Although the incandescent light bulb has dominated lighting applications since its introduction in the late 19th century, its efficiency has remained at less than five percent. Worldwide ecological awareness and the search for more cost-effective lighting options have made the light bulb the target of technology, resulting in the arrival of white LEDs (Light Emitting Diodes). LED efficiency is five to eight times higher than that of an incandescent lamp, and these semiconductor devices offer several advantages – such as longer life time and instantaneous full brightness – even when compared to other high-efficiency illuminants such as fluorescent lamps.

To achieve optimal usage of the LED’s advantages, it takes some smart electronic control that considers the requirements of the LED itself, as well as the regulations regarding issues such as harmonic distortions of the current drawn from the supply.

The article describes how to reduce power dissipation of the LED driver to avoid additional thermal load for the LEDs. The first example shows a combination of two buck converters, one to drive the LEDs, the other one to provide the IC’s supply, completed by a fast start-up circuit. The second example covers an LED driver with power factor correction, applicable for high numbers of LEDs, such as for uniform illumination of areas or as a replacement for fluorescent tubes. Powered from a 120 V or 230 V AC supply line, both concepts make use of the same universal switched mode power supply IC.

Basic Considerations
LEDs, like other semiconductor diodes, feature a characteristic that shows an abrupt current increase after a certain threshold voltage has been exceeded. Furthermore this characteristic is pretty much temperature dependent. Consequently, it is recommended to drive LEDs with constant current rather than from a voltage source to achieve a well defined operating point. To avoid sacrificing the high luminous efficiency due to poor efficiency of the drive circuit, DC-DC converters are the best choice to provide electrical power to the LEDs, especially when supply voltage is significantly higher than LED forward voltage. But even when employing a DC-DC converter, total efficiency may vary significantly with the topology and choice of components.

Buck Converter Approach
The step-down or Buck converter is commonly used to drive LEDs due to its simplicity and some inherent advantages:

- Constant load current can easily be achieved by peak current switching and a constant off-time; no control loop is needed.
- The current setting is basically independent from voltage variations, as long as the supply is higher than the LED forward voltage.
- The Buck converter is inherently open load protected; in case of on open failure of an LED or the wiring, it simply stops operating.

Figure 1: Buck converter driving LEDs with constant current

Figure 1a shows the basic Buck converter topology powered from a line voltage of 110 V or 230 V AC. The line filter and bridge rectifier with a ceramic bypass capacitor are followed by a block named “Valley Fill”, which will be discussed later. The converter itself, consisting of an inductance, a power transistor, free wheel diode, the control IC, and a few passive components, provides constant current to the LEDs.

Each electrical component dissipates power. However, a closer analysis shows that the main sources of losses are the inductor, the free wheel diode, the power transistor, and the linear IC supply. So what can be done to optimize efficiency?

First, consider the topology of the LEDs (generally spoken assuming a given output power). A chain of LEDs with high forward voltage and low current yields better efficiency than a low voltage, high current solution. Main reasons are the free wheel diode and the switching transistor. The influence of the inductance is small because higher current allows higher current ripple and thus lower inductance, and which scales nicely with the DC resistance. Reducing the inductor’s specific losses (i.e. increasing L/Rdc) means: increasing its size.

The transistor causes static and transient losses. Static power dissipation is proportional to Rds,on, but low Rds,on causes high gate and drain charge. Gate charge times switching frequency is the average current that has to be provided by the control circuit, and the product of drain charge, DC supply voltage, and switching frequency is dissipated by the transistor itself. Therefore the design target should be to minimize the total power dissipation, especially considering that the transistor’s on-time duty cycle is typically pretty small.

A common proposal for reducing the control IC’s supply losses is the application of a transformer -- instead of the inductor -- with a secondary winding to supply the IC. This mostly application-specific device increases system cost. Figure 1a shows an alternative solution, deriving the IC’s supply from a second, but very small inductance. This “auxiliary converter” operates in discontinuous conduction mode, and the energy it provides can easily be calculated by:

W = ½ Ipeak x La x fs,

with Ipeak: switching current, La: auxiliary inductance, fs: switching frequency. A few microhenries are typically sufficient to provide several microamps for the control circuit.

The required start-up circuit employs a DIAC (as used in TRIAC dimmers) that turns on when the voltage on capacitor C1 exceeds around 30 V, and discharges it into C2, providing a voltage according to the capacitor ratio, high enough to start operation. Power dissipation in the charging resistor Rs can be kept very low, and still the circuit operates in a wide supply voltage range with low turn-on delay.

Power Factor Correction
Even though Power Factor Correction (PFC) is not required for low power (<25 W) lighting equipment, it makes sense to employ at least some kind of current wave shape improvement. “Valley Fill”, as shown in Figure 1, is the most basic passive kind of PFC. The current shape does not at all look like a sine wave, but the rectifier’s conduction time is significantly increased, and the harmonic content is reduced accordingly. Even though the additional three diodes cause extra power dissipation, system efficiency may be improved for two reasons:

1. The lower average DC voltage reduces switching losses in the power transistor.
2. The reduced harmonic current content reduces losses in the line filter.

Figure 2: Comparison of line current IAC a) without b) with “Valley Fill”

Figures 2a and 2b show the line current of a 10 W LED load with a single bypass capacitor of 10 µF/400 V and a (22 µF + 22 µF)/200 V Valley Fill approach, respectively.

For higher power ratings, active PFC is recommended. The easiest way to achieve a good approximation to sinusoidal current wave shape is to have a boost converter operate in critical conduction mode with constant on-time. Peak switching current is thus proportional to the momentary supply voltage, and off-time varies with the difference between supply and output voltage. The switching transistor is turned on after detection of the inductor current’s zero crossing. Average current of a switching period is one-half of the peak current.

The usual topology is to place a Buck converter behind the PFC booster that provides constant current to the LED chain. A different approach is possible when the LED chain is long enough to obtain a forward voltage higher than the maximum peak line voltage. For an AC voltage of 120 V, considering a maximum of 140 V, this peak can reach nearly 200 V. Assuming a minimum LED forward voltage of 3 V, it takes at least 66 LEDs in series. These are quite many, but for uniform illumination of an area, it certainly makes sense. Alternatively multi-chip LEDs, such as four chips in series on a single device, can be used. This reduces the number of LEDs to 17, and would be suitable, for example, for fluorescent tube replacement.

In such a case, the PFC converter’s output can directly drive the LED chain, if it provides constant average current or power rather than voltage.

Figure 3: PFC converter for long LED chains with Vf > VAC,peak

Figure 3 shows the schematic of such a solution. The circuit features open loop control without any stability issues. In order to compensate the effect of power supply variation on LED current, the peak supply voltage is stored on capacitor C1 and used to reduce peak switching current. The voltage dependency of output current is non-linear, but with proper dimensioning this linear compensation leads to a current variation of <1 percent for a line voltage tolerance of plus/minus 15 percent.

Zero crossing of inductance current is detected using resistor Rshunt; thus no transformer is required.

Boost converters generally need some open load protection. The divider R1/R2 feeds back the output voltage to turn the converter off as soon as a typical limit of 280 V is exceeded.

The single converter approach offers PFC in combination with excellent efficiency at low cost. A few numbers may serve as an example:
68 (17 x 4) LEDs operated at typically 225 V/150 mA = 33.75 W
efficiency: near 94 percent, power factor related to harmonics: PFh = 0.998, 3rd harmonic’s amplitude: <3 percent

Optimization of LED lighting applications is a combination of suitable LED topologies and the best fitting DC-DC converter technique. A universal switched mode controller helps support many different architectures to achieve optimal performance at low cost. The two examples presented in this article demonstrate techniques to achieve cost and power efficient solutions for a variety of LED lighting configurations, ranging from simple LED strings for incandescent bulb replacement to LED arrays as an alternative to fluorescent tubes and for larger area illumination.