LESSON 08 · AMPLIFIERS · EXTRA SOUND THAT WASN'T IN THE RECORDING

What do amplifier distortion ratings actually tell you?

See distortion curves, understand clipping, and explore what audiophiles mean by second- and third-order harmonics.

EXAMPLE SPECIFICATION · AMPLIFIERS 0.09% THD · at 1 kHz

One fictional THD reading at one test frequency. See this exact 1 kHz value plotted below ↓ The output power and load are not specified by this example.

START WITH THE SPEC SHEET

Three distortion ratings. Three very different kinds of information.

These are common ways manufacturers express measurements. The graphs show exactly what each fictional statement can tell you, and just as importantly, what it cannot.

01 · A FULL-RANGE CLAIM?0.005% THD20 Hz–20 kHz 0.0010.010.11201k20kTHD (%)Frequency (Hz)0.005%?

Only if this number applies across the stated test-note range, the dashed line represents a maximum claimed distortion, not an actual measured curve.

Without the output power, load and exact wording, the claim is incomplete.

02 · ONE TEST NOTE0.09% THDat 1 kHz 0.0010.010.11201k20kTHD (%)Frequency (Hz)0.09%??

The dot shows a single reported test result. Nothing in the stated number tells us how the amplifier behaves at other frequencies.

The result still needs a power level and load to compare fairly.

03 · RATED POWER WITH CONDITIONS80 watts / channel 8 Ω · 20 Hz–20 kHz · ≤0.08% THD · both channels driven 0.0010.010.11201k20kTHD (%)Frequency (Hz)≤0.08%

The dashed line illustrates the stated distortion ceiling of this power rating, assuming it applies across the frequency band. It does not reveal the true measured curve.

More complete conditions make a rating more useful, not necessarily the better-sounding amplifier.

Read these graphs: Left is deep notes, right is higher notes; up is more harmonic distortion. Dashed lines are stated limits, not actual measurement traces. The dot is a single reported point. These invented examples do not represent real amplifiers. Why the test conditions matter ↓

LOOK AT THE GRAPH

See how distortion changes when an amplifier is pushed

Follow the examples in order: first, low and flat distortion; next, an earlier rise with increasing output; finally, two different mixtures of added harmonics even when the overall distortion figure is identical. Moving right means more output power, and upward means more distortion.

FOLLOW THE THREE STEPSLow and flat → early rise → different harmonics
QUIETER OUTPUTMORE OUTPUTNEAR THE LIMIT

Higher on the graph means the test detected more extra sound. This doesn't by itself tell us if you'd hear it.

0.0010.010.111011050100150 Output power (watts/ch) Added sound (THD %)
Showing: Stays clean longer
How to read it: Amplifier output power (watts) · illustrative 8 Ω test. The dashed curve is the first example for comparison. The harmonic-mix example deliberately has the same THD curve, while the inset shows different harmonic combinations. All examples are fictional. The numbers are graph labels, not a score.

Low and flat

Low added sound until near its limit

This example keeps distortion very low through most of its power range. The line only climbs steeply close to its limit, where clipping or limiting can begin. That sharp climb can indicate clipping or limiting.

You may hear little or no difference at comfortable levels, but pushing beyond the clean range can make loud passages rough or harsh.

A low THD curve doesn't prove two amplifiers will sound identical. It is one piece of evidence.

ANOTHER VERY COMMON AMPLIFIER MEASUREMENT

THD versus THD+N: what does the “+N” change?

Imagine feeding an amplifier one pure test note. If extra tones come out at two, three or more times that note's frequency, the amplifier has added harmonic distortion. THD adds up those unwanted harmonic components. THD+N also counts background noise, such as hiss or hum. The “+N” matters, so don't compare the two figures as though they are identical.

THD or THD+N (%) 0.0010.010.110.010.1110100 Output power into 8 Ω (watts, logarithmic scale) Noise is a larger fraction
THD: added harmonicsTHD+N: harmonics + noise

Original hypothetical curves, not real amplifier measurements. In this simplified example, noise remains fixed at 80 µV RMS, the test note is 1 kHz, the load is an ideal 8 Ω resistor, and harmonic distortion rises only near the output limit.

01 · QUIET OUTPUT

Noise takes up a bigger share

When music is playing quietly, even a small, steady background hiss can be a comparatively large part of the weak test signal. The THD+N percentage can look higher, even though the amount of hiss hasn't increased.

02 · MODERATE OUTPUT

The two curves move closer

As the signal gets stronger, the same amount of noise becomes a smaller percentage of it. THD+N may fall toward the THD value even though the amplifier hasn't suddenly become quieter in absolute terms.

03 · NEAR ITS LIMIT

Harmonic distortion can take over

As the amplifier approaches clipping or another output limit, distortion may rise sharply. Both readings can then increase together because harmonics, rather than background noise, dominate.

What audiophiles say

“THD+N is the more honest number.” It often provides a useful overall check of unwanted output, but one percentage still doesn't identify whether the result comes from harmonics, hum, random hiss, or some combination.

What the evidence tells us

The measured noise depends on the measurement bandwidth and filters used. A reading from 20 Hz–20 kHz cannot automatically be compared with one using a much wider bandwidth. A cleaner-looking THD figure isn't proof that noise is low, and a higher THD+N figure isn't automatically audible.

When comparing equipment: check whether the manufacturer reports THD or THD+N, along with test frequency, output level, load, analyzer bandwidth and any filters or weighting. Look for a THD+N versus power curve if you want to know how the result changes with listening level.
Why do the percentages behave this way? Technical detail

In a simplified case, when harmonic components and background noise are uncorrelated, THD+N is roughly the square root of THD² plus the squared noise-to-signal ratio, with both expressed as percentages. The same fixed output noise voltage represents a larger fraction of a lower signal voltage. Real THD+N measurements may also include other residual signals such as hum, and analyzer details affect what counts. This graph is for learning, not a prediction of an actual amplifier.

A COMMON SHOPPING QUESTION

What those ratings still leave unanswered

The graph comparisons above are not interchangeable tests. A single result, a maximum allowed percentage, and a rated-power condition all answer different questions. What else would you need to know to make a fair comparison?

These figures are illustrative, not claims about real amplifiers. The 0.005% figure is 18 times smaller numerically than 0.09%. That does not mean the music will sound 18 times cleaner, or that either level of distortion would be audible under normal conditions.

The more complete example: 80 watts per channel, 8 Ω, 20 Hz–20 kHz rated power range, ≤0.08% THD, both channels driven. This gives you a test load, output rating, frequency range and distortion ceiling, but still not a measured distortion sweep.

WHAT A 1 kHz NUMBER CANNOT TELL YOU

Two completely different curves could share the same 0.09% figure.

Both invented curves pass through 0.09% at 1 kHz. The blue example stays constant, while the orange example worsens toward the deepest and highest notes. Neither can be inferred from the single reported 1 kHz measurement.

0.0010.010.11201k20kTHD (%)Frequency (Hz)same 1 kHz point
Example: stays flatExample: rises at the extremes

Important: A line showing a maximum allowed distortion is not the same thing as a plotted measurement. A one-frequency specification reveals exactly one point. A power-rating statement reveals its stated test conditions and limit, not the amplifier's full distortion spectrum.

Four things to check before comparing distortion claims

  1. How hard was the amplifier working? Was the test at 1 watt, 20 watts, rated output, or near clipping?
  2. Which notes were used? A single 1 kHz note tells you about that test point. A properly specified range can tell you more, but only if you know how it was tested.
  3. What was connected? At the same voltage, a 4 Ω test load draws twice the current of 8 Ω. Compare the same load and number of driven channels.
  4. What's included in the number? THD reports harmonics. THD+N includes noise as well. Measurement bandwidth and filtering also affect the reading.
One subtle trap: “20 Hz–20 kHz” may describe the full-power frequency range rather than a claim that every frequency was tested for exactly the same distortion value. Read the manufacturer’s full specification line, not just its smallest percentage.

FIRST, THE EVERYDAY EXPLANATION

Imagine a guitar playing a clear note. An amplifier is meant to make that electrical signal stronger, but it can also add tiny extra tones that weren't present at that level in the original.

A THD percentage reports added harmonic tones; THD+N includes noise as well. Either reading may change as the amplifier works harder, and two amplifiers with similar total values can add different kinds of extra sound. The power, speaker load and test note are important context.

Near its limits, an amplifier may stop reproducing the peaks of the signal cleanly. When the top and bottom of the waveform are restricted, this is called clipping, and the character of the added sound can change dramatically.

CONNECT THE GRAPH TO THE MUSIC

What could you actually hear?

At extremely low levels, different harmonic mixtures may not be noticeable in music at all. At greater levels, especially with simple sustained tones, coloration may become easier to hear.

The tone that seems “smooth” to one listener might seem dull or altered to another. It is safer to compare actual recordings at matched levels than to infer a listening experience from a single THD figure.

WHAT AUDIOPHILES SAY

“Second harmonic is warm and musical; third harmonic is harsh.” That familiar explanation contains a useful starting idea, but it isn't a universal rule. Both added tones are musically related to the original note, and audibility depends on more than the order.

WHAT THE EVIDENCE TELLS US

A second harmonic of a 200 Hz note appears at 400 Hz; a third appears at 600 Hz. Their measured levels and combinations matter. The sum called THD loses information about which harmonics dominate, how the distortion changes with frequency and power, and what happens with several musical notes playing together. Tube or transistor labels do not supply that missing evidence.

ONE NUMBER, TWO POSSIBLE SOUNDS

Second versus third harmonic: what changes?

Play a clean test note at 200 Hz. If the amplifier adds a second harmonic, an extra tone appears one octave higher at 400 Hz. A third harmonic appears at 600 Hz. The percentage called THD rolls different extra tones into one figure.

Example A: mostly second harmonic

400 Hz · H2
600 Hz · H3

Example B: mostly third harmonic

400 Hz · H2
600 Hz · H3

The bar lengths show relative emphasis within each invented distortion pattern, not measured sound levels. In this illustration the combined distortion amount could be matched even though the harmonic mixtures differ.

A common claim is that H2 creates warmth while H3 creates hardness. That is not a universal rule: the levels, listening material, other distortion components and audible thresholds all matter. Music can also hide low-level harmonics. Purposefully added saturation in music production is a different choice from an unwanted hi-fi playback error.

Two real bench tests and what they tell you

In Stereophile’s Yamaha A-S3200 test, the amp reached 110 W per channel into 8 Ω with both channels driven at 1% THD+N. Its Benchmark AHB2 test reported clipping at 108 W into 8 Ω with only the left channel driven and a different procedure that summed harmonic distortion separately from noise. The two results answer different questions. A fair comparison would use the same load, frequency, driven channels, output level, bandwidth and distortion metric.

Three things to look for on the original plots

  • Read the left side at realistic output. THD+N combines distortion and noise. At very low output a roughly fixed amount of noise can make the percentage rise as the signal gets smaller. Look for residual noise measurements if you’re concerned about hiss on sensitive speakers.
  • Look for the steep rise near the right edge. The rise tells you where the amplifier approaches its output limit in that test. If loud peaks stay below that region on both amps, the higher clipping number alone gives little reason to expect a different sound on those peaks.
  • Check the spectrum and the analyzer’s floor. Stereophile found predominantly third-harmonic distortion in one Yamaha test. The Benchmark’s below-clipping distortion was close enough to the analyzer’s own residual that the lab used a different procedure to examine it. A harmonic spectrum shows which tones were added and at what level. Whether you can hear them depends on their level, the music and the listening conditions.

What makes an amplifier worth more to you?

Yamaha’s A-S3200 illustrates an integrated amp with source and volume controls, a phono stage and a dedicated power-amp input. The Benchmark AHB2 illustrates an engineering focus on very low noise and distortion in a dedicated power amp. If a tube amp is described as “warm,” check its harmonic spectrum and its response into a speaker-like load. Stereophile’s Cary SLM-100 test shows how a high output impedance can change frequency response with the load. That tonal shift and the added harmonics are separate effects.

Pay for the difference that solves a problem in your system: less audible hiss, clean peaks on a demanding speaker, or the inputs and routing you need. See how integrated amps, receivers and separates route the signal.

TAKE THIS TO THE SHOP

Before you decide what to buy

  1. Compare figures only after checking output power, speaker load, test frequency or range, whether they report THD or THD+N, and the measurement bandwidth or filtering.
  2. Look for distortion-versus-power and distortion-versus-frequency curves, when available, instead of judging an amplifier by its smallest percentage.
  3. Treat claims about warm second-order or harsh third-order distortion as hypotheses unless an actual level and spectrum are provided.

What else can change the outcome?

Output level, frequency, speaker load, recording complexity, harmonic levels, intermodulation, amplifier circuit and feedback, ear sensitivity, and any distortion intentionally added during production.

Want the engineering explanation?Advanced detail

THD is a ratio of harmonic components to the fundamental, expressed as a percentage. THD+N also includes noise in the defined measurement bandwidth. A lone percentage doesn't specify signal frequency, power level, load impedance or measurement bandwidth, and a power-band rating is not equivalent to a complete THD-versus-frequency plot. Harmonic order is the integer multiplier of the original test tone: H2=2×, H3=3×. A conventional THD-versus-power graph commonly uses a logarithmic vertical axis; the example displayed here is entirely synthetic.