Taking the Smoke and Mirrors Out of AC LEDS, Part 1

We explain what Seoul Semiconductor has wrought with its AC LED technology, and how useful it is for applications such as high-bay lighting.

In this two part series, we'll address the following questions:

1. What ways are there to power an LED (a DC device) from 120VAC?


2. Why choose one way over the other? functional pros and cons


3. What's the basic idea behind the Acriche AC LED?


4. What's different about original Acriche?


5. What is Seoul's new and improved AC-LED approach using a "floating" IC driver?


6. Can we make a cheap, energy-efficient, 200-600 Watt commercial high-bay luminaire using AC LED technology and eliminating the power supply completely?

Part 1 will address the first four questions, and in Part 2 we'll deal with the final two tomorrow.

Figure 1

No. 1) How many ways?

Since LEDs went mainstream in the 1970s (with 99.9% of them rated no more than 0.1 Watt for the next 25 years), many have known how to operate them directly from 120VAC in a relatively efficient way. Figure 1 shows two inverse-parallel strings of a large number of LEDs in series with a power resistor. Figure 2 shows only a single string but accompanied by a diode-bridge rectifier.

Figure 3 achieves a similar result but with a single (non-polarized) capacitor. Notice it can operate with only a single LED.

Figure 2

The idea in Figures 1 and 2 is that each LED (if red) has a typical voltage drop of about 2 Volts (higher for other colors) so you string enough of them to somewhere between 100-120V and then put in a series "ballast" resistor, R-1 in Figure 1, to limit the final current. It is a sort of trial/error process to get the right brightness without blowing out LEDs or requiring an overly dissipative power resistor. Simplistic, but it absolutely works. Obviously, there are many nuances. "Why 120V when peak AC sine wave is 120 X 1.4 or 165V?" you may ask. Never mind.

Figure 3

These circuits have for decades been a simple way to create non-white, "never-burn-out," 1-2 watt, plug-in night lights.

No. 2) Pros and cons

It is silly (if we ignore some obscure and virtually irrelevant possible benefits) to use the back-to-back strings of Figure 1, since one string is always off. The single string of Figure 2 lets the same string operate each half-cycle and performs the identical function of Figure 1 with half as many LEDs. In the simple circuit of Figure 3, the capacitor C-1 has what is called "capacitive reactance" and, in the presence of 60Hz AC, acts just like the current-limiting power resistor of Figures 1 and 2, except that the capacitor "magically" dissipates almost no power. White-light night lights, using almost this exact technique, are for sale at Home Depot and elsewhere as you read this.

Figure 3 is very interesting in that it can use just a single LED, not needing large number of series LEDs or heat-creating power resistor to limit current. It can use one of today's single 1-2 Watt high-efficacy white LEDs and at only a Watt can function as a rather bright nightlight. But the capacitor in such a circuit begins to do bad things above a couple of watts. At turn-on, the capacitor acts almost like a short circuit for a millisecond, and if turn-on happens to occur at sine-wave peak, arcing can occur in a switch because of high short-duration current spikes similar to a 300W incandescent bulb. The power factor of such a capacitor-regulated light is terrible, but at such low power, this is of no consequence.

These "no power supply" approaches have power usage and brightness levels which vary all over the place depending on high or low line voltage. If the voltage gets much under 120V, brightness can drop off dramatically because LED brightness is not proportional to line voltage. Or if the line voltage goes high, as it often does in many parts of the US, it can overheat and damage the ballast resistor.

A real problem for circuit of Figure 1 or 2 is that, unlike for an incandescent lamp -- or any CFL/LED lamp with a driver / power supply in it -- absolutely nothing happens each half cycle of the sine wave until the sine wave reaches an amplitude equal to the sum of all the LED turn-on Vf levels (about 2.2V).

Figure 4

No. 3) The Acriche AC LED's basic idea

See Figure 4 for how the current is flowing as the line voltage goes up and down at 60Hz. With a basic AC LED (like the original Acriche), rectified (unidirectional) current only flows in that period when AC voltage is greater than about 70-75V, resulting in the distorted current waveform shown. That distortion reduces power factor but, more importantly, sharply increases the total harmonic distortion (THD). At a Watt or so, this is unimportant. Over the last several years, there has been an increased awareness of what bad THD can do, especially where higher-power commercial lighting is involved. Consequently, circuits such as those in Figures 1, 2, or 3 are a no-no in mainstream large-building commercial lighting, where, unlike perhaps in residential lighting, power distribution anomalies can be a very big deal.

No. 4) Where does the Acriche AC LED fit in?

From day one, the Seoul Semiconductor Acriche AC LED approach, and all its many patent application variations, have been based on Figures 1 or 2. The only difference is that instead of connecting a string of physically separate LEDs, they interconnect a string of LEDs which has been made as a quasi-monolithic IC. Their first patent applications were filed over eight years ago, but their US Patent application 20,120,305,951 is recent and is a more clearly written representation of all their efforts. They ask you, the customer, to add the bridge rectifier and current-limiting resistor.

No. 5) The original Acriche process

In a conventional monolithic IC, all of the dozens, thousands, or millions of diodes and transistors are separated by a special back-biased PN junction. Hence, 99% of all ICs are called "junction isolated" ICs. Such ICs are made as a single silicon element and there are no interconnecting wires among any diodes or transistors, only wire connections to the outside world.

In the early days of military electronics (and even today) there were "hybrid" integrated circuits in which many tiny chips were put on a substrate and hundreds of ultrasonically welded/bonded, .001 gold wires then interconnected to them -- a tedious and expensive process

Seoul's process uses fewer and simpler chips and their process borrows a bit from both the traditional "hybrid" assembly process and one of the simpler initial stages of the "monolithic" process.

When LEDs (or any diodes or transistors) are made, they are tested while still in wafer form and then are all physically separated into chips.

Seoul, with a technique sometimes used by some other non-LED semiconductor companies, separates groups of them (where a "group" has been separated from the wafer but some chips still are not separated from one another) so that they can be mounted in a package like a single chip. But then Seoul has to wire-bond-connect all the individual LED anode/cathode points on the top side of electrically distinct, but physically not separated, chips because they don't have a process to properly "junction isolate" each LED from others (nor does anyone else). Such a process would be needed before you could interconnect all elements via a surface metallization pattern as is done with regular ICs. Such a profoundly useful junction-isolation process is compatible with silicon semiconductors but not with non-silicon processes at this time.

Yes, it is a "multi-junction" component. But this is just semantics. Any time you series- or parallel-connect any diodes you have multiple junctions. It's just a matter of whether you connect one to the next with a wire or via a surface metal film pattern.