Most of the latest mobile phones have built-in cameras with megapixel resolution. To enable users to take pictures in a darker environment, additional lighting is required, but standard flashlight solutions, as used in digital still cameras, are typically bulky and therefore not suitable for the small size of a mobile phone.
Recently introduced high power LEDs used in automotive applications do, however, offer an attractive solution to this problem, as they can easily be built into small packages and do not require high voltages. Such a complete flashlight solution can be very cost and space efficient. In addition, the same LEDs can be used to provide a torch function or, a movie light, by simply keeping them illuminated for longer at less current. To produce the white light required, white LEDs are normally used. Unfortunately, white LEDs have forward voltages in the 3.6 to 3.8V range, so they cannot operate directly from a single Li-Ion cell, which is typically used in a mobile phone.
Generally, the light emission characteristic of LEDs depends strongly on the LED's semiconductor material, which means that it can vary from part to part in manufacturing. Light emission usually depends on LED current, and the ratio of emitted light to LED current is one of the major parameters which varies in production. Thus, in general, LEDs are pre-selected in different groups with similar current to light conversion ratios. Among these groups of LEDs, the forward voltage can also vary.
To use the LED as a light source to illuminate a scene for a picture, light pulses have to be generated at the same time as the image sensor is active. To optimise the system as regards power consumption and heat dissipation, the LED should be on for as short a time as possible, so synchronisation with the image sensor is necessary. Typically, light pulses of 100ms or more are used, depending on the LED light emission capability and the way the image sensor and flashlight are operated together.
Hence, there are two things to solve: the LED current has to be controlled and synchronisation of the light pulses to the image sensor has to be facilitated. Simple circuits which directly control the LED current may be the first choice and offer a high degree of flexibility, allowing different LEDs from different suppliers to be used. To regulate current, it must first be measured, and the simplest and cheapest way to measure LED current is to use a shunt resistor in series to the LED. So the total required output voltage of the converter is the sum of the forward voltage of the diode and the voltage drop across the shunt resistor. To keep the circuit simple and cheap, it is best to design the shunt resistor so that the voltage drop across the resistor is the same as the feedback voltage of the converter selected.
To minimise losses, the lowest voltage drop across the shunt resistor is desirable, since higher voltages just add losses. As the sum of the forward voltage of the diode and the voltage drop across the shunt resistor is usually higher than the battery voltage of a mobile phone, boost converters have to be used. Assuming a converter with a feedback voltage of 500mV is used for our example, a total output voltage of 4.1V to 4.3V is required. Such a circuit is illustrated in Figure 1.
This looks simple and cheap, but there are tradeoffs. As already calculated above, the minimum required output voltage is 4.1V, which means, in a fully charged Li-Ion battery, the input voltage can be higher than the output voltage required. When assuming a voltage drop across the rectifying switch of a minimum 100mV, this is right on the limit where output current can be controlled. If the output voltage drops below 4.1V, a standard boost converter will no longer be able to control the LED current. This can be solved by adding resistors in series to the LED or, by using different and more expensive Buck Boost conversion topologies to ensure high efficient flashlight supplies under all possible input voltage conditions.
When triggering the flashlight by enabling the boost converter, time is required to build up the output voltage. During this time, the LED already starts conducting and shows a heavy load to the boost converter during startup, which makes startup even slower. This also causes high current pulses in the battery, which may cause additional system problems as well. The waveforms in Figure 2 show this in detail. The upper curve CH4 shows the flashlight trigger signal, the second CH3 is the input current, the third CH1 is the converter output voltage, and the last CH2 is the LED current. As can be seen from these curves, high input currents of more than 1.5A are needed for several 100us to control around 300mA of current through the LED.
The third and last major drawback of this topology is the missing protection of the output when the LED is disconnected. There is no inherent protection against output voltage going too high under this fault condition.
A voltage-controlled topology does not have this problem. A simple circuit operating this way is shown in Figure 3. The boost converter simply regulates a fixed output voltage, for example 5V, and is already enabled before the flashlight is triggered. The current is controlled by a resistor in series to the LED. To enable the flashlight, an additional switch can be used as shown. Starting the flashlight at the gate of this additional FET can be achieved with almost no delay. Waveforms are shown in Figure 4. The arrangement of the channels and the related signals is the same as already described with Figure 2. It can be seen that the LED current rises faster and the input current loading is kept to a minimum.
With high voltage already available at the output capacitor, LED current is present immediately. There is a slight drop of the output voltage due to the slower reaction time of the boost converter, which depends on the capacitance of the output capacitor. An additional advantage of this approach is a stable regulated output voltage rail, which can also be used to power different parts of the circuit, taking benefit from the voltage rail with higher voltage than the battery voltage.
A significant tradeoff with this approach, of course, is that no current regulation is implemented. Changes in the forward voltage of the LED directly translate into a change of the LED current. Over the production cycle, this change can be very high, so it might be necessary to adjust the LED series resistor to the forward voltage of the current LEDs. Pre-selection of LEDs with the same forward voltage and forward current characteristics might solve the problem as well, but would definitely increase costs.
Taking both the approaches above together could result in a circuit as shown in Figure 5. This topology incorporates current regulation as well as overvoltage protection, and the more complex network on the feedback divider takes care of that. The resistive divider connected to the shunt resistor in series to the LED builds the voltage sensing part of the circuit. If properly designed, this limits the maximum output voltage of the circuit to the maximum allowed voltage at the converter and the connected circuits.
The shunt resistor, which has a significantly smaller resistance to that of the feedback divider, hardly changes the voltage division of the divider at all. As soon as a higher current flows through the shunt resistor, there is an offset voltage in the feedback divider, which lowers the output voltage of the converter. In this way, the LED current is regulated as long as the sum of the voltage drop across diode and shunt resistor is lower than the voltage programmed with the resistive divider.
In case the shunt resistor has a value, which just allows low current to flow through the LED, the startup time can be lower than in the simple current controlled architecture discussed above. Placing the shunt resistor in parallel with a lower value resistor switched by a FET, increases the programmed LED current immediately. If this is carried out after the output voltage of the boost converter has been built up, delay time between triggering the flashlight and LED flash can be decreased. Using the lower LED current mode, for example, as a torch or a movie light, which is active after enabling the boost converter, adds an additional feature without any additional costs.
Startup waveforms of the improved circuit are shown in Figure 6. Again, the arrangement of the signals is the same as in the diagrams shown in Figure 2 and 4. It can be easily seen that input current and flashlight trigger delay time are well controlled.
Although this topology can control current and voltage, the accuracy of the current control is worse than in the simple current control architecture in the first example. The TPS61020 from Texas Instruments, which is used in the example circuits above, operates well in all of these circuits. In addition, it is capable of down regulating the output voltage, which makes it suitable for all these topologies even when the sum of the LED and shunt resistor voltage is lower than the input voltage. This means the circuits can be used with new higher efficiency LEDs as they become available.