Audio Amplifier Design: Power, Impedance, and Noise

Power Amplifier Fundamentals

Audio power amplifiers do something deceptively simple: take a wimpy line-level signal (usually around 1 Vrms, which is 0 dBV if you're keeping score) and drive a speaker load — typically 4 to 8 Ω — hard enough to actually move air. The real trick is pushing tens or even hundreds of watts through those voice coils while keeping distortion low and not turning your amplifier into a space heater.

Most engineers underestimate just how much current you need to move at the output stage. It's not just about voltage swing.

---

Power Output Calculations

Let's talk numbers. For a class AB amplifier, maximum output power follows this relationship:

In practice, you can't actually swing rail-to-rail. Class AB designs typically get within about 10% of the supply rails, so $V_{peak} \approx 0.9 V_{supply}$ is a reasonable estimate. Push harder and you'll start clipping ugly.

Here's a worked example with a ±18V supply (that's 36V total across the dual rails) driving an 8Ω load:

Not a ton of power, but enough for near-field monitors or a bedroom setup. If you need to figure out where your amp starts clipping or what your actual peak voltage swing is, the Amplifier Clipping calculator will save you some algebra.

Voltage gain in power amps usually sits between 26 and 34 dB — enough to take a line-level signal up to speaker-driving levels without excessive noise. Double-check your gain budget with the Power Amplifier Gain calculator before you commit to resistor values.

---

Speaker Impedance Matching

Amplifiers get rated into specific loads, and this matters more than most people think. Drop the impedance and you're asking the output stage to source more current for the same voltage swing:

Say you've got an amp rated for 100W into 8Ω. That means it's swinging $V_{peak} = \sqrt{2 \times 100 \times 8} = 40V$ peak. Now connect a 4Ω speaker. Same voltage swing, but power doubles to 200W because the current also doubles — you're now pulling 10A peak through those output transistors.

Your transistors need to handle that current or they'll let out the magic smoke. Check the datasheet's safe operating area (SOA) curves. I've seen too many fried output stages because someone assumed "it'll probably be fine."

Speaker sensitivity is the other half of the loudness equation. It's usually specified in dB/W/m — how loud the speaker plays at one meter with one watt of input. The SPL at your listening position works out to:

where $S$ is that sensitivity rating. A typical 90 dB/W/m speaker fed 100W will hit 110 dB SPL at one meter. That's loud enough to cause hearing damage pretty quickly, by the way.

Want to predict how loud your setup will actually be? Plug your numbers into the Speaker Sensitivity calculator and adjust for your actual listening distance.

---

Amplifier Classes Compared

Different amplifier topologies make different trade-offs. Here's how they stack up:

Class AB: The Standard

Class AB is the workhorse of audio amplification. You bias the output transistors with a small quiescent current — typically 10 to 50 mA per device — just enough to eliminate the crossover distortion you'd get with pure Class B. The result is efficiency way better than Class A (which wastes power like it's going out of style) while keeping distortion respectably low.

Here's something that trips people up: power dissipation in Class AB is actually lower at maximum output than it is at moderate levels. The worst-case dissipation happens around $V_{out} = 0.64 V_{supply}$, not at full power. Size your heatsinks for that condition, not for the peak power rating.

Class D: The Modern Choice

Class D amplifiers use pulse-width modulation to switch the output transistors hard on or hard off. No linear region, no massive dissipation. Typical efficiency runs 85–95%, which is why every portable Bluetooth speaker and car audio system uses Class D now.

The Class D Efficiency calculator will estimate your efficiency based on MOSFET RDS(on) and quiescent current. Switching losses matter too, but for most designs under 500 kHz switching frequency, conduction losses dominate.

The catch: you need an output LC filter to reconstruct the audio from the PWM signal. That filter adds cost, board space, and a bit of complexity to the design. You're also generating RF hash at the switching frequency, which means careful PCB layout and sometimes additional EMI filtering.

Integrated Class D chips like the TPA3116 or MAX9744 handle most of this for you — they include the output filter and have been optimized for EMI. Unless you're building something really specialized, start with an integrated solution.

---

Headphone Amplifiers

Headphone amps face a completely different design challenge. You're driving high-impedance loads (anywhere from 32Ω for consumer cans up to 600Ω for studio monitors) from relatively low supply voltages. The good news is you need way less power. The bad news is output impedance and noise become much more critical.

Let's work through an example. Say you want 110 dB SPL from a 300Ω headphone rated at 100 dB/mW sensitivity. Required power:

That's not much power, but the voltage swing is significant:

You can work this out for your specific headphones with the Headphone Power calculator. It'll tell you both the voltage and current requirements.

Output impedance matters a lot here. The classic rule of thumb: keep your amplifier's output impedance below 1/8 of the headphone impedance to avoid frequency response deviations from the impedance curve interaction. For 32Ω headphones, that means $Z_{out} < 4\Omega$. Most op-amp output stages get you there easily, but discrete designs need careful attention to this.

---

Noise Floor and SNR

The noise floor sets your dynamic range ceiling. Signal-to-noise ratio is straightforward:

A really good audio system hits 120 dB SNR. That's state of the art — it means your noise is literally one millionth the amplitude of your full-scale signal. Most consumer gear sits around 90–100 dB, which is still perfectly acceptable for most applications.

Sources of Noise

Three main noise sources will bite you:

Johnson noise comes from every resistor in your circuit. It's fundamental physics:

where $k$ is Boltzmann's constant, $T$ is temperature in Kelvin, $R$ is resistance, and $B$ is bandwidth. Higher resistance means more noise. Keep your impedances low in sensitive stages.

Op-amp input noise shows up as both voltage noise (specified in nV/√Hz) and current noise (in pA/√Hz). The voltage noise adds directly to your signal. The current noise flows through your source impedance and creates a voltage noise term proportional to that impedance. High source impedances make current noise worse. Power supply noise will couple into your signal path if you're not careful. Use proper LC filtering on the supply rails and add local bypass capacitors — a 10 μF electrolytic in parallel with a 100 nF ceramic works for most applications. The ceramic handles high-frequency transients while the electrolytic provides bulk capacitance.

The Audio SNR calculator will crunch the numbers if you know your signal and noise levels.

Op-Amp Selection for Audio

For low-noise preamp stages, these are the usual suspects:

The NE5532 is the classic choice. It's been around forever, costs almost nothing, and delivers 5 nV/√Hz input noise. Bipolar input stage means you'll see some input bias current, but it's a solid performer.

The OPA2134 uses JFET inputs for extremely low input current and very low distortion. Input noise is 8 nV/√Hz — a bit higher than the NE5532, but the JFET inputs mean almost no current noise. Great for high-impedance sources.

The LM4562 is the low-noise champion at 2.7 nV/√Hz. It's more expensive, but if you need every last dB of SNR, this is where you go. I've used these in precision measurement preamps where noise really matters.

---

Protection Circuits

Every power amplifier that's going to see real-world use needs protection. Here's what you can't skip:

DC offset protection is non-negotiable. If your output stage develops a DC offset — maybe from a failed transistor or a power-up transient — you'll pump DC current straight through the speaker voice coil. That'll either burn it out or at least shift the cone position and cause distortion. Use a relay that monitors the output and disconnects the speaker if DC offset exceeds about 50–100 mV. The relay stays open for a second or two at power-up to let things settle. Thermal protection keeps you from cooking your output devices. Mount a thermistor or temperature sensor on the heatsink. If temperature exceeds around 80°C, either reduce the gain or shut down entirely until things cool off. I've seen amplifiers without this literally desolder themselves from the board. Short-circuit protection saves you when someone plugs in a bad cable or a speaker wire touches the chassis. Implement current limiting in the output stage — if output current exceeds your safe limit, reduce drive or shut down. Some designs just use fast-blow fuses at the output, which works but means you're replacing fuses after every fault. Tweeter protection is specific to multi-way speaker systems. Put a series capacitor in line with the tweeter to create a first-order high-pass filter. This blocks low frequencies that could damage the tweeter or cause excessive excursion. Size the cap based on your tweeter's impedance and desired crossover frequency.

---

Practical Design Checklist

Before you commit to a design or order boards, walk through this:

• [ ] Calculate maximum output power from your supply voltage and load impedance — be realistic about voltage swing
• [ ] Verify your transistors or IC can handle the peak current with at least 1.5× margin
• [ ] Set your voltage gain (typically 26–34 dB for power amps) and pick resistor values that won't add excessive noise
• [ ] Check slew rate — you need enough bandwidth for full power at 20 kHz without clipping
• [ ] Size your heatsink for the worst-case dissipation, which for Class AB happens at about one-third full power, not at maximum output
• [ ] Calculate your noise floor and verify SNR exceeds 90 dB (that's −90 dBV noise floor)
• [ ] Add DC offset protection with a relay and monitoring circuit
• [ ] Decouple your supply rails locally — 10 μF bulk plus 100 nF ceramic at every IC and high-current stage

Miss any of these and you'll either blow something up or spend hours debugging why your amp sounds terrible.