The vast majority of light bulb replacements are AC mains powered. Due to their low-power consumption, there is no legal requirement for power factor correction (PFC). The law requires lighting applications with 25 W or more power to have a good PFC, with a power factor of 0.95 or more. However, as described in Chapter 8, some local regulations and “green” organizations do have some more onerous PFC requirements (power factor = 0.7–0.9) for all LED lamps, regardless of their power consumption.

Tube lamps need the drive electronics to be housed inside the tube, but in such a way that light output is not obscured. Due to the tubular space, it is common to find a PCB strip across the center of the tube, with LEDs on one side and drive electronics on the other side. One potential problem is power dissipation because the circuit may draw 10 W power or more. A continuous glass tube may no longer be possible. The tube may have to be glass or transparent plastic on the LED side and metal or head-conducting opaque material on the other side, as shown in the cross-sectional diagram in Fig. 17.1.

Streetlights can be in the 25–250 W power range, with 100 W being common. In some cases the LEDs are required to be isolated from the AC mains, for safety reasons. For example, in Italy, it is a legal requirement that all streetlights are isolated. An isolated fly-back converter can be used in lower-power lamps up to 50 W, but fly-backs are not so efficient and a forward converter may be needed for 100 W or more.

As the power levels are needed by streetlights, they must also have good power factor. One nonisolated solution is to have a boost converter with constant on-time, producing ∼400-V output, followed by a number of buck converters, each driving a long string of LEDs. By having several LED strings, there is some safety redundancy; if one fails, others will continue to produce light.

An example of where good PF may be required in low-power applications is street sign lighting. Street signs with “stop” or “give way” warnings often have a small lamp above them, so they can be easily seen at night. As the local authority has many of these lamps across a town or city, the combined power is considerably more than 25 W. Hence the local authority may demand a good PF for each lamp, so that the system as a whole has a good PF.

The standard control protocol for stage and theatre lighting is DMX (see Chapter 16). The two main reasons for using DMX is these applications is the wide-dimming range with 65,536 levels and the fast response (23 ms maximum, 2 ms minimum). The problem for the LED driver is that such a wide-dimming range is very difficult to achieve.

Consider the dimming range of a simple buck circuit; the inductor is usually chosen so that the circuit operates in continuous conduction mode. This means that once the MOSFET switch turns off, current continues to flow through the inductor for perhaps four cycles before the stored energy is dissipated and current stops. The switching frequency of a buck controller is perhaps 1 MHz maximum, so a 1-μs switching period, which means that the smallest current pulse is about 4-μs long. To get 10,000:1 dimming, the longest pulse is 40-ms long. This means a PWM dimming frequency of 25 Hz would produce visible flicker. Generally, the PWM dimming frequency is 200 Hz (100 Hz absolute minimum).

One solution to give a high dimming range, if a buck circuit was the only option, would be to operate the buck convert continuously and use switches to bypass the LEDs when they need to be off. A short-circuit bypass would not be practical because this would imply zero duty cycle of the buck controller. There is a minimum switching time/minimum duty cycle to consider. However, it would be possible to have a “dummy load” whose voltage drop is less than the minimum forward voltage of the LEDs, but sufficiently high to meet the minimum on-time requirement. Such a scheme is shown in Fig. 17.2, where a normal P–N junction diodes acts as the dummy load. The disadvantage of this scheme is lack of efficiency, as it will draw moderate power when the LED is not lit.

A boost circuit with an auxiliary MOSFET switch in series with the LEDs provides excellent dimming range. When the PWM dimming signal is low, the series MOSFET turns off and LED current stops instantaneously. To achieve good dimming, the boost circuit itself needs to have the switching clock synchronized with the PWM dimming signal, so that only whole switching cycles are possible. If the switching were asynchronous, it would be possible for fractions of a switching cycle to occur. For example, suppose that at the first-PWM dimming signal there were 2.5 switching pulses, but only 2.1 on the second-PWM dimming signal. As it takes a few cycles to fully charge the output capacitor, the current would change with each PWM pulse.

In Fig. 17.3, I show how fractions of a switching cycle can occur in nonsynchronous systems, if the PWM dimming signal rises or falls midcycle. In the synchronous system, the switching pulses will only be an output if the PWM dimming signal is high at the clock signal rising edge. Also, if the PWM dimming signal falls is the middle of a clock signal, the switching pulse remains high until the end of the switching cycle, so partial switching pulses cannot occur.

The ideal position would be for the output capacitor of the boost circuit to be kept fully charged by periodic switching, even when the PWM dimming signal remains off. This would prevent the LED current changing during partial switching cycles, caused by asynchronous PWM dimming. The difficulty in keeping the output capacitor charged is determining the output voltage; it should be just sufficient to achieve the correct output current when the PWM dimming signal goes high. In boost converter ICs, like the Microchip HV9912, there is an analog sample and hold function built in, which “remembers” the last output voltage. But this only holds up the reference voltage for a limited time. This is one case where a microcontroller wins over an analog system; the memory can be almost indefinite.

General stage lighting lamps do not need wide-dimming ranges. They can be either AC mains powered or battery powered, depending on the exact application. Those applications powered by AC mains may be nonisolated and drive lamps directly, or may be isolated using a built-in AC/DC converter and so effectively powered by a relatively low-voltage DC supply. Lighting a stairway, for example, may use a low-voltage DC supply because of the risk of being in contact with liquids.

One advantage of LED lamps is that they run cooler than incandescent lamps and, as a result, plants need less watering! Even so, the LED lamps can be moderately high power, say up to 200 W, using many 1-, 3-, or 5-W LEDs in a large array. The highest power LED lamps for plant growing are about 1 kW, but lower-power lamps using more efficient LEDs are now available, so such power levels are very rare.

A disadvantage of using LED lamps with just blue and red light is that gardeners cannot tell if the plant is healthy. There is no green light to be reflected by the leaves; some growers insist of having a few green LEDs added, to give a little green light, for this reason. Blue LEDs with a wavelength around 460 nm are preferred for leaf and stem growth, but red LEDs with a wavelength around 610 nm are preferred for flower and fruit growth.

The demands on the LED driver are quite simple, as no dimming is required and there can be some tolerance on the LED current. The power levels required for the lamp will probably mean that AC mains power is necessary, with PFC (see Chapter 8). The humid atmosphere in a glasshouse will probably require the lamp to have AC mains isolation. A few specialist noncommercial applications, such as space travel, may use battery-powered LED lamps to grow plants.

Inside the pool itself, the regulations specify the use of 12 V DC maximum supplies, with low-leakage current and a power source 35-m away from the pool edge. Having an AC mains power supply so far from the pool edge means that a heavy-duty cable is required on the 12 V DC side, so that there is very little voltage drop along the cable. As 12 V DC is the maximum voltage at the pool, lamps are normally wired in parallel. Each lamp would then use a buck converter to drop the output voltage to a lower level, so typically the underwater lamp would contain only one or two LEDs.

Outside the pool, but within a few meters of it, safety electrical low-voltage circuits are a must. However, the voltage can be up to 50 V DC. Again, the power source must be 35 m from the pool edge. The AC mains supply to the power source must go via an 30-mA residual current detector circuit breaker, which disconnects the AC supply if there is an imbalance in the live and neutral current (i.e., an earth leakage).

To get the most life out of a nonrechargeable battery, boost or boost–buck circuits are used. As an example, suppose that the output of a boost converter, driving the LEDs, is also used to provide power for the converter’s PWM controller. The PWM controller would take power from the battery initially, but then take power from the higher-voltage output once the switching has started. The circuit would then be able to operate continuously until the battery voltage is almost zero.

A LED light that uses a single high-power white LED needs a forward voltage of about 3.3 V to operate, let us assume that the circuit is being powered by a battery of two alkaline cells (typically 1.5 V, but 1.6 V when new, and 1 V when spent). What are the driver topology options? A buck could not be used because the output voltage is slightly higher than the input. A boost could not be used because it needs to have some margin, because of duty cycle constraints, typically the output voltage needs to be 1.2 or more times the input. So a boost–buck would be a good choice in this case.

Cell phones usually use Li-ion batteries, where the cell voltage is about 3.6 V. Unlike alkaline batteries, the voltage across Li-ion types is only slightly lower until the energy is almost depleted. In other words, the terminal voltage during discharge does not change much. Note that rechargeable Li-ion batteries should not be operated with very low-terminal voltage because they can be damaged. Often an undervoltage lock-out circuit is used to disconnect the load, before the battery voltage falls low enough to cause damage.

In cell phones, the display (and sometimes the keyboard) is backlit. Due to the stable battery supply voltage, it is possible to use a white LED with just a series resistor as current limiter (Nokia used this technique). The brightness does not change very much as the battery charge reduces. However, for this scheme to work, the forward voltage drop of the LED is critical and LED manufacturers have to supply LEDs with a narrow-forward voltage drop. If the LED forward voltage drop is accurately controlled, LEDs can be connected in parallel.

Many cell phones now use boost or boost–buck topologies. This allows lower-cost LEDs to be wired in series, so the forward voltage matching is not required. They can be supplied from a number of different manufacturers, so long as the brightness levels are similar. Brightness is measured in lumens. The eye has a logarithmic amplitude response, so one LED can have almost double the lumen output of its neighbor before the difference is noticeable.

Sports and diving watches have backlights that often use a boost–buck circuit. The circuit has to be extremely small, but fortunately the power levels are very low and no large components are needed. The controller ICs include the MOSFET switches and are in small wafer-scale ball grid array or tiny DFN-type packages. The largest component is usually the inductor.

Channel lighting requires an isolated driver for safety reasons because of the metal channel housing. Most often, engineers use a standard AC/DC-isolated power supply with 12- or 24-V DC output. Then the LEDs are connected in a short string of three or six LEDs with a constant current driver in series. Some manufacturers produce such LED strings, with the LEDs and constant current circuit mounted on a short-PCB strip. The PCB strip can be either fiberglass board, such as FR4, or a flexible PCB that can be stuck directly onto the housing back panel.

Cars have traditionally used a 12-V lead acid battery as the main power source, but this has problems for design engineers. First, the 12 V is a nominal value and can be at 13.5 V during vehicle operation when the alternator is providing power. Voltage spikes can appear on the power rail too, due to inductive components being switched, but these are usually clamped to about 30 V by a transient suppressor fitted in the vehicle. During “cold-crank,” when the vehicle is started on a cold morning, the battery voltage can drop to 6 V or lower. Due to the wide-battery operating range, boost–buck circuits are commonly used.

Some car-lighting systems use a boost converter, to raise the nominal 12 V up to a higher voltage, such as 36 V. This creates a lighting power bus. The system then uses multiple buck converters, one for each lamp, with the lighting power bus providing a power source. This technique allows low-cost buck converters to be used. Energy storage on the lighting power bus holds up the voltage during cold crank, so the lights do not flicker during starting.

Head-up displays use LEDs or laser diodes to project an image onto the inside of the windscreen. The image is then reflected back from the glass, for the driver to see. The images are to aid the driver, perhaps highlighting an area where children are playing near to the road, or to highlight a road junction ahead. At least one company was projecting the image from an infrared camera, so that people could be seen on a foggy day.

One LED/laser driver for head-up display systems is the ISL78365 from Intersil. It has four-constant current sink circuits; in a color system the four channels could be assigned to red, green, green, and blue (RGGB). The reason for the two green channels is that green LEDs are less efficient and so two are needed to attain the same light levels as red and blue. To make this driver efficient, the system is intended to have individual buck converters for each channel, to provide four positive supply rails. Each converter’s output is controlled from the ISL78365, so that the headroom voltage on the current regulator is kept to a minimum and no energy is wasted. Fig. 17.4 shows this.

Recently, cars using a dual-battery system have started to appear. These use a 48-V battery and a 12-V battery. The two batteries can exchange power in either direction. These systems are likely to be found in cars that have stop–start motor control, where the engine is turned off when the car is stationary and the brake has been applied for a few seconds. The starter motor can be operated from the 48-V supply, to reduce the current drawn and hence voltage drops in the power cables. However, much of the car electronics needs 12 V and so the second battery provides this. But the lighting system could be powered from the 48-V battery, to allow simple low-cost buck converters to be used.

Trucks and buses can have 12- or 24-V batteries; 24 V is common in Europe. This means that electronic circuits used in a truck or bus has to withstand much higher-voltage transients. When the alternator is running and the battery is disconnected, a condition known as “load dump,” the voltage can reach 120 V. The higher voltage helps in that a simple buck circuit is often sufficient, but it must be able to withstand much higher voltages.

Train voltages are variable. For example, metro trains are typically 110 V DC nominal (77–138 V). In Europe, many trains use a 72-V supply. Some diesel-powered trains have a 24-V supply. For the majority of lighting requirements, a high-voltage buck controller (like the HV9910 or its clones) could be used here.

Forklift trucks and golf carts have very large-lead acid batteries and voltages are typically in the 24–120 V range. The voltage can be much higher during charging, but as they do not have a starter motor, cold crank is not an issue. Of course, when motors are operated, such as lifting the “forks” in a forklift truck, there can be considerable transients. Again, a high-voltage buck controller would be a good choice here.

Aircrafts have an AC power supply, typically 110 V AC at 400 Hz. Cabin lighting and emergency exit signs using LEDs are now commonly used. Note that space-grade (radiation-hardened) components may be necessary because the aircraft flies at high altitude for most of its life. Cosmic rays/heavy ions can cause damage in integrated circuits, but thick metal enclosures can be used to give some protection.

Emergency exit strips in aircraft traditionally used along the side of the aisle are electroluminescent, with a phosphor material between two conducting layers (the top layer being translucent) and directly powered from the 400 Hz, 110 V AC supply. Some LED replacements have been used more recently and would typically use an AC/DC converter. However, photoluminescent strips that do not need any power source are also now being used.

In large systems, the rechargeable battery can be held centrally within the building, with a low-voltage bus connected to each emergency light. Smaller systems have a low-voltage bus with a current limiting circuit that prevent the bus voltage from being pulled down by any one of the emergency lights in the system. Thus the bus trickle charges the batteries continually and the batteries supply the power to the light for a short period only. A similar system is used in fire alarms, where the alarm sounder and the flashing beacon are powered from the battery, which has been trickle charged over the low-voltage bus.

Some alarm clocks have a light, instead of a buzzer or radio, to wake the user. Such alarm clocks have a special dimming requirement. As they are used in a darkened room, typically a bedroom, the dimming range has to be very wide. This is similar to the theatre lighting requirement and a dimming range of 10,000:1 or higher is required. When the light first turns on it must be very dim and gradually brighten to wake the user. A step effect in brightness is not acceptable; it must brighten smoothly like light from a rising sun. My comments in Section 17.4 about theatre lighting are also relevant here.

The “dawn” effect in an alarm clock is helped if the light is initially a yellow–white color or “warm white.” Then, as the light gets brighter it should become a blue–white color or “cold white.” This can be achieved by using warm-white LEDs and then adding blue light by switching blue LEDs into the circuit at high-brightness levels.