Nowhere is the motivation to convert all light sources to light-emitting diodes (LEDs) more evident than at the local home improvement store. Even a cursory glance at the LED bulbs on the MR16 shelf is likely to overwhelm consumers and lighting professionals alike. MR16 is a standard format for halogen multifaceted reflector light bulbs — MR stands for multifaceted reflector, and 16 is the number of eighths of an inch the front is in diameter. With the staggering assortment of lamps that are currently being produced and the equally staggering claims and metrics that are being used to entice you to buy one over the other, consumers are faced with the question: How did buying a replacement for my burned-out bulb become so difficult, and why do all of these “replacements” produce significantly less light than their halogen predecessors?

Figure 1. Controls that define the Bezier reflector profile (black).
Recently, I set out to discover the answer to this question. Are the LEDs not effective for such lighting applications? Does the problem lie with the lighting designers, or is it their design tools and software? Is it simply a matter of cost that prevents designers from using LEDs that can make a true MR16 replacement? I’ll take a look at the problems that are faced when designing an LED MR16 flood replacement lamp. For this article, I’ll concentrate on designing a simple reflector and then look at a plastic optic to see if there is an optical reason why it’s hard to find good replacement lamps.

Figure 2. Beam angle as a function of control point weight and position for a fixed reflector length and slope. The design goal is highlighted.
The first problem I encountered was choosing a source. A survey of MR16 lamps from various manufacturers conducted by the Lighting Research Center found that the average lumen output for MR16 lamps on the market is 625 lumens. There are very few LEDs on the market that produce that much light, let alone have room to allow for any losses in the system. There are two choices: use more than one LED, or use a high-brightness LED package that contains many LED dies. I chose the Cree MP-L LED. This LED has 24 dies in the package that combine to produce minimum flux levels in the 800 lumen range and provides some room for losses; however, the cost is high for this choice.

Figure 3. Field angle as a function of control point position and weight for a fixed reflector length and slope. The design goal is highlighted.
After I chose the source, it was time to concentrate on the reflector. Designers have been designing reflectors for filaments for a long time, well before software became a designer’s tool. Unlike a filament lamp, however, where the light source is a distance inside the reflector, the LED sits at the reflector base. A parabolic reflector might not work as well. For my system, I used a Bezier profile revolved around the axis of the reflector. The Bezier reflector is effective because it allows me to define a curve that has a small number of controls. Since the diameter of the reflector at the base and exit aperture are well-defined, I can design the reflector based on the length, position, and weight of the Bezier control point, as well as the slope of the reflector at the exit aperture, as shown in Figure 1.

altNow, I need to decide on design metrics. How do I know when my design is complete? The flood lamp has a beam angle (full width at half maximum intensity) in the 35- to 40-degree range — this is a good starting point, but is it enough? For this design study, I looked at beam angle, field angle (full width at 10% intensity), and efficiency. Figures 2, 3, and 4 show these metrics as a function of the control point position and weight for a single reflector length and slope. For all three metrics, any of the designs that use the parameters in the upper left corners will suffice. The change in this region is very slow, which leads me to believe that for this length, finding a reflector that meets the design goals is trivial.

Figure 4. Efficiency as a function of control point position and weight for a fixed reflector length and slope. The design goal is highlighted
With optical feasibility established, I wanted to explore another design space. The aluminized reflector in the previous design peaks at roughly 84% efficiency. By in creasing the efficiency, the LED can be run at a lower power level, thus reducing the heat that needs to be dissipated by the package. To that end, I set up a single plastic optic that surrounds the LED. Part of the optic uses total internal reflection to collect the sidelight, while the rest focuses the remaining light into the beam pattern. The same design metrics are used, but the variables are considerably different and there are more of them, as shown in Figure 5.

Figure 5. Controls for the plastic lens.
Instead of looking directly at slices through the design space, I let an automatic optimization routine try to find a solution. After four iterations, the optimizer converged to give me the design shown in Figure 6. The efficiency is approximately 91%.

Figure 6. Final lens design form and intensity distribution.
These findings are both encouraging and discouraging. With this particular LED, it is possible to design a single-LED MR16 replacement lamp with a total efficiency in the 83% range with a reflector, and approximately 91% with a lens. Both of these designs easily meet the requirement that they produce a similar level of flux as the average filament MR16 lamp. Clearly, I’ve chosen to make the design easier at the expense of part cost. For optical professionals, it’s easier to justify a sharp increase in the cost of the lamp if you compare the lifetimes of the lamps, but for the average consumer, this is a tough sell. Based on these findings, the industry needs to improve the lumens-per-dollar value for their high-brightness LEDs while continuing to push the flux output into the many hundreds-of-lumens range per package.