Light-emitting diode (LED) bulbs offer the potential of up to 50,000 hours of operating lifetime, or nearly 25 years of typical usage. This is a 50x improvement over incandescent-equivalent technology. With demand for LED lighting growing rapidly, a key issue that could hold the industry back is if solid-state lighting (SSL) bulbs do not achieve the promise of long operating life. The obvious design considerations for solid-state lighting are efficiency and cost. But, thermal management is just as vital as any other design criteria, because too much heat can impact operating life, not to mention bulb safety. The energy savings of solid-state lighting gains the most over the full operating lifetime potential of the actual luminaire. While the LEDs offer the promise of this long lifetime, additional components required in the LED driver circuit can dramatically decrease luminaire operating life if intelligent thermal management is not implemented.
It is easy to overlook some important thermal aspects of LED design -- issues that can result in potentially catastrophic luminaire failures. An LED bulb can be used in an enclosed lighting fixture or a fixture that is open to normal air circulation. The thermal conditions in these two cases are radically different, but the bulb in both instances has the same physical and electrical design. The temperature inside a closed lighting fixture can rise quickly to levels above 60°C, which subsequently causes the temperature inside the light bulb to exceed 90°C. In open-air fixtures, the temperature inside the bulb itself can be as much as 30°C lower than its closed fixture counterpart.
An LED-based bulb with no thermal protection whatsoever used under conditions where there is near zero air flow could result in a thermal runaway condition. Figure 1 shows the construction of a typical A19 retrofit LED bulb and the confined space in which the driver circuit needs to operate. This tight space exacerbates the temperature issues. Early examples of poorly designed LED luminaires include devices that failed after 1,000 hours, just like the incandescent bulbs they were intended to replace, and even a design where the bulb itself experienced thermal runaway, melting the casing and posing a potential fire risk. The end result was a costly recall of a large number of bulbs. These early models did not take into account the importance of thermal design on the overall quality of the LED bulb. A simple solution is to integrate a basic thermal shutdown circuit, something that is already very common in IC technology.
Click on image to enlarge.
Figure 1. Typical construction of an LED-based solid-state lighting luminaire.
Most LED drivers used in solid-state lighting contain a straightforward thermal protection circuit. Most power-management ICs employ a simple thermal shutdown function where the output of the regulator shuts down to protect itself when a maximum temperature is reached. This does protect the main IC, but when applied to an LED lighting circuit, it presents two critical problems. First, the output of the LED driver shuts down completely, eliminating the light. The output doesn't turn on again until the thermal event clears and the temperature of the IC drops below the hysteresis point in the thermal shutdown circuit.
I agree that replacing T8 flourescent bulbs with LEDs is a formidable challenge, especially considering the lumens/W rating of a typical 32W flourescent bulb. The cost to implement an LED solution that has a comparable lumens/W rating is much higher than that of a standard flourescent bulb. That is why, as of today, there is little support material created for this application as the LEDs have yet to reach the level of efficiency to make a solution that makes financial sense. But, LED technology is improving very fast as is the driver technology, and eventually we'll be able to create something that makes sense. Check back from time to time on our website for additional information.
Take a look at this brief article from 2010. http://ledsmagazine.com/features/7/6/6
I am wondering why there is a need to have the secondary-side electrolytic (or any capacitor) at all? If the flyback circuit is operating in a constant-power mode at tens or hundreds of KHz, is it not OK to allow the LEDs to "flash" at that rate? As long as the average LED current is constant, doesn't our eye integrate the light and ignore high-frequency flicker?
Hi Hubertus Notohamiprodjo,
Please can you consider writing article on direct replacement LED fluorescent without ballast transformer T8 versions. I am sure there many design challenges; especially, thermal design and safe operations which many of us would like to know more. I know there are many building owners and facility managers whom eager switch but there are concerns for security lighting uses and reliability. I liked you last article on LED lamps very much and thought you extend your effort to direct replacement LED fluorescent without ballast transformer T8 versions.
Merv Perry MBA MSc CISSP
Thanks for your nice comment on the article and for posting the link to your detailed article about electrolytic capacitors and their use in solid state lighting applications. The intention of this article is to bring to light the fragility of the electrolytic capacitors, the potential impact on the operating life of an LED-based luminaire due to this fragility and ways where active solutions can be implemented to maximize that operating life. Your article nicely covers the end-of-life characteristics of the electrolytic capacitors and it is important for engineers to understand the impact of the reduced capacitance rating as it reaches its end of life. This part of electrolytic capacitor characteristics was not covered in my article as it would have made for a much longer and more extensive article than was originally intended. Luckily, the iW3626 LED driver mentioned in this article has a patented power factor correction circuit that allows the designer to customize the design for high power factor, low output ripple current or a balance of the two, unique in single-stage, low cost off-line LED drivers. LED bulbs using this product can meet power factor and output ripple current requirements over the full operating life of the LED luminaire while protecting all components in the system thermally.
With regards to the exploded-view of the A19 type bulb in figure 1, take into consideration that this is a graphical rendering of an A19 bulb and not an actual photo, hence the reason why it may not look exactly like any A19 bulb that you are familiar with.
Nice article pointing out the key, and often overlooked, issue with electrolytics in LED lighting.
The lightbulb shown is one I am very familiar with. It is not an A19 but rather is an R20 - the smallest of the "CAN" lighting bulbs used for down-lighting. Note that the enclosure is a heat sink with a fair amount of surface area due to the fins. This was a substntial improvement over earlier LED light bulbs that had no fins thus causing the case temperature to be significantly higher.
The LEDs themselves, the largest source of heat, are mounted on the single metal piece that is the heat sink. This arrangement dissipates the heat pretty effectivey to the outside world.
The circuit in this particular unit I am familiar with - or at least the version that was in prouction a couple years ago. It used no electolytics. Also only components that were rated to 125C or higher were used.
I have a few of these in my house in ceiling cans - the worst thermal environment - and none of them have failed in over 3 years, including one that is on all the time - that is about 30,000 hours so far.
You have hit on an important point about electrolytics - the 2x lifetime for 10C.
however there are some more subtlties to the 10C vs. 2x lifetime relationship that are covered in my recent artilce in EETimes sister publication, EDN; http://www.edn.com/design/analog/4411475/Ensure-long-lifetimes-from-electrolytic-capacitors--A-case-study-in-LED-light-bulbs
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.