What’s all this stuff about capacitive power supplies for LEDs?

A while back a viewer of our 60-W LED bulb teardowns posted a comment about the complexity of the power supplies found in the bases of these bulbs. We talked about some of the reasons these bulbs contain switching supplies in a previous article. But our latest post elicited this comment from another reader:

Well, some LED bulbs do not have a switchmode in them. They just have a simple capacitive dropper and a rectifier. If you have a much higher voltage than you need to drive the LED and a high value resistor, the output to the LED becomes nearly constant current. If we are using AC we can replace that resistor with the reactance of a capacitor and achieve a similar result with very little energy wasted across the cap. So why bother with the complexity of a switchmode, when it’s usually the first thing to fail in an LED bulb?

Once again adopting the style of the late great Bob Pease, we’ll try to explain some of the reasons you could indeed use a capacitive supply for an LED bulb, but you might not want to.

The idea of a capacitive supply is well known enough to have its own page on Wikipedia. If you refer to that page you will find that the basic idea is indeed to use capacitive reactance as a means of lowering the voltage from the mains to something lower, without having to use a transformer or some kind of semiconductor switch to do it.

The Wikipedia page also contains a photo of an LED bulb with what to all appearances seems to be a capacitive supply. There is no reference given as to who’s LED bulb this is, but the text mentions a 1.2 uF capacitor providing 90 mA powering 48 white LEDs, apparently divided into four branches of 12 that each use 20 mA.

There is not exactly a plethora of references for real capacitive supplies on the Wiki page – The one main reference when we looked was in German. One can, however, find an app note from Microchip that discusses basic capacitive and resistive supplies. But the application Microchip seemed to have in mind in its note was powering low-power microcontroller chips, not LEDs.

So we can use the capacitive supply circuit that Microchip gives in its app note as an example of what a capacitive supply for an illumination-grade LED might look like. Most of the circuit calculations center on figuring out the value of the capacitor C1 providing the reactive resistance for “dropping.” For the illumination grade LED, we picked a Luxeon Z high-brightness device from Lumileds. This LED wants to see something over 500 mA of drive current and it will have a forward voltage around 2.8 V at that current.

If you run the numbers for the size of C1 you’ll need to get 400 mA, you come up with a capacitor in the 13 µF range having a voltage rating of at least 250 V. To squeeze a cap of this value into the base of an LED bulb requires that use of capacitor technology that is space efficient. And that likely means using an aluminum electrolytic capacitor. The problem is that though illumination-grade LEDs are more efficient than CFLs or incandescent lamps, they still give off significant heat — Junction temperatures exceeding 90°C are often the norm.

Unfortunately, the rated lifetime for electrolytic capacitors tends to lie in the 20,000-hour range (about three years), and it goes down when the cap sees extended high ambient temperatures. For comparison, the rated lifetime of the LED bulb itself is about 45,000 hours using the Energy Star lumen maintenance criteria for bulb end-of-life. The rated life of the switching-supply controller chip is well over 100,000 hours, as is the rated life of other semiconductors in the LED supply.

The reason high temperatures shorten electrolytic cap life is that the dielectric material in the cap evaporates over time. The evaporation rate accelerates at higher temperatures. Effective series resistance declines with rising temperature, and capacitance rises a few percent.
High-temperature failure modes can be a bit hard to predict, however. You can wind up with a shorted cap, with catastrophic consequences, or a high resistance that just prevents that LED from turning on. But the main point of note is that the electrolytic cap is the most likely component to fail in an LED power supply.

Another point to note is that the dropping cap is the component that sets the current through the LED. So changes in its value, as can arise over time from heat effects, can change the LED light output in terms of both lumen output and sometimes in terms of color temperature.

Finally, the behavior of capacitive droppers tends to be suboptimal when there is a light dimmer in the wall switch. The chopped ac waveform from the triac dimmer contains a variety of harmonics. In that capacitive reactance varies depending on frequency, the dropper cap will present a different reactance to the different harmonic frequencies involved.

All in all, the drawbacks of the dropper cap approach make it clear why LED bulb makers typically elect to go with a power supply design that is more elegant.