An embedded gradual dimming function provides an easy way to generate the special illumination sequences being requested by portable goods manufacturers, so that they can differentiate themselves from their rivals.
Not only will the LED peak current I(LED) tend to be fully programmable, but it will be expected that each LED can be dimmed to any value between zero and its maximum specified level.
Generally speaking, LED drivers provide 
a constant current to bias the LED in appropriate conditions. In a typical portable system, the power source is a battery with an output voltage ranging from 2.8 V to 4.2 V (assuming a standard Li-Ion battery is used).
Since the forward voltage of the low power LEDs currently on the market varies between 2.8 V and 3.5 V, depending upon bias current and room temperature, an interface is necessary to ensure the LED is properly biased during normal operation.
This 
is the purpose that the driver IC serves, and the first thing to be considered is the voltage span of the current control system. The next decision the design engineer needs to make is whether to connect the LEDs in parallel or series. Both of options have their advantages and their drawbacks.
In colour applications the capability to independently and dynamically adjust the brightness of each LED is highly desirable.
Although it is possible to use a boost structure, with switches controlling each LED, the series arrangement is not the preferred solution, as a parallel structure is far easier to implement. The charge pump is the most appropriate type of DC/DC converter to generate a low voltage while keeping EMI issues to an absolute minimum.
However, using multi-mode operation (1X, 1.5X, 2X) provides a net efficiency improvement, saving energy and extending battery life.
The next key parameter to be considered is the current matching between the LED emitters. An RGB structure cannot accommodate bias current differences between the LEDs, since such differences would affect colour rendering. The problem is solved by using a set of accurate current mirrors.
To achieve precise and stable forward bias conditions in the LEDs, a reference current is generated by means of the external resistor and a constant voltage sourced from a band gap reference. Transistor Q2, associated with operational amplifier U2, outputs a constant voltage at the Vref pin.
The external resistor, connected across Vref and ground, creates a constant current flow through transistors Q1 and Q2. This current is mirrored and amplified by the set of transistors Q3-Q7, connected via switches S1-S5, and summed by transistor Q8. Finally, transistor Q9 copies the reference current into LED1.
Such a structure is duplicated for each LED, the layout of the chip optimising the matching between them. As a consequence, every LED emitter shares the same I(LED) peak and extra electronic circuits are necessary to independently control the brightness of each LED. This is achieved by using an independent PWM modulation for each emitter.
The switches S6-S8, controlled by the digital signals PWM1-PWM3, turn ON/OFF the associated current mirrors, thus generating a brightness control for each LED. A constant peak current is realised in the LEDs, ensuring the colour rendering is not hampered by the brightness control. The operating point for each LED stays in the reference colour defined by the standard colour map.
The waveforms, coming from a typical application (see Figure 2), illustrate the behaviour of the three PWMs. The LEDs are controlled by a common low frequency clock with a duty cycle set for that specific application. It is possible to independently decrease/increase each PWM, from 0-100% duty cycle, with the I(LED) peak being constant.
For digital control, LED current is preset via the I²C port and the PWM.
Author is Michael Bairanzade, application engineer and Marie-Therese Capron, director, low voltage power management at ON Semiconductor