artnoiDecember 17, 2017
Last edited: Sep 8, 2025
This article is part of my audiophiles series of articles
Here, Ethan Winer defines audio fidelity for you:
Sound, for all the nuance and excitement it offers, only operates in two dimensions: Time and intensity.
In turn, only four parameters are needed to assess everything that affects audio fidelity.
Sound is 2D wave, and we usually only need 4 parameters to judge overall performance.
Anything not present in the source material is called noise.
If the input is silence, but there’s a hiss in the output, that hiss is noise. If the input is a 0dBFS 1kHz tone, but the output has some other components like sidebands at 5khz and 2kHz, the sum of those components constitute noise.
Noise is bad because it coexists with the main signal, degrading it by masking.
We can measure noise in dB below our main signal. Ideally, we want a system with noise level ~100dB lower than the main signal. This means that we’ll have to crank up the volume over 100dB just to hear 0-1dB of noise.
Most audio devices these days are very good in the noise department, thanks to improvements in engineering and prevalence of other noise-sensitive electronics that required engineers to fix noise issues.
Only truly shitty audio products have noise level lower than 80dB.
A flat response means that your system, when fed signals of varying frequencies but of the same amplitude, will output the same amplitude for all frequencies.
If the response is not flat, then the output signal will have different amplitude. If there’s a dip, then the amplitude at the dip frequencies will be lower (quieter). If there’s a peak or a boost in the bass range response, lower frequencies will be louder, resulting in warmer, bassier sounds.
The ideal frequency response is is a flat line across audible range. Most audio devices today are flat, except for transducers like speakers and headphones.
How audio signal may distort, i.e. incorrectly manifest. There’re many types of distortion, but we’ll focus on two: harmonic distortion and intermodulation distortion (IMD).
Harmonic distortion is when the output features some weird tones at frequencies that are multiple to the input signal. Most harmonic distortion is “2nd harmonic”, i.e. at 2x the input frequency.
Due to how our brain and ear works, harmonic distortion may sound ok or even pleasant to the ear, because it’s somehow musically related to the input signal.
We can sum up all of those harmonic distortion into Total Harmonic Distortion (THD), usually expressed as percentage e.g. 1% THD. Some measurements also show THD and noise as a parameter: THD+N, usually expressed as dBFS.
Intermodulation distortion is a different beast. IMD is a result of amplitude modulation of different input frequencies.
Unlike harmonic distortions, IMD manifests across the frequency band and sounds very bad because the noise is musically not related the input tone. IMD is, like THD, expressed as THD.
A 1% THD is much better than 1% IMD.
Since sound is naturally a longitudinal wave with time as one of its axis, large time-based errors will affect other parameters.
An example would be analog audio’s wow (cassete tape standard ~ 0.08%) or digital audio jitter (although I personally treat jitter as noise because of how fast the timing errors are).
“These four basic parameters define everything that affects audio fidelity. As powerful as it is, there’s no secret “magic” to sound, no unknown parameters that “science” hasn’t yet learned to identify, as is sometimes claimed.
The closer an audio signal is to the source across these four parameters, the higher its fidelity.” - Ethan Winer
We want our audio gear to have the best fidelity possible, so we focus on minimizing noise, distortion, and time-based errors, while keeping the response flat.