Interview with Paul Arden of Luminova

Interview with Paul Arden of Luminova

CGA talks to Paul Arden about his role at Luminova and his views on the architectural CG industry.


CGA: Could you introduce yourself and tell us about your role at Luminova?

PMA: My name is Paul Arden and along with Chris Blewitt, I am a partner in Luminova. I suppose you could say I am the Technical Partner. Most of my time is spent
finding ways to improve our services through the application of the latest technology in global illumination rendering and real-time vr.

CGA: Tell us about your background. How did you get into the Architectural CG industry?

PMA: I studied Computer Science at La Trobe University here in Australia. Through my association with Chris I developed an interest in computer graphics, from the
perspective of the lighting industry. For some time I had been telling Chris of the significant potential of global illumination technology. The opportunity soon arose to exploit this technology as a business in its own right so we took it.

CGA: You deal primarily with the technical side of business at Luminova, how have you seen developments in the CG industry affect your business?

PMA: Unfortunately architectural visualization does not drive the CGI industry. Animation, effects and games have a much greater influence on the course software development takes. As a result, most systems (even those using global illumination) are not physically based. They have elements of a physically based system, but not an end to end solution which works in natural real world units.

Having said this, the industries mentioned above often help push CGI technology to new levels. While we are required to adapt their methods rather than use ready
made solutions, they are still responsible for a great deal of the advancement in CGI. Fortunately these industries are starting to see the benefits of global illumination and
are beginning to find ways to make it more usable in their fields.

We are often forced to take off the shelf products and heavily customize them to our needs. None of our software tools can function alone, they require the
connective tissue we have developed to create a coherent system.

The many great advances in speed and quality that have been in today's graphics software is often only part of the solution. Finding ways of utilizing these advances
in our production and rendering pipelines is where the real payoff is realized. The best technology in the world will mean very little until you use it on a real world
project. Without this there is no means to demonstrate its potential to those who don't understand the underlying technology.

Probably the most significant change in CGI occurred some time ago when Windows NT started to emerge as a viable graphics architecture. Things have come a long way since then, I remember our first OpenGL graphics card. It cost about $US2,000 and could draw about 750,000 poly/sec, now the latest Nvidia based cards cost as little as $US200 and can do many millions poly/sec. Processor speed has gone from
200Mhz - 1700Mhz in only a few years and memory and storage have seen similar improvements. Before these changes CGI applications were limited to high end SGI or Sun platforms pushing their use out of the reach of many users.

CGA: What types of hardware resources are used at Luminova to ensure that your clients get quick turn around times?

PMA: The vast majority of our hardware are reasonably high end PC machines. We typically use Nvidia based graphics cards in our machines and make sure they have plenty of RAM (usually 1G) and plenty of processing grunt. We make heavy use of SMP machines, you can never have too many processors.

Our Render Farm currently consists of 100 processors and is the primary means by which we produce all of our output. We usually double the size of our render farm every 6 months or so to deal with demand and increased complexity.

Naturally we have plenty of servers and storage to manage all of this as well.

CGA: What advice would you give to a firm looking to set up an in-house rendering farm?

PMA: Don't fool yourself into thinking that setting up render farm is just a matter of buying lots of machines and installing software. This may be the case for a very small farm of 10 or so machines, however once you start dealing with 100 or more
things start to get very interesting. First up, most software packages (with the notable exception of mental ray) out there have very limited support for render farms. Usually you will need to take this further with your own tools and
modifications. Even software with network rendering features often has trouble working on a large number of hosts. You will also run into many other issues such as network topology, load balancing and simply attempting to manage such a large
number of computers. You need systems to perform software updates, periodic restarts, monitoring and job scheduling. To successfully set up a render farm you will need to become an expert in many areas which you may not have been involved with previously.

CGA: What have been your biggest challenges both past and present,with regards to computer renderings?

PMA: Developing a modeling process and system which allows radiosity ready models to be constructed rapidly in a team environment. Cracking this nut was nearly our death at the start since we were not aware of just how much impact modeling has on the entire process. Once we realized the effect we put all of our effort into resolving this problem.

Establishing the infrastructure for our render farm was also quite a challenge. As I mentioned earlier it requires a very broad knowledge and is not a simple as it sounds. It took some time but we now have a system which we can expand at will and
manage effectively with very little technical input.

CGA: What are your thoughts about emerging rendering technologies? What technologies do you see leading the way in the architectural field?

PMA: Unfortunately I am unable to discuss many of the very latest technology in detail due to confidentiality. However I think at the end of the day we will have much faster rendering tools which can handle much, much larger scenes. I think everyone is starting to jump on the global illumination bandwagon now,
however the early starters such as Lightscape and mental ray will have the upper hand due to their maturity as production tools. The latest craze of skydome images is something I find quite bizarre since Lightscape has had this ability since its
first versions. However at the very least the increased interest in global illumination will hopefully allow more resources to be committed to its development, rather than been seen simply as an add-on it needs to be an integral part of the rendering process.

CGA: What software do you currently use and have you used in the past for computer renderings and why have you chosen those particular applications?

PMA: Currently our primary rendering tools are Lightscape and mental ray. Both have strengths in different areas and allow us to deal with a wide variety of situations. We have developed tools which allow us to move our model, texture and lighting detail freely between Lightscape and mental ray allowing us to apply different technology to different situations.

For modeling we primarily use FormZ for polygonal modeling and Rhinoceros 3D for freeform NURBS based modeling. FormZ in particular is well suited to making radiosity-ready models, which is very difficult with most other modeling tools. Rhino is the only NURBS based modeller I have seen which is easy to use and correctly supports most types of trimmed surfaces as well as more straight forward NURBS surfaces. Both of these tools are also extremely reliable which contributes greatly to their productivity.

We are also currently testing many prerelease products for various vendors. I think there will certainly be some interesting unveilings at SIGGRAPH this year.

CGA: What do you feel are the current limitations of current rendering technology?

PMA: This is simple to answer: speed, quality and complexity. No matter how good it gets we will always want our rendering technology to be faster, produce higher quality images and allow greater scene complexity. Apart from these major points,
I think the main limitation of today's rendering software is a lack of a coherent physically based pipeline which takes you all the way from material and lighting definition to final rendered image. Many renders still work in clipped RGB colour spaces and take little notice of real world units. Tone-mapping is often performed in a trivial manner, even when in a high dynamic range scene, this can be one of the biggest factors in obtaining high image quality. Most renderers also still only support trivial light source distributions such as cone and point, for our work it is absolutely essential to have the ability to use arbitrary photometric distributions for light sources.

CGA: What project are you most proud of and Why?

Probably the Myer Southland Cosmetics area. This is an extremely large model which required significant automation to prevent individually modeling each piece of merchandise. It is also an excellent example of how far you can take Lightscape renderings.

CGA: What do you not like to see in computer generated architectural work?

PMA: Output which is not rendered using a physically based global illumination process. It is extremely difficult, if not impossible to produce convincing output without it. The whole point of the exercise is that it communicates to the people who matter (those spending the money) what their final product will look like, not what an artist (3D or otherwise) thinks it will look like.

CGA: What tip(s) can you give our readers to improve their architectural renderings?

PMA: Assuming you are using a physically based rendering tools such as Lightscape, I would say the construction of the model has more to do with the final quality than any other factor. If you have a poorly constructed model, no amount of material
refinement and creative texturing will make up for the shadow artifacts, light leaks and jaggy shadows. In my experience if you have an efficient system, modeling correctly does not take any longer and in the end saves considerable time by
reducing the amount of tinkering that has to go on in the rendering stages.

CGA: What is your favorite link to visit on the web? (not necessarily CG related)

PMA: http://www.rendermania.com/

CGA: Which/What web based resources that you have found the most informative?

PMA: Being a member of ACM SIGGRAPH (http://www.siggraph.org) and Eurographics (http://www.eg.org) I tend to spend some time trawling through the latest papers and historic ones when I have a specific need for them. Obviously I frequent the web
sites of all of the major CGI software makers, however since much of the emerging technology is only discussed in academic circles I also tend to spend time looking at University web sites for the latest info.