Include 3D Geometry - why do edges sometimes not get converted to splines correctly/accurately ?

[chrisQX8DX]

Hi there, i've noticed on several occasions that when you include an edge from a 3d object the resulting curve is not always exactly replicated. Here is an example from a model that was created using a network surface patch. The model is a bit dodgy in places but it machined without problems that can't be sanded away. I presume that the cause of the wrong spline creation is because of the dodgy/kinky surface on the rounded corner but i still don't understand why that would prevent Fusion from creating this edge as it is on the model and as i can select it.

Here's what it looks like when this happens, the red outline and blue curve is how Fusion generate the 'included' edges of the geometry:

I've attached an example file to play around with (the issue is the edges connecting to the rounded corner). I've tried some workarounds but even unstitching the model doesn't help whereas i though that at least on a single unconnected face the edge should be 'included' as is, even if the face is somewhat twisted.

PLEASE NOTE: this is not about fixing the model, there just wasn't any other way to create the model (the attached file is only a cutout of a bigger model) and it machined fine. But i now need to machine another part that should fit onto this model so i can clamp it while glueing and for that i would need the EXACT edge in a sketch to model the clamping caul and not an approximation

thx for any help or enlighteenments about this strange issue 😁.

[TrippyLighting]

Looking at the curvature combs confirms the "dodgy" nature of the geometry. Usually that sort of geometry is the result of inadequate modeling techniques and it isn't a surprise that such bad edges don't project well onto sketch curves.

If you can share the complete model with timeline we might be able to identify the root cause.

[chrisQX8DX]

hey Peter, thx for your reply. I know what i did wrong in modeling but as i said, the model is valid for production, unfortunately i can't share the whole model for IP reasons.

However, i'm not trying to project the curve, i am trying to use 'Include 3D geometry' so the thing i don't understand is why the edge is not replicated 'as is'.

[TrippyLighting]

ge witin some internal tolerance and I cannot answer the real reason for that, because I have no insights of how those algorithms work.

What I observe is that Fusion 360 approximates that edge with a control points spline with some number of control points. You can see that spline when you "break link" and make that sketch curve editable (but loose the parametric link to the edge).

I find that I very rarely need "include 3D" as I prefer to work with edges and surfaces directly.

If you are running into a modeling  due tom this behavior, please explain what you want to do. Perhaps there is anotehr way to accomplish it.

@jeff_strater might be able to provide some insights.

 

[laughingcreek]

---

@chrisQX8DX wrote:

...i'm not trying to project the curve, i am trying to use 'Include 3D geometry' ...

---

just semantics,  but include 3d geometry is a projection by definition.  

[chrisQX8DX]

not sure i understand, i thought that when you use 'include 3d geometry' the edge is just extracted/duplicated standalone ... that's why i was surprised that the edge you can select on the model is changed when you include it and not the same as the one highlighted when you select it on the model

[TrippyLighting]

---

@chrisQX8DX wrote:

not sure i understand, i thought that when you use 'include 3d geometry' the edge is just extracted/duplicated standalone ... that's why i was surprised that the edge you can select on the model is changed when you include it and not the same as the one highlighted when you select it on the model

---

You can't just "extract an edge" that is not how those algorithms work 😉

[laughingcreek]

in blender you can have an "edge" existing on it's own without a face associated with it.  That's not what happens in a nurbs/berp modeler like fusion.  an "edge" isn't a separate thing, it's just where a surface ends. it's edge.  when you "include 3d geometry" the edge into a sketch your creating a sketch article (more specifically a control point spline) that approximates the edge. 

see @TrippyLighting explanation above.

[TrippyLighting]

---

@laughingcreek wrote:

in blender you can have an "edge" existing on it's own without a face associated with it.  That's not what happens in a nurbs/berp modeler like fusion.  an "edge" isn't a separate thing, it's just where a surface ends.

---

Almost all NURBS models contain trimmed NURBS surfaces. The trimmed edges are created by trimming loops, which are also curves. However, I don't know if it is even technically possible for a sketch engine to directly access a trimming loop. I would assume that the sketch engine samples the edge and creates a curve based on the sampled data.

[chrisQX8DX]

hmm ... i knida understand what you're all sayin ... it just seems kinda weird ... i mean this 'edge' is obviously defined in fusion since i can see and select it on the object. and when you project or include i get that fusion will recreate or aproximate this 'edge' as a spline curve. what's odd is that Fusion seems to decide on it's own with how many control/spline-points it will aproximate this 'edge'. i guess there isn't a setting in Fusion that would change this 'resolution' ?

theorethical musings aside .. am i correct in assuming that the better the curvature (so clean curvature graphs in U and V direction) the more precise the aproximation of the included edge will be ?

[jeff_strater]

the short answer is:  @TrippyLighting is right (which is usually the case...).  There is an approximation step, even when "including" an edge like this.  A sketch object is a different data model object from an edge.  That is an implementation detail that you shouldn't need to know, but that is the underlying reason.  Most times, Include is pretty accurate.  I don't know, in particular, what the cause of the inaccuracy here is.  Peter's theory about curvature of the edge is certainly a good guess as to the cause.  We do have a project, somewhere in the backlog, to try to improve the accuracy of these operations, but, as you can guess, there is a tradeoff here - more accurate usually means more control points, which can lead to its own set of headaches.

One question that I don't think has been asked here is:  What are you intending to do with the included edge?  Many times, you can reference the edge directly, without having to go through a sketch.  For instance, a sweep path can be a set of connected edges, it does not have to be a series of sketch curves.  That will be 100% accurate, since you are referencing the edge directly.   There is no approximation.  Of course, that doesn't help in all cases (which is why Include is there in the first place).

[chrisQX8DX]

thanks again to everyone for the replies, i'm always keen to understand the underlying technology even if it's not always relevant/necessary to understand it for modeling. And yes Jeff, i'm always amazed at the depth of Peters knowledge 😊.

so maybe to give a bit of background to what my problem is ( you can skip this part if it's not interesting):

i am trying to make a clamping caul for glueing an ebony backstrap to a guitar headstock. it's a really tricky part to glue which is partly due to the geometry of the back of the headstock but mainly due to the problem of heat bending the ebony acurately and asymetrically and the fact that i work with hot hide glue which doesn't have a long open time so i need to move quickly. i've been doing this for quite a long time so it's not a show stopper when things go wrong but i'm always dreading the glue up and i've tried a variety of techniques for making a clamping jig so now i'm trying to make one in Fusion.

Here's what the final glued up version looks like and because of the clamping/heat bending problems i cracked the backplate ... again ... (which with ebony isn't really a problem and easily repaired with a splint and sawdust for an invisible fix)

the problem of the kinky geometry isn't a problem when making the neck since in wood most of the times close enough is close enough and i usually do a fair bit of hand shaping to get the final shapes i want instead of spending ages modeling and machining a perfect version. However in this case as you've seen the dodgy geometry becomes an issue when trying to use it as a template for making the clamping thingy.

to be honest i can't really remember if i modeled the final neck-headstock transition in Fusion or in Grasshopper (i always use grasshopper for my necks since i built a parametric patch a couple of years back which alows me to set a few parameters with sliders and it spits out an almost finished neck to client specs) but the neck/headstock transition is a tricky bit to model and getting a perfect curvature on this part is really difficult so i often resort to the 'close enough is close enough - fix it in the mix' technique which works fine).

i attached a file with the clamp i modeled by converting the backface to a tspline with the curvature option set really tight and then thickening it and cutting of one side straight (i need the top of the caul to be a flat face so the clamps will hold). i haven't machined it yet but the problem i see with this model is the downward drooping bits:

which will break the plate again when i clamp it since the part i'm glueing on is a bit oversized and the downward pointing bits will for sure crack the ebony plate. it would be necessary for the continuation of the clamp caul to continue in a straight plane (not sure if i'm explaining this well).

i think the main problem of all this is that the upwards slope is not symetrical (because of the asymetric headstock design) and the fact that i modeled a bit of a spine into the middle (to add a bit more strength to the headstock). The spine/raisedmiddlebit doesn't really do much so for the future i'll probably remodel this with a flat curved plane:

so the reason why i wanted an exact replica of the faces edges is so i could use them to extrude them upward and cut out just the part of the clamp that fits the headstock back and not have these downwards drooping bits.

Projecting the outline and using it to cut out a straight profile like this doesn't work (results in an error) which i guess is because either the curve isn't completely inside the object or not a closed loop (or possibly both, i haven't had the time to look at this in detail yet), but i also cannot extend/offset/presspull the front face so it doesn't overlap (also gives an error)

i also tried various ways of modeling this by using plane intersection curves and lofting them but in the end the tspline conversion is the best i could come up with.