No matter what portable power source you use, the lower the starting voltage your circuitry operates at, the better. A lower startup voltage also maximizes runtime. Furthermore, to completely discharge the power source, circuitry must run on ever-lower voltages and currents.
Existing boost circuits can start up and drain a power source down to 1 V, but that still leaves too much unusable energy in a battery. Other power sources, like solar cells or micro-turbines, need circuitry to start up at much less than 1 V. For instance, a single solar cell has a full-sun output of only 0.58 V.
The circuit in Figure 1 targets this problem. It performs a dual-inductor step-up conversion at a startup voltage as low as 260 mV. An inductive dc-dc boost circuit outputs a higher voltage than its input. A germanium-transistor balanced-boost circuit is simple, using only two NPN transistors, but it unconditionally starts at a very low voltage. Previous silicon-transistor boost circuits required approximately a volt to start and many more components.
The circuit operates like a free-running multi-vibrator. The repetitive cycle starts with VIN slightly above Q2's VBE. This creates a positive Q2 base current (IB = (VIN VBE)/R1) flowing through L1, and Q2 turns on, which switches inductor L2 to ground. Q1 is off, and L1's current flow is very small. D1 and D2 are off. Energy stored within L2's magnetic field builds as L2's current rises with a positive di/dt. As this current rises, it also flows through Q2's RSAT. Q2's collector voltage becomes sufficiently large to turn on Q1.
R2 connects Q1's base voltage to Q2's collector. R2 also limits Q1's base current. With Q1 turning on, the previous base drive to Q2 is now shunted to ground, and Q2 turns off. Switching off Q2 allows L2's flyback energy to forward-bias D2 and flow into the load (R3) as L1's magnetic field collapses. D1 remains off. With L2 discharged, D2 turns back off. L1's magnetic field has been building as L1's current rises with a positive di/dt. This current flows through Q1's RSAT. Q1's collector voltage becomes sufficiently large to turn on Q2.
Q2's base voltage is connected to Q1's collector by R1, which also limits Q2's base current. With Q2 turning on, the previous base drive to Q1 is now shunted to ground, and Q1 turns off. Switching off Q1 enables L1's flyback energy to forward-bias D1 and flow into the load (R3) as L1's magnetic field collapses. D2 remains off. With L1 discharged, D1 turns off.
This self-oscillating action repeats until the battery voltage falls below Q1's or Q2's VBE. As the input voltage increases, the stored energy in L1 and L2 increases, and thus the average voltage on R3 increases.
The inductance of L1 and L2, the RSAT of Q1 and Q2, and the switching characteristics of Q1 and Q2 determine the period and duty cycle of the self-oscillation. The circuit can be optimized for a specific load and input source with adjustments to both L and R. As shown, the typical switching frequency at the output is 88 kHz (with VIN = 0.5 V). Inductances of 100 H would yield 60 kHz, and 39 µH would give 152 kHz.
The advantage of this dual-boost configuration versus a single-ended boost configuration is that the output ripple contains lower noise and the input source is not off during flyback. For a solar cell or micro-turbine input, an off cycle is less than optimal.
Figure 2 shows the circuit's input/output transfer characteristics. The voltage gain (step-up action) is shown for several resistive loads. Note that this boost circuit is open-loop, so it doesn't regulate the output voltage or current. Some applications, however, don't require regulation.
For example, this circuit could drive an LM2901 quad comparator and LM2902 quad op amp directly. Other applications (logic circuits) require only limited upper voltage regulation that could be accomplished with a shunt regulator or zener on the output.
But for the highest efficiency, use this boost circuit only to temporarily power up a full-feature, high-efficiency switched-mode power-supply (SMPS) IC, connecting the transient boost's output to the IC's low-current VCC input. Once the IC starts up, the boost circuit can be turned off. One way to do that would be to substitute P-channel JFETs (NTE326) for R1 and R2, then pull their gates above the input voltage (VIN + 1.2 V).
Also, the input voltage is limited to 2.0 V. A larger input voltage will cause excessive current to flow in the base of Q1 and Q2, which are connected through R1, R2, L1, and L2 directly to VIN.
A slightly different transfer characteristic results if this circuit drives a white LED rather than a resistive load. A white LED typically requires 3.6 V at 20 mA for proper operation—thus, the need for a boost circuit if an alkaline battery is the power source. An LED's brightness depends directly on the average flyback current flow through D1 and D2.
Nominal LED current measurements with an alkaline battery were 3 mA at 0.53 V, 14 mA at 0.95 V, 26 mA at 1.19 V, 31 mA at 1.27 V, and 50 mA at 1.53 V. Those results were with Coilcraft DO1608C-683 inductors and a Nichia BSPW500BS LED.
With many portable electronics (toys, PDAs, etc.), you must discard a single alkaline as "dead" when the cell voltage is about a volt. But such batteries can provide LED illumination with this circuit, as well as a more complete battery discharge.