Of course, we're not there yet. Solid state lighting is just reaching a point where it can cost effectively fill some key niches in the general lighting environment, almost exclusively in near 24x7 applications such as hotel lobbies and elevators, and outside where the higher efficacy cool white devices can really "shine". So what are all those products doing in all those booths? Well, frightening the heck out of the customers is one possibility. Lighting specifiers, and other sane mortals, have to find themselves asking, "With all these LED products out here, am I crazy for not using them? And do any of them really work because it's obvious there's plenty of junk as part of the show and tell?" Common enough questions, and just in case you didn't find the answers, no, you're not crazy and yes, some of them are really work. So far, the specifiers aren't being widely fooled by the pretend products, which is a good thing. In addition, the first key SSL specifications have arrived on the scene and customers can now start asking for the spec data to help them separate the wheat from the chaff (more on that next time). With nice rational fears in play, we're also seeing a few irrational ones as well.

In a pre-exhibition session, one lighting designer shared the technology lessons he had gained from his experiences, which included outfitting at least one skyscraper in the Middle East with a color changing exterior LED lighting scheme. He had implemented a number of creative solutions in response to the hot desert environment, and experienced a lot of frustration with installation issues, such as stretched interconnects that led to shorts and failures, but it had led him on a bit of quest to understand the full variety of challenges that were lurking in the LED technology. His investigations led him all the way down to the epitaxial process at the pre-LED wafer level, and to what he perceived as the "Achilles' heel" of the technology. You grow crystals on a substrate and, he asked, "How many of us have had those crystal kits or grown sugar crystals on a string? We all know how random that is. That's the same basic thing that happens here, so you will never get very predictable results... that's why they do something called color binning." The audience understood him to then extend that premise to a firm belief that once an LED chip is characterized at the chip level (during a blink-of-an-eye duration test, in which it never heats up beyond a comfortable room temperature), that once it is installed into some type of package and allowed to heat up to a more normal electronic chip kind of temperature, that the color could shift in a random direction, and to a random extent. Hence, he concluded, it's unlikely we would ever get predictable side-by-side color samples, especially from white LEDs.

Could this be true? New mission: Find the people that have the answers to the inconsistency concern. At the Osram Sylvania booth, my host had procured the right person to discuss color shifting and staunchly refused to give him up, even during an attempt to get him over to the DOE group that was on tour. (Sorry Jim, I got him first... which is lucky as we needed a column. We'll cover the Round 5 CALiPER results next time). The answer to the question was what I suspected. Yes, they do shift substantially under temperature. While their binning has been narrowed down to 3 MacAdam ellipses (basically undetectable to the eye), the color shift at temperature can be as much as 7 MacAdam ellipses, BUT it does it predictably and they have the curves to back it up. Discussions elsewhere confirmed the same type of thing, and with the addition of a great point from Philips Lumileds. That paraphrase is that when customers ask about "the heat problem" or "the binning problem", they're only problems if you aren't engineering your product. We don't call it a "problem" that you need to put a heat sink and fan on a Pentium processor chip. We don't spend much time concerned about Intel's inconsistent manufacturing processes, which are all built on variable crystal structures. They call their binning fancy names like "2.8 GHz Pentium Duo" and "3.0 GHz Pentium Duo", and sell the faster bin for a higher price and the slower bin for less (how sneaky is that!). It's kind of how things are handled with semiconductors, and why there is still some confusion. We're seeing the tremors that naturally come when technology areas gently collide -- semiconductors meet lighting and a new knowledge base is required to bridge the gap.