The conventional boost-converter topology has the potential trap of no short-circuit protection (Fig. 1). Typically, a converter is safeguarded by placing an electronic fuse in front of it. This fuse can also act as a hot-swap controller (Fig. 2). During live insertion, the hot swap gradually turns on the MOSFET, preventing current surges into the load capacitance, CIN. In normal operation, if the current passing through RSENSE into the boost converter exceeds the preset value, the MOSFET instantly shuts off.
Regularly used boost controllers offer a wide range of operational input voltage. For example, National Semiconductor's LM3488 begins operation when the input voltage reaches 2.9 V, which can happen long before the hot swap is completely turned on. As a result, CIN, the reservoir for the boost converter, isn't fully charged, and excessive current will be needed through RSENSE.
Another factor greatly contributing to the current surge during startup is the controller's attempt to bring the partially charged COUT up to the required VOUT. The combination of these current surges can easily turn off the fuse.
These well known startup hurdles are commonly treated with a soft start. This method helps, but it can't provide 100% reliable operation for all possible output voltage-load capacitance combinations.
The solution shown in Figure 3 (unrelated parts excluded) delays the controller activity until the hot swap is entirely on, CIN is charged, and the voltage on COUT almost equals the input voltage. The principle of operation is as follows:
For normal controller operation, the frequency-setting resistor (RF) must be pulled to ground. If the switch (transistor Q) is turned on, pulling RF high, then the boost's oscillator stops. The switch remains on while the difference between VIN and the voltage on CIN stays higher than about 0.7 V.
Only when the current through the hot swap brings the voltage across CIN close to VIN (less than about 0.7 V) does transistor Q turn off, allowing the boost controller to operate. At that moment, the voltage across COUT (charged through the inductor and diode) is lower than CIN by just a diode voltage drop.
Capacitor CDEL is optional and grants (in conjunction with R1) some extra time for Q to stay on. This lets CIN and COUT be charged even closer to VIN (taking away the aforementioned 0.7-V difference). Note that this method isn't a panacea. Still, it works very well with the correct combination of a soft start and properly selected input and output capacitors.