complex that the scene is going to get quite large and quickly if using many different complex models due to MS's requirement that at least 1 cell definition needs to be in the file."

I3loom

Maybe the way around this is to deal with this recursively, as Modo does with replicas..... I think. Or a compression app like WInZip. Given that ideal conditions for the display system is to have the maximum number of identical instances warping through the pipeline, is there a way to decompose and sort the elements for the maximum efficiency? Is there a way to produce the variations to cover the end of the 'long tail' progressively, like preview rendering ? Procedural decompression.. based on an L-system for trees, or a split grammar for buildings like CityEngine or CSG for TF or GC-generated solid.

Maybe the proxies should have multi-representational libraries, like VRay.

1. Minimal representation is a point with good floating point accuracy. I guess linear elements would have two points.. etc.

2. Preview options could include a medium resolution mesh, or a RPC-style billboard, or a partially constructed L-System / CGA-generated mesh, or a TF / GC-generated BRep, or a mesh with pre-baked UV texture, normal, displacement etc maps. Shrink-wrapped as required.

3. And the final 'beauty pass' fully detailed item.

The cool thing would be the different representations in the library updating automatically based on changes to any one of them. Mstn would load the best proxies based on the hardware and editing context.

"The problem is though even using with shared cells approach for such geometry, these objects are so complex that the scene is going to get quite large and quickly if using many different complex models due to MS's requirement that at least 1 cell definition needs to be in the file."

I3loom

Maybe the way around this is to deal with this recursively, as Modo does with replicas..... I think. Or a compression app like WInZip. Given that ideal conditions for the display system is to have the maximum number of identical instances warping through the pipeline, is there a way to decompose and sort the elements for the maximum efficiency? Is there a way to produce the variations to cover the end of the 'long tail' progressively, like preview rendering ? Procedural decompression.. based on an L-system for trees, or a split grammar for buildings like CityEngine or CSG for TF or GC-generated solid.

Maybe the proxies should have multi-representational libraries, like VRay.

1. Minimal representation is a point with good floating point accuracy. I guess linear elements would have two points.. etc.

2. Preview options could include a medium resolution mesh, or a RPC-style billboard, or a partially constructed L-System / CGA-generated mesh, or a TF / GC-generated BRep, or a mesh with pre-baked UV texture, normal, displacement etc maps. Shrink-wrapped as required.

3. And the final 'beauty pass' fully detailed item.

The cool thing would be the different representations in the library updating automatically based on changes to any one of them. Mstn would load the best proxies based on the hardware and editing context.

The most annoying part is when you create a scene with cell library window. Refresh the view, no one knows what has to go to the end, if you accidentally pressed something that starts from the beginning. As you move the window to refresh starts again, if you write the cell as shered it again. This is some horror if you load a file size of 30MB (the tree). Maybe we do something about it. I noticed that there is still a view phong. In views, as others have said, should be minimal linear representation of complex elements.

kopernik123