Fusion 360s T-Spline performance

[TrippyLighting]

The image below shows a redering of the object/file in question that brings Fusion 360 to it's knees.

You can read more about it here and perhaps this helps to explain that the perfomance problems are a result of attemting to do "real" stuff 😉

The curved arm with the solid surface spiraling around the grid structure were created in Blender by replicating a simple low polygon mesh (array modifier) along a spline (Curve modifier). The resulting mesh has 19k quad faces and was exported as a .obj file. This was imported into Fusion 360 and converted into a T-Spline. 

1. The initial conversion of the imported mesh into a T-Spline took 5 minutes. 
2. Loading the file in it's currnt state takes 5 minutes. Importing the 
3. right-clicking on the LED arm in the browser slecting"Save as STL" it takes 5:30 minutesfor the STL coversion window to pop up. The default from having exported this previously was "Medium" for refinement.
4. Once the pop-up window came to life, actualy exporting the data as a .stl file took another 5 minutes.
5. Saving a new version to the cloud takes 5 minutes.

When performing solid modeling edits and simply zooming in/out and navigating in the viewport  Fusion quickly warns that Graphics memory is low (machine specs below) and turns of the antialiasing, which is the only eye candy I had enabled.

The T-Spline cannot be edited as it brings Fusion to an absolutely grinding halt. It does NOT crash Fusion 360 though and that is relly a good thing. Fuisnon 360 simply becomes unresponsive.

If generative designs are to happen in Fusion 360 we need much better performance for T-Splines.

I am on a Charter Communications cable modem with 4-5Mbit/sec upload rate and 60-65 MBit dowwnload rate (as measured by speedtest.net). Loading and saving files to teh Fusion cloud should really not take an eternity!

[jeff_strater]

Hi @TrippyLighting.  I don't disagree.  However, you are pushing the limits of TSplines a bit with 19k faces, so the performance does not surprise me too much (nor the file size - I suspect it's pretty big).

In my mind, this is a job, not for TSplines, but for native mesh modeling (which is what Blender is).  TSplines is really meant for a low-density mesh/scaffolding, where you are relying on the smoothing algorithm in TSplines to get you the nice shape you want from this scaffold.  The good news is that we are working on a mesh modeling environment.  The bad news is that it probably won't be available as soon as we all would like.

Jeff Strater (Fusion development)

[TrippyLighting]

While it is great to hear that mesh modeling will at some point be part of Fusion 360, that's not the limitation with this model.

I was perfectly fine modeling this with nother application 😉 The .obj file ist fine good enough for exchanging this sort of data.

The desired end result of this was to 3D print these relatively small parts and in fact  that worked perfectly fine with the STL exports from Fusion.

I would also have been perfectly happy to export a subdivided mesh to get enough mesh resolution for  smoothly printed result, something the base(control) mesh was not able to provide.

The real limitation for this sort of thing is not Fusions lack of mesh modeling abilities but it's dismal mesh handling abilities. The current limit to convert a mesh into  b-rep is 10k faces.

The imported low resolution control mesh had already 19442 qaud faces. I needed at least one level of subdivision (77756 quad faces) or even 2  (622084). That's what Fusion 360 is up agains when having to deal with meshes! Perhaps we need Fusion 360 internal subdivision before any modeling abilities so we don't have to exchange that excessive amout of mesh data.