MergeSrf gives open edge at symmetric surfaces


Hello, I have very strange problem with open edge…
I have symmetric surfaces, I apply BlendSrf with G2, Mirror.
Untitled.3dm (507.4 KB)

First question!
Why G2 at edges is replaced with G1?

Second question!
Symmetric and it seems with a good accuracy, but MergeSrf the open edge - accuracy perhaps weak gives surfaces, after application of MatchSrf - there are no open edges.

Help to understand these questions please :worried:

(David Cockey) #2

I tried repeating your steps and the edges of the blend surface are G2. See attached file. ModelerDC1.3dm (731.2 KB)

I also had MergeSrf report the edges are two far apart. Checked with CrvDeviation and the differences were 10E-8 so well within the tolerance. Don’t know what the problem is.

(Pascal Golay) #3

Hello - the edges of the blends are G1 to each other as the file opens and the edges of the input surfaces are G2 across the centerline. If you join the blends together and the outer surfaces together first then Join the results together the state of the two edges there depends upon the order of Join - one way the the G2 edges will be pulled to the G1 edges and the result is G1, the other way the opposite happens.
If you MatchSrf the blends to one another for G2, with Average, then it all stays G2 on Join, and MergeSrf works as well.



Ok, but the problem with open edge nevertheless isn’t solved :slightly_frowning_face:
Maybe somebody knows: @TomTom, @Micha, @Lagom

(Tom) #5

as David said: there is no problem.

Edges having g1 is totally okay. That doesn’t mean that two surfaces cannot have g2 when their edges just have g1.For curves its always the third control point which determines g2.
So if you create so many cps it can happend that there is a non-visible unsmoothness resulting in g1 only. But besides that, how do you measure g2, meaning for which tolerance it is not g2? Besides that, you can match symmetrical surfaces with g1 and they are always g2. Its just the flow which may look weird. So for symmetry you go for g1 or g3, but g2 is none-sense.


And why so it turns out?

But GCon all the same gives G1

Well, but even if I use G3 - in the analysis I see all the same G1

(Tom) #7

As I said, its very close to g2 but the analysis tells you it is not. So what? But this analysis doesn’t tell you anything. If you create tons of cps its very likely that the cps become unsmooth resulting in not having g2 edges. But again who cares if you maintain good highlights. This isn’t a problem at all.

PS: if you have g2 but your edges are wavy, what does it help? Better having nice edges rather than g2 edges with sinus looks


Ok. Means to me you shouldn’t lean on GCon?
By the way, if the analysis sees G1 instead of G3 - the deviation has to be serious. Or not?

(Tom) #9

well g2 isn’t everything. There is no value telling you it is g2 or it is not. if the curvature from one shape minus the other curvature is smaller than the tolerance, it depends on the tolerance. Rhino completely misses an analysis which yields graphical and numerical deviations from points measured. As long as you not doing final class A models you can go for :" If it looks good its good". And if you care, you definitely need a costly high end modelling software, where you can display local discontinuities.

(Tom) #10

Its not fully serious. This is what David and I saying.


Well, but I still don’t understand - why CrvDeviation can’t find overlapping if surfaces symmetric and without serious deviations?

(Tom) #12

I don’t know, but if you match a curve with g2 you can match them de facto exact with very low deviation. So crvDeviation probaly has very strict tolerances. However if you match two surfaces you cannot match them as accurate as two curves, because of having two principal curvatures involved. So the command returns g1 although it satisfy g2.


GCon and CrvDeviation - are almost useless because of tough admissions…

(Tom) #14

i would say they fulfil other purposes…


And when they can be useful?


Your initial surfaces are overly complex. You achieve more predictable results with less complexity (= easier to modify/adjust). That would mean building a simple quadrant of your doubly symmetric shape and make sure the curves are degree 2 or 5 and good looking (use the curvature comb); and only then build the G2 transition.

An alternative method, and probably faster, would be to simply build your main large surface, add some CPs towards the centre, and then pull them in the Z direction (upwards) to sculpt your shape. It will then be inherently smooth.


Have dealt with the analysis…
I have noticed such piece here - the open edge after MergeSrf appears only at the processed surface - Rebuild probably spoils accuracy…

1.3dm (411.0 KB)

It is all about accuracy!
Initial surface - Plane + the shift of control points. Normal accuracy!

In the second case a lack of control points - inaccuracy and it has affected the following smoothing surface. Bad accuracy!

In the third case normal accuracy.

And the accuracy of edges at two last surfaces identical…


Curve in 5 degrees, but as a result of all the same G1


Here’s a quick example of what I mean regarding quadrant and simple surfaces. No isocurve mess. No internal wobbles. Easier to manipulate by fine-tuning CPs. Hope it helps a bit…

Untitled.3dm (392.3 KB)


@Lagom What was your workflow to manipulate smoothly all those 64 CP? Move UVN? Dragmode control Polygon?