Talking about Sheetmetal working on a part sometimes get this problem a corner on a seemingly simple fillet fails no matter what I do to the solid.

If someone could shine some light on why this sometimes happens - here is part other corners filleted without a hitch.

Just replace your object on the gride!
Now you are out of the space in the stars' field! :)
So all will works fine! 0 to 50 fillet possible! :)

Hi Alex, I'm sure it's some kind of bug in the "intersect with side faces" step of the filleting process.

One thing I noticed when examining the surfaces is that they are generic cubic spline surfaces with 4 points in one direction, note this when turning on control points:



So that means that operations involving those faces are not able to go through any special case plane handling which usually will improve accuracy.

I tested replacing those 2 surfaces with simple 4-corner-point planes and that seems to then work better, see attached file. I replaced them by deleting them one at a time and then selecting the object and running Construct > Planar to fill in the planar hole.

How was the shape originally created? Because all the edges that look like lines are not simple lines but generic cubic curves now, is this the result of running Flow on an original object? The downside of Flow is that every surface and edge in the original object is going to get turned from an analytic curve into a generic curve so that it can be potentially warped into a different shape. This may be degrading the analytic quality of your object if you're doing this often.

I don't particularly see anything actually wrong with the object though, it may be possible with a lot of detective work to fix Fillet to work with your original.

- Michael

So you have 2 dieffrents solutions!
That's paradisiac! :)

Thanks Pilou for noticing that!

Yes - the location far away from the origin is a contributing factor.

If your object has larger coordinate value it essentially steals some precision away from the fractional accuracy of calculations. I'd guess that this combined with using Flow maybe is preventing the fillet calculation from getting a precise enough result in one of its steps and that's what causes the failure.

It's probably something that could be fixed to work if I could find the particular step and loosen a tolerance value for these conditions which is probably currently too tight. But this is unfortunately easier said than done and it's easy for such a modification to have negative side effects in other cases.

But it looks like if you avoid either one of your current conditions, either have analytic planes or move the object to the origin the same fillet should succeed.

- Michael

Hi All,

Thanks for speedy replies - the shape is created by using planar surfs so I create the outlines fill in with planar - join the surfaces together apply radius on edge then shell - must be something in between the application of radius or shelling - I will check out further but strange it happens differently in locations as the base outline is the same all the way - a continuous polyline. I dont use flow as using this method I know the parts unfold no problem.

Hi Alex, how do you build the curves that you use planar srf on?

Most likely what look like lines there are actually 4 point generic curves rather than simple 2 point line segments.

- Michael

Hi Michael,

The parts are away from the origin due to the large assembly they belong too. I cut and copy this position from my master wireframe and then I work on the part alone saves hiding many parts and isolating one part. When I finish the part I paste back into the assembly really easy no need to position.
This probably explains the bug and when I do a standalone test usually on a fresh part close to origin doesnt happen.


With the construction as it a sheetmetal part the work flow I use revolves around using planar surfs - I usually create a series of rectangles to give me base shape which I trim if needed I then join them and then apply radiuses as required and finally apply shell command to desired thickness works as a recognizably sheetmetal part in other programs like Solidworks Sheetmetal and Autopol.

In the many parts that I usually create the radius issue pops up every now and then with most of the radiuses being completed and only one random corner misbehaving.

thanks for your time - the ARC thickness script would certainly make Sheetmetal more streamlined within MOI.

Michael,
So for these "far away from origin" issues, does a "cplane" bring the object to a local coordinate system or do the underlying calcs still look at "world coordinates"?

Hi Burr, all actual calculations are done in world coordinates. Setting a cplane just changes the UI layer, everything that you create is still actually stored in world coordinates.

But it could be possible to internally move objects to the world origin, do the fillet calculation and move it back again transparently. That's probably something I ought to do for v4.

- Michael

Ok thanks michael!
I guess the other side of that is peoples particular setup and job at hand...
So i work in "inches".... But if i start modeling a 175 foot object (or i guess sci fi guys modeling 1000 foot spaceships), things start moving away from world zero...

Does changing my unit system to feet or meters or miles change that initial interaction with "how far away from zero" i am?

Is there something like a "scaled environment" where i could start a project with an initial declaration that i will be modelimg at "a larger scale"? (I guess that would mean i couldnt model detail then, because it was expecting BIG stuff, side effect......)

Hi Burr,

> Does changing my unit system to feet or meters or miles change that initial
> interaction with "how far away from zero" i am?

Yes, if it means you'll be using small numeric values, like say x = 10 in meters rather than x = 10000 in mm. But the only thing that matters is the number itself, that you're using less digits in the number values.


> Is there something like a "scaled environment" where i could start a project with an initial
> declaration that i will be modelimg at "a larger scale"? (I guess that would mean i couldnt
> model detail then, because it was expecting BIG stuff, side effect......)

Well there is an automatic mechanism that adjusts calculation tolerance to be a percentage of the object size so it adapts to the current object being operated on. But that's a somewhat different issue than losing fractional precision in arithmetic because of larger numeric values being used in coordinates.

- Michael