CAA V5 based or ACIS kernel?

Paul,

I went to the link and read some of it. I have no idea what CAA V5 is but it looks like enough material to keep Scott Adams in business for years. Please help a simple person understand.

The other Paul

Paul Salvador wrote:

Paul,

Neither I hope.

The rebuild times on Catia 5 are dismally slow, so is ACIS.

Mark

Paul,

From what I understand, it's kernel or component technology? Scott Adams,.. as in,.. Dilbert!? 8^)

..

Well, I can't compare but from the past, that is probably still true? BTW, one client I had (was working with 5K unique prismatic parts), choose UG over Catia just because of comparable slowness issues (not just large assembly use).

..

I am assuming CAA is a modeling kernal. If it is so slow how in the world did Boeing make the 777 with it?

Paul,

They didn't, They used Ver4 on UNIX clusters that utilized every idle CPU on the network.

Version 4 isn't an "exact" modeler so the data is easier to handle.

Mark

I'm not sure what version 5 release Mark saw or uses but from what I was told, the latest V5R14 or the last few releases have been faster. And, the client (to be fair, an ex-UG user who was looking for something faster and had the capital to do it) I was talking with said Catia modeling was a bit faster than SW. BTW, UG modeling rebuilds are faster, the problem I saw with UG was the menu structure/steps which was suppose to improve more (few steps) with NX3.

Also, btw, from what I know about the Catia and UG world, it's a different mind set,.. most are still not into parametric/feature based data in the way we SW and Pro/e users are,.. and most Catia/UG users don't like how parametric/feature based modelers limit their workflow (get it done, hybrid approaches). Anyhow, what's interesting is it's difficult to compare SW with them unless the data being compared is equally parametric/feature based. Without getting into a debate with what approach is better, it's best to say they all have good approaches and disciplines and it can be difficult to weigh which modeler is faster, because of the disciplines being used.

..

..plus, the data that was being used is reference graphic data sets, polymesh data sets, not the whole data file(s).

Well, surfaces are usually represented (tessellated) using polymesh data so you would have both in memory. What I'm saying is there is only polymesh data and most polymesh data or vertices are scalable to meet the graphical/visual/detail needs (or you access other data sets). And yes, it's a representation/approximation. Modifing the original data is managed seperately. UG does this as well and most 3D cad programs do this as well now, i.e., to reduce overhead. Polymesh data is not that taxing on most systems/graphics cards today and the graphics cards used back then for the 777 project would/should have been just as good or better than what we user have now using high end PC gfx cards.

..

So we are, as a group, heading for a performance wall which CATIA v4 can deal with but the technology we have and that which might be foisted on us doesn't have. Seems kind of strange to me.

Although, I was always of the impression that it was the parametric side of things that really brought down SW. After all SW has to build and tear down bodies from the first feature on if I don't miss my guess, whereas the non-parametric modeler is alway living in the here and now.

The other Paul

MM wrote:

Cliff.. I could give you a response (not always right or right for you) for just about every question (more or less experience responses) and I'm sure you will respond with an other(s). Mesh data sets, by themselves, are not an overhead which we/I have issues with, the surface/solids are, and most do not work on fully resolved data sets, they manage segments of the design (sub assemblies with references). And, this is not about manufacturing (cutting/tooling off polymesh data) but about data management (PLM). Some systems store the mesh data with the solids/surfaces or with properties and some separate linked data for reference/management and some do not. Man, you really like to get off on tangents, be argumenative or pose other questions,... because,.. you're bored? I'm trying to be constructive in understand where this is going, kernels and PLM, because I don't have all the answers, just more questions. But, Cliff, I don't care to play Q&A games to fill your time, ok? Thoery and what was and what is, is all relative so, I think for you to understand more (you're out of touch, imho),.. you have to actually use/demo the programs related to these ng's. Otherwise, enough of the know it all approach, it's been getting stale,.. .. 8^)

Cliff,

I think what Paul was trying to describe is the way they utillize a combination of "resolved" and polymesh (what Catia calls "mock up" ) data in order to facillitate the process of design. Catia 4 achieved this by allowing you to generate a seperate associated mesh file for each component. At least this is my understanding from people I've talked to at Boeing. We didn't have the mock up module at my last job.

A fully resolved airframe is a huge data set. It's not necessary to have the whole wing resolved in order to design the load ribs on the front of the spar. All you need resolved is the spar itself, and the leading edge of the wing. All other geometry can be represented as a shaded mesh.

This is probably not the most straight forward way of doing this, but it was one of the first. Solidworks has "light weight" parts where only the display list is loaded. Parts can be resolved, or made light weight, interactively as you work. Pro-E has "simplified reps" which is similar. I would imagine UG also has an anolog to this as well.

You were right about the graphics. In 1987, the highest end PHIGS, and OpenGL subsystems (massive amounts of high wattage silicon) weren't nearly as powerful as todays PC cards. That's just progress.

Regards

Mark

Jon,

From experience, Ironcad, and an eval version of Concepts Unlimited. It's been well documented for years. It also uses allot more recources than PS.

CAD CAM publishing did a test about a year ago comparing the latest version of Inventor, with Solidworks. The test model was an automotive grill modeled native in both systems. When a change was made that required a rebuild, the difference was 8X to 10X slower for IV. I don't kow how much was ACIS fault, or the incompetence of the Adesk programmers. Probably a bit of both

I would hope that they've improved on this, and I'm sure they have to some extent. ACIS does have some interesting features. And it may not bother some people, but it'd sure bother me. Waiting for rebuilds really messes with my concentration. Probably just gettin old....

Mark

Jon,

Nope,, takes to much time (and patience, of which I have precious little)

Having used their product since 87 or so, my guess is they'll follow the same pattern they've followed up to now. It will be "usable" six to nine months after release, but they'll get it right eventually.

It will probably be a bit tougher this time. Since they've gone native Windows, I can only assume that they'll be using the windows memory mangler. One of the things that made the program so stable ( the most stable I've ever used) was the fact that "you" could control how much physical memory it used. Upon launching, MC would grab that chunk and control it.

We'll see,,,,,,

Regards

Mark