8 Replies Latest reply on May 29, 2009 5:04 AM by Valery Rutkovskiy

    when is a curvature continuous surface really c2?

    Glen Walters
      I have a SW surfacing question for the SW gurus

      It seems the continuous option SW offers for the various surface types is not really continuous or perhaps there is a problem with the porcupine curvature display?

      I attached a simple surface model to show this.
      I created a simple lofted surface with compound surface. I cut square hole thru this surface. I created surfaces to fill the hole using each surface tool SW offers for doing such things .....fillled surface, boundry surface, and lofted surface. I created an intersection curve thru the center of these filling surfaces and turned on the curvature display for each. The curvature display indicates that the surfaces are not curvature continuous.

      Is this really the case or I am I missing something?

      Cheers
      sw 2008 sp5
        • when is a curvature continuous surface really c2?
          Matt Lombard
          Glen,

          Great question. To measure continuity, I don't think we have great tools. The Deviation Analysis helps you measure tangency, and I think your continuity test has to use deviation analysis, because its possible that surfaces could have the same curvature value at a shared edge, but be tangent to different directions.

          One problem with your test. I'm not completely sure that two different splines are going to use the same curvature scale. I'm not sure it won't either. Another tool that you could use to test this is to make the splines individually, and click on the last point of the spline and look at the propertymgr values. It should show you numerically what the curvature or radius is. You'll have to create the splines independently to avoid sharing that point.

          Where you place the line might have something to do with it too. I've done the same test, and surfaces tend to fail in the corners. This is why I recommend making patches like that with spline loops, no sharp corners.

          Another test you could use would be to knit the test face into the rest of the body and turn on the Curvature display (View, Display, Curvature). If the color does not change abruptly over the edge, you're good.

          If you zip the SW part you will be able to upload it. Very interesting that on a SW forum you cannot upload SW parts directly.
            • when is a curvature continuous surface really c2?
              Neil Larsen
              the nature of curvature comb scaling is something I've asked about in the past - of course with no answer..
              I would still like to hear how genuinely it represents 1/R and satisfy myself that making adjustments to get nice smooth combs isnt a delusional exercise.
              The scale adjustment might be linear but it could be a power relation.. who knows?..is one spline scale consistant with another? is a small wave in the comb actually quite good enough or definitely an undesireable bump?
              as Matt points out deviation analysis can reveal that what you thought was a sweet blend may actually contain objectionable excursions... to this effect I have also asked how accurate the deviation analysis is as well...again with no answer..
              also it would seem that curvature/zebra display is based on the graphics mesh normals/tangency so I'm not too sure how accurate that is..curvature output from SW as vrml or stl isnt necessarily very smooth or accurate even at max resolution - in fact in some places the way the mesh is triangulated produces rather ugly faceting and obvious deviations

              without getting obsessive about these matters it really would be good to have some more detail to be confident about the indicators or perhaps some other type of solution or hybrid would be better?
              in general I dont know that what we are seeing is as useful as we might hope in determining the true shape or sheet continuity
              most likely its in the realm of near enough but just how near would be nice to know
              personally I would like to see this matter covered in some specialty help notes for surfacing but perhaps Matt can find out via his inside contacts and put something in his next book
              I would suspect we will never know from SW themselves, which is very poor tech support but consistant with their fluffy marketing practices
                • when is a curvature continuous surface really c2?
                  Matt Lombard

                   

                  ...perhaps Matt can find out via his inside contacts ...

                  ah, Neil, if only you knew how funny that is.

                  I think the evaluation tools are going to lag one step behind interest. I do see Solidworks surfacing becoming a more and more visible topic, and I think with more interest we will see better tools and more information. I think the deviation analysis is the most accurate tool we have, but it only measures tangency. Tangency is important, but it's only half the story, with continuity being the other half. I would like to see some development in the quantitative analysis side of things, because it's always nice to know how good is good enough.
                    • when is a curvature continuous surface really c2?
                      Neil Larsen
                      I have been known to write amusing things at times..
                      'the turd and the fly' apparently is not one of them

                      however, there is certainly 'lag' - its not just one step behind though
                      I asked those same questions above maybe 4+ years ago
                      I also asked about curvature comb smoothing tools back then.
                      I still want to know why we cant have precise loft connectors and boundary curves placement which I also first asked about now years ago
                      all of these deficiencies are readily apparent to anyone who makes active use of the tools but all we experience from SW is a reluctance to hear, say and do
                      its very frustrating to ask and ask for obvious and sensible things and get nothing or something else 'you didnt know you wanted'
                      I think most likely I will step back for another few releases and see if anything significant has changed by then -wouldnt count on it though
                      The good part for you is I will probably need to buy your book again to have it explained what got reworked for no better purpose and to discover the stuff that isnt 'discoverable'
                      cheers
                • when is a curvature continuous surface really c2?
                  Glen Walters
                  thanks matt I'll try your suggestions. I think SW did a great thing with the surfacing features introduced in 2007. Seems they haven't done much to improve the tools since.
                  • when is a curvature continuous surface really c2?
                    Kevin De Smet
                    You have to keep in mind that even in dedicated surfacing applications such as AliasStudio, measuring continuity is still very difficult. Positional and tangency can be directly measured but continuity is much less defined.

                    From what I've read (I don't claim to know all the answers)
                      • when is a curvature continuous surface really c2?
                        Neil Larsen
                        ok well probably no one gave some creative thought to it yet..

                        here's a first idea I thought of to throw out there for discussion
                        see attach - sorry about the messy illustration and scan

                        it is the usual deviation analysis + edge curvature from both directions, combined.
                        the colours I used are just to emphasise the different arrow sets.
                        the existing colour spectrum for dev would remain and I was thinking the curvature could be black arrows one side of the edge and white the other.
                        curvature arrows point in to the edge - dev upwards
                        the density of the arrow display would be by slider as per usual.
                        the other possibility is to show just the differential between radius from each direction so the 'dis-continuity' arrows would snake from one side to the other

                        hmm..hope that makes sense