2010 was the year the warm white LED stole Christmas. Suddenly they were everywhere in the ornaments. Just the year before, all you could get was cool white, ice cold lights -- in fact, sellers even tried to make the best of it by calling them “ice lights.” It was a banner year for LED, capped by the recent Cree XM-L device, which offers almost 140 lumens/watt in a cool white, and begun by Philips and the $10 million L-Prize.  But this too is a good time for taking stock of issues and challenges.

Let’s pause to savor the triumph. The XM-L reconfirms Cree as a best-in-class industry leader, along with Philips in a situation that’s arguably analogous to Sunpower or First Solar. While only cool white is initially offered, the literature projects that warm white lights will be close behind, continuing the previous XP-G family trend of narrowing the gulf between cools and warms. Unnamed sources whisper future bins up to a startling 170 Lu/W.

Photons by the pound! (They’re only massless below the speed of light.)                                                                                                     

At 140Lu/W, the XM-L is at about 60% of the endpoint of Haitz’s law. We are not only way beyond fluorescents by now (and bye-bye mercury), but also well beyond the 100-ish Lu/W typical of other high efficiency whites such as arc and discharge lights (orange low-pressure sodium, at 200 Lu/W, is a special cheater’s case). “Not as bright” criticisms of LED tubes will soon be history.

I built an XP-G array (along with Teknowlogica and Integrated Controls Inc.) that could scorch your shirt and could only be watched through welding glass -- in fact, it almost melted the glass. At the low end, I put an XP-G on a controllable current source, blacked out the room, and cranked the dial down to the minimum threshold I could see.  It was 80 NANOAMPS -- 80 billionths of an ampere. And my eyes aren’t 25 years old anymore.

Since the millennium, LED efficiencies have risen abut 400%; cost and output size have improved much more. In that time,solarcell efficiency has risen 100%, with great fanfare. The hybrid-electric car, maybe 30% over simple ICE. The commercial airliner (Boeing 787 Dreamliner), 20%. The combined-cycle power plant (GE arduously steam-cooling the first stage blades), only 15%.

Rising efficiencies deal a double whammy to the old LED bugbear of heat. Incandescents can radiate their heat as infrared; LEDs have less heat but you have to extract it, a la Pentium, because the light is cold. Here’s the upside double whammy: As efficiency goes up, not only are you putting less power in, but more of it goes as photons and less as heat.  Examples help. Start with the limiting case:  A 100% efficient LED would always have zero cooling load, although the power input would not be zero. 

Prediction: The big fin ring behind the LED lamp will vanish. And it was this (plus cost), that had kept the L prize unclaimed.

We are onto something special.  We can clobber the lighting load. Of course, the pundits long since invented a Moore’s law analogy, Haitz’s law.  So often are such analogies rather forced, and I’ve actually found that it’s more of a lurch and-stall.

So what are the issues facing LED now? Here is my individual, but hopefully useful, collection:

1)      There isn’t much more efficiency to gain. From 60% of the Haitz endpoint, there is probably not another doubling. Somewhere between 180 and 200 Lu/W in a white, we hit physics limits.

2)   Cost, cost, cost. The L prize is finally claimed. But LED bulbs cost $20 or more; long-tube fluorescent replacements can cost up in three digits. How long to get the price down? Hopeful estimates say $10 by end 2011, but I am skeptical. CFL history may provide a model. Electronic CFLs arrived circa 1990 at $30. Post inflation, this is similar to the current LED bulb price. After 20 years, CFLs are down to $4.  However, much of the reduction was electronics deflation, plus offshoring—both already factored into the LED bulb price.  At first sight, economies will have to come from the LED itself.

Hope springs eternal. Minimalists such as Acriche (a Seoul semiconductor offshoot) feed the LEDs almost raw line AC, abandoning the control electronics. I was skeptical, especially regarding how the part would handle AC line variation (diodes have abrupt voltage thresholds).  However, when bench tested, the Acriche part exceeded hopes. An onboard thermistor heats up at higher line voltages, restricting the current. Acriche also has a better power factor than most CFLs.  Phase dimming is explicitly unapproved, but worked surprisingly well—better than many first generation phase-dimmer LED lamps (more later).

3)  Patience, patience, patience—likely to be needed. Engineering is incremental; little irritations will gate a new technology for years. It has taken the CFL twenty years to get where we are (and the circuit boards still overheat when mounted tube down).

Early CFLs had color problems. One maker’s Perfect Solution came out slightly green, another’s was purplish. You could not mix them. Replacements in an array were a nightmare. This mirrors today’s hodgepodge of LED color and output bins. The new warm white strings on our Christmas tree didn’t play together. The LEDs were tightly matched for color temperature within each string, but not between strings. We had three strings, and got some degree of visual discord for the rest of the Holidays. Fluorescent mismatches had been worse; hopefully, LED makers will tighten their batch spreads (a la CPU clock speeds). However, this may take time and investment.

Touchy users: A remodeling contractor suggested LEDs for the ceiling can-mount dimmers. The cost was nothing compared to the carpentry, and the LEDs would run much cooler. The homeowner agreed—but tore them out soon after commissioning. He expected dimming to color shift towards the orange, based on incandescents. But LEDs stay about—about—the same light color. Sours the romance. Blending in amber LEDs may run into patent trolls in the color-mixing area (as well as the need for special functions in the driver chips). Leading to…..

4) Technology fracturing and patent trolls. Technology fracturing is nothing new. During the invention of TV, RCA had patented the best picture tube, and Philo Farnsworth the best camera tube.  Farnsworth refused to sell out.

Today, top dogs such as Cree have world class efficiency and a plain LED chip. We have others with innovative, integrated topological architectures—and varying but generally lower efficiency.  Examples are LEDEngine, Bridgelux, Luminus (large dice limitations), and Acriche.

Patent trolls are sometime LED pests.  Mr. Harvey’s patent got Apple sued for putting an LED in a power adapter.  I encounter cases privately.  A minor-legend video developer abandoned a better-mousetrap LED camera light, beset by the trolls. The litigators showed no interest in developing the product in question, and far as I know the world lives to this day without it.  The threats were based on sweeping early patents.

Hopefully, industry collegium would apportion for everyone’s benefit (as with the IC patent dispute). There has been licensing but not what one would desire.

5)  Dimming.  The LED itself dims more gracefully than almost any light known (even this statement is subject to messy wrinkles). Dimming is critical to early residential LED deployment, dodging around CFL price.  But LED drivers circuits are not naturally compatible with the billions of installed phase dimmers, from many makers.  What you do is cheat the circuit to make it phase dim. Every lighting designer and his dog just released their first LED phase dimmer, and everyone claims his own really works. But it’s tricky to make them work well, at EVERY dimming level, with EVERY dimmer. There’s always a flicker point somewhere.  I have spent weary hours tuning driver circuits. I fear a backlash if consumers have trouble, although smart CPU control offers hope.

There are new standards for LED dimming, sometimes chez power standards. The winners are yet uncertain.  Analog or digital, AC or DC, high- or low-voltage, high- or low-frequency? Phillips created something sensible with DALI. Friends have a startup, QML, for DALI commissioning tests, maybe a positive sign. But there is still a hurdle from here to there.  And the optimal systems for warehouse, naval ship, and bedroom set are different.

6)  Finally, LED should avoid the CFL era subsidy pitfalls. Heard of households that buy subsidized bulbs and install them in a closet? There is worse. The Marina Motel, near the Golden Gate Bridge, put 100 CFLs in a winding outdoor courtyard. A glance showed these were indoor-only lamps, with circuit boards visible through open heat slots. The owners said the bulbs had been put in by bundling subcontractor, who took the PG&E subsidy and split (apparently no audit). When the rains began, the explosions began. One brave bulb was still running after three winters.

The above challenges are solvable. The weekly miracles of LED, has been like a time warp back to Silicon Valley 1999. With each new project, you throw out your old LEDs and driver chips for something newer.  Working out all the applications will be an ongoing process.  Consider the potential for increasing world food supplies (or at least world flower supplies), by stretching the growing day with photosynthesis-tuned LED greenhouse lights.  A startup on Treasure Island, Lumigrow, is attempting just that (disclosure, my customer). More fun will follow.

I have no idea whether Cree or Philips have yet produced a pound of photons. Maybe the marketing VPs can announce it and ring a gong on the day. And I hope a warm white holiday was had by all!

Doug Widney is a San Francisco-based energy consultant with a diverse electric Greentech practice.  Besides strategic advising, he remains an active circuit designer. He can be reached at dougwidney@yahoo.com.