Buck or LDO: Choosing a Power Topology

Two numbers decide most of it: the voltage difference and the load current. Their product is the heat, and the heat decides the topology.

Quick Answer

(Vin − Vout) × Iout below ~1 W: an LDO is simpler, quieter and cheaper.
Above ~1 W: a buck converter, because the heat has nowhere to go.
Noise-critical rail: two-stage — buck to an intermediate rail, LDO for the last drop.
Check the headroom: an LDO needs a minimum Vin − Vout; below it, nothing works.

1. Efficiency decides first

Efficiency of an LDO versus input voltage compared with a buck converter — Figure 1 — LDO efficiency is Vout/Vin; a buck stays near 90 % across the range.

At 5 V input and 3.3 V output, an LDO is about 66 % efficient — acceptable. At 12 V it is 27.5 %, and at 24 V it is under 14 %. The rest becomes heat in a package that was never designed to dissipate it.

2. Heat decides second

Power loss versus output current for LDOs at 5, 9 and 12 V input and a buck converter — Figure 2 — Loss grows linearly with current for the LDO; the buck stays low.

P = (Vin - Vout) x Iout

12 V -> 3.3 V at 0.5 A: 4.35 W
5 V  -> 3.3 V at 0.5 A: 0.85 W
12 V -> 3.3 V at 0.1 A: 0.87 W

The 1 W rule is a starting threshold, not a physical law: it corresponds to the point where a small SMD package starts to need copper, vias and a derating table. Below it, an LDO is usually the right engineering choice; above it, the buck pays for itself in heat.

Rail	LDO is fine up to	Above that
5 V → 3.3 V	≈ 0.5 A	Buck
9 V → 5 V	≈ 0.25 A	Buck
12 V → 3.3 V	≈ 0.11 A	Buck
24 V → 12 V	≈ 0.08 A	Buck

3. What an LDO buys you

Property	LDO	Buck
Output noise	Low, no switching	Ripple at fsw
EMI	None	Requires layout care
External parts	Two capacitors	Inductor plus capacitors
Transient response	Fast, limited by bandwidth	Good, with more design
Efficiency at high Vin	Poor	Good

For an analog front end, a sensor reference or a PLL supply, the LDO's quiet output is worth more than the efficiency. For a digital rail feeding an MCU and a display, the buck's efficiency usually wins.

4. The decision map

Load current versus input-output difference with a 1 W boundary — Figure 3 — Below the curve, an LDO; above it, a buck. The curve is P = 1 W.

The curve is just (Vin − Vout) × I = 1 W. At 12 V to 3.3 V it crosses at 115 mA; at 5 V to 3.3 V it crosses at 590 mA. That single line answers most of the question before cost or noise enter.

5. Cost and area

Relative BOM cost and board area for LDO, buck and two-stage — Figure 4 — The buck costs more parts; the difference is small next to the thermal design it replaces.

A buck adds an inductor, more capacitors and layout work. At low current, that is more than the LDO costs. At higher current, the LDO's copper and thermal vias can cost as much as the inductor — and the buck removes the heat rather than spreading it.

6. Two-stage

Two-stage power design with a buck to an intermediate rail and an LDO for the final output — Figure 5 — Buck for the drop, LDO for the clean output. The LDO now dissipates a small, known amount.

Example: 12 V to 3.3 V at 0.5 A. Buck 12→5 V at 90 %, then LDO 5→3.3 V. Total loss is about 0.18 W in the buck plus 0.85 W in the LDO — still 1 W, so the LDO still needs copper. Move the intermediate rail to 3.6 V with a tighter buck and the LDO drop becomes 0.15 W.

Intermediate rail	LDO loss at 0.5 A	Comment
5.0 V	0.85 W	Still needs copper
4.0 V	0.35 W	Check LDO headroom
3.6 V	0.15 W	Tight but common for 3.3 V

The lower the intermediate rail, the less the LDO dissipates — but the more the buck has to regulate accurately, and the less headroom the LDO has when the load steps. 4.0 V is usually the practical compromise.

6b. Checklist

Check	Pass
(Vin − Vout) × I	Below 1 W for an LDO
LDO headroom	Vin − Vout above dropout at min temperature
Thermal	Tj below 100 °C at max ambient
Noise	Buck ripple acceptable, or two-stage
Enclosure	Ambient measured inside, not on the bench

7. FAQ

Can an LDO replace a buck at 1 A?

Only if the voltage difference is small. 5 V to 3.3 V at 1 A is 1.7 W — too much for most SMD LDOs without significant copper.

Is a buck always noisier?

At the switching frequency, yes. With good layout, a well-chosen inductor and output filtering, the ripple can be tens of millivolts — fine for digital, not always for analog.

Why not always use a buck?

Cost, complexity, EMI and board area. For a 20 mA load, an LDO is the better engineering answer.

What about headroom?

An LDO needs a minimum Vin − Vout; below it, the output follows the input minus dropout. Check the datasheet, especially at low temperature.

Can I parallel LDOs?

Not reliably; they do not share current. Use a single part rated for the current, or a buck.

Does switching frequency matter for the choice?

It sets the inductor size and the ripple. Higher frequency shrinks the inductor but increases switching loss and EMI.

8. Sources

Linear regulator and buck converter datasheets — dropout, efficiency curves, thermal resistance.
Application notes on LDO vs switching regulator selection (TI, Analog Devices).
IEC/CISPR emission standards for the EMI side of the comparison.