how do u make this kind of surface in rhino

[JimCarruthers]

The "level 2" training pdf that comes with rhino has a section on various ways of making that general type of shape. Myself, I would make a large surface for the top, then trim it off and fillet or blend it.

[andrewjohn81]

If you want 1 surface, then you must push/pull CV's. If that surface was 1 nurbs object you would have quite a few problems with it. Since it has no corners, it can not be manipulated at the "corners." I know that sounds strange. The rounded corners you have there would cause multiple CV's at the same origin. This is fine asl long as all the cv's connecting remain completely tangent. As soon as one of them moves, that geometry will no longer be renderable. It will have nasty black things at the corner, for lack of better words.

If you just need the shape, then you can do what Jim said.

I would just create the curve network and use the network curves, sweep and/or loft (as another method)

[JimCarruthers]

You can make it a valid single surface using RailRevolve, that's another way.

[andrewjohn81]

that's not a bad idea, but I was assuming he wanted to have similar parameters as the picture. Otherwise I don't see why you would want it in 1 piece.

Rail revolve would be pretty hard to control there also. I don't use rail revolve much, but wouldn't the radius be longer on the long side of the object? From my understanding it basically scales the profile from the origin point to match the rail that you gave it. This would change the shape of the radius unless it was a perfect circle. It would "cap" the end of something, but I believe that it's meant more for organic shapes like the ones used in the examples.
It also works nice if you just need ot cap the end of something that isn't perfectly round quickly. I'm sure I have yet to use it's full potential. I'll play around with it some more today to see what other kinds of options there are.
_________________________________________________________________

Added after trying:

It actually did a much better job that I expected. If it were a perfect square it would be fine, but since I used a rectangle it did as I expected. The radii on the longer side are shorter that the others.

For this example, I wouldn't use rail revolve. The geometry it creates is not very good for that type of shape.

[tilite]

i have to disagree.

i just tried the same rail curve on a few shapes... it seems to work perfectly. On both square and rectangular profile curves. Maybe i have confused the intentions but i would definetly recomend atleast an attempt.

[7Stones]

Trying to create this particular surface has long bothered me as well.

A rail revolve will get you close if you start with the proper shaped profile curve and rail, but in the end it just doesn't get you close enough. At least it doesn't for me.

Cutting the domed top surface of the box out of a larger surface or shape doesn't work either unless you can figure out how to properly shape the surface or object you are cutting. Note: The domed box top is not nor can it be cut from a perfect elipsoid. If an elipsoid is used then the top edges of the box will be arched and not straight.

The quickest method I know of for creating the domed box top surface is to draw two straight lines that represent opposite edges of the box top. Next, draw a curve mid way between and parallel to the two straight lines which represents a profile or cross section of the domed box top. With the lines and the curve drawn you then can loft them to create the domed top, join it to the box's sides and bottom, and then fillet all the edges. This gets you close, but still no cigar.

I'm going to have to keep experimenting with this one myself. I'll let you know if I come up with a solution.

Wish I could get my hands on the Level 2 Training PDF Jim mentions to see how it solves the problem, but I don't have my Rhino CDs handy to see if I have it.

[andrewjohn81]

tilite, if you create a rectangle, then create a recognizeable/measureable object on one of it's edges at a midpoint. Create a line that follows that straight to near the center.
If you do a rail revolve using that you will notice that the edges aren't all the same radius/shape.

It's confirmed. The way it works is it scales the object at every point within a certain tolerance to follow the rail. If that is so, then the further it would be from the origin point, closer or further, it would scale the entire object thus scaling the radius too.

If you wanted an even radius, that would be a bad method to use.

If it were me creating this object I would use multiple surfaces.
It sounds like most people are having problems creating the bent surface, a surprise to me.
The easiest way to do this is to first build the surface at a parameter that is only 1 span of degree 3 surface. Simply pull up the 4 center cv's evenly and you have yourself a perfect domed surface with straight edges at the corners and edges.

I'm just curious why you wouldn't want to pull/push cvs? You could do the same thing creating curves, but then you'd just be pushing/pulling cv's of curves first. It's pretty much the same thing. The trick is to get yourself as few cv's as possible. That just makes it alot easier to control the results.

[Dimos]

I managed to do the surface in two ways, without pulling CVs, just by designing curves and applying either "Patch" either "Sweep2Rail". I 'll post all the images concerning the method with "Patch" and one concerning the second method.

I drew the curves and a rectangle and I applied "Patch"

http://img134.echo.cx/img134/5498/surface17nr.th.jpg (http://img134.echo.cx/my.php?image=surface17nr.jpg)http://img202.echo.cx/img202/4242/surface47fz.th.jpg (http://img202.echo.cx/my.php?image=surface47fz.jpg)http://img202.echo.cx/img202/5105/surface29in.th.jpg (http://img202.echo.cx/my.php?image=surface29in.jpg)

Then I extruded (without cap) the base of the shape and I applied "FilletEdge" at the highlighted edges

http://img172.echo.cx/img172/747/surface51kc.th.jpg (http://img172.echo.cx/my.php?image=surface51kc.jpg)http://img172.echo.cx/img172/6383/surface69gk.th.jpg (http://img172.echo.cx/my.php?image=surface69gk.jpg)http://img172.echo.cx/img172/5349/surface77lw.th.jpg (http://img172.echo.cx/my.php?image=surface77lw.jpg)

The other way is to make a "Sweep2Rail" using as rails the long highlighted curves and as cross sections the sorter ones, as it is shown at the picture below. Then, I mirrored the surface and I applied "Matchsrf" with curvature continuity and all boxes ticked on (on the main window). Finally I merged (smooth=no) the two surfaces and I continued as in the first method.

http://img142.echo.cx/img142/4185/surface94sr.th.jpg (http://img142.echo.cx/my.php?image=surface94sr.jpg)

[Xist]

Dimos: Nice to see your approach. Before seeing your patch method I would have done it the 2 rail sweep way (how I do most things, really). I just gave the patch method a shot and had my surfaces in about a minute..

It's always nice to learn something new =)

[7Stones]

Hehehe... Dimos, I get exactly the same surface lofting three curves that you get via a patch with the curves and a rectangle. Either way works just fine. Its good to know about both methods in fact.

What bothers me about the surface generated via these two methods, however, and maybe this bothers Saiko as well, is that the curvature of the domed surface reverses as it approaches the corners of the rectangle. If I'm reading the shape that Saiko posted correctly, then the curvature of the surface shouldn't reverse. The entire surface should be convex and not a mixture of convex and concave. That's what I was after when I was trying to make a similar shape a while back. Tricky.

[Dimos]

What bothers me about the surface generated via these two methods, however, and maybe this bothers Saiko as well, is that the curvature of the domed surface reverses as it approaches the corners of the rectangle. If I'm reading the shape that Saiko posted correctly, then the curvature of the surface shouldn't reverse. The entire surface should be convex and not a mixture of convex and concave. That's what I was after when I was trying to make a similar shape a while back. Tricky.

In the beginning I was thinking the same as you, but I looked again on the image that Saiko posted and I thought that the surface is with reversed curvature near the corners, but you are right, I am wrong. I will keep looking for a method.

[7Stones]

OK... This will be my first attempt at posting images to a CGtalk Forum. Let's hope I don't screw it up.

I'm not happy with the following method, but here is what I did to create the convex box top:

Image 1 - Instead of starting with a standard rectangle I created a rectangle with rounded corners by joining lines and quarter circles. The red rectangle also shown will be used to trim the convex box top surface after I create it, so I wanted the black rectangle with the rounded corners to be outside the red rectangle.

Image 2 - After drawing two diagonal lines across the rectangles I used them to split the rounded rectangle into four curves.

Image 3 - Snapping to the mid points of the rounded rectangle's sides and a point 1.5 units up from the origin with the "Curve>Arc>3 Points" command I created to two arcs shown in red.

Image 4 - I then rebuilt the arcs, changing their point count from 3 to 5. After rebuilding the arcs and turning on their control points I selected the control points show in yellow and 2D Scaled them form the origin in the top viewport by 1.2.

http://img292.echo.cx/img292/1037/convexboxtop011iv.th.gif (http://img292.echo.cx/my.php?image=convexboxtop011iv.gif)http://img292.echo.cx/img292/361/convexboxtop025an.th.gif (http://img292.echo.cx/my.php?image=convexboxtop025an.gif)http://img292.echo.cx/img292/1222/convexboxtop037kj.th.gif (http://img292.echo.cx/my.php?image=convexboxtop037kj.gif)http://img292.echo.cx/img292/8776/convexboxtop040or.th.gif (http://img292.echo.cx/my.php?image=convexboxtop040or.gif)

Image 5 - Next I created the convex box top by using the "Surface>From Curve Network" command and selecting the four curves that make up the rounded rectangle and the two red arcs between them.

Image 6 - To tidy things up a bit I rebuilt the convex surface, cutting down on the number of its control points, and then trimmed it with the red rectangle in the top viewport.

Image 7 - After creating the bottom half of the box with its fillets I created blend surfaces between the box's sides and its convex top.

Image 8 - To create the filleted corners on the top of the box I like to use the "Surface>From Curve Network" command with the blend curves shown in yellow and the edges of the surrounding surfaces. You could also use the Patch command to create the corner fillets, but don't use the rail command. If you do use the rail command the resulting surface won't share tangency or curvature with the surfaces around it. The yellow blend curves were created by extracting curves from the surround surfaces or by projecting curves onto the surfaces and then blending between them.

http://img292.echo.cx/img292/5460/convexboxtop055ur.th.gif (http://img292.echo.cx/my.php?image=convexboxtop055ur.gif)http://img292.echo.cx/img292/1599/convexboxtop062kx.th.gif (http://img292.echo.cx/my.php?image=convexboxtop062kx.gif)http://img292.echo.cx/img292/407/convexboxtop074rw.th.gif (http://img292.echo.cx/my.php?image=convexboxtop074rw.gif)http://img292.echo.cx/img292/4631/convexboxtop088wu.th.gif (http://img292.echo.cx/my.php?image=convexboxtop088wu.gif)

Image 9 - And that completes my attempt at creating the box object with an entirely convex top. In the meantime I'm still in search of a quicker and simpler way to achieve the same results without the problem of the box top corners looking concave instead of convex.

http://img292.echo.cx/img292/7427/convexboxtop093iz.th.gif (http://img292.echo.cx/my.php?image=convexboxtop093iz.gif)

[andrewjohn81]

if you use the pulling CV's method it actually does not create a concave/convex. It does produce a flat tangency to the edge though. That is, however, what was produced in the first image that was provided for the quasi-challenge. I think this is actually ending up more like some kind of challenge, and a good turnout at that.

The method that 7stones used, although way more complex than needed, isn't a bad approach. The tradeoff here is that using that method you will get variable radii. If that is the desired result then that is perfect. I just wanted to point that out.

If you need a fully convex shape at the top and don't want any variable radius then the four sides will have to be arched at the top. I usually prefer the shape that 7stones used. I might have just pulled the top and side cv's at the same time off of one exploded box. Then I would trim away, probably using pipes and a single plane for the bottom to keep it flat, the extra pieces. Then I would just fill in the gaps using the appropriate methods, such as blend surface, or creating some Sweep2 surfaces. I usually try blend first.
I almost forgot. I usually shrink the surfaces after trimming them so tangency is easier to get.

[7Stones]

andrewjohn81... Thank you for pointing out that the method I used will yield variable radii. I think it is important to realize the the object Saiko posted does indeed use variable radii to achieve the the convex box top. At least I'm pretty darn sure it does. (However, I believe it might be possible to use a perfect radii between the mid points of the convex box top's sides and still not have the surface change to meet flat tangencies at the corners. How to achieve such a surface "easily" and "accurately" is a question I'm still seeking to answer.) :)

Sorry if my concave/convex description caused any confusion. It was my poor attempt to describe the "flat tangency to the edge" problem as you put it. If you project a curve diagonally across the box and down onto its top the resulting contour curve has to change the direction of its bend to meet the flat tangencies at the corners, thus my use of the terms concave and convex to describe the surface.

This thread is looking like a challenge? lol... I suppose it is, but I'm hoping the emphasis is more on the learning aspects rather then the competative nature of a challenge. :)

[andrewjohn81]

it would be interesting to see someone prove me wrong and shut me up, but I don't think it is possible to not have flat tangency at the corner without having variable radius.

If you examine the image in the first post, even there, you can see that the tangency is much flatter at the corner than at the center. With the way it is built, I would say, it most certainly has a flat tangent right at the corners.

If someone proves me wrong on this though, please post the file. I would be VERY interested to see how this is done.

[jamesvalue]

it would be interesting to see someone prove me wrong and shut me up, but I don't think it is possible to not have flat tangency at the corner without having variable radius.

If you examine the image in the first post, even there, you can see that the tangency is much flatter at the corner than at the center. With the way it is built, I would say, it most certainly has a flat tangent right at the corners.

If someone proves me wrong on this though, please post the file. I would be VERY interested to see how this is done.

could it be a solution?

(sorry . but I didn't follow this thread , completely)

Extruding a arc using another arc , creates a concave NURBS surface (whith the corners lowered twice)

Bye

see the attachment for more infos

[Dimos]

could it be a solution?

Extruding a arc using another arc , creates a concave NURBS surface (whith the corners lowered twice)


It is not a solution because the edges of the extruded surface are not straight lines.

[jamesvalue]

It is not a solution because the edges of the extruded surface are not straight lines.

Maybe you mean U & V lines/points count?

If so you can easely rebuild your surface adding or removing points (and lines)

Bye

[TheRainKnight]

it would be interesting to see someone prove me wrong and shut me up, but I don't think it is possible to not have flat tangency at the corner without having variable radius.

If you examine the image in the first post, even there, you can see that the tangency is much flatter at the corner than at the center. With the way it is built, I would say, it most certainly has a flat tangent right at the corners.

If someone proves me wrong on this though, please post the file. I would be VERY interested to see how this is done.

you re right. it's mathematicly impossible, if the border of the surface is flat and rectangular (important, wrong if the corner are rounded), you'll never have anything else that a flat tangent in the corner. Or the surface is not smooth....

but i tried this. build a patch surface on the curves on the left. the resultant border is not flat, whatever stiffness you specify. The filletting work give some non constant radius as expected.
http://cheval32.perso.cegetel.net/forums/test_cgt.jpg

[7Stones]

I was writting the following the same time The Rain Knight posted his solution above.

Good work The Rain Knight! That is close to what I was after.

~~~

I think I better try and clarify some notions (right or wrong) about the way I'm interpreting the shape of the box's convex top...

First, I have to recognize that 2D wireframe images like the one posted can be very misleading and hard to interpret. Everyone will see something slightly different with regard to the shape depicted. Subtilely curved surface can be misleading without orthographic views for comparison.

If a flat rectangle, defined by straight lines, is used to create the convex box top then the tangency must be flat at the corners of the resulting surface, making the surface appear flat at the corners. If the corner tangencies are not flat then the rectangular surface won't have edges defined by straight lines. However, who said the lines defining the edges of the rectangular convex surface had to be straight along all axis?

I believe it is necessary for the box top surface edges to posses a little bit of a variable arc to them to achieve the desired rectangular, convex shape. The arc or curve in the lines that make up the surface's rectangular edges are just so subtle that they look straight even though they are not, and by not being straight they eliminate the flat tangencies at the corners. This is how I'm interpreting the convex surface based on the image provided and a real world object (plastic soap box) I'm using as reference. I'm not seeing flat tangencies hidden behind fillets in the corners of the object's convex surface, nor am I seeing perfect arcs used anywhere in the surface. I'm seeing a flat, rectangular surface on a filleted box being pushed out as if there was pressure behind it. As the flat surface is being pushed out both it and the fillet surfaces around it are being distorted. The fillets start off with a straight edge were they join the box's flat sides, but the edges of the fillet surfaces that join the convex box top have a very slight curvature that looks like a straight line. So this is the type of convex box top surface I'm trying to find an easy way to make and by doing so I hope I'm helping out Saiko with his object also (If I'm not helping or wrong about the shape you are trying to create please let me know Saiko).

If Saiko's object really does have a convex top with flat tangencies in the corners then I would say the best way to create it is to draw the appropriate curves and loft or patch them as previously describe (I'd go with a patch myself). The resulting surface can then be joined to the rest of the box and filleted. Easy... Nothing to it.

[jamesvalue]

The filletting work give some non constant radius as expected.


As usual (I would say!) (LOL!)


Nice work!
Bye

[Glyptic]

The "level 2" training pdf that comes with rhino has a section on various ways of making that general type of shape. Myself, I would make a large surface for the top, then trim it off and fillet or blend it.

Looking at the Level II manual is a very good idea. Its one of the best investments I've made in Rhino's learning materials.
Also look at Network Surface for your top surface. Its tailor made for this kind of thing. Getting those rounded edges could be a Sweep, or a Blend. Does the top need to be a different Layer like you have here with the two colors? If not you could make your rectangular box with a Solids tool, Explode, delete the top surface and replace it with a Network Surface based on Arcs. . Then Fillet the edge for some roundness or Blend into the lower box.
Good luck,
larry

[saiko]

thanks eveyone for ur effort to solve the problem....
thanks 7Stones for elaborate post, thanks Dimos too for ur effort.....but that one i had also tried but it does not give desired result....

this kind of i had first noticed in the Porsche Cayenne dashboard, and then in some other electronic eqipment,
i tried to generate the same in rhino but failed to get the exact one as i told u,

then i used 3dsMax to get it, i got it in 1 minute with polygon modeling, without any struggle.....
thats what i had posted at the begining.

thanks again for all ur effort....

[saiko]

Hi 7Stones
i was trying your method out....
but in step-5 i am having some problem

Image 5 - Next I created the convex box top by using the "Surface>From Curve Network" command and selecting the four curves that make up the rounded rectangle and the two red arcs between them.

i find 8 curves in the rounded rectangle, 4 straightlines and 4 arcs at the corner:sad:, and i failed to make surface from curve network...

could u please crarify this a bit more...

[jamesvalue]

Hi 7Stones
i was trying your method out....
but in step-5 i am having some problem

i find 8 curves in the rounded rectangle, 4 straightlines and 4 arcs at the corner:sad:, and i failed to make surface from curve network...

could u please crarify this a bit more...


I run into the same problem !
So first you have to (using the "join command") transform your rounded rectangle from 8 to 4 sides .(3curves up=1side , 3curves down=1side , + 1left and 1right segments , total makes 4 sides).

[jamesvalue]

I found this "interesting" italian- written tutorial , that more or less reach the same kind of surface , using the network srf comand (so I try to study this comand a bit better)

Bye

[JimCarruthers]

Doing a surface like that as one using Networksrf is a bad idea, the result is the adjacent edges of the 'square' surface are tangent to each other, at which point the geometry is "undefined" and subsequent operations in those areas will be screwed up.

[andrewjohn81]

Thank you Jim. I didn't want to say anything again because I've said those kinds of surfaces aren't very good to use several times already.

Also, you have very little control over the radii at the corners.

[7Stones]

Saiko... To convert the rounded rectangle into four curves that can be used with the Surface>From Curve Network command you must first join together the lines and arcs that form the rounded rectangle. Next, draw two lines diagonally across the round rectangle. Start and end the diagonal lines from were the corners of the rounded rectangle would be if they were not rounded. Use the diagonal lines to "slpit" the rounded rectangle into four seperate curves/sections (the four seperate sides of the rectangle). Now you can use the resulting curves with the Surface>From Curve Network command.

JimCarruthers & AndrewJohn81 (and Saiko)... You won't get any arguement from me that "creating a surface like that using Networksrf with the curves I described is generally a bad idea". However, It is still the only way I have found to achieve the specified surface with the least amount of trouble. With that said, though, I must say that I would not use the resulting trimmed convex surface in the final model. What I would do is take the resulting trimmed convex surface and project a grid of lines onto it and recreate the surface from the resulting projection curves and the edges of the surrounding blend surfaces (pseudo fillets) using the Surface>From Curve Network command. This will yeild a much better (and untrimmed) surface that is less likely to get you into trouble. There just are some surfaces that there is no easy way to create with NURBS. I've tried all the other methods listed in this thread and they "will not yeild" the desired results. They come close, but not close enough.

[andrewjohn81]

I really have to disagree with you. There have already been several descriptions of ways to get good surfaces that describe that piece without a problem. And they produce the final model without having to create it once, then disect and rebuild it.
The problem with that method, as I described, is you don't get the proper radii. So, it really doesn't give you the correct model at all. I did that, disected it, just to make sure I was correct, and the radii were different sizes all over the place. Although everything is tangent, you can't really control all aspects of the model that way.

[Dimos]

Saiko... To convert the rounded rectangle into four curves that can be used with the Surface>From Curve Network command you must first join together the lines and arcs that form the rounded rectangle. Next, draw two lines diagonally across the round rectangle. Start and end the diagonal lines from were the corners of the rounded rectangle would be if they were not rounded. Use the diagonal lines to "slpit" the rounded rectangle into four seperate curves/sections (the four seperate sides of the rectangle). Now you can use the resulting curves with the Surface>From Curve Network command.


You can do it using points to split, as an option of the split command itself. It snaps at the mid as well as at the end of the rounded corner.

[7Stones]

I really have to disagree with you. There have already been several descriptions of ways to get good surfaces that describe that piece without a problem. And they produce the final model without having to create it once, then disect and rebuild it.
The problem with that method, as I described, is you don't get the proper radii. So, it really doesn't give you the correct model at all. I did that, disected it, just to make sure I was correct, and the radii were different sizes all over the place. Although everything is tangent, you can't really control all aspects of the model that way.

Hmmm... You say there have already been several descriptions of ways to get good surfaces that describe the convex box top without a problem. Are you just refering to the ways listed here, in this thread? If so... Like I said... I've tried all the methods listed here and a few others without success. The results are always surfaces with flattened corners or surfaces with sides that look pushed in midway between the corners.

I'm not after a just a good surface. And I'm not after a surface with perfect radii. I'm after the exact surface shown, with all its variable radii. Also I'm after a clean surface with the same viriable radii from one side to the other (mirrored halfs in a single surface). And I should say that from my own experimentation a Patch worked much better then a Surface From Network Curves when using the projected curves to rebuild the surface. Ewww, the results from a network curve surface were awful in this instance. ...Could be I'm after the impossible. :)

I have nothing against being proven wrong. I'm far from being any kind of expert and I'd love to see what you've come up while trying to recreate the convex box top. It seems to me we may be after different results.

I'd have to agree that there is no way to control all aspects of the surface I'm after.

~Cheers

[andrewjohn81]

I'm afraid the surface that you are after may be completely abstract and impossible in reality. I haven't found a surface, in reality, that isn't possible to create in the computer as of yet.

If you'd really like to problem solve this one then model it out of clay, sculpi, playdo, or similar stuff. Then if you still can't solve it, take a few pictures of it and post it up. Almost any problem can be solved this way because it is easier to visualize how it is created, although I wouldn't recommend doing it for every surface of course. I have used it as a learning tool though in school and it works great. Many times you visualize, in your head, the object but negate only one part of it. Because of that you may waste your time in the computer.

For this surface I think there are few possible shapes, with Many possible ways to create them. Some are more difficult and completely unnecessary, but work nonetheless.
I will create most of the possible shapes next week and post them, unless someone else does it first. I'm going to be gone all weekend and my computers are busy for today that have rhino. Hopefully then someone can pick the one that they were looking for, we can discuss how it was created, and can be done with this topic.

[andrewjohn81]

Ok, one of my computers wasn't being used for a few minutes, so I did a few little boxes.

The two on the left are ones that I just thought hadn't been explored yet. I didn't bother cleaning them up though because I didn't think that's what anyone was looking for. The one all the way to the left has constant radius, but of course it has those arches...not like the pic.

The one in the middle is actually similar to a few I've seen, but it only took me roughly 10 clicks and a few keyboard commands. The explainations I saw were a bit more complex than I thought they should be.

The one on the right took the longest. I tried to make it resemble the pic a bit more, but not as far as scale. I had to wrap it up quickly at the end and was in a rush, but I think every surface is tangent. Notice the lines look straight, as in the picture, but they actually are quite curves if you look at different angles. I shot that angle as an example to show.

You can download the file here (http://andycg.com/temp/3shapes.rar)

[7Stones]

AndrewJohn81... The far right one is the closest to what I'm after personally. If you could give a brief description of how you made it that would be great. I am pretty certain I know how you made it but I'm curious anyhow. (I'm using Rhino 2 SR3)

The other two were obvious examples which I had already tried and not at all what I was after.

BTW: I am working off a couple of real world examples (as I mentioned one is a soap box, the other is a wood jewelery box). Sorry I can't get my hands on some photos to post.

Please forgive me if I'm dragging this out.

[andrewjohn81]

this is what I did:

-create a box
-explode
-delete the top
-dup top edges
-rebuild those curves to 4 points degree 3
-shape all curves with slight arc, opposite curves need to be the same. Dup if necessary
(you could also just build your own, that's cool too)
-create the shape you want your top arc to be by snapping midpoint to each line.
-create network surface
-rebuild that surface to 4u,4v 3u,3v

-join the 5 bottom surfaces
-fillet edge on the four edges
-explode
-rebuild the radii to be 4u,4v 3u3v (only needs to be degree 1 in one direction, but I don't know which direction you shape would be)
-shrink all surfaces

-create circle and extrude along each edge of bottom, excluding the corners. Only do the straight ones.
-trim the top surface using those four 'pipes'
-delete the 'pipes'

-use any method you wish to recreate the top surface that shape, but not trimmed.
-create a tangent radius curve for one side of the edges
-sweep2 to get a surface there
-do this for an end and side

-network surface to create the corner

-rebuild/reconstruct/tweak surfaces to force tangency because the match tools never do it quite right.

-mirror the pieces and move them into place.
-Join

-Done

[7Stones]

Thanks!

Here are two images of a box I made yesterday using a method similar to what you describe. It's getting closer to what I'm after.

http://img193.echo.cx/img193/3533/convexbox2wire8ck.th.jpg (http://img193.echo.cx/my.php?image=convexbox2wire8ck.jpg) http://img193.echo.cx/img193/5876/convexbox2render9xz.th.jpg (http://img193.echo.cx/my.php?image=convexbox2render9xz.jpg)

I will give the method you descibe a go.

[saiko]

I'm afraid the surface that you are after may be completely abstract and impossible in reality.
but if u make it in any polymodeling software it looks perfectly ok!
should i post the max file?

and 7stone, though your last one is also having some inverted surface but u seem to have achived the closest surface what i was looking for,

http://img292.echo.cx/img292/2934/box5yi.th.jpg (http://img292.echo.cx/my.php?image=box5yi.jpg)
i'd like to know how u did it....



check out these examples:
http://www.watch.impress.co.jp/av/docs/20050406/sony1_02.jpg

http://www.watch.impress.co.jp/av/docs/20050406/sony1_08.jpg

look at the buttons, though not exactly the same what we are trying but the middle one give food for thought....

[andrewjohn81]

I'm fairly sure the radius in the center of each fillet is slightly larger than the ones at the end. You can see described by the highlight on the middle button fairly clearly. It bends closer to the center, and pretty evenly at that. It also doesn't look like a perfectly uniform surface because it's probably a silicone based button, but I see where you're going with it.
Good find on those shapes.

After speaking with one of the material engineers here also, he said that is almost certainly a manufacturing "problem" that they use to their advantage on occasions such as these.

[7Stones]

Saiko... You picked out exactly what is still bothering me about my last attempt at the surface. The corners do still look like they are made of inverted or concave surfaces and they do still look strange.

I need to find some time this week to do a little more experimenting and then I'll post what I found to be the best method for creating the surface.

[andrewjohn81]

I just noticed a bit more of the post.
Please, Saiko, post the Max file.
If not that, then export it into Rhino, which is what I was going to do.

That way I would use it as a template for the shape you are actually trying to create.
I think you may find that it only "looks" like what you are looking for because of normals, but we'll see.

[saiko]

now i feel andrewjohn81 is right!
i examined the max file again, and it looks the same as rhino

http://img.photobucket.com/albums/v737/saikob/buldgecopymax.jpg

(though in the wirefreame it looks 'perfect' ).....
http://img.photobucket.com/albums/v737/saikob/sasacopy.jpg

and this is the image of the porsche dashboard where i had seen this kind of surface for the first time but surprisingly here also the surface is looking inverted towords the corner!
http://img.photobucket.com/albums/v737/saikob/porsche.jpg


what i feel now is, in rhino also the surface is not getting inverted but becoming planer toword the coner,
which somehow, in the render looking inverted.....

that way 7stone has already achieved the desired surface.


the 3DSmax file attached.

[KBOC]

I don't know if this was mentioned, but NETWORKSRF works.

I just did it.

Here's the file

[andrewjohn81]

you should probably read the rest of the threads KBOC. The top is suppose to be curved. The argument is about the shape of that. Not that network surface fails or anything. It's the shape it will create in the corners.

[7Stones]

Saiko... lol... Like I said earlier... the images we are working with here can be very deceptive. It's not always clear what the actual geometry is and one must really either work from a real world object or carefully dissect a CG version of it to determine its true geometry, as Andrewjohn81 suggested.

At times, the shaded view in Rhino can also be a little deceptive in itself (just a little). It presents objects in a lighting situation not often seen in the real world, which can, on very rare occasions, make objects appear different from what you would expect. In some situations it is also good to check your render mesh settings to make sure they can accurately recreate the surface in polygons when shading, but you probably knew that.

I'll keep working on reproducing the real world surface I'm after and let you know what the results are and how I achieved them.

[KBOC]

The top is suppose to be curved.

Exactly, you can get the shape he wants with networsrf. I posted a file example.

[andrewjohn81]

the file you posted had a flat top and wasn't the shape that people have been going after.

[KBOC]

Flat top? What are you talking about?

[andrewjohn81]

the file you posted had some radii, but no curved top....it was planar.

What people were trying to do was create the top of the box/button/whatever with a curved top and have a radius that went all the way around while still keeping tangency and the same radius all the way around.

After being established that what we all initially thought we saw in the Max image was impossible, they are now trying to find the best way to create a similar shape shown in the pictures.

The main problem was, if you curve/bulge, the top of a box then you either will have arcs at the edges, or the surface has a planar tangency at all four corners giving the appearance that the surface is convex in the center and concave in the corners.

The object you created, however, was planar on the top, neither convex nor concave.

[7Stones]

OK… here is a mini tutorial of sorts on how the domed box top “might” be created, which is inspired by the ideas posted by everyone in this thread. Thanks to everyone who posted ideas! And thank you Saiko for starting this thread!

I must say that my early attempts at recreating the domed box top were flawed and misguided because I didn’t truly understand the surface and hadn’t studied it enough. I even stated that the method(s) I used early on were not ones I was happy with, nor was I recommending anyone else use them. Those early attempts did, however, give me a better understanding of the surface shape I was after.

I should also mention again that personally, I wasn’t just seeking to recreate the domed box top surface. I was also seeking a logical, quick, and predictable method for doing so that would offer some degree of control over the resulting surface’s shape. And the particular domed box top surface I was after was one that did not have flattened corners, but yet, it did seem to have straight looking edges.

Oh… One last thing… I’m using Rhino 2 SR3 to create the surface with the Patch command and there do seem to be differences in the resulting surfaces created with this command in Rhino 2 versus Rhino 3. I’ll talk more about these differences below.

~~~
Part 1

How to Create a
Rectangular or Square Convex Surface
with edges approaching straight lines and
without flattened corner tangencies.

First, let me quickly review the problem and what is meant by “flattened corner tangencies:

If one creates a rectangular or square convex surface with either the” Patch” or “Surface from Network Curves” commands by using lines forming a rectangle or square and arcs drawn between the mid points of the lines, as shown in Image 1, the resulting surface is always going to be a convex surface with corners that look flattened, as shown in Image 2. This is a result of using perfectly straight lines for the edge of the surface, which join to form sharp corners. The straight lines and sharp corners cause the surface curvature to flatten out.

http://img198.echo.cx/img198/4716/conveximage017db.th.gif (http://img198.echo.cx/my.php?image=conveximage017db.gif) http://img198.echo.cx/img198/6224/conveximage021tz.th.gif (http://img198.echo.cx/my.php?image=conveximage021tz.gif)

If the arcs are drawn diagonally, however, from corner to corner, across the rectangle or square created by the straight lines, as shown in Image 3, and then the Patch command applied, the resulting convex surface won’t have flattened corners. This does introduce another aesthetic problem, though. The resulting surface, as shown in Image 4, will tend to have edges that look pushed in mid way along their length. This is a result of the surface edges curving upward at the corners and then down as they travel in from the surface corners to the surface edge mid points. So the edges may look straight from a distance but they are actually wavy.

Note: In Rhino 2 it might be best to create the Patch surface using a square rather then a rectangle. The resulting square surface can then be scaled into the desired rectangular shape. If the surface is created from a rectangle and diagonal arcs in Rhino 2 then the surface’s control points have a tendency to run perpendicular and parallel to the diagonal arcs rather then the lines of the rectangle, which will become an important factor when pulling points later. This may not be the case in Rhino 3.

http://img198.echo.cx/img198/9098/conveximage039if.th.gif (http://img198.echo.cx/my.php?image=conveximage039if.gif) http://img198.echo.cx/img198/2938/conveximage049nz.th.gif (http://img198.echo.cx/my.php?image=conveximage049nz.gif)

In the end I found it best to combine the use of arcs that run both diagonally across a square and perpendicular between the mid points of its sides. I also found that if I drew the square with the box’s final edge fillet radii instead of sharp corners I would get a much more pleasing surface with a little less need for point pulling later on. So what I did is used lines and arcs to draw a square in the top viewport with rounded corners equal to the radii of the fillets on the box I want to create. I then placed a point in a side viewport above the center of the square, were the apex of the box top’s domed surface should be. Using the Curve>Arc>3 Points command I snapped to the opposing mid points of the square’s arcs and lines and the point centered above the square to create the arcs that bisected the square both from side to side and diagonally from corner to corner, as shown in Image 5. Selecting all of the lines and arcs shown in image 5 I applied the Surface>Patch command to them to create the Patch Surface shown in Image 6. When creating the Patch Surface I used the following settings: Sample point spacing 0.1; Surface U spans 4; Surface V spans 4; Stiffness 0.001. Using 4 for the surface span settings gave me a surface with just enough control points to control it but not so many as to be confusing.

http://img198.echo.cx/img198/2821/conveximage059ig.th.gif (http://img198.echo.cx/my.php?image=conveximage059ig.gif) http://img198.echo.cx/img198/9450/conveximage067fa.th.gif (http://img198.echo.cx/my.php?image=conveximage067fa.gif)

Turning on the Patch Surface’s control points I selected the control points shown in yellow in Image 7 and pulled them downward in a side viewport to make the surfaces edges look like they were arched upward and not so up and down wavy.

Note: This is the time to keep editing the surface’s control points, selecting them in groups of opposing pairs, and moving them together to keep the surface’s symmetry while altering its shape until it looks the way you would like it to.

Once the surface was shaped appropriately I scaled it in one dimension, changing it from a square to a rectangle as shown in Image 8.

http://img198.echo.cx/img198/2940/conveximage070fs.th.gif (http://img198.echo.cx/my.php?image=conveximage070fs.gif) http://img198.echo.cx/img198/8710/conveximage084gv.th.gif (http://img198.echo.cx/my.php?image=conveximage084gv.gif)

[7Stones]

Part 2

How to Create a
Rectangular or Square Convex Surface
with edges approaching straight lines and
without flattened corner tangencies.

Next, I untrimmed the surface as shown in Image 9 and then trimmed it in the top viewport with an ordinary rectangle the same size as the surface was before it was untrimmed. Performing this step gives the surface perfectly straight edges with sharp corners in the top viewport.

With the surface re-trimmed I quickly created the box sides by drawing a corner-to-corner rectangular plane as shown in Image 10. Joining the corners of the rectangular plane to the corners of the patch surface with lines I used the lines and the edges of the surfaces to create side surfaces using the Surface>From Planar Curves command.

http://img198.echo.cx/img198/7868/conveximage092op.th.gif (http://img198.echo.cx/my.php?image=conveximage092op.gif) http://img198.echo.cx/img198/2326/conveximage107ul.th.gif (http://img198.echo.cx/my.php?image=conveximage107ul.gif)

With all the boxes surface made I then joined them together and tested the resulting object for errors and open edges. After the object was checked out I filleted its edges as shown in Image 11. Now, I could have stopped here with regard to creating the domed box top but the resulting fillet surfaces were not very clean and symmetrical. The fillets also were not quite the right shape either, so I exploded the object and deleted the fillet and side surfaces. Using the extrude command I selected the edge curves of the box bottom surface to create new side surfaces with straight, untrimmed edges as shown in Image 12. After the side surfaces were extruded to the same height as the corner fillets the corner fillets were deleted and recreated with the extrude command as well. The idea here is to keep the object’s surfaces as simple and clean as possible rather then use trimmed surfaces.

http://img198.echo.cx/img198/1693/conveximage118ea.th.gif (http://img198.echo.cx/my.php?image=conveximage118ea.gif) http://img198.echo.cx/img198/530/conveximage120sl.th.gif (http://img198.echo.cx/my.php?image=conveximage120sl.gif) http://img198.echo.cx/img198/1028/conveximage136at.th.gif (http://img198.echo.cx/my.php?image=conveximage136at.gif)

This brings up yet another problem… When the fillets were created the patch surface was trimmed but the trimming operation was not very clean or accurate. You can divide the surface up into quarters and the edges of each quarter will be different. So what I did is duplicate the edges for one quarter of the patch surface as shown by the yellow curves in Image 14. I then mirrored these curves and joined them to form a new and symmetrical rectangle. After un-trimming the patch surface once again I used the new rectangle to re-trim it as shown in Image 15.

http://img198.echo.cx/img198/707/conveximage145kl.th.gif (http://img198.echo.cx/my.php?image=conveximage145kl.gif) http://img198.echo.cx/img198/3589/conveximage156hu.th.gif (http://img198.echo.cx/my.php?image=conveximage156hu.gif)

Almost done… The next step was to create blend surfaces between the patch surface and the side surfaces as shown in Image 16. Then the corner surfaces were created with the Surface from Network Curves command. The vertical size of the box lid's side surfaces were then adjusted and the rest of the box created. The resulting surfaces where then joined, tested and shaded.

http://img198.echo.cx/img198/2990/conveximage161ro.th.gif (http://img198.echo.cx/my.php?image=conveximage161ro.gif) http://img198.echo.cx/img198/9741/conveximage170jw.th.gif (http://img198.echo.cx/my.php?image=conveximage170jw.gif) http://img198.echo.cx/img198/83/conveximage187ew.th.gif (http://img198.echo.cx/my.php?image=conveximage187ew.gif)

Tah-da! That completes one method for making a rectangular convex surface or a box with a domed top. The model shown here isn’t perfect, but with a little more work during the point pulling stage it could be.

Comments, criticisms, corrections, and objections welcome. If you have a better way to create the surface in question… PLEASE… post it!

[KBOC]

the file you posted had some radii, but no curved top....it was planar.
doh, I get it now... alright... I still think this can be done with networksrf.... let me try again.

[7Stones]

KCOC... Yes, it can be done with a network surface once you've found a way to come up with the proper curves, not that it is too difficult to come up with the curves.

Personally, I found it easier and quicker to use the Patch command in this instance.

[andrewjohn81]

That is a very unique way of going about that shape. It's good you check for bad geometry/errors throughout the process. I like that.

My possible critique though is that it appears that you have something opposite of what the very first problem was.
At first, the problem was that the corners appeared concave, although they were just flat. Your corners appear to be much more convex than the sides, so you look to have the opposite problem. This may just be the shading, but that's what it looks like in the image.

I'm also not horribly happy with the topology of the object. Sometimes you've gotta do some crazy looking pieces to get things done, but if this was the start of a more detailed object, then it wouldn't be very nice to work with.

It generally isn't a good idea to make an object corner to corner like that. Better to keep the UV coordinates flowing in the same directions. What you have done there is turned them nearly 45 degrees. This makes meshing quite difficult, and working off of that surface can also be hard. There are ways to make that surface flowing in the same direction. That would be my only advice.

Very well thought out though. Nice contribution.

[7Stones]

Andrewjohn81... Yes, the corners are "more convex than the sides", but I don't see this as a problem. It is in fact a necessity (to get the particular shape I'm after) to have the corners more convex then the sides to achieve the rectangular convex surface "with edges that 'appear' to be straight" but "without flattened looking corners" (I don't want even the slightest hit of flatness in the corners, otherwise I like the last method you listed for making the surface). What is important to me is that the entire surface is convex and appears pushed out in the same direction from edge to edge and corner to corner. To do this I knew that the curvature of the surface would have to be greater in the corners.

I agree about the problems with the topology and the UV coordinate direction of the surface. It really bothers me that I can't tell the Patch command in what direction the UV coordinates should be plotted. I may make one more, much smaller post about a method to create the surface with better topology and UV plotting.

I've gotten in the habit of always checking my objects for errors and have been doing so for some time now. It saves a lot of time and frustration. :)

Thanks for the imput! It's appreciated.

[andrewjohn81]

Ok, I understand more, then, what you were going after.
You could, fairly easily recreate the shape, once trimmed, by taking some curves off of that surface and rebuilding it from those. It may change the shape ever so slightly, but you probably wouldn't notice if you did it right. It wouldn't be visually different, so I think that's fine in this case.

I would probably create some planes where I would want curves, then get the curves from the intersections. Rebuild the curves, this is where you may loose a small amount of the shape, and try network surfaces first to see what you get. It doesn't always work perfect for more complex shapes, but sometimes it'll do a great job.

7stones, please don't be insulted, I always just put extra info for the people who may be reading all of this just to learn. This is the way I really learned, even though I went to school for it first. Too bad some of my teachers were fairly useless.

I seem to remember you Did want the look of the top surface being pushed out from behind. I did that to a box and it does seem to want to flatten out more on the sides than the corner. So I think your shape really hit right on.

[7Stones]

Andrewjohn81...

Insulted? Me?

Never!

I hope I didn't say anything to make you think I was insulted. I'm here to learn and value everyone's comments.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.