Incorrect fit point spline update bug ?

[martinYFAFK]

I'm having issues with unexpected behavior of a fit point spline. The shape of the spline is dependent on the order of operations when applying dimensions and constraints. The sketches in the pictures be below have the exactly the same dimensions and constraints, yet they look very different. In the first picture I put the 50mm vertical dimension before adding the fit point spline, in the bottom picture I first add the spline and constrain it, and then added dimension for the 50mm vertical line. Adding the vertical dimension rescales the sketch lines but doesnt seem to update the spline. This is really weird behavior and totally unexpected. 

On top of this patterned splines wont update if I do a circular pattern and change the original spline. I manually have to go in and edit the circular pattern and update the sketch. This is really bad behavior since all other sketch elements will update automaticly. 

I have attached the sketches to the post to verify 

[martinYFAFK]

Here's an overlay of the two pictures to highlight the differences

[TrippyLighting]

Yep, this is known behavior. If you can answer more about what you want to model I cal likely show you workflow(s) that prevent that work around this behavior. I personally avoid mirroring and patterning in sketches as much as I can.

[martinYFAFK]

Wow, ok. It blows my mind that it can be known and not fixed. To have geometry that is considered fully defined, yet clearly isn't in a professional cad package is just unbelievably bad. How am I to trust the geometry created by the program ?

[TrippyLighting]

---

@martinYFAFK wrote:

Wow, ok. It blows my mind that it can be known and not fixed. To have geometry that is considered fully defined, yet clearly isn't in a professional cad package is just unbelievably bad. How am I to trust the geometry created by the program ?

---

Using a different tchnique that can be trusted would be one way 😉

If you can share what you want to design I can probably help you.

[martinYFAFK]

Thanks for the help, but lets say I have designed a product with a few splines. During the validation phase we discover that some dimensions needs to be adjusted. The correct solution is then to delete all fit point splines, update the dimensions and re-draw all of the splines, even if nothing is patterned ? 

The issue here is not to get to where I want - I have solved that already after a few hours of debugging. The issue here is that the program is not behaving in a predictable way. To get different results weather I put down one dimension before the other cannot be considered anything other than a bug. A fit point spline that is fully defined with all the degrees of freedom cannot mathematically yield different results, it's just wrong. 

[davebYYPCU]

To say, that you have to delete (anything) for updating, 

is just a wrong workflow.

To avoid expert advise is not a good look either....

[martinYFAFK]

Thats exactly my point. I shouldn't have to delete anything to update the geometry, what I'm seeing, and trying to report here is that the spline isn't updating correctly. Maybe I'm not being clear enough, so here's a picture to explain it better:

Pulling the handles of the first spline to force it to update doesn't change anything. 

(edit: splitting up pic for better viewability)

[TrippyLighting]

Let's say that you have discovered a bug.

What are you going to do ?Get all upset and stop designing or find a work around ? OK easy answer correct ?

Because even if they'd try to fix it right away its might take longer then you can wait for completing your design.

An, maybe, just maybe you'll learn a few things in the process and then figure out that the workaround is actually not a workaround but a more robust workflow.

[wersy]

When I read this, in retrospect, it gives me the idea why I maybe had the same problem.

It concerns the hull of this model (simplified shown in this picture)

I drew the outer contour with a fit point spline and mirrored it. One sketch for horizontally and one for vertically. And in order to be able to adjust the shape afterwards, I parameterized the values.

After about 2 weeks of detail work and preliminary inspections in many places, I found that the hull was taller in height than in width. I could have lived with that, but it required a lot more effort to always mirror parts several times instead of just using circular pattern.

Since it was getting a bit cramped for the mechanical parts in the rear of the fuselage, I finally decided not to continue working on it, but to make a new model 20% bigger right away.

In my case, too, it probably had something to do with the order of the measurements. Before I continued working on the new design, I checked that everything was correct.

As long as it is not improved, you have no choice but to check it.

[TrippyLighting]

@wersy non need to mirror splines.

[martinYFAFK]

My intention was first and foremost to file a bug report to help improve the software I'm using and paying for. This forum was apparently the preferred way for Autodesk to get that sort of feedback, at least according to this help article: https://www.autodesk.com/products/fusion-360/blog/how-to-best-report-issues-and-get-help/

Maybe I should have been clearer with that. 

The spline is needed for a product that is going to be injection molded. I use the spline to define a quarter of the path for a sweep. Once I have done the sweep I mirror the body 2 ways to go from quarter to full body. The Industrial design defined the base shape of the product with a spline and I want to keep that same curvature, hence the need to use this specific spline. I don't see how there could ever be a workaround to get the same exact same shape as a spline without using a spline, or at least not anything that would not entail a ton of extra steps. But if you know a way, then you are clearly the better fusion 360 user. 

The issue was that the behavior was unexpected and it took a lot head scratching and unnecessary effort to figure out what was wrong, once I knew what was causing it I changed my workflow to accommodate the bug. I don't wish that sort of frustration to my worst enemy so I thought it'd be prudent to report it. 

The thing with the circular pattern was just a bonus bug report, as all other primitive shapes are automatically updated when you do a circular pattern it seems like a dangerous behavior from a UX perspective to have most, but not all, shapes auto updating. If that is intentional then there should be a huge warning flag when trying to pattern a geometry that isn't auto-updated

[TrippyLighting]

OK. That bug is indeed many years old and has been reported and argued many times. So perhaps this might be added to it.

I did not say not to use a spline. I suggested not mirroring or patterning a fit-point-spline.

Whether or not I am a better Fusion 360 user can be debated. I have more than 3 decades of professional experience in a number of CAD applications, 20 years of that in 3D.  Fusion 360 is only one of the tools I use.

My recommendation for splines is to work with Control Point splines for this sort of thing and not with fit point splines.

If you create 2 separate control point splines in the same or different sketches and dimension the control points identical then the splines will be identical.

This appears to work across CAD applications (or at leas the one I used in that case). I used another CAD app and created a control point spline there and the surfaces create from Fusion 360 and the other CAD application were identical.

As someone who can 

[martinYFAFK]

It'd be interesting to see the code behind this, testing a bit more I'm throwing a wild guess that the tangency weight (i.e. mapping of authority of each handle to a certain dimension) is calculated at creation of the spline, where the dimension of the handle is scaled to the overall dimension of the spline. Upon resizing of the spline this tangency weight is not updated, so you have to scale the handle dimensions with the same ratio to keep the shape constant.

[TrippyLighting]

Very interesting observation!

I'll share this on the Autodesk Expert Elite channel. Maybe this will help identify what can be done to get this behavior fixed!

[MichaelT_123]

Hi Mr MartinYFAFK,
The spline definition requires not only stating coordinates of its fit points but also establishing boundary conditions.
Dependent upon a level of a spline's degree it involves setting values of n-derivatives on both ends (for open curve only).
It can be done explicitly (perhaps by purist mathematicians) or implicitly with some nonchalant sloppiness of CAD designers. In the latter case, the process is mostly automated by applying an arbitrary algorithm. It is the pragmatic approach, as it would be quite cumbersome to apply all derivative vectors to the let’s say fifth-degree spline. F360 just does that automatically but it also sensing (my guess) a design context selecting the appropriate algorithm. It is also worth noting that the UI has no provision of dealing with higher degree derivatives which also influence the shape of a curve.
I reckon this is the slope you bounced into 😉. Is it a bug? It depends... how one looks at it.
Perhaps it would be good to expose better F360 ‘spline machinery' to unsuspecting CADmen.
Regarding the auxiliary predicament, who knows, it might be related. However, personally, I wouldn’t gamble with the continuity of mirrored/patterned splines.
Construct the curve in one go... and more likely... you will be in charge of the outcome.
Regards
MichaelT

[jiang_peng]

Actually there is an internal scale factor of each fit-point spline. The scale factor is determined by the size of the spline at the first time of activation of the tangency control(Dragging it or adding dimension to it). The purpose of the scale factor is that we want to "unify" the size of the tangency control and the spline. For example, we will have a longer tangency control on a larger spline even if the value of tangency at that fit point is small.

I'd prefer following approaches in this case:

• Using a Control point spline, Control point spline does not have any scale factors.
• Fit point spline, after the shape and size is done, use Copy/Paste to create a new spline. The scale factor will be also copied to the new spline.

Sorry for the inconvenience!