Electronics

Power Supply Efficiency

Linear vs switching regulator efficiency, load dependence, and losses.

Topology comparison

TopologyTypical ηNotes
Linear (LDO)(V_out / V_in) × 0.95Dissipates difference as heat
Zener shunt< 50%Toy / reference — wastes energy
Charge pump70–85%Discrete ratios (×2, ÷2)
Buck85–95%Efficient step-down
Boost85–93%Efficient step-up
Synchronous buck90–96%Replaces diode with MOSFET
Flyback75–88%Isolated; low-mid power
LLC resonant92–97%Isolated high-efficiency

Losses

Conduction lossI² · R in switches / inductor DCR
Switching lossDepends on V × I overlap during transitions
Diode dropV_f · I — use synchronous or Schottky
Core lossHysteresis + eddy currents in magnetics
QuiescentIC operating current — matters at no-load
Copper / PCBTrace resistance; significant at high current

Load dependence

  • Efficiency is usually peaked around 50–80% load.
  • At very light load, switching and quiescent losses dominate — pulse-skipping mode helps.
  • At heavy load, conduction (I²R) losses dominate.
  • Report η at multiple loads (10%, 50%, 100%) — peak alone is misleading.

Rule of thumb

LDO efficiency= V_out / V_in (e.g. 3.3 V from 5 V ≈ 66%)
Switcher efficiency> 90% across useful load range
Heat dissipationP_diss = P_in · (1 − η) = P_out · (1/η − 1)

Worked example: 5 V to 3.3 V at 0.5 A

The load takes 3.3 V × 0.5 A = 1.65 W. A linear regulator passes the same 0.5 A from the 5 V input, so input power is 5 × 0.5 = 2.5 W. Efficiency = 1.65 / 2.5 = 66%, and the 0.85 W difference (5 − 3.3 = 1.7 V × 0.5 A) is heat in the regulator. A small SOT-223 package at roughly 60 °C/W would warm by 51 °C above ambient.

A buck converter at 90% efficiency needs 1.65 / 0.90 = 1.83 W in, and dissipates 1.65 × (1/0.90 − 1) = 0.18 W, which is a 4.6 times reduction in heat. The input current drops from 0.5 A to 1.83 / 5 = 0.37 A. For battery operation that is a 34% longer run time for the same load, since 2.5 / 1.83 = 1.36.

Practical points

  • The linear regulator is still the right answer when current is tiny, noise matters (RF, precision analog) or the dropout is small. Below about 100 mA the saved watts are negligible.
  • Light-load efficiency decides battery life. A device that sleeps most of the time spends its energy at 1% load, where a converter with 100 microamps of quiescent current may be only 50% efficient. Compare light-load curves, not the peak number.
  • Efficiency ratings include the whole chain. A laptop adapter specified at 90% at full load may be 80% at 10% load. Standards such as the US DOE and EU ErP rules define average efficiency at several load points.
  • Heat is what limits power. A 5% efficiency difference on a 200 W supply is 10 W of heat, enough to change the cooling design.
  • Input and output ripple matter too. Lower efficiency at higher switching frequency is the price of smaller inductors and capacitors.
Was this article helpful?