I created a sketch, then patched it, and when I tried to extrude a surface from one of its edges, its normal is set in opposition to the first surface:
It would make much more sense for it to be on the exterior side. The weird thing is that, when I first tried creating what I need (I have to re-create it due to reasons), it worked correctly:
What could be the reason?
Surface normal is somewhat arbitrary. Though it may be obvious to you what "exterior" is, for a simple L-shape like that you could interpret it differently for different final shape. It is not "incorrect".
At any rate, if you don't like the orientation, you can reverse it:
@iulianOnofrei wrote:
Makes sense, but then, if I create the same thing (albeit with different dimensions), with the same steps, in the same order, is it expected for the normal to still be different in some cases?
Yes. I'm not sure what you mean by "different", but, as I said before, normal is assigned somewhat arbitrarily, and does not understand your intention. For instance:
It seems the normals are not created randomly.
In the case of extrusions from single lines, the outer side is always created on the right in the direction of travel.
In the case of extrusions from angled lines, the outer side is always on the outside.
@iulianOnofrei wrote:
> Though it may be obvious to you what "exterior" is, for a simple L-shape like that you could interpret it differently for different final shape.
So, @wersy's reply completely invalidates this. It IS obvious for the app as well. Meaning, for it to keep a continuity of the same normal directions for linked surfaces.
Jeff is the current director of software development at Autodesk. He played a major role in developing Fusion and, prior to Fusion, worked on developing Autodesk Inventor. A key point in learning is not just learning to listen, but also who to listen to!
@iulianOnofrei wrote:
I don't believe in titles, I believe in arguments.
I don't believe in titles either, but I respect experience. I've met Jeff in person and exchanged many messages with him on the forum and in a few Zoom meetings over the last 10 years.
He is a software developer with decades of experience developing the software we are using.
@wersy provided feedback based on his experience and observations. However, unlike Jeff, he has no insight into how the geometric kernel used in Fusion works or how Fusion is programmed.
I've worked with CAD professionally for over 30 years and in 3D since 1998 (SolidWorks). I've worked with Fusion for 10 years this November. Surfacing is my specialty in Fusion, and I am sure @wersy can attest to that. I have not found a repeatable pattern of how Fusion decides on surface normals. Whenever I think I did find a pattern, the next surface model makes me reconsider.
My journey into computers began with programming some 35 years ago, and I still do a little embedded programming. I am convinced that it isn't completely arbitrary or random, but I don't assume that it would help or even be relevant to modeling.
Traditionally, the main purpose of CAD software is to design and manufacture physical artifacts. As such, the desired result is a solid body, a BRep, and not an infinitely thin surface model. When you stitch surfaces that enclose a volume watertight, that creates a BRep. A BRep has a defined inside and outside, unlike an unstitched surface model, and Fusions will flip the normals as required to make a valid BRep.
If a surface normal points in a direction that isn't useful for a particular model, use the Reverse normal command in the surface tab.
I found it to be a very rare occurrence when I have to use it to create a successful model from the surface.
I now understood your point of views. Since I was not explicit enough in my question, I will try to explain it better.
---
I understand that a single surface could have an arbitrary normal, either like this:
or like this:
(However, note that from my tests, the normal's direction was always towards the bottom, probably due to the existing context of the default origin.)
But, a second, perpendicular and connected surface, has no reason to ever be created like this:
or like this:
It has enough context to easily be created like this:
or like this:
---
Hope this clears all the confusion related to my question.
First, a couple of clarifications: @wersy - I never said that surface normal direction was "random", I said "arbitrary". There are a definite set of rules which are followed. Curve direction is one of those, but there are more. I just cannot, without hours of investigation, tell you what, exactly, those rules are. But, it is definitely intended to be deterministic. Splines, especially, are fiddly when looking at surface normal.
@iulianOnofrei. if you have a case where the exact same set of steps, on the same geometry, produces different results, I would very much be interested in seeing that case. That would be a bug.
Second, in your example, I assume that the "bad" surfaces, such as this one:
were created as two separate surfaces, not one. If this is a single surface, then I agree, that is a bug, and please share a design or a repeatable set of steps to reproduce it. If not, and you then stitch these together, I suspect you will get a self-consistent answer.
Regarding: "It has enough context to easily be created like this:", if we are talking about two separate surfaces, then, yes, technically, there is enough information to make some assumptions. But, that would requires a pretty complex set of logic. On the second extrude, Fusion would have to:
Given that you can flip the surface normal, and that stitch will make them consistent, I feel confident in saying that Fusion will never go to those lengths. Also, there will be cases where there is no right answer, like in the video below. That third surface cannot be created "correctly", because either orientation would be inconsistent with one or the other of the adjacent surfaces.
@iulianOnofrei wrote:
Exactly, it looked like a bug, ...
Thanks for reporting!
This isn't one, but in your journey with Fusion you will likely discover some bugs. Please don't hesitate to report them.
Even if it has already been reported, the more cases and data that can be added to a ticket, the likelier it has a chance of getting fixed!
Right. These are two surface bodies:
and, if you stitch them together, the result is consistent:
Thanks for asking the questions, though. There may well be bugs in this area.
It looks like I might have to eat my words!
Sometimes it's the simplest of models that highlights a bug.
In this video I took a model found on the German side of this forum. The first surface extrusion is created at an angle. I change it to a two-side extrusion, also at the same angle. It creates the 2nd surface with a normal opposite of the other surface. This is a single surface body with two surface normals pointing in opposite directions.
I would have expected both normals to point int he same direction.
Actually I would not even have expected to have a second surface, but a single surface given it is extruded at the same angle.
So now i have to unstitch the two surfaces and then stich them. That corrects the issue, but that should not be necessary.
