Concept renderAn LDO burns the voltage difference as heat, so its efficiency cannot exceed Vout divided by Vin. It is quiet, small and simple. A buck converter switches through an inductor and stays efficient across large drops, with ripple, EMI and extra parts as the cost. Most Li-ion devices use both, with an LDO on noise-sensitive rails.
Every battery device needs a voltage regulator between the cell and the chips. The choice between a low-dropout regulator (LDO) and a buck converter sets battery life, board area, heat and the noise that reaches your radio and microphones. This guide is for the founder or engineering lead who signs off on the power tree of a single-cell Li-ion product.
Part figures come from vendor product pages and datasheets, as of November 2026. The parts named are examples that show the trade-offs. They are not recommendations for your design.
LDO vs buck converter: what is the difference?
Both are step-down voltage regulators. They differ in where the extra voltage goes.
| Question | LDO | Buck converter |
|---|---|---|
| How it steps down | A pass transistor drops the extra voltage and turns it into heat. | A switch chops the input and an inductor and capacitor smooth it into a lower voltage. |
| Battery current for a given load | Load current plus quiescent current. The input current never drops below the output current. | Roughly output power divided by efficiency and battery voltage. It can be well below the load current. |
| Efficiency limit | Cannot exceed Vout divided by Vin. | Set by switching and conduction losses. Read the curve for your load in the datasheet. |
| Output noise | No switching ripple. Low-noise parts filter ripple from the input. | Ripple at the switching frequency and its harmonics, plus fast edges on the switch node. |
| Parts and area | The regulator and its input and output capacitors. | The regulator, an inductor, input and output capacitors, and often a feedback divider. |
The second row matters most for battery life. TI defines LDO quiescent current as input current minus output current (SLVA079, as of November 2026). So a 100 mA load on an LDO draws at least 100 mA from the cell. A buck trades voltage for current, so the cell can supply less.
How efficient is each regulator on a Li-ion cell?
TI gives LDO efficiency as output current times output voltage, divided by the sum of output current and quiescent current times input voltage. Its dissipation is (Vin minus Vout) times the load current (SLVA079, as of November 2026). With quiescent current near zero, the best case is Vout divided by Vin.
A buck has no such ceiling. Its efficiency depends on the part, the inductor and the load. TI's buck power-stage note uses estimates of 85% and 90% in its worked formulas and points to the datasheet's typical characteristics for real values (SLVA477C, revised September 2026, as of November 2026).
Worked example: 3.3 V and 1.8 V rails from one cell
Take a load of 100 mA on each rail. For the cell, use 4.2 V, the default charge regulation voltage of the BQ25120A charger (datasheet SLUSD08A, as of November 2026), and 3.6 V as an example of a part-discharged cell. Your cell's discharge curve gives the real range. For the buck, assume 85% at this load, the estimate TI uses in its note. Battery current through the buck is output power divided by 0.85 and by the cell voltage.
| Rail and cell voltage | LDO efficiency ceiling | Battery current, LDO | Battery current, buck at 85% |
|---|---|---|---|
| 3.3 V at 100 mA, cell at 4.2 V | 78.6% | 100 mA | About 92 mA |
| 3.3 V at 100 mA, cell at 3.6 V | 91.7% | 100 mA | About 108 mA |
| 1.8 V at 100 mA, cell at 4.2 V | 42.9% | 100 mA | About 50 mA |
| 1.8 V at 100 mA, cell at 3.6 V | 50.0% | 100 mA | About 59 mA |
Two results follow. For the 1.8 V rail, the buck roughly halves battery current. For the 3.3 V rail, the gain is small at a full cell and reverses at 3.6 V, where the assumed 85% buck draws more than the LDO. Read the efficiency curve for your exact part and load before you decide.
Heat in a small enclosure
The 1.8 V LDO from 4.2 V dissipates (4.2 − 1.8) × 0.1 = 0.24 W. TI's note bounds dissipation by maximum junction temperature minus ambient, divided by the package's junction-to-ambient thermal resistance, and gives 285 °C/W for a 5-terminal SOT-23 as an illustration (SLVA079, as of November 2026). At that figure, 0.24 W means a rise of about 68 °C. Check the thermal resistance for your part and board in its datasheet, and measure inside the closed enclosure.
Sleep: quiescent current or efficiency?
At a few µA of sleep load, regulator overhead can be as large as the load. Quiescent current spans orders of magnitude between parts. TI's 1999 LDO note works an example with the TPS76933 at 17 µA maximum quiescent current (SLVA079). TI now lists the TPS7A02 LDO at 25 nA typical and the TPS62840 buck at 60 nA typical operating quiescent current (product pages, as of November 2026).
Low-power bucks use light-load modes to keep efficiency up. TI states 80% efficiency at 1 µA output for the TPS62840, from 3.6 V to 1.8 V. TI says the BQ25120A buck extends light-load efficiency down to 10 µA load currents, with 700 nA typical quiescent current when the buck is enabled with no load (product page and datasheet, as of November 2026). Compare parts at your real sleep load, then measure the whole board. Our guide to testing battery life honestly covers how to turn those numbers into a claim.
What happens near dropout as the battery drains?
Dropout is the input-to-output difference at which an LDO stops regulating. For an LDO with a PMOS pass element, TI models it as load current times the on-resistance of that transistor, so it grows with load (SLVA079, as of November 2026). For example, TI lists the TPS7A02 at 270 mV maximum dropout at 200 mA with a 3.3 V output (product page, as of November 2026). At that worst case, it needs about 3.57 V at its input to hold 3.3 V at full load.
A buck has the same limit in another form. When the input falls toward the output, it runs at full duty cycle and passes the battery through. TI lists a 100% duty-cycle mode for the TPS62840 (product page, as of November 2026). Either way, a 3.3 V rail from one cell sags late in the discharge. Espressif says the ESP32-C3 digital and analog pins work at 3.0 to 3.6 V (ESP32-C3 schematic checklist, as of November 2026). If your chips cannot tolerate the sag, use a lower rail or a buck-boost, and set the low-battery cutoff to match.
When does a buck converter need a low-noise LDO after it?
Radio front ends, audio codecs, microphone bias and sensor references are sensitive to supply ripple. TI's LDO note says supply rejection between 100 kHz and 1 MHz matters most when a switch-mode supply feeds the LDO, because that is where the switcher's ripple typically sits (SLVA079, as of November 2026).
PSRR varies widely between LDOs. TI lists the low-noise TPS7A20 at 60 dB at 100 kHz and 7 µVrms of noise. The nano-power TPS7A02 lists 35 dB at 100 kHz and 130 µVrms (product pages, as of November 2026). Of these two, the part with the lower quiescent current is the weaker ripple filter, so choose each rail's regulator for its job.
Light-load modes and noise
Many low-power bucks change how they switch at light load. The TPS62840 moves between PFM and PWM on its own, and TI also offers forced PWM and a STOP mode for noise-sensitive measurements (product page, as of November 2026). If your device records audio or receives RF at light load, measure the noise floor in each mode the firmware can select.
When should a PMIC replace separate regulators?
A power management IC (PMIC) puts the charger, one or more regulators and control logic in one package. TI's BQ25120A is one example made for wearables. It combines a linear charger from 5 to 300 mA, a 300 mA buck with a 1.8 V default, and a load switch or LDO rated at 100 mA, all set over I2C. It lists under 50 nA ship-mode current and a 2.5 mm by 2.5 mm package with six external components (datasheet SLUSD08A, as of November 2026). If the charger input comes from a USB-C port, USB-C power delivery for device design covers when a plain 5 V sink is enough. For cell protection and fuel gauging, see battery management systems for small devices.
The trade is flexibility. Each output has a fixed current rating. The BQ25120A datasheet lists its LDO at 100 mA under some conditions and only 10 mA with a 2.2 V input and 0.8 V output. A Wi-Fi SoC may need more peak current than a wearable PMIC rail provides. When the ESP32-C3 runs from a single power supply, Espressif recommends 3.3 V with an output current of no less than 500 mA (schematic checklist, as of November 2026).
How does the regulator change layout, EMI and the bill of materials?
A buck adds an inductor and a loop that switches hard. TI's layout article for step-down converters calls the input capacitor the most important component for reliable operation. It says to place it first after the IC, as close to the IC as manufacturing rules allow. It names the switch node as the main generator of EMI and says to keep its copper area small, with the inductor close to the IC (SLYT614, Analog Applications Journal 1Q 2015, as of November 2026). The same article says copying the evaluation board layout is the lowest-risk option when the board allows it.
For trace widths, ground returns and keep-outs around antennas, see PCB layout rules for small battery-powered devices. A buck's switching edges also show up in emissions testing, which our EMC testing guide covers.
Use this table as a starting point for each rail.
| Situation | Usual choice | Why |
|---|---|---|
| Rail sits just below the cell voltage, such as 3.3 V from one Li-ion cell, and the load is modest | LDO, if dropout holds at your lowest operating battery voltage | A buck saves little here and can draw more battery current near the bottom of the discharge. |
| Large step down, such as 1.8 V or 1.2 V from one cell, at tens of mA or more | Buck | An LDO wastes half the energy or more, and the heat stays in a small enclosure. |
| RF front end, audio codec, microphone bias or sensor reference | Low-noise LDO, often fed from a buck | Check the LDO PSRR at the buck switching frequency. |
| Wi-Fi SoC with transmit peaks | Buck or LDO sized for the peak, with the bulk capacitance the vendor asks for | For an ESP32-C3 on a single supply, Espressif recommends 3.3 V and no less than 500 mA. |
| Charger, two or more rails, ship mode and button reset needed in a ring or band | A wearable power management IC | Fewer parts and one I2C-configured device, with fixed limits on each rail. |
What should you ask your hardware team about the power tree?
Ask for a one-page power tree that shows each rail, its regulator, its loads and its current at sleep and at peak. Then ask:
- What is the lowest battery voltage at which each rail must stay in regulation, and does every LDO hold its dropout there at peak current?
- What is the battery current, not the efficiency, of each rail at the sleep load and at the active load?
- Which rails feed RF, audio or sensor references, and what PSRR does their regulator have at the buck switching frequency?
- Can each regulator supply the peak current of the radio or SoC it powers, per that vendor's hardware guide?
- How hot does each LDO get at its worst-case input voltage and load, inside the closed enclosure?
The power tree is part of the electronics work. For how that work is organized, see engineering.
Questions about LDOs and buck converters
Is an LDO always less efficient than a buck converter?
No. An LDO's efficiency approaches Vout divided by Vin, so it does well when the two are close. In the example above, a 3.3 V rail from a cell at 3.6 V draws less battery current through an LDO than through a buck at an assumed 85% (formulas from TI SLVA079, as of November 2026).
Why put an LDO after a buck converter?
The buck does the large step down and the LDO filters its ripple for a sensitive load. TI's LDO note says supply rejection between 100 kHz and 1 MHz matters most here because that is where switch-mode ripple typically sits (SLVA079, as of November 2026). Pick the LDO for PSRR at your switching frequency.
How can Duobay help with the power design of my device?
Duobay develops custom devices with a Shenzhen engineering team that covers electrical, firmware, mechanical, app and test work. The power tree is part of the electronics design, and we build samples and small batches. Source files are assigned to you at every paid gate. Initial calls are free. Tell us about your device.
If you want a power tree reviewed against your battery, radio and audio needs, tell us about your device.
Sources. All pages accessed November 6, 2026. Vendors revise datasheets, so confirm current values before you decide.
- Texas Instruments, Understanding the Terms and Definitions of LDO Voltage Regulators, SLVA079 (October 1999) (link)
- Texas Instruments, Basic Calculation of a Buck Converter's Power Stage, SLVA477C (December 2011, revised September 2026) (link)
- Texas Instruments, Five steps to a great PCB layout for a step-down converter, SLYT614, Analog Applications Journal 1Q 2015 (link)
- Texas Instruments, TPS7A02 product page (datasheet Rev. C, September 2022) (link)
- Texas Instruments, TPS7A20 product page (datasheet Rev. H, July 2024) (link)
- Texas Instruments, TPS62840 product page (datasheet Rev. D, April 2020) (link)
- Texas Instruments, bq25120A datasheet, SLUSD08A (May 2017, revised January 2018) (link)
- Espressif, ESP32-C3 Hardware Design Guidelines, Schematic Checklist (link)
Planning a device like this? Tell us what it needs to do. A Duobay engineer replies within 1 business day. Send a project brief →