For example, a traditional LED Flash driver uses a boost converter in current-controlled mode as shown in Figure 1.Consider the case of driving 1A each through 2 LEDs in parallel, such as the Lumileds Luxeon PWF1, which would generate approximately 20 lux at 2m using a high uniformity optic (Lumileds Technical Datasheet DS49). The output voltage at the boost converter = LED forward voltage + voltage across current sense resistor. The maximum forward voltage = 4.8V and assuming 200mV across the current sense resistor, then the boost converter output voltage = 5V. Assuming the camera phone battery voltage = 3.3V under load, and the boost converter is 85% efficient, the battery current in this case would = 5V/3.3V/85% x 2A = 3.6A. This exceeds the typical phone battery’s capability.

Figure 1: Current controlled boost converter as LED Flash driver

The alternative solution is to use a xenon flash, but this requires
1) a storage capacitor that is very bulky for mobile phone form-factors and
2) high voltage, resulting in circuit and safety issues.

Also, in order to fit into a camera phone the smallest diameter 330V electrolytic capacitors, 5.5mm or higher, have an order of magnitude less energy than found in a typical digital camera so the total light energy is lower than a supercapacitor LED solution. Finally, with LED flash, the same LED circuit can be used at lower current for video capture/torch function.

To overcome the power limitations in current LED flash systems, some camera phone suppliers have used long flash exposure times to compensate for the lack of light, thereby increasing the total light energy, but this results in blurry photos. Solutions
Design solutions:
To provide enough LED flash power to eliminate dark and blurry photos camera phone designers can use a high capacitance (0.4F to 1F), low ESR (less than 100 milliohms), thin (1 to 3 mm) prismatic supercapacitor that is compact enough to fit inside a camera phone. The supercapacitor supports the battery and delivers the pulse power to drive an LED to full light intensity.

There are two possible supercapacitor-based solutions: * Solution 1 places a dual-cell supercapacitor at the output of a buck-boost converter, offering maximum power and supporting a flash photo up to 3 meters.
* Solution 2 uses a less expensive single-cell supercapacitor in series with the battery, yielding an effective flash up to 1.5 meters.

Part 1 of this series outlines solution 1, and part 2 will outline solution 2.

Solution 1, Supercapacitor at the output of a boost converter or charge pump
Figure 2. shows a block diagram of the dual-cell supercapacitor solution while Fig 3 shows its circuit implementation. A small, low-cost, current-limited charge pump pre-charges the supercapacitor to ~5.5V. Once the supercapacitor is charged, the current switch is enabled to deliver a high current flash pulse, with the energy and power coming from the supercapacitor rather than the battery and charge pump. During the flash pulse the charge pump can either be enabled or disabled. The charge pump is current limited to ~300mA. In Torch mode, the charge pump is left enabled and the battery and charge pump can deliver a constant current which is less than the charge pump current limit.

Figure 2 High Power LED Supercapacitor Solution Block Diagram

For the reference design, CAP-XX chose the highest power bright LEDs, Lumileds LXCL-PWF1, which can handle a peak pulse current of 1A for < 200msecs. The company drove 4 x PWF1 at 900mA each. The total LED current of 3.6A was limited by the Micrel MIC2545 current switch, which was chosen for its current capacity and relatively small size.

Circuit implementation
Refer to Figure 3, a circuit implementation for solution 1. When power is first applied, the Flash/*Torch select must be low or floating so that U1 (SP6685 charge pump) is enabled. This turns M1 OFF and U1 pin 5 (EN) is pulled high through R6. Depending on the size of the supercapacitor you will need to wait 10 – 15 secs before the supercapacitor is fully charged from 0V. There is a high inrush current when the supercapacitor is at 0V since until it is charged to a voltage ≥ Vin which makes the supercapacitor look like a short circuit on the output of the charge pump. To address this issue R11 was added to the circuit to limit the initial inrush current to < 750mA.

Note that the SP6685 is only used to charge the supercapacitor so it is always in its Torch mode (pin 4, Flsh connected to Gnd).

Once the supercapacitor has been charged, select Flash or Torch mode. When this signal is Hi (Flash mode), M2 is ON which sets the current setting resistor for U2 (MIC2545) = R9//R10. This sets the LED Flash current.

The LEDs are on while the Enable input (U2 pin1) is held Hi. This turns on U2. Due to the very large capacitance of the supercapacitor chosen, the flash pulse discharges the supercapacitor only a relatively small amount, typically < 1V. This means the time to re-charge the supercapacitor between flash photos is relatively short, typically ~2 secs and is less time than most LEDs require between flashes to prevent thermal issues. Figure 5 shows the supercapacitor voltage, battery current and LED current during and after a flash pulse. D6 was added to prevent the supercapacitor discharging into the battery through U1 when U1 is disabled.

Supercapacitor C and ESR are selected as follows:

Total LED current (I_LED) = 3.6A for a 150msec flash pulse, denoted by PW_FLASH.

* From the Lumileds datasheet, at 0.9A nominal LED forward voltage = 3.75V, allow 4.2V
* From the Micrel datasheet, the R_DSON resistance < 50mΩ, so the voltage drop across the MIC2545 current switch < 180mV
* Therefore the minimum voltage at the supercapacitor at the end of the flash pulse must be ≥4.2V + 0.18V = 4.38V ≈ 4.4V
* Vout (charge pump voltage) is set to 5.3V, therefore the total voltage drop allowed at the supercapacitor, Vd = 5.3V – 4.4V = 0.9V
* Supercapacitor voltage drop, Vd = I_LED x (ESR + PW_FLASH/C)
* Or, re-arranging terms, C ≥ I_LED x PW_FLASH /(Vd – I_LEDxESR)
* In the above example, C ≥ 2A x 0.15s/(0.9V – 2AxESR)
* Assume a supercapacitor ESR = 100mΩ, then C ≥ 2A x 0.15s/(0.9V – 2Ax0.1Ω) = 0.43F. Select a supercapacitor with ≈ ½ the assumed ESR to allow for ageing over life. We chose a 0.55F, 50mΩ supercapacitor.

Figure 3: Dual-Cell Supercapacitor Solution Circuit Implementation

Note that two supercapacitor cells are used in the circuit to achieve the necessary voltage rating of 5.5V maximum voltage. 100m is a good starting guess for ESR. Designers may need to iterate between C & ESR to find a suitable supercapacitor. Set the charge pump output voltage to the lowest possible value while still having sufficient headroom for the solution.

The supercapacitors chosen for this solution have very low leakage current, typically < 1μA. However, when two cells are used in series, a cell-balancing circuit is required to ensure that any difference leakage current between the two cells does not cause the midpoint voltage to drift such that one of the two cells goes over-voltage. The simplest cell-balancing circuit is a pair of balancing resistors, which are shown in figures 2 and 3. For a cell phone camera flash implementation, where the supercapacitor will be charged to > 5V prior to a flash pulse, a suitable value would be 22KΩ resistors. This would result in a total leakage + balancing circuit current of ~80μA if the supercapacitor were normally held at better voltage (~3.6V). If the supercapacitor circuit is also kept at power, this may be too high for good battery standby times. Possible remedies for this are: 1) disable the charge pump when the phone is not in camera mode – this means there is no supercapacitor + balancing circuit leakage drain while the phone is in standby, or 2) use an active balancing circuit using a high impedance low current op amp – a reference design is available from CAP-XX with total current < 2μA.

Charge pump selection is not critical. We selected the SP6685 for its small size. Note that the soft start function of most charge pumps will not properly handle a supercapacitor at the output due to the high start-up current mentioned earlier, but adding a current limit resistor at the input to the charge pump (R11 in Figure 3) is a simple solution, if this is an issue. Design solution
Results, Solution 1:
CAP-XX was able to place two supercapacitor cells, the circuit of Figure 3 and four replacement LEDs in a leading brand camera phone and put it back together with no changes in external appearance. Figure 4 shows photos taken using the unmodified phone and the phone modified with two supercapacitor cells and four replacement LEDs. The unmodified phone delivered 1W of flash power for 160ms while the modified phone delivered 15W of flash power for 160ms.

Fig 4a: Photos in low light with normal phone

Figure 4b. Phone modified with supercapacitor-based solution

Figure 5 shows the battery current, LED current and supercapacitor voltage during a flash pulse and supercapacitor re-charge after the pulse. Note that battery current never exceeds 300mA even though the flash pulse is 4A. The supercapacitor provides the 3.7A difference.

Figure 5: Solution 1 Battery, LED Flash current, Supercapacitor voltage during and after a flash pulse

About the author
Pierre Mars is the VP Applications Engineering for CAP-XX Inc. Based in Sydney, Australia; CAP-XX has sales offices in the USA and Taiwan and is represented in Europe by ACTE Components Ltd. The company can be reached at sales@cap-xx.com. Design tools, application notes and other details are available at www.cap-xx.com.