When is an approximation good enough?

[Anonymous]

(Apologies if in the wrong forum – it could go in a few!)

When is an approximation good enough?

With the recent spate of discussions generated by the Maxon Freeform plug-in, I thought it might be time to again raise a nagging question I have about ArchiCAD.


Draw a circle in 2d. You can dimension to it, accurately work out intersection points with lines etc.
Now model a 3d curved surface – for simplicity part of a cylinder.
Look at it in 3d, and you’ll see it has been “dumbed down” into facets. We all know you can increase the facet number to make it look acceptable in a rendering. But try opening a section, and all you get is facets; the geometric description of the object – that it is part of a circle, has been lost.
For rendering who cares? – I don’t. But try documenting a project and you’ll be employing lots of work arounds – drawing circles and using other overlaid 2d drawn geometries to arrive at an acceptable result.

Now it is getting messier.
You have a model that is an approximation, and 2d lines conveying the “real thing”.
Let us say you need to share your building model with a steel fabricator. However their draftspeople can draw real 3d geometries – I’ve seen their drawings (no doubt you have too) and so they need to start from scratch. Fine, it is probably best that they do from a liability standpoint (more on that later*).

However let’s say I use the freeform modeller to do a more complex shape. I can only ever convey that geometry as facets! Our steel draftsperson now only has facets to see how I arrived at the form, even if I used simple geometries and transformations (say a tilted cylinder)!

What has happened to the virtual model of the building that everyone interacts with? The concept looks a little sick if one cannot communicate the desired geometries with confidence …..

<<Aside: Revit users – can you comment on this? What happens when you section round elements? Facets? I’d love to know….>>

So how do Graphisoft get out of jail?
(*) At the moment (here in Australia at least) there is the issue of liability, and so we have the RAIA getting Architects to issue disclaimers as to the quality and use of CAD data we provide to third parties.
Hence Graphisoft’s oft touted “Virtual model” is implemented less than they might like, and we (Architects) generally encourage industry to recast our drawings rather than use what we provide, so as to shift liability.
Yes, it is inefficient.

Looking ahead.
I’ve been reading a bit about Boeing’s use of Catia on the 787 – all manufacturers use the same toolset and model. Fantastic for coordination and accuracy. Surely this is where we are headed (although given the scale of our industry maybe IFC gives us the glue to communicate together?).

So, how do Graphisoft address this? I suspect this would involve a foundation level change, but it is one that is needed ….

I’d be interested to know what you think….

[Rob]

Robert,

I agree that AC modeler is not one of the cutting edge ones and there is a lot of unanswered questions in it but introducing very accurate and complicated geometry engine would increase file size, computing time etc. I am not really sure if I could stand waiting on my already chocking computer for any longer and believe me it's got enough grunt to cope with any serious software.

Common mate, comparing Catia and AC is a bit too far fetched, firstly you're talking about industries with fundamental differences in the scale, forms and precision, secondly a budget for development of software and the end user resources to buy the final product are way apart. However, seeing that as the future of AC - oh, yes please! but we should be realistic and don't try to overkill it. Freeform modeling is beautiful to some extent and if you don't have clients of Frank Gehry ( in that case you could afford Catia or customised version of Microstation, I suppose...) it's highly unlikely you would use wild forms in exceptional precision on every day basis.
Don't get me wrong I've got my whinge as well like curved ramps for example, it's just pathetic to create a proper one without segments and facets.

[rm]

Rob wrote:
Robert,

I agree that AC modeler is not one of the cutting edge ones and there is a lot of unanswered questions in it but introducing very accurate and complicated geometry engine would increase file size, computing time etc. I am not really sure if I could stand waiting on my already chocking computer for any longer and believe me it's got enough grunt to cope with any serious software.

Robert K.,
I don't think Robert W. is really asking anything exceptional about the need of accuracy. His example of a simple circle not being translated into a 3d surface without facets is a ligitimate question and concern. I dislike having to clean up the differences between 2d & 3d also.

At our price point, we should not have to have a discussion about accuracy , or quite frankly the need to go into Maxon to design a sloped cylinder. The strategy of GS to try to get us to buy into using Maxon for "free form" architecture is ridiculous.

How much do we pay for this software, and we still can not do lathes and sweeps, and forget nurbs.

Architecture requires accuracy, just ask your lawyer, and I'm sure you know this too.

As for "choking" your computer, computers made within the last 2 years be they Mac or Wintel are pretty fast machines, with plenty of memory capacity and great graphic cards.

ACs code is old, and fat. I don't know anything about the dynamics of writing code and carrying over legacy code from one version to another. But from demos I have seen of many 3d architecture modelinig programs, when it comes to generating sections and 3d views, AC usually is at the bottom of the heap in generation time. To be fair every demo shows an optimized model, and perhaps because of the amount of information contained in an AC model I am not comparing Apples to Apples.

Rob wrote:
Common mate, comparing Catia and AC is a bit too...

My sense is Robert W. is noting that these things can be done, and your right that Catia and AC are different beast. But do you think there is much in AC that can not be done in Catia?

I really hope in AC 10 the Maxon addon is dropped and we get see some really improvement in modeling accuracy and tools.

Robert W.....good food for thought you put out there!

[Rob]

rm,

I didn't say that releasing Maxon addon (and therefore passing the bug onto different application for freeform modeling) is a great idea. I do not like that either. My point was that extra superb accuracy in the geometry core would be a big blast to AC speed. Now, realistically can you tell me where and how often you are using nurbs, I mean how the hell would you document such a thing mate?

Architecture requires accuracy, just ask your lawyer, and I'm sure you know this too.

We are talking about millimeters/fractions of inch with AC not bloody microns with Catia and I don't have my lawyer for accuracy, thank you.

As for "choking" your computer, computers made within the last 2 years be they Mac or Wintel are pretty fast machines, with plenty of memory capacity and great graphic cards.

...and point is???

As I said, I agree that AC modeling capabilities have to be expanded but comparing that with Catia or Boeing is nonsense.

[rm]

Robert K,

Sounds like I got under your skin... did not mean too, sorry if I did. I actually went to a demo of Catia at the AIA ( American Institute of Architects ) national convention.

They actually build with a 3mm tolerance, not microns. I think that would be pretty hard to do.

As for nurbs, you got me on that, not sure how to dimension to them, with ArchiCAD that is. ArchiCAD has enough trouble performing simple diagonal dimensions. But I guess if you need to now how, look no further than FormZ, or Catia.

My point is that I was agreeing with Robert W. that there SHOULD be accuracy between the 2D and 3D model. If anyone wants to settle for less, power to them.

My comment about attorneys, was not a slam on you, it was actually a slam on the attorneys especially in the US. God forbid you sneeze wrong on a set of CDs and your in court......it really sucks.

Yet ALL software companies get away with these ridiculous license disclaimers that essentially say, ANYTHING that goes wrong with the software is not our fault. Can you imagine as architects what response we would get from the public if we operated in that environment. Now when it comes to accuracy, your darn right that I expect it, down to the 1/64th of an inch. I said it before, I will say it again, we pay way too much for the software to expect anything less.

Again, sorry if you took my comments as negative toward you, never crossed my mind to do so.

Be sides, I'm expecting you to show me around if I can every make to Australia.

Cheers!

[Rob]

rm,

I didn't want to sound offensive (apparently I did though).

I just didn't catch your drift in some moments so I've pushed it (a bit too much as I should've, I suppose.)
Hopefully, you'll get over it...

[gerd]

hi,

I also agree: Its not necessary to reach the precision and the construction abilities of catia for architecture.

but on the other hand:
- its more calculation for the program, but the machines get faster, so in future this will be possible without slow down the working too much
- full precision for sections
- much less lines (1000 lines --> 1 arc or 1 elliptical bow); millions of lines slow down archicad very much, so refresh of screen could be faster!
- possibilities to realize level of detail for objects and general archicad geometry dependent from distance (for faster rendering or real time walk through or faster working in 3d window)
- architecture gets more organic; archicad should be able to draw this
- catia may be best construction program for airplane construction engineers, but not for architects.
- maxon form is a possibility, but the working method differs very much from working in archicad. so I hope to get real archicad modelling tools

but of course: these are wishes for future, would be nice in next version, but I think it will take much more time.
best regards, gerd

[Scott Davis]

<<Aside: Revit users – can you comment on this? What happens when you section round elements? Facets? I’d love to know….>>

No facets when sectioning round items. Here's a quick screen shot of a section through a half-dome shaped wall.

[stefan]

Gehry's office have transformed CATIA into a Building Modeller 😉 It's not purely aircraft-software anymore. This kind of applications stores much more then geometry: it stores how the geometry is created, e.g. a sweep stays a paramtric sweep. Then end result are accurate surfaces or volumes, but at any time the design history of the geometry can be edited. That is a whole different league then FormZ, which is a nice modeller, but which lacks all design (modelling) history.

But remember that the Bilbao Guggeheim was modelled in clay after Gehry's sketches and this clay model was digitized and transformed into NURBS geometry in CATIA. The digital model was used for pre-fabrication and on-site assembly, but not for design.

So there is at least some comparing here ...

Sure, most geometry in ArchiCAD ends up as faceted meshes, but I think, inherently, the GDL-model has most required info included. In fact, using MaxonForm doesn't change that much in this comparison: it's pure, simple geometry and, unless you are using CNC manufacturing, most contractors are not equiped to do more things with it then from a regular 2D drawing.

[Anonymous]

Thanks for the feedback – I find it interesting when people get defensive about software with statements like “introducing very accurate and complicated geometry engine would increase file size, computing time etc.”.

I’ve been using ArchiCAD most of my career and I have learnt that a lot of issues are based around ‘legacy code’ and fundamental decisions made when the program was first written. Issues of speed and accuracy are not always hardware based. There is no point being defensive – I was trying to convey real world problems I have encountered and wondered how others have dealt with them, and also raising awareness regarding what I see as a significant limitation in the software that could hinder further development.

It is interesting to see Revit, a program written from the ground up more recently than ArchiCAD, appears to be more advanced in this area.
What isn’t clear from Scott’s response (thank you Scott) was whether the lack of facets was just a display thing, or whether the curve info was retained – i.e. could the curve in section be dimensioned as a curve without employing work arounds?

And is it so unreasonable to use CATIA as a comparison?
I think it is prudent to look around at other CAD software, as to what is reasonable to expect, and what is still beyond most of us.
I too have read about Gehry’s development of CATIA for architecture, and while it may seem irrelevant now, who knows what the future holds…?
Here in Melbourne we have Mark Burry using it to document geometries for the “Sagrada Familia” in Barcelona ….

Which brings me to Stefan’s comment “Sure, most geometry in ArchiCAD ends up as faceted meshes, but I think, inherently, the GDL-model has most required info included. In fact, using MaxonForm doesn't change that much in this comparison: it's pure, simple geometry and, unless you are using CNC manufacturing, most contractors are not equipped to do more things with it then from a regular 2D drawing.”. (end quote).
So if we accept CNC machining is beyond most of us mere mortals, we need to convey data in a regular 2D drawing?
It appears so.
From ArchiCAD I cannot give them the facetted 3D approximation even for the most basic transformations. Instead I am left communicating them with a whole lot of 2D lines and notes, from which they reconstruct the item – which gets back to my original issue. If we loose the geometric fundamentals in our facetted world, what can we accurately communicate to others?