
Transistors Q1 and Q2 in Fig 1, develop a feedback signal of the proper positive polarity for regulator IC1 from the circuit's negative output. You can use this simple technique to sense any negative-voltage output.
The regulator drives a charge pump comprising L1(Sumida, Arlington Heights, IL, (708) 956-0666), C1, C2, D1, and D2. When the regulator's switch is on, current builds in inductor L1. At the same time, the charge on C1 transfers to the output capacitor, C2. During the switch's off time, energy stored in the inductor charges capacitor C1.
The level-shifting feedback circuit begins with Q1. Under normal conditions, Q1's base is about 0.6V above ground. The negative output voltage sets the current through R1. Assuming that Q1's base current is negligible, R3's current also flows through R2, biasing Q2's collector at a positive voltage proportional to the negative output.
Q2 is diode-connected and compensates for the change in Q1's base-emitter voltage and collector current with temperature. Both transistors see the same collector current, and their base-emitter voltages track well. Because their base-emitter voltages cancel, the voltage across R2 also appears at IC1's feedback pin.
The resulting output voltage is
VOUT = VFB(R3 / R2) - VBE,
where VFB is IC1's internal 1.244V reference, and VCBE is Q1's base-emitter voltage (~0.6V). The VBEterm in the equation implies a minor output-voltage dependency on input voltage and temperature. However, the variation because of this factor is usually well below 1%.
Essentially, Q1 holds its collector voltage constant by changing its collector current and functions properly as long as some collector current exists. This condition put the following limitation on R1: At minimum input voltage, the current through R1 must exceed the current through R2.
R1 <(vIN MIN - VFB - VBE) / VFB.
If the input voltage drops below the specified limit (for example, during a slow start-up) and Q1 turns off, R3 provides IC1's feedback pin with a positive bias, and the output voltage decreases. Without R3, the feedback pin would not get a sufficient positive signal, forcing IC1 to provide excessive output voltage and possibly damaging the circuit. (DI #1627)