3.3 V from a LiPo cell: dropout, buck-boost, and what actually works

A LiPo (or any lithium-ion) cell charges to 4.2 V, sits around 3.7 V for most of its discharge, and is considered empty at about 3.0 V, where a protection circuit usually cuts it off. A 3.3 V rail sits inside that range, near the bottom. While the cell is full there is almost a volt to spare; towards the end there is nothing.

A linear regulator can only take voltage away. Its dropout is the smallest gap between input and output at which it still regulates, and it grows with load current. For the AP2112K-3.3, a common choice on single-cell boards, the datasheet gives:

At 600 mA, the regulator’s full rating, the worst case needs 3.7 V in: the cell’s nominal voltage. At light load it regulates almost all the way down to 3.3 V. Below those points the regulator does not switch off. It goes into dropout and its output follows the cell, a little below it. Whether that matters depends on the parts on the rail: many 3.3 V microcontrollers and modules are rated down to 3.0 V, so check the minimum supply voltage in each datasheet.

Average current is not the problem; peaks are. A Wi-Fi radio draws short bursts of several hundred milliamps while it transmits. The dropout is largest at exactly those moments, the cell’s own internal resistance pulls its voltage down at the same time, and near the end of the charge the two together take the rail below what the chip accepts. Its brownout detector resets it, and the symptom looks like a firmware bug: the board reboots whenever it tries to connect.

1. A low-dropout regulator, and accept the tail. Simplest and quietest. Choose one whose dropout at your peak current is small, put bulk capacitance at the load, and set the firmware’s low-battery cut-off above the point where the rail starts to sag.
2. A buck converter with a 100% duty cycle mode. The TPS62743 (2.15–5.5 V in, 300 mA) switches efficiently while the cell is full, then holds its switch on and passes the cell through as it nears 3.3 V, much like an LDO in dropout. Better battery life than an LDO at high load, the same tail.
3. A buck-boost converter. The TPS63001 takes 1.8–5.5 V and holds 3.3 V across the whole cell, stepping down while the cell is above 3.3 V and up below it. Up to 800 mA in boost mode. It costs an inductor, a few more parts and some switching noise, and it is the only option here that uses the whole charge at a steady 3.3 V.
4. A lower rail. If every part on the rail runs at 3.0 V, a 3.0 V regulator keeps regulating for longer down the cell, for the same parts count as option 1.

Before dropout, check that the regulator runs from one cell at all. Many popular buck converters are rated for 4.5 V input and up, which a single cell never reaches: the XL1509-3.3, for example, is rated 4.5–40 V. In the session in our MCP guide, exactly that part was offered as one of two substitute regulators for a single-cell board. The session took the other, a step-down converter that cannot hold 3.3 V as the cell runs down, and the board was routed and passed the design-rule check. A design-rule check measures copper; it does not read the regulator’s datasheet. Take the input range from the manufacturer’s datasheet, not a parts summary.

With a simple charger such as the MCP73831 and the regulator wired to the cell, the board draws from the charger’s output while it charges. The charger cannot tell your load from the cell. It ends the charge when its current falls to a set fraction of the programmed current, and a load that draws more than that keeps the current from ever falling far enough, so the cell can sit at 4.2 V instead of finishing. Microchip’s AN1149 shows the fix: a load-sharing path with a P-channel MOSFET, a Schottky diode and a resistor, which runs the board from USB while it is plugged in and from the cell when it is not.