Somewhere on nearly every board, a voltage has to move. A 12 V rail needs to become 5 V for a microcontroller, or a clean 3.3 V has to drop to 1.8 V for an ADC that will not tolerate more than a few millivolts of noise on its supply. Two families of parts handle this job, and they behave so differently that choosing the wrong one costs you either efficiency or signal quality.

The decision rarely needs to be difficult. A handful of numbers — input voltage, output voltage, load current, ripple budget, and board area — will point you to one answer almost every time.

The Core Trade-Off in One Equation

A linear regulator, whether a classic 7805 or a modern low-dropout (LDO) part, works by holding a pass transistor in its linear region and continuously adjusting it to absorb the difference between input and output. Current in equals current out, so its efficiency is roughly Vout divided by Vin. The power wasted as heat is simply:

Pdiss = (Vin − Vout) × Iout

A switching regulator takes the opposite approach. It chops the input, stores energy in an inductor and a capacitor, and delivers it to the load in pulses at tens of kilohertz to a few megahertz. Because the pass element is either fully on or fully off most of the time, losses stay low even when the input is far above the output. Efficiencies in the mid-80s to mid-90s percent are routine for a well-designed buck converter.

That single difference drives almost everything else: heat, noise, complexity, and cost.

Where Linear Regulators Still Win

Linear parts get dismissed as inefficient relics. That is a mistake. In the right spot, they are the better engineering choice.

  • Small voltage drops at modest current. Dropping 3.3 V to 2.5 V at 200 mA dissipates only 160 mW. A switcher here adds an inductor, more capacitors, and switching noise for a few points of efficiency.
  • Noise-sensitive analog rails. Preamplifiers, voltage references, VCOs, and precision ADCs care about supply cleanliness more than they care about a few hundred milliwatts. A good LDO can reject 60 dB or more of input ripple at low frequencies and produces no switching harmonics of its own.
  • Always-on standby rails. Many LDOs draw quiescent currents in the microamp range, which matters in a device that sleeps for weeks on a coin cell.
  • First-pass simplicity. Two capacitors and the part itself. Nothing to compensate, no loop stability to verify on the bench, no EMI to think about.

The catch is dropout voltage. An LDO needs some minimum headroom — often a few hundred millivolts, sometimes less than 100 mV for specialized parts — so check the dropout specification at your actual load current, not at the headline number on the front page of the datasheet.

When a Switching Regulator Is the Only Sensible Choice

Reach for a switcher when any of the following is true:

  • The input-to-output ratio is large, or the load current is high, or both.
  • The input varies widely — a 6 V to 18 V wall adapter, or a battery that swings from 8.4 V down to 5.5 V.
  • You need to step up. A linear regulator cannot produce an output above its input, and charge pumps are only practical at low currents.
  • You need an inverting rail, or a rail that tracks a battery voltage with high efficiency.
  • Thermal limits make a linear solution impossible, which is more common than people expect.

Heat: The Number to Check Before You Commit

Take a real case. You have 12 V available and need 5 V at 500 mA. A linear regulator dissipates (12 − 5) × 0.5 = 3.5 W. That is not a rounding error; it is a small space heater inside your enclosure.

Now check the package. A tiny SOT-23 part might have a junction-to-ambient thermal resistance of 200 °C/W or more, so 3.5 W would push the junction hundreds of degrees above ambient. Even a DPAK soldered to a generous copper pour, sitting near 40–50 °C/W, gives a rise of well over 100 °C. In a sealed box at 40 °C ambient, you are already past the limit.

The estimate is quick: Tj ≈ Tambient + Pdiss × RθJA. Keep the result comfortably below the maximum junction temperature in the datasheet, with margin for a hot day and a poorly ventilated enclosure. If the arithmetic fails, the answer is a switcher — or a much larger package with a heatsink, which is usually worse than simply using a buck converter.

Noise, Ripple, and Sensitive Analog Circuits

Switching regulators are not inherently dirty, but they demand respect. Expect output ripple in the tens of millivolts, plus harmonics that extend well into the megahertz range and a magnetic field around the inductor. If your signal chain has gain at those frequencies, you will see it.

Practical ways to clean up a switching rail

  1. Keep the hot loop small. The input capacitor, the high-side switch, and the ground return form a loop that radiates. Minimize its area on the layout and place the inductor close to the switch node.
  2. Add a second-stage filter. A ferrite bead and a ceramic capacitor, or a small LC network, knocks down high-frequency content. Damping matters; an undamped LC can ring and make things worse.
  3. Cascade an LDO. Set the switcher slightly above the target voltage — say 5.4 V — and let an LDO produce the final 5.0 V. You keep most of the efficiency and hand the noise problem to a part designed to reject it.
  4. Separate the grounds thoughtfully. Return the noisy switching currents on their own path back to the input capacitor, away from analog ground.

Note that an LDO's ripple rejection falls as frequency rises, so cascading helps most in the band where the regulator still has gain. Above a few megahertz, filtering and layout do the work.

A Decision Checklist

  1. Write down Vin range, Vout, and maximum load current.
  2. Compute the linear dissipation and check the junction temperature with your actual package and copper area.
  3. Estimate switcher efficiency and compare total loss, including inductor and switching losses.
  4. Ask what ripple the load actually tolerates. Digital cores are forgiving; a 24-bit ADC front end is not.
  5. Count the cost of complexity: board area, inductor height, EMI testing, and the time to debug a layout that misbehaves.
  6. If both options are close, pick the linear part for simplicity, or a switcher followed by an LDO for a clean compromise.

Mixing Both on One Board

Most well-designed systems use both. A buck converter drops the main input down to an intermediate rail with high efficiency, then LDOs fan that rail out to the analog sections, the PLL, and any rail that needs to be quiet. The switcher does the heavy lifting; the LDOs do the tidying up.

Keep the two apart physically. Put the inductors and their input capacitors near the board edge or in a corner, keep switching nodes short and away from sensitive traces, and give the LDOs a clean input with solid decoupling. When you prototype, measure the output ripple with a proper short-ground-tip probe before you trust it — long scope ground leads pick up more noise than the regulator produces. Then load the rail to its worst case and watch the temperature for ten minutes. Between those two checks, you will know whether your choice was right.

Photo: blickpixel / Pixabay