Lifetime and reliability in high-brightness LEDs, and why SoC designers should care

One of the wonderful things about this job is the chance to sit in on papers unrelated to my beat. The subject of this posting, a paper at EDN’s Designing with LEDs seminar, is a case in point. By just sitting there for 45 minutes I heard some hard-earned lessons from a different world, but with important implications in the SoC space.

The presenter was Geof Potter, a power technologist at Texas Instruments and long-time hand in power-supply design. After a quick review of what goes into an LED lighting assembly—basically, a power supply and a bunch of LEDs—Potter started to decompose the assembly—called a luminaire, by the way—into its components and to discuss the impact of each on lifetime and reliability.

The two terms are different in Potter’s world, by the way. Lifetime refers to the length of service before the luminaire will have fallen below some determined fraction of rated light output—usually 70 percent. Reliability refers to the probability that the luminaire will require repair or replacement during its rated life. So it’s entirely possible, for example, for a luminaire to have a lifetime of ten years, but that someone will be pulling the thing out and fixing it every three weeks.

Back to the question at hand. It seems obvious that the shortest-life, least-reliable component in the luminaire would be the LED. But it turns out that’s only true if someone has been sloppy. LED lifetime decreases rapidly with temperature, and less rapidly with voltage. So run the little guys conservatively, cool them aggressively, and they will be fine. There are other issues in the LED assembly, such as aging in the encapsulant and lens, reducing transparency; or aging and loss of light output in the phosphors used as secondary emission sources in many lamps, but they also are not usually primary factors in luminaire life.

The next most obvious culprit is the fluid-filled aluminum electrolytic capacitors in the power supply. Everybody has had bad experiences with e-caps: aging, degrading electrically, and finally giving up the ghost with a distinctively bad aroma. Wrong again, Potter said. When used conservatively, extended-life e-caps actually have a working life that can exceed that of the LED assemblies. Similarly, the opto-couplers in the supply have bad reputations. But Potter said paying extra for modern, high-reliability couplers and using them appropriately pushes these devices out of the running as prime culprits too.

Of course lighting manufacturers may not choose to pay more for high-lifetime components, and assembly houses may not choose to actually use them even if the OEM is paying for them. So here is one quick lesson for the SoC world: if it’s possible to design around an external component that is subject to inappropriate cost-cutting, do so.

But if LEDs, e-caps, and opto-couplers aren’t the worst problem, what does that leave? Solder joints. You might think that the industry would have solved this problem in the last 50 years or so. But according to Potter it has reappeared in the LED lighting space. And the reasons are unfortunately relevant to many SoC designs.

The key issue is quality of soldering, Potter said. All solder joints age as a function of electrical, mechanical, and thermal-mechanical stresses. Aging accelerates greatly if the joint is compromised by poor mechanical connection, contamination, or insufficient heating. All this is well known in the electronics industry. But the lighting market tends to be highly cost-sensitive, with a supply chain heavily outsourced to low-labor-cost areas and more familiar with mechanical assembly than electronics fabrication. And luminaire designers may have little insight into just how and by whom their designs will be assembled.

Furthermore, luminaires can live in high-stress environments. Think of an automobile headlamp eyebrow, that cute little string of white LEDs that makes this year’s luxury sedans distinctive in the dark. The front of the car may be moving at 100 km/hr into a -10C evening when the lamps come on, quickly self-heating to +70C, all the while enduring road shocks and engine vibrations. Maybe Bosch is comfortable assembling a fixture for this environment, but is your low bidder in Southeast Asia? Potter’s suggestions are to ensure manufacturing quality and minimize thermal and mechanical stress. But since these may be largely beyond the control of the design team, he double-underlined another point: use the smallest practical number of solder joints.

That’s where SoCs come in. In the SoC world we also face growing cost pressures, uncertainties about our customer’s supply chains, and increasingly, the need for our designs to operate in hostile environments. Networking gear hanging on a tower, automotive sensors mounted in a bumper, and even a handset sitting on the dashboard of a car parked in Phoenix are all facing harsh environments. Potter’s data suggest that by doing whatever is possible in the architecture and implementation of the system SoC to minimize the number of solder joints in the end-system, we can have a dramatic impact on system reliability, and on total cost of ownership. Ideas such as the highest feasible level of functional integration, absorbing or designing-away external passive components, and minimizing both the number of pins on the SoC and the number of lines that must pass through connectors are not just good practice. They are becoming differential advantages than can sell—or doom—an entire SoC family.