Issue #1,112 | Inside the Business of CAD | 8 November 2021
Guest Editorial by John Callen
In the upFront.eZine article on Solid Edge 2022, Dan Staples makes the following comment: “I do think [convergent modeling] is a final frontier here, because meshes for a long time were the domain of the film industry and character modeling.”
It turns out that faceted models were the very first b-rep [boundary representation] structures, referenced in 1977 by CMU’s GLIDE, a polygonal modeler.
The Start of It All
GLIDE stands for “graphical language for interactive design” and was a research program in the architecture department at Carnegie Mellon University. The GLIDE program was funded by the Army Corps of Engineers who wanted a system to model their various facilities.
It was headed up by Chuck Eastman, who is generally acknowledged as the father of BIM [building information modeling]. Kevin Weiler, who originated non-manifold topology b-reps as his PhD thesis at Rensselaer Polytechnic Institute, was one of the members of Eastman’s GLIDE team. I was a member of the GLIDE team 1979-1981, where I was responsible for implementing the Euler operations from the original BDS [Building Design System], which were written in BLISS, to GLIDE’s Pascal.
GLIDE was an extension to the higher-level programming language, Pascal. Solid modeling operations were incorporated into an interpretive version of the programming language. Procedural constructors were developed to model aspects of buildings.
GLIDE pre-dated non-polygonal surface representations. Booleans were performed with the faceted representation; the intersection code was relatively straight forward. Later on, b-rep modelers introduced non-polygonal surface types.
Romulus. Being developed in parallel to GLIDE was Ian Braid’s BUILD project at Cambridge University in the UK, another b-rep program. The BUILD research was commercially spun off into Shape Data, which was the precursor to ACIS and Parasolid.
Romulus introduced analytics and had discrete intersection routines to do surface/surface intersection calculations based on surface type.
Only later were NURBS surfaces introduced and there was a big debate over whether analytics should be converted to NURBS equivalents and all intersections done in NURBS — or whether to continue to maintain the discrete surface types for precision and efficiency.
Performance and Interactivity
Kepler. After graduating from CMU, I joined the Kepler project at Evans & Sutherland. Kepler was an application development environment front-end to Romulus that provided a design space, giving designers a degree of interactivity not possible with solid modelers at the time. (Kepler really wasn’t an acronym; the project lead chose the name due to the historical individual’s role in astronomy.) Objects were built in the Kepler environment, and then the corresponding modeling commands were passed to Romulus to instantiate the design, often taking many minutes to calculate on a DEC VAX 780 mini-computer.
When the Kepler project’s proof-of-concept was completed, it was decided to no longer pursue the approach.
D-Cubed. Shape Data Ltd was the company that originated Romulus and later Parasolid. John Owen at Shape Data developed his dimensional constraint technology, which is the basis of D-Cubed . It can trace its roots to the earlier D/T-D [dimension/tolerance-draughting] work in Romulus.
After Kepler, I ended up relocating to Shape Data and led a project to implement encoding dimensions and tolerances in Romulus and automatically generating engineering drawings, easily a decade or more before PMI. The D/T-D function was released with Romulus v5.2.
During a design review I had with Bob Sproull, Bob pointed out that the D/T-D framework we had implemented in Romulus could be extended to be a parametric modeling system.
The company, unfortunately, had some very strong positions on how a solid modeler ought to be used and, as a result, did not realize advances like parametric modeling and PMI when they presented themselves. The D/T-D function was later removed from Romulus.
Evans & Sutherland was a distributor of Romulus, and then later acquired Shape Data. This was just the first of a long series of acquisitions passing Shape Data from one company to another, ultimately ending up at Siemens.
ACIS. The managing directors of Shape Data split off to form Three-Space Ltd, the company that created ACIS.
The Continual Development of Representations
Gradually, b-rep modelers expanded their geometry classes to support analytics (cylinders, spheres, toroids, and so on), and eventually a variety of surface representations, such as NURBS.
Facets became the primary method of representing scenes in image generation, such as for flight simulators, where the image generation pipeline was built on facets for performance. This need for image generation performance then moved over to the film industry and character modeling. Highly detailed visuals required many, many facets, though. Eventually facets were replaced by texture maps with which image generation reached new levels of realism.
Non-manifold topology (NMT) b-reps introduced the next level of representation, which allowed for interim Boolean results to exist (but not necessarily persist). ComputerVision’s Liberator project of the early 1990s, headed up by Gary Crocker, implemented NMT and supported mixed model representations — wires, sheets, and solids. An example was a chair with wire legs, sheet back, and a solid box seat.
Unfortunately, a real-world application never materialized. At one point, I proposed that NMT be used to model machineable models and associated machining processes, but CAD companies were only focused on design, not manufacturing.
Meshes are again making a resurgence with 3D printing. Again, this is based on the basic technology used to process STL files for 3D printing. By the way, most CAM systems operate off of a faceted representation. This is, again, due to the nature of the algorithms not operating directly off of the source geometry and the fundamental geometry of most CNC controllers, which are predominantly lines and arcs. The ability to specify NURBS curves to a controller is a fairly recent and limited advancement.
Interestingly enough, it seems that modeling systems cover the two representation extremes (b-reps and facets), but to my knowledge have pretty much skipped over cellular (homogeneous or adaptive) representations, such as voxels [3D pixels] and dexels [depth pixels]. These are gaining some pickup through today’s scanning and 3D printing technologies.
Through all this time, it has been interesting to see the advances on-going in modeling technology. Probably the most compelling is advancing the model beyond just a geometric representation, but also encoding all the corresponding product data. Who knows where this all will lead? Maybe someday geometric tolerances might be encoded in the solid model similar to how RESABS [numeric precision] and RESNOR [angular resolution in ACIS] are maintained.
- - -
There is so much that isn’t known about the early days of the industry. And there was so much shelved back in the day, because the hardware wasn’t up to it. Imagine if someone dusted off that stuff and migrated it to today’s hardware so that, as the saying goes, “Everything old is new again.”
[John Callen is an architect by training, and spent his 40-year career in marketing and engineering positions at high-tech companies. Much of his career was with mechanical design and manufacturing systems, and now he has returned to his architecture roots as Director of eTools Marketing at Lutron Electronics.]
And in Other News
Companies like PTC and Facebook just know that moving to the next level of abstraction gives them an advantage over competitors. They’ll be spending years and billions on Atlas and Metaverse. Companies should understand what their core is, according to Daniel Jackson.
They’ll fail, because full MCAD doesn't belong on the cloud and people don’t wear shoes on their faces. Better technology is not better humaneness. As technology becomes more abstract, Romanticism is the natural backlash.
Take, for example, the natural technological superiority of ebook readers over dead-tree books: even so, ebook titles now cost more, and the digital format is sinking in popularity to paper.
- - -
Cubic Orb’s plugins for land surveying work with most CAD programs, including those based on AutoCAD, IntelliCAD, ARES, Microstation, BricsCAD, and TurboCAD:
-
KaliBro - georeferences, edits, manages raster images in CAD
-
GeoView - operates on sets of coordinates in CAD
-
Wms2Cad - displays maps from WMS/TMS and WMTS services
-
TranMap - transforms CAD drawings between coordinates systems
-
And more.
Demo versions are available to download at cubicorb.com.
- - -
Lantek celebrated its 35th anniversary and reported that Q3 sales of its machining software were 30% higher than a year earlier. It has big plans for the future: increase R&D spending by 70% (from what, we don’t know), and adding 130 more employees from the current 260. www.lantek.com
- - -
Open Design Alliance adds a validation engine for IFC [Industry Foundation Classes] files to its SDK [Software Development Kit]:
-
Performs multi-level validation on IFC models
-
Customizable for a range of tasks
-
Supports simple low-level syntax checking
-
Custom validation defined in IFC files
-
Higher-level checking specified by the IDS [Information Delivery Specification] and MVD [Model View Definition] standards from buildingSMART.
The validator is part of Open IFC Viewer v22.9, which is available free from openifcviewer.com.
- - -
LEDAS worked with AWV Informatik to develop a 3D machine learning system that identifies machine tools with 95% success rate. ledas.com/news/884-3d-machine-learning-by-ledas-and-awv-achieves-95-success-rate
- - -
Graphisoft updates ArchiCAD 25 with a preview of Redshift renderer, more import and export formats, a preview of partial overlap checking, and more: graphisoft.com/contact-us/press-relations#/pressreleases/archicad-25-update-2-enhances-powerful-collaboration-design-documentation-and-visualization-capabilities-3140031.
Letters to the Editor
Re: Bricsys Conference
You wrote that the BricsCAD acquisition was “For Hexagon to get deeper into the AEC space.” Did you choose not to comment on this, to give us readers time to mull this over?
I’m interested in knowing what you think about that statement. It's been three years since the acquisition [of Bricsys by Hexagon] and nothing has really jumped out at us yet.
- Name withheld by request
The editor replies: Here is my initial take on the statement. From what I see (I am a beta tester of BricsCAD), Hexagon is putting a lot into BricsCAD, and that is the good news.
Hexagon’s origin is in CCM [computer coordinate measurement], which is a post-MCAD process, so they don’t have experience in AEC [architecture, engineering, construction]. They see BricsCAD as the path to entering the market, surrounded by all their other products, like point cloud acquisition and plant design software
Their mirror is Trimble, which started with surveyor’s equipment and is also trying to get into AEC, starting with the acquisition of SketchUp nearly a decade ago. We’re not, however, seeing any impact on the industry, as SketchUp is not a great entry point for a corporation trying to target a discipline dominated by Vectorworks, ArchiCAD, and Revit; SketchUp began as conceptual design software, not as design design software.
BricsCAD has a reasonable chance of making an impact in AEC, but continues to suffer from the same problem it’s always had: lack of mindshare.
I would say the most important aspect of BricsCAD is that it also works as a DWG platform, which is why the CADWorx division of Hexagon fostered the acquisition of BricsCAD in the first place. I am watching to see if BricsCAD-as-platform spreads to other areas of Hexagon, keeping in mind that Hexagon already owns several other CAD packages.
Re: Rådgivende Ingeniørers Forening
Your conclusion could also be “forming purchasing pools” to buy stock in the design software companies, because those companies listen far more to shareholders than to users.
It is also true that Norway has a $1.3 trillion sovereign wealth fund, which is an investing arm of the government: forbes.com/sites/palashghosh/2021/04/07/norways-sovereign-wealth-fund-makes-first-investment-in-renewable-energy-infrastructure
I would guess that the Norwegian government has sentiments similar to the RIF, so perhaps they could pool resources to buy a larger share of Autodesk stock, which would give them a louder voice in how things work at that company.
- Peter Lawton
The editor replies: Good idea, but.
There have been suggestions in that past that users buy up a majority of ADSK stock, but that would never happen. Sovereign wealth funds do not interfere with the running of companies.
Mr Lawton responds: Some of the stock-purchase suggestions to which you refer have come from me, as I have been championing that cause for the past 15 years, at least. 😎
However, to state flatly that it ‘would never happen’ is rather defeatist and I would respectfully submit that big companies count on that very attitude to continue to get away with being poor corporate citizens. “What are those two million frustrated users gonna do about (our latest rip-off maneuver), hunh? Fire the board?!”
Well, if those two million frustrated users are also shareholders, the board will be far less dismissive of our requests. Does that make sense? Not everyone connects those dots.
It will take time, perhaps a decade, but if enough of Autodesk’s corporate victims pool enough resources to purchase a significant portion of Autodesk stock, the company will have to listen.
Also, it may be true that sovereign wealth funds don’t deliberately ‘interfere’ with corporate governance, but they do influence it, if only by virtue of their purchase of stock. If RIF writes a letter to Autodesk, and Norway’s SWF (in conjunction with RIF) is holding 5% of shares (about $3.4B today, or 2.6% of the SWF), how do you think Autodesk will react? It’s just a matter of time and money, and there are many parties interested in getting Autodesk to change its ways. All those parties need is organization and patience.
Notable Quotable
“A model is not evidence. It is a theory expressed in arithmetic terms.”
- John Hinderaker
Thank You, Readers
Thank you to readers who donate towards the operation of upFront.eZine:
-
Stephen Schuller
-
Nicholas Busigin: “Thanks for the great eZine. I enjoy reading it every week.”
To support upFront.eZine through PayPal.me, then I suggest the following amounts:
-
$25 for individuals > paypal.me/upfrontezine/25
-
$150 for small companies > paypal.me/upfrontezine/150
-
$750 for large companies > paypal.me/upfrontezine/750
Should Paypal.me not operate in your country, then please use www.paypal.com and use the account of [email protected].
Or ask [email protected] about making a direct bank transfer through Wise (Transferwise).
Or mail a cheque (US$ or CDN$ only, please) to upFront.eZine Publishing, Ltd., 34486 Donlyn Avenue, Abbotsford BC, V2S 4W7, Canada.
*4700
Comments