Tracing the rays
LEDS are replacing traditional light sources because they are cheaper and more energy-efficient, but also because they can be very stylish. The combination of growing numbers of LEDs and increasing demand for better aesthetics has led to a rise in the use of optical modelling software, so that LEDs can be ‘tailor-made’ for specific applications. But, as the hardware evolves, the software needs to catch up and improve its ease-of-use, processing speed, and standardisation.
The LED market has been driven in part by government requirements for more energy-efficient lighting in homes, offices, and city streets. ‘In the USA, they have outlawed 60 and 100W incandescent light bulbs, and the government is really pushing people to buy LED lamps,’ explained Rich Pfisterer, CEO of Photon Engineering. And, a desire for sleek, attractive lighting has heightened the demand. ‘Aesthetics is becoming more and more of a differentiator in the market,’ added Pfisterer. ‘The drive to make chic lighting for housing is driving the price of LEDs down − so there is more opportunity to use them in cars, architecturally, and for room lighting.’
The optical properties of LEDs mean that, unlike other light sources, the surfaces that the light interacts with can be designed in such a way that the beam pattern can be controlled. ‘One of the great strengths of the LED is that it is a very small source compared to other light sources − and, because of that, it can be shaped very efficiently,’ according to Dr Jake Jacobsen, technical marketing manager of the Optical Solutions Group at Synopsys. ‘You can really shape the light in a way that you just couldn’t with other types of sources.’
As the hardware has evolved, the LED market has driven the use of optical modelling software for freeform design. ‘Many software vendors have added the capability to design 3D freeform optics specifically for LED emission,’ said Michael Gauvin, vice president of sales and marketing at Lambda Research. This freeform design capability is being exploited for many lighting applications. ‘One of the big trends in the industry right now is to work on freeform surfaces,’ added Dr Bill Cassarly, senior scientist of illumination engineering at the optical solutions group at Synopsys. ‘It’s used in lots of applications, including street lighting, room lighting, and headlamps for cars.’
The first stage in designing an optical system is to assemble the requirements, which can be fairly general, or in some cases, very specific. ‘With streetlamps you don’t want − or it’s not allowed − for you to put light into certain areas,’ explained Gauvin. ‘For instance, you do not want light going into an oncoming driver’s field of view which creates glare problems, or behind street lamps situated on roadways that create unwanted light into residences and buildings. There are specific street lamp standards that say where you can and cannot put light.’
Once the requirements and regulations are known, it is important to decipher whether the proposed lighting system is feasible, according to Synopsys’ Jacobsen. In some cases, it is not: ‘A lot of times, customers come in and say: “I want a small optic that collimates the light from a rather large source.” And, sometimes, what customers are trying to do is physically not possible.’
The next step is to identify and model the LED light source that will be used in the system. ‘The most important thing about getting accuracy in an illumination system is how accurately you model the sources,’ said Dr Mark Nicholson, vice president of the Zemax Group within Radiant Zemax. ‘We use a goniometer, which photographs a light source such as an LED from all angles. That builds up the three-dimensional radiance of that source in both the near field and the far field. The measured source model can then be used directly in the software.’
Then, everything that the emitted light will interact with in the surrounding area is measured. For example, if a company is re-designing its office lighting with LEDs, the workplace is modelled by the software to analyse how the emitted light will be distributed. ‘We build models of the room with the right dimensions and we have everything in the room characterised, so we understand how they scatter light,’ explained Pfisterer. ‘I have a box here from a company with several ceiling tiles, floor tiles, and paint samples. Once we have characterisation data, we then build the model and predict the room lighting from there.’
Programmes such as TracePro from Lambda Research, LightTools from Synopsys, FRED from Photon Engineering, and OpticStudio 14 from Zemax, use ray-tracing simulations to model how light travels through an optical system and how it interacts with different materials. ‘If [the rays] hit a mirror they just reflect; if they hit something made of plastic, they refract [change their angle],’ according to Nicholson. ‘And, if there are rays of different wavelengths, they get bent by different amounts because the materials have different properties at different wavelengths.’
After the software has visualised the results of the proposed lighting system, decisions can be made as to whether the desired outcome has been achieved. If modifications are needed, this is where the design stage comes into play: a process known as optimisation. ‘If you imagine you have a source and detector, and some sort of mirror that is reflecting the light, you might want to try bending the mirror into different shapes and seeing what happens until you reach the desired goal,’ explained Nicholson. Optimisation can be carried out by hand by manually perturbing the parameters; however it is usually carried out automatically by the software itself. ‘You can tell the software what it is you want it to achieve, and then the software can go through the process of adjusting all of the parameters of the system in order to get the optical system that best meets your needs,’ he added.
Although existing software can carry out design optimisation with minimal input from the user, this process is not instantaneous. ‘The Holy Grail that everybody is looking for is the ability to do the calculations in real time,’ explained Pfisterer. ‘So, the user turns the LEDs on, in the software, and then immediately you can see all the room lit up. If you could rotate the mouse and watch the light change, it would be exciting − doing that right now would take days.’
A real-time capability would allow for faster product design and therefore a shortened time to market. ‘I think the biggest trend is getting designs out faster − time to market is important, so the faster you can get your design done the better,’ said Synopsys’ Cassarly.
To advance the software to the point where this can be realised, the processing power needs to increase to reduce the calculation time. Zemax’s Nicholson says: ‘Processing power has had a huge effect on the industry. In 20 years of selling Zemax [software] we have seen ray tracing speeds going from tens of ray surfaces per second, up to hundreds of millions of rays per second. Optical and illumination design has really been enabled by the sheer horsepower that computers bring.’ And, compared to other fields which use the freeform software, such as imaging, the power of the computer is particularly important for lighting design: ‘In illumination, it is always many millions of rays being traced in order to get highly accurate results,’ he added. ‘You need the horsepower to be able to blast as many of these rays through the system as possible.’
Software companies are starting to take advantage of computational ‘accelerators’ – graphics processing units (GPU) – whereby both GPUs and CPUs are used to increase calculation speed. However, to use GPUs effectively requires writing programs differently, so the software has to be rewritten. Nonetheless the advantages of faster speeds will prompt companies to modify their software so as to be compatible with GPUs in coming years, according to Pfisterer: ‘One of the things you’re going to see in the next couple of years is analysis software moving onto the GPUs to take advantage of their speed. The GPUs are exciting, because you can buy boards with thousands [of cores] on one board, and the idea of doing calculations on thousands of these simultaneously is tremendously exciting from an illumination standpoint.’
Massively parallel computing, where several computers are used in parallel, is another approach that can be used to increase the speed of ray-tracing calculations. However, this method is not deemed as practical, according to Pfisterer: ‘Some companies are looking at massive distributed computing, but I think the technology that will win is the GPUs because virtually everyone has them on their laptops right now, they are affordable, and they are very fast.’
A current solution that addresses the issue of speed for optimisation is a programme from Zemax, OpticStudio 14, for engineers wanting to visualise an illumination system before they commit to its full design. The software carries out estimated ray trace calculations to give designers an initial idea whether it’s worth going further with the design process. ‘We have just introduced LightningTrace, which makes a lot of illumination design interactive,’ said Nicholson. ‘So, you can sit at the computer and use slider tools to adjust system variables, and the whole illumination factor changes before your eyes − and that helps you get into the ballpark of a good design. Then, you can switch over to the conventional ray-trace, where you need to trace millions of rays and get really accurate results.’
Although processing power is a major factor, to really allow for developments in the software and make the design process faster, standardisation is key, according to Pfisterer: ‘International committees are trying to develop standards for passing characterisation data back and forth between software products. And, there is a lot of work being done so that vendors who make LEDs can create common formats that anybody’s software can take in and manipulate,’ he said. ‘That’s really the big activity − the standardisation of the data and of the measurements.’
A continuing challenge is to help engineers who are less familiar with optical design theory to design optical systems successfully. ‘One of the things we see in the industry is that there are people who do not have formal optics training, but who are asked to do optical design because the company may not have trained optical engineers,’ said Synopsys’ Jacobsen. ‘Or, we see that people who are doing the mechanical and electrical design are also charged with doing the optical design.’ Pfisterer, of Photon Engineering, agreed that this is becoming a problem within industry: ‘My concern is that you have companies who just find somebody and say “you’re our lighting engineer now” when they do not have any experience – it’s craziness.’
Though desirable for both software companies and customers alike, modelling design software will never replace a trained engineer because of the sheer complexities involved when designing an optical system, according to Pfisterer: ‘The interface for the software is something we continue to evolve − but it’s never going to take the place of someone who is educated and trained and understands how to do these things because the issues can become very complicated. In an LED, it isn’t just the optical standard that is necessary. You have also got to look at thermal problems, an electrical circuit, and packaging and mechanical design as well.’
Some software vendors are offering training or consultancy to customers in order to prevent companies from falling into this trap. ‘We do a lot of training with customers,’ said Zemax’s Nicholson. ‘We would much rather not sell a copy of Zemax to somebody that is then not going to be successful with it.’ Other companies take on full projects, which companies often opt for after realising that they haven’t got the correct personnel to carry out the specified design. ‘You’ve got more and more difficult requirements coming from customers, and if they don’t have the experience or training they are just falling flat, because they have no clue what to do,’ said Photon Engineering’s Pfisterer. ‘So they’re coming to companies like us to ask: “How do we do these things?” – that’s where the expertise comes in.’
Although the software will never take the place of a trained engineer, further intelligence in the software will help to overcome the issue of lack of training. ‘We are trying to build the expertise of a trained optical engineer into some of our systems,’ explained Synopsys’ Jacobsen. ‘From an optical software point of view, we want to make the system as easy as possible for novice users. This way, a user who isn’t trained in optics will have a much better chance of success.’
Companies are focusing on making the user interface as easy as possible. ‘There is a huge push to make sure the user interface is easy to use − so it checks to make sure if you are doing something wrong. It gives you feedback and allows people who are not experts to produce good designs,’ added Synopsys’ Cassarly. Ease of use is becoming more and more important, not just because of the shortage of trained engineers, but because freeform design is being utilised in an increasing number of applications. ‘The biggest hurdle that we have to face is the fact that we have so many different markets in the optical field,’ said Gauvin of Lambda Research. ‘And, each designer seems to have a different idea of how they want to implement these new freeform optic capabilities. So we have to create an interface and a product that addresses those issues for each one of these different markets.’
About the author
Jessica Rowbury is a technical writer for Electro Optics, Imaging & Machine Vision Europe and Laser Systems Europe.
You can contact her on email@example.com or on +44 (0) 1223 275 476.