Voltage Regulators: 7805 vs Buck vs Boost, Compared

Every student electronics project starts with a power supply and ends with a power problem. A 7805 that ran cool on the bench overheats in the enclosure. An Arduino that worked on USB resets every time the WiFi transmits. A sensor that should read 0 to 5V reads garbage because its 5V rail sags to 4.2V under load.

These are not mysteries. They are voltage-regulator choices made without the numbers. A regulator is not a magic black box that turns 12V into 5V — it is a device with an input range, an efficiency curve, a thermal limit, a dropout voltage and a ripple spec, and your project fails the moment any one of them is exceeded.

This guide covers the three regulators students actually use — the 7805-style linear regulator, the buck (step-down) converter and the boost (step-up) converter — how each works, the math for picking one, and the thermal and layout rules that keep the 5V rail where it should be.

The three families, honestly compared

	Linear (7805, AMS1117, LM2940)	Buck (step-down switching, e.g. LM2596 module)	Boost (step-up switching, e.g. MT3608 module)
What it does	Drops voltage by burning the excess as heat	Switches fast and stores energy in an inductor to step down efficiently	Steps voltage UP from a lower source
Efficiency	Vout/Vin — e.g. 5V from 12V = 42% efficient	75–95%, often ~90% in the middle of its range	75–90%
Heat	The whole problem: heat = (Vin − Vout) × current	Much less; inductor and switch warm slightly	Moderate
Output noise	Very clean — low ripple, no switching spikes	Switching ripple (~tens of mV), occasional spikes	Switching ripple, higher than buck
Cost/effort	Cheapest, simplest, no inductor	Cheap modules; needs inductor selection if designing	Cheap modules
Typical student use	5V from 7–9V for small logic loads; clean analog rails	5V from 12V/24V; battery-powered systems; motor supplies	5V from a single Li-ion cell (3.0–4.2V); portable gadgets

Linear regulators: the heat equation you must do

A linear regulator's power loss is elementary and unforgiving:

P_dissipated = (Vin − Vout) × I

Drop 12V to 5V at 1A and the regulator burns 7W — more heat than the entire circuit it powers. A bare 7805 in a TO-220 package has a junction-to-ambient thermal resistance of about 65°C/W, so 7W pushes it roughly 455°C above ambient, which means it will go into thermal shutdown (or die) in minutes.

Do this before every design:

Compute Pd = (Vin − Vout) × I for your worst-case continuous current.
With no heatsink, Pd must stay below about 1W for a TO-220 (2W if you accept the case getting uncomfortably hot). Small SOT-223 regulators like the AMS1117 (common on dev boards) should stay below ~0.8W without a heatsink.
If the number is bigger, your options are: lower Vin (use a 7.5V wall adapter instead of 12V), add a proper heatsink (recompute with the heatsink's °C/W added to the package's junction-to-case value), or switch to a buck converter.

Dropout voltage is the second linear-regulator gotcha: the minimum Vin−Vout gap the regulator needs. A 7805 needs ~2V of headroom (7V in for 5V out); a low-dropout (LDO) part like the AMS1117 or LM2940 needs only ~1V or less. The classic student failure: powering a 5V 7805 from a 6V battery pack and wondering why the output is 4.2V.

Capacitor rule for linear regulators: put ~10µF electrolytic on the input (more if the supply is far away or noisy) and the datasheet's recommended capacitor on the output — usually 0.1µF ceramic close to the pins plus a 10µF for the AMS1117. Output caps on LDOs are a stability requirement, not decoration: the wrong ESR can make some LDOs oscillate.

Buck converters: when efficiency actually matters

Switch a buck converter when your (Vin − Vout) × I exceeds what a linear regulator can comfortably shed, or when you run on batteries and every milliamp-hour counts. Examples:

12V lead-acid/solar to 5V for an ESP32 + sensors at 500mA: linear wastes 3.5W; a buck wastes ~0.3W.
24V industrial supply to 5V for a logic board: linear is a 19W heater; buck handles it cold.
Battery node that must last weeks: the wasted power is lost runtime.

What to check on a cheap LM2596-style module before trusting it:

Set the voltage with no load attached, with a multimeter — modules ship set to arbitrary voltages. Turn the potentiometer while watching the meter.
Ripple: expect 30–100mV of switching ripple. For analog sensors and audio, add an LC filter or a post-regulator LDO (buck to 5.5V, LDO to clean 5V — a common, legitimate pattern).
Input capacitor: if the module's input leads are long or thin, ripple current heats the input cap; keep wires short and thick.
Switching frequency noise: keep the module away from sensitive analog front-ends and radio antennas where possible.

Boost converters: stepping up (and the trap)

Boosts are for single-cell portable builds: 3.7V Li-ion to 5V for a display and microcontroller. The trap students fall into is current: a boost delivering 5V at 1A from a 3.7V cell draws roughly (5 × 1)/(3.7 × 0.85) ≈ 1.6A from the battery — inductor saturation current and the battery's discharge rating both matter. Size the cell and the inductor for the input current, not the output current.

Also note: a boost cannot step down. If your Li-ion charges to 4.2V and you boost to 5V, fine — but if you ever connect a 6V source, the output follows the input straight through the diode. Never assume a boost module limits its output voltage against overvoltage inputs.

The 10-minute regulator selection recipe

List your rails: voltages, worst-case continuous current per rail, and which rails are analog/radio-sensitive (want clean power).
List your sources: wall adapter, USB, Li-ion pack, lead-acid, solar — with their voltage ranges (a "12V" lead-acid is 10.5–14.4V).
For each rail: if (Vin − Vout) × I < 1W and the source never dips below Vout + dropout, use a linear regulator (simple, clean). Otherwise use a buck.
Check dropout at the worst-case source voltage, not the nominal one. A regulator that works on a fresh battery fails on a discharged one.
Add protection: a fuse or polyfuse on the input, a reverse-polarity diode or MOSFET where students will handle the connectors, and a TVS diode if the supply sees inductive spikes (motors, relays nearby).
Decouple every IC: 100nF ceramic at each chip's supply pins, plus bulk electrolytic (47–470µF) where current pulses happen (radio transmit, motor start).
Test under full load before boxing it up: measure the output voltage while the heaviest load runs, and touch-test the regulator after 10 minutes. If you can't hold your finger on it comfortably, it needs a heatsink or a redesign.

Five mistakes that kill student power supplies

Feeding 12V into the Arduino's barrel jack while drawing 800mA through the onboard regulator — the board's tiny AMS1117 cooks. The barrel jack is fine for the microcontroller alone; off-board loads get their own regulator.
No common ground between supplies — two regulators, two "5V" rails, no shared ground reference: sensor readings float and communication fails. Tie the grounds together (single-point, at the power entry).
Undersized input wires to a buck module — thin long wires plus switching current equals voltage sag and ripple the converter can't fix.
Boosting a dead battery — a boost at 3.0V input pulling 1.6A accelerates the cell toward over-discharge; use a low-voltage cutoff or the battery's protection board.
Ignoring inrush: big bulk capacitors on the output draw a start-up surge that can trip polyfuses or reset the source. Stage the capacitance or accept a slower start.

FAQ

Can I just use the 5V pin of a USB charger for everything?
For logic and small sensors, yes — a decent 5V/2A charger is a fine bench supply. For motors, GSM/WiFi transmit bursts, or anything drawing over ~1.5A total, give the heavy loads their own regulated feed with bulk capacitance, and keep the microcontroller's rail clean.

Is the AMS1117 on my dev board enough?
For the board itself plus a few sensors, yes — that is what it is designed for. It is not a power supply for peripherals. Anything with a motor, a display backlight, or a radio beyond the onboard one gets its own regulator.

Buck or linear for an analog sensor rail?
Linear, or buck followed by an LDO. Switching ripple shows up directly in ADC readings as noise you cannot filter in software — a clean 3.3V LDO fed from a buck's 5V output is the standard professional answer.

Why does my 7805 output drift when it gets hot?
Thermal drift plus the beginnings of thermal shutdown. Measure it: if the case is above ~70°C, you are outside the comfort zone. Reduce Vin, add a heatsink, or move to a buck.

Can I parallel two 7805s for more current?
Not directly — their outputs fight. Use a single higher-current regulator (LM338, buck module) or a proper current-sharing design. Paralleling linear regulators without ballast resistors is a classic way to burn one of them.

Limitations

This guide covers the 95% case: low-voltage DC rails for student projects. It does not cover mains AC power supply design (transformer, rectification, safety isolation — treat mains with the respect it demands), multi-phase converters, battery charging circuits, or EMI compliance testing. High-current designs (tens of amps) need proper inductor and MOSFET selection with thermal simulation, not module shopping. Always verify your chosen regulator's datasheet limits — dropout, thermal resistance, and maximum input voltage — against your actual worst case.

Where to go from here

Power Supply Design: Linear vs Switching Explained — the deeper theory behind this selection guide.
Decoupling Capacitors: Why and Where — the 100nF rule and bulk capacitance, in detail.
How to Use a Multimeter for Circuit Debugging — measure your rails under load instead of guessing.
PCB Layout: 10 Rules That Prevent Failures — power routing, ground planes and thermal reliefs when you commit the supply to copper.
More electronics topics in the Electronics / E&TC branch hub.