Some surfaces behaviour makes me confused

[Jacoolo]

amazing

anyway, I returned to my old logitech mouse design. I felt not happy without final good model.

So, I run three approaches to bite the "thing":

in the first approach I made split curve by sketch as shown below



bottom curve



intersection curve



It can be noticed the curvature at "s" shape is far from desired. More - there is curvature discontinuity in intersection curve even though both layout curves seem to be continues



then I made top surface as VSS and trimed it



in the end draft surface by VSS and Boundary Blend



far from success though

 

[Jacoolo]

now second appraoch

main curves, sketched ones were replaced by "hybrid" curves



I had a hope I could avoid mentioned in previous post curvature discontinuity in intersection curve, but no...



so this time I tried to make it by boundary blend. Since there is a problem to set a drafti.e with 3 deg, because bottom end curve is not tangent to such angle, so decided to handle this with two steps - main BB and COS for trim and final side curve creation



final side BB with draft



in the end "spine" surface



but his appraoch also provides not desired results at end part of model

 

[Jacoolo]

third appraoch - not finished yet

same layout curves as in previous appraoch, but this time main surface is generated from start to the end by one feat - VSS with tangency. However since - as I mentioned before, such angle can not be introduced at whole curve long, so I decided to make multi angle draft surface. At the rear part there is 0.5 deg which finaly comes to 3 deg as shown below



I haven`t had time toi set this with evelgraph so it this surface is made by three ordinary VSS, which one is controled by trjapar.

Anyway this appraoch gives the most eye candy result so far.

Still I do not quite know how to perfect finish rear and front(nose) part of this nouse



I also assume side surface is dived on to, and tjis one close to draft surf is made as separated one with curvature transition to the main one, as in appraoch one and two.

Still I do not why I obtain this curvature discontinuity in intersection curve when all curves seem to be continues

 

[Jacoolo]

please do not hesitate to quote any of this three approaches

second - please excuse me my grammar. I was in hurry when writing these posts )))

 

[Jacoolo]

third approach as far seems to be right one, but...

the quality of rear part of the surface is far from perfect. With a closer look in seems to be concave





horizontal curavature seems to be good but, that one in opposite not(is it a curvature or radius?)

I was curoius where is a reason for such VSS behaviour. The angle diversity for draft surface was right choice, but it occured to not be the last one. I was thinking where does the problem lay?

So I start to consider if the one of trajectory does ahve an impact. I use 3 traj in this example



The qeustion was - how section behave along normal and chain 1 trajetory? To investigate this I created special sketcher with a normal line to see how the normal look like



it can be noticed there is a moment where normals intersect each other.

 

[Jacoolo]

the next step was to change normal and chain 1 trajectory to ensure better normal arrangement



that was the right directon







this is the same VSS but without additional normal trajectory

 

[Jacoolo]

in the end it can be summarized in following sentence - greater problem seems to be, simplier solution occurs in the end

the point was I fixed to much on split curve from real mouse, trying to make it same with one step

this "S" shape should come now not at the beggining

in addition I encountered funny error, I`ve seen first time



Have You seen something like that before?

Now I have to put some attention for fron part of the mouse and the "spine" surface

any input is pleased to come, so do not hesitate!

 

[Jacoolo]

two cents more, while I feel lost in the maze

There is a question - how should be draft surface prepared in proper way?

example 1) - third approach for mouse model

the draft surface is made by VSS through split traj with a setting: Normal to Projection - planar datum



then I use this draft surf for main surface VSS, where normal dir for section is controled by special traj



the results are not so nice though



example 2)

draft surface is made by VSS but without any normal control - simple normal to origin



then main surf by VSS with tangency





much, much better results

so, in the end, what is the proper way to build draft surface?

 

[jeff4136]

I've been trying to remember which of the old cartoons I watched
as a kid had the cat saying; "I hate mieces to pieces!".
(mieces being: mice)
- - - -
Thinking outside the 'box' and outside the boundaries ...
Following a school of thought that says a network of freehand curves
(no matter how smooth or 'right looking' they appear to be) is
usually a poor foundation for a smooth complex surface logically
leads to the same conclusion for complex boundary curves, in this
case the top view profile or top / bottom parting line. 'Coherent'
surface shapes will intersect to form coherent boundary curves that
can't be guessed at and created freehand. That's the thought. Here's
the model. Have fun playing with both. ;^)
2008-03-25_025949_outside_the_box_mouse--wf2--.zip

 

[Jacoolo]

Thx Jeff for the model

I like the most the way You set spline for main traj, great tip.

The rest - well I like all Your models, this is either, however it introduces only general appraoch for such design. For me lacks some great details I try to address - i.e, your models has no control over shape seen from top view. You set is as rectangle, so then You can simple trim it with ordinary curve-trim-approach. This is nice while You avoid problems with "tricky" intersection curve which curvature is sometime hard to control. Two, You avoid problems which I met while using such curves, plus special normal curve plus tangency to draft surface. I bet, less is better, and maybe I chose wrong direction. However till this time, I miss clear statement when exactly draft surface should be used.

ok, two models from my side for exchange


2008-03-25_053242_no_spec_norm_traj_good_curv.prt.rar

2008-03-25_053309_worse_results.prt.rar


2008-03-25_053329_mouse.rar

Check rear surface with curvature > surface to catch what I mean by "worse" results

 

[jeff4136]

> your models has no control over shape seen from top view


Sure it does. You're thinking 'inside' the box. Depending on what the
desired shape is you may have to create the basic surface(s) in a
completely different manner. (This attachment is a simple modification
of the previous version.)
2008-03-25_080219_outside_the_box_mouse--wf2--a.zip



> some great details I try to address


Part of what I'm trying to address with this train of thought; those
'details' may be incompatible with what we want the 'basic' shape to
be. By trying to force the surfaces we end up creating a hard to
manage and esthetically unpleasant mess much like trying to
compensate for incompatible blend boundaries by adding intermediate
curves to force the shape.


I realize this is probably going off on a tangent and I'm not sure it
addresses the specific problem you're having. It's past my sleepy
time but I'll take a look at your models tomorrow.


> I miss clear statement when exactly draft surface should be used.


I'm not sure what's being asked here. Maybe it'll become apparent
after looking at the models.

 

[Jacoolo]

yea, I see it now

sd9 = d32 - .75 * sin(90 * trajpar) + plus project curve

but then, without those two additional things, included feats were not able to address this.

I try to set thing first manualy to play around a little with "curves". I consider relations in VSS for "polishing" state, so it is still ahead of me though.

anyway jeff, thx for input

 

[Jacoolo]

last iteration of mouse model for this day



I wanted to attach model but my upload to Mcad Central fails each time I make an attempt to do this

any way, the red surf is made by VSS with Tangency option to red and yellow surf, section profile is done by conic arc

 

[jeff4136]

> ... but I'll take a look at your models tomorrow


I had a chance to look at the models and examining the surfaces I'd
say the quilt created with three trajectories (fid_7096) is the
better surface set. The knot spans are larger, more evenly
distributed and it is the smoother of the two. I think the 'simpler'
surface definition is probably due to the planar origin trajectory.
I'm not sure what criteria you are using for the 'worse' judgement.



> when exactly draft surface should be used


I'm still not clear about the "draft surface". I'm assuming you are
referring to "Var Sect Sweep 16" (which controls the tangent
direction for both versions)? Irregardless, the only one size fits
all answer has to be; when it helps you achieve a goal. I'm wondering
if the real question you're asking isn't more related to choice of
origin trajectory and section plane orientations.


Why don't you post some additional pictures of the mouse (ortho views
to help determine the shapes involved) and we can make a project of
it?

 

[Jacoolo]

I suppose all depends from different taste of how the shape goes. I will try to explain as best as I can why I see "this" surface better than "other". First I want to focus on reasons for my "quality" interpreatations

So as main origin of my modeling appraoch stays this layout:



where:

1)is the main surf
2)is the side surf - curvature transition to 1) and tangent to draft surf as I suppose
3)spine surf builded over main surf 1)

taking this, surf 2) should be drafted



The model which I attached before contains only main surf 1) without surf 2) and prepared for it draft surf. I gonna try to attach this model today.

Apart of reflection I used curvature analisys to control surf quality. Specialy for rear part of main surf 1)





Worse surf for me is that one which contains many steps i curvature plot.

I know that the biggest problem for me to address is proper shape interpretation and explain other what I mean. Since I am surfacing noob it comes for me not so easy though. Second I rarely use shaded curvature analisys(Gaussian?), while I have problem with correct it interpretation

I will try to intoduce more pics and - I hope - more models in next posts

 

[jeff4136]

> Worse surf for me is that one which
> contains many steps i curvature plot.


Ok, I see.
First; the 'high frequency' ripples are not possible within a single
knot span so are probably curvature graph function related. Next; the
ripples are much less (?) than 10% vector length. Cause for concern
and a second look using another function? Maybe, if smooth is
critical. Cause for rejection? Probably not. I'd reduce the graph
scale. Almost any graph will look nasty if you scale it up enough.


Interpretation of analyses isn't black and white or easy. And, FWIW,
Pro/E's analysis functions are not the best around (tho' not 'bad' by
any stretch either).



- --


I guess it's worth noting; my 'smooth' assessment was a composite.
The reason I showed the shaded analysis was to indicate that there
was no appreciable difference between the two versions.
Edited by: jeff4136

 

[mcgowanp]

Jacek, the one piece of advice i would give you here is simplify the surfaces down, it looks to me like you are try to get far to much geo changes in one surface and this is where your getting problems in the surfaces.


I work with logitech also but different company, infact working on a mouse file now, I have quite a few surfaces in already and it still doesn't not resemble in any manner a stitched up quilt let alone a mouse.


Simpify your surfaces, and dont be afraid to have lets say you side surface as a combination of a number of surfaces, not just one surface.


I'm on a deadline for the next day or so but i'll try and have a look at your model later.


Paddy

 

[Jacoolo]

ths guys

I promised models here. That said I gonna do everything to make it true, but I am afraid it can be hard today. I was busy with my own web page update. A little one but devil hides in details, and these details I was trying to address. Anyway, I am confident this model will be taken to better state and then the way published in my page and down here either.

"I'd reduce the graph
scale. Almost any graph will look nasty if you scale it up enough." - well I`ve done in opposite, while I found it as main reason for bad results

 

[Jacoolo]

my consideration about handling main surf with one VSS feature - proposal from jeff

well this way one can save many steps and deacrease features count, however, it is hard to obtain right shape for first hit



now, trimming for the desired shape seen from below - projected curve



this is fine, however form side view it is not so nice though



the same for triming using curve from side view



nice but...



tunig is necessary - an addition relations to VSS must be added



with this approach you can cheap on feat count and curves number - no intersection curve. However it is hard to tune surface by

sd25=45-40*trajpar
sd19=27.5+10*trajpar

and obtain desired shape with 100%. You can only come close to curves layed in sketches(paper one or fotos)

These are my 2 groszy(or cents if You like) for this moment. I do not promise I will stay with it all time long, my opinion could be changed)))

 

[design-engine]

I learned that dell peripherals are developed in their pan Asian area. FYI

Edited by: design-engine