Back about 10 or 12 years ago I had imagined that one day LEDs would actually be useful for more than indicator lights, but I thought that day might be decades in the future. Now for years I had used LEDs as indicators in electronic projects. I did notice that the newest ones were much brighter than the ones from years past. Sunlight viewable was a big thing back then.
Anyway, with that thought in mind, I set about making my first LED application where the LEDs would actually light something up. This was HO train passenger car lighting. I had some yellow-green LEDs which were used for taximeter backlights. By today's standards they were very dim, about 10 mcd, although with a wide 100° or so viewing angle. They probably put out at best 1/10 of a lumen (didn't know how to do the calculations back then). I made a circuit to charge a small NiCad battery which powered the LEDs when the train was stopped or moving slowly. At normal voltages track power charged the battery and lit the LEDs. The actual lighting level was rather dim, even using 7 LEDs driven at 15 mA. Sure, it was noticeably lit in a dark room, but that was about it. As for color, well, that left something to be desired. In truth the yellow-green color was an attempt to imitate fluorescent lighting, but it was even greener than the crappiest cool white tubes imagineable. I lit a few cars and decided that the results didn't justify upgrading my entire fleet. In case anyone is wondering about why I did this at all instead of using incandescents, the reasons were many. Incandescents in models create too much heat, are difficult to change when they burn out, don't give much light for the power (the grain of wheat bulbs probably don't do much better than a few lumens per watt), and only give one kind of light (crappy yellow-orange). I had given up long ago using them to light trains. The less than satisfactory results and maintenance gave me a bad headache.
After my feeble attempt at lighting trains I did use LEDs for one other lighting project. I lit the interior of my thermoelectric temperature chamber with 40 surface mount yellow-green LEDs. Power consumption was nearly 2 watts and light output was lousy at best. Still, it was better than nothing. I remember at the time wishing that LEDs were brighter, and also that they came in white similar to fluorescent lights. Of course, I had no idea how someone would actually make a white LED since LEDs are by definition narrow-band emitters, but I could still wish for one, couldn't I?
I shelved my LED lighting projects and pretty much forgot about LEDs as anything other than indicators until the late 1990s. I remember by then that blue LEDs had finally come down in price from $30 each a few years earlier to about $3. I purchased a few to play with. Interesting but very dim. A year later I purchased my first white LED. Again, interesting, and actually about as bright as a grain-of-wheat bulb. Nice bluish-white color, too. I thought of course of the train lighting project again but at $3.50 a pop and requiring about 5 or 6 per coach it would just be beyond my means. I figured within a few years like everything else electronic mass production would bring down the price. Over the next few years I purchased a few more white and blue LED samples, but the prices were just too high to do anything useful with them.
This brings me to mid 2003. My brother had asked me to find some white and red LEDs for lighting die-cast cars. I was all set to order some whites from an online store at $1.09 each in 100s (a good price I thought) and on a whim decided to check eBay. I was surprised to discover LEDs of similar brightness going for 1/3 that price. I took a chance and ordered some. I tried a few vendors and was fairly happy with the price and brightness. Finally, I could get white LEDs in mass quantities without breaking my budget! I never would have dreamed of this only a few years back. And I could get the other colors as well, all very bright compared to what I was used to. I immediately made an LED bikelight, redid the lighting in my temperature chamber, tested the new lights in my trains. I combed the Internet for LED-related information. I learned of the planned advances for LED lighting in the decades ahead. I was surprised LEDs had progressed this much in so short a time. Later in 2003 I bought a few Luxeons to play with. Amazing I thought, and although still pricey at $10 each I said these will be great once they come down to $3 or so (guess what-that's what they cost now!).
My reawakening to LEDs in mid-2003 wasn't without drawbacks. I noticed another problem. LED brightness was increasing, seemingly overnight. Today's hottest LEDs were yesterday's old news. 8000 mcd was great in mid-2003. A year later 16000 mcd was tops. I decided to not bother stocking up on LEDs, but would rather just get 100 every now and then for evaluation purposes. I also shelved the train lighting project again for two reasons. First of all, although LEDs made huge leaps and the whites were great, I suddenly became much pickier. Remember that 12 years ago I had actually considered sickly yellow-green acceptable. Now I was not only looking for white, but I was aiming to match the 4100K to 5000K tubes commonly used as closely as possible. Most of the available white LEDs were just too blue. At first I mixed white and amber. That did the trick but I figured why not just wait until I could find an LED with better color? Late last year I tested BestHongKong's UWLC whites. Bingo! This was what I was after.
That brings me to the second problem. I didn't want to light my fleet with LEDs only to have ones two or three times as efficient come out in a few years. For even lighting I needed four LEDs per car. I could get satisfactory light from the ~14000mcd UWLC series driving them at about 10 mA (in series via a step-up converter). However, in a few years I might be able to get the same light driving at only 2 or 3 mA. Besides longer life, the lower power requirements made it easier to power the LEDs during times of low or zero track voltage. If I could get the power requirements down enough, then instead of a small NiCad battery I could just use a supercap. Also, supercaps would undoubtably increase in capacity over the next few years anyway, giving me all the more reason to play the waiting game. Anyway, that's where LEDs have brought me on this project to date. I don't want to commit to relighting the fleet until LED efficiency starts to level off. So far this shows no signs of happening for at least another few years. My best guess is that once cheap eBay LEDs are at 150 lm/W there's not much point in waiting.
The moral of this long story? Don't wish too hard for something. You just may get it. Today's LED brightness and prices are something I couldn't have dreamed of only a few short years ago. However, with these advances have come even higher expectations for the future. It's hard to start a project when you just know parts twice as good are practically around the corner.
What's the next big thing? After nearly 50 years with little efficiency advances thermoelectrics seem poised to follow the same curve LEDs are. I recently read about some new thermoelectric materials 2 or 3 times better than the long used Bismuth Telluride. Once these make it out of the lab watch out. The days of compressor-based cooling will be numbered just as the days of the incancdescent light are now drawing to a close.