AC LEDs

Trick question: What are the characteristics of a driver that would work best for a "driverless" AC LED?

There's been a lot of discussion recently, including in this forum, about what is or isn't an "AC LED," and whether they are useful or not. In this blog, I instead am going to take a look at AC LEDs from a driver standpoint, and make some suggestions on how they could be made more useable.

When AC LEDs first came out, it was pretty clear that the name described them. A string of LEDs was placed across the AC line from hot to neutral, with a second string in anti-parallel from neutral to hot. The forward voltage of the string was tailored to be very approximately equal to the line voltage, and so the LEDs would conduct the right amount of current averaged over a line cycle at nominal line voltage to produce the desired light output. (The string need not be a set of packaged LEDs in series; it could be a single package with a set of junctions whose forward voltage is suitably high.)

However, this configuration is notably inefficient of die area: Only half of the LEDs are conducting at any given time, and so the die has to be twice as big as it would be if all of the LEDs were on simultaneously. It appears that this configuration, which used to be sold by Seoul Semiconductor, is no longer available.

It doesn't take a lot of thinking to realize that a straightforward fix for a die area problem is to add a rectifier bridge in front of the LEDs. The bridge changes the line voltage to a single polarity, which can then be applied to a single string of LEDs, which are then operated during both polarities of the line cycle. This seems a very attractive configuration: No power supply is required, just a bridge. There's no EMI, no EMI filter, no ICs or inductors, power factor and THD are (at least potentially) within limits. With no power supply, reliability is improved, and best of all, the "driver" is cheap.

Now, should this be called an AC LED? After all, we distinguish an AC capacitor from a normal capacitor by whether or not it has both positive and negative voltage applied to it, which is to say, whether it is in front of or behind a bridge. My judgment call: If it doesn't involve high-frequency switching, I'm going to call it an AC LED.

So now we have a string of LEDs with up to 168V across it (120Vrms times root 2). What happens when the line voltage is 10 percent high (132Vrms)? Clearly, the LEDs conduct more current. What is wanted is that the forward voltage should rise pretty steeply with the current, so that the increased voltage doesn't cause the LEDs to conduct excessive current and fail. Here is where there is room for improvement from the LED dice manufacturers. A lot of work has gone into reducing the forward voltage of LEDs since this helps efficacy; and this has also resulted in a much steeper I-V curve, so that the forward voltage of the LEDs is much the same at low and high currents.

But for AC LEDs, we want exactly the opposite. We want a big change in voltage to produce only a small change in current. That way, the LEDs will conduct approximately the same current at 120V as they do at line peak at 168V. And they won't blow up at 185V. Of course, the powerwill increase at higher voltage, and thus so will the light output. But if the current is constant, the power will increase linearly with the voltage (power = voltage times current, voltage increases, current is constant), rather than quadratically (both current and voltage increase).

We want less like the left figure, and more like the right. (Source: LG Innotek)

With a device like this, AC LEDs could finally become a serious player in the world of SSL. And LED lighting would become more reliable and cheaper.