As the cost of energy continues to rise, significant attention is being paid to using more efficient and more cost-effective lighting sources. A major application is large-area lighting, using metal-halide high-intensity discharge (HID), fluorescent and high-brightness white LEDs. Intelligent high-brightness LEDs are looking to oust commonly used incandescent and mercury vapor light sources, as used in streetlighting. In addition to their maintenance requirements, the latter source has lower efficiency, contains mercury, and requires a significant warm-up and restart period (see “Energy Efficiency Is a Global Issue”).

High-brightness LEDs have different characteristics and requirements from other lighting sources, posing new challenges for their driving circuits. They feature low forward voltages of 3 V to 4 V and require constant-current drive for optimal operation. They're also typically low-power devices (1 W to 3 W) with drive currents in the range of 350 mA to 700 mA. The latest high-brightness LEDs have flux levels ranging from 80 lm to 110 lm, with efficacies of 70 lm/W to 90 lm/W.

But, they need to be used in an array to obtain the same light output possible from other sources. For example, to get the same light output as a 100-W metal-halide HID lamp, an array of 30 LEDs to 80 LEDs is needed, depending on the LED drive current and flux output.

Intelligent Power-Supply Architecture

Modern high-brightness LEDs for streetlighting require a flexible and intelligent power supply that must meet several requirements. The supply must be energy efficient and needs to meet harmonic-content or power factor correction (PFC) requirements like the European Union's IEC61000-3-2 specification. It also must support a wide range of input ac voltages, regulate a constant current to an LED string and provide dimming control.

It should be noted that digital pulse-width modulation (PWM) techniques are normally used for dimming LED streetlights. In the United States, streetlighting is not normally metered, and the power is managed by the local utility company. The utility is interested in obtaining the most efficiency from its power grid, so a power factor of 0.9 or greater is generally required.

A two-stage architecture with a power factor boost stage is followed by a flyback stage (Fig. 1). This can satisfy all the aforementioned requirements for intelligent LED streetlighting. While there are single-stage PFC/flyback topologies such as that implemented in ON Semiconductor's NCP1651 controller, the requirements of digital dimming of the LEDs favor a two-stage approach.

The bandwidth of traditional PFC control loops is normally on the order of 10 Hz to 20 Hz, and LED dimming frequency is above 100 Hz to avoid visible flicker. Thus, a single-stage architecture for digital dimming is not an option.

In addition to the blocks in Fig. 1, a small low-power auxiliary supply is used to power the secondary stage. That stage consists of the current regulation loop and the microcontroller and communication interface circuitry needed for network control.

For traditional streetlighting, different bulb/ ballast combinations are used depending on the area being illuminated. This is a function of the height of the fixtures, the spacing between the fixtures and the light output required on the ground. In the case of streetlighting, the amount of light on the pavement and the light pattern are determined by established standards depending on the expected traffic flow.

Numerous parameters factor into determining the number of LEDs required for a particular application. These include the type of power LEDs, the LED drive current, the expected operating temperature conditions, thermal management techniques and the optical design. The approach used for this power supply was to create an architecture that would accommodate various LED string lengths up to 60 LEDs and to support the use of a common ballast.

To achieve long lifetimes for LEDs, it is important to maintain the LED junction temperature at a reasonable level in the range of 80°C to 90°C. This is a vendor-specific parameter. LED lumen output is reduced as the temperature increases. So, even though high-power LEDs can be rated for currents of greater than 1 A, this design is focused on providing a constant current in the 250-mA to 400-mA range. This reduces the internal heating of the LEDs.

Boosting PFC

The power requirement for most LED streetlights is on the order of 60 W to 150 W. As a result, a critical-conduction-mode PFC topology is used, employing the NCP1606 controller. Historically, most lighting ballasts have been optimized for one region, but for flexibility, this design has a standard universal input of 90 Vac to 265 Vac to support North American as well as international ac power sources (Fig. 2).

The output voltage of the boost stage is 400 Vdc nominal and is designed to deliver 100 W. Since the NCP1606 implements a boost-conversion stage, this output voltage is determined by the maximum ac line voltage, and the characteristics and accuracy tolerance of the controller.

In the United States, it is also common to power streetlights and commercial lighting from 277 Vac. This voltage is derived from one phase of a 3-phase 480-Vac circuit. To modify the design to support this requirement, a higher dc bulk voltage at the output of the PFC boost stage is required. This involved changing the feedback voltage to generate an output voltage of 480 V. It also may increase the maximum-rated voltage of the bulk capacitor and power MOSFET.

Following the PFC stage is a fixed-frequency isolated flyback stage and a linear current regulator that regulates a constant current to a string of LEDs (Fig. 3). The regulator maintains a constant voltage drop across the linear current-regulator MOSFET (Q2).

The flyback controller is the ON Semiconductor NCP1216, a compact 8-pin controller with a current-mode architecture. This controller is based on a proprietary high-voltage process that allows the device to start up from the high-voltage bulk and contains dynamic self-supply circuitry to power the device in normal operation.

The heart of this circuit is the secondary-side control block, which provides the constant-current regulation for the LED string. It maintains a constant drop across the current regulator MOSFET regardless of the output load voltage. It also accommodates a fast logic-level PWM dimming input signal needed to adjust the luminaire light output.

By using a series-pass linear regulator to control the LED current (U5A, Q2, R10), fast response to the PWM LED dimming control signal can be achieved at pulse frequencies up to 1 kHz, with reasonably fast rise and fall times. This could not have been done very effectively with the main flyback feedback loop without fairly low frequency constraints on the PWM input signal.

Despite the use of a dissipative current regulator, very low dissipation in Q2 is achieved by regulating the voltage drop across the series-pass MOSFET to about 1.5 V (power dissipation is 550 mW) with the flyback feedback loop. This provides maximum efficiency and versatility to deal with different output voltages due to varying LED forward drops and different LED string configurations. It also allows the flyback transformer to be designed to accommodate the maximum-desired output voltage yet handle lower voltages without dissipation or circuit change issues. That would occur if the output voltage were tailored for a specific LED configuration and directly regulated.

Regulation of the voltage drop across Q2 via the flyback converter is accomplished by a sample-and-hold circuit and error amplifier composed of D6, R24, C12 and U5B. This also allows fast PWM gating of the current amplifier without significant voltage excursions on the input voltage to Q2.

Reference amplifier U4 (TL431) provides the reference voltage for the current regulator (U5A) and the feedback error amp (U5B). Feedback to the NCP1216 current-mode controller on the primary side is with a conventional opto-coupler (U2). In the event of an open LED string or no load on the power-supply output, a simple voltage-clamp feedback circuit is provided by Z1, Z2, Z3 and additional optocoupler U3.

To further minimize the size of the transformer, a 130-kHz switching frequency was selected. The NCP1216 controller also features frequency jittering capability to reduce the EMI signature.

An 800-V power MOSFET was used in this circuit to accommodate the reflected voltage from the secondary. Note that, depending on the maximum forward voltage of the LED string used, a higher or lower voltage MOSFET could be used.

The flyback transformer T1 is designed for discontinuous-conduction-mode operation with a maximum output of about 250 V. The core is a standard E21 type ferrite core. However, other core geometries, such as ETD, PQ or EC types, could have been used depending on the required maximum output voltage.

Keeping the primary and secondary wound on single layers is important to minimize leakage inductance effects. If smaller cores are desired, multiple layer windings can have potentially detrimental effects on circuit parasitics and subsequent voltage spikes and ringing.

Fig. 4 and Fig. 5 illustrate the current switching waveform as a PWM signal is applied. The upper trace is the current through a sense resistor in series with the LED string. The lower trace is the PWM input-signal waveform. The Fig. 4 waveform shows a normal case with 50% dimming, while the Fig. 5 waveform shows an extreme case. They illustrate the symmetric rise and fall times, which facilitate linear dimming control.

Network-Controlled Dimming

The example given illustrates a circuit configuration to drive long strings of LEDs from the ac main while offering excellent power-factor performance, dimming capability and fault protection. Overall, efficiency driving a string of 60 LEDs at 350 mA was approximately 85%.

One interesting aspect of having network-controlled dimming is that a microcontroller function can be embedded in the luminaire. This could allow enhanced intelligence functions such as active temperature monitoring to compensate light output based on the temperature of the LED array, which can vary dramatically by season. In addition, it can prove safety monitoring in case of a fault.

The combination of high-brightness LED technology, efficient power-supply design and intelligent control provides interesting opportunities to reduce energy consumption and lifecycle costs. Moreover, since LED efficacy and lumen output continues to advance at a brisk pace, this design was made to accommodate the fact that the number of LEDs required for a given light output will continue to fall as LED technology improves.