What is SNR and Why it matters?
Signal-to-Noise Ratio (SNR) is a measure of how strong a desired signal (like music) is relative to background noise. A higher SNR means a signal stands out more clearly against noise, which generally improves clarity and intelligibility.
In audio systems, SNR also relates to dynamic range, the difference between the quietest and loudest sound you can reproduce. It’s expressed in decibels (dB). The ear’s dynamic hearing range from threshold of hearing to threshold of pain spans roughly 140 dB, though not all at once due to auditory compression mechanisms in the ear.
In speech and communication research, SNR plays a critical role in how well listeners can understand spoken words in noise. Psychoacoustic studies measuring just-noticeable differences (JNDs) in SNR find that normal listeners can detect relatively small changes in SNR in speech contexts, on the order of a few dB.
| Reference SNR (dB) | Average SNR JND (dB) | 95% Confidence Interval |
| 0 dB | ~2.9 dB | 2.5 - 3.3 |
| +6 dB | ~3.5 dB | 3.1 - 4.0 |
This means people generally need about a ~3 dB change in SNR before they reliably perceive a difference in clarity of speech in noise.
In practice, this means that even if a system has a high technical SNR spec, the audibility benefit in everyday environments can be limited by masking and cognitive factors (how the brain separates speech from noise, attention, etc.). So while an SNR of 80–90 dB might be technically excellent, real-world perceived improvement may be smaller once masking and room noise are considered. A high technical SNR (e.g., 110+ dB) may not sound significantly clearer than 90–95 dB SNR in real music listening contexts, because the ear & brain mask background noise.
What is THD and When you hear it?
Total Harmonic Distortion (THD) quantifies how much a signal’s harmonic content deviates from a pure original. In ideal systems, harmonics only reflect what’s musically meaningful; but distortion introduces additional harmonics that weren’t in the source. THD is usually expressed as a percentage, although they are converted from dBc.
A common belief is that humans cannot hear THD below some fixed percentage, for example: 1%. But psychoacoustic research shows this idea is oversimplified. The detectability of distortion depends on:
- Signal type (pure tones vs. music)
- Frequency
- Masking and context
- Listening environment
Rigorous psychoacoustic studies demonstrate that distortion detection thresholds increase (i.e., become less detectable) when masking sounds like background noise or reverberation are present, and when signals are more complex than pure test tones.
Researchers also note that conventional THD measurements (based on single tones) do not always correlate well with subjective perception of distortion in music or real audio content. Complex signals have masking patterns that can hide or reveal harmonic distortion unpredictably. Modern perceptual models (such as ITU-based PEAQ standards) therefore combine multiple psychoacoustic cues, not just THD, to better correlate with human judgment.
High-order harmonics as low as −60 dB relative to the fundamental can be audible if they fall into sensitive regions of the ear (e.g., 2–5 kHz) and are not masked by the fundamental tone. Even systems with objectively higher THD can sound subjectively better than systems with lower THD depending on the type of distortion and how masking shapes audibility.
Thus, low THD specs (e.g., <0.01%) are reassuring, but THD measurement alone is not a robust predictor of perceived fidelity, therefore, data such as the one in the table below tends to be more of a selling point than actual audio system performance vs. the person listening to that audio system.
| Distortion Component | Audibility Tendency |
|---|---|
| −40 dBc (≈ 1%) | Often audible |
| −60 dBc (≈ 0.1%) | Sometimes audible depending on masking |
| −80 dBc (≈ 0.01%) | Hard to hear in music |
| < −90 dBc (≈ 0.003%) | Generally inaudible |
Putting it all together: What humans actually hear?
| Metric | Typical Perceptual Threshold | Notes |
|---|---|---|
| SNR | ~3 dB change needed to notice changes in noise, although no absolute threshold | Based on speech-in-noise tests |
| Harmonic Distortion (%) | No absolute threshold, depends on frequency & masking | Perceptual models required |
| Distortion audibility | More noticeable around 1–5 kHz | RTINGS research |
Nevertheless, very trained ears can hear very minute changes in distortion and overall audio quality, although not all the time (ref).
Thus, numbers will only be numbers unless one is listening to specific things within the output of the sound system. Which combined with psychoacoustics and each human ear performance (unique to every single human being life style, health and genetics), basically shows that SNR and THD data, measured by precise tooling (oscilloscopes, spectrum analyzers, etc.) is not equivalent to what human can actually hear, and therefore enters more the territory of "a selling point" than actual human usable data.
References (select PDFs, peer-reviewed paper and related blog posts)
- McShefferty, D., Whitmer, W., & Akeroyd, M. (2015). The Just-Noticeable Difference in Speech-to-Noise Ratio. Trends in Hearing.
- Temme, S., Brunet, P., & Keele, D. B. Jr. (AES papers on harmonic distortion audibility) psychoacoustic methods related to PEAQ.
- Ken W. Grant & Therese C. Walden, Understanding Excessive SNR Loss in Hearing-Impaired Listeners — J Am Acad Audiol.
- ITU PEAQ (Peer-Reviewed Objective Audio Quality Measurement Model) — ITU-R BS.1387.
- RTINGS.com Distortion Audibility Research. (Technical Overview).
- NCBI Bookshelf Auditory Perception & Masking Mechanisms.
- ISO 226/Equal Loudness Contours (ear sensitivity vs frequency).
- Geddes & Lee The Perception of Distortion Theory.
- Application notes - Audio Distortion Measurements (bo0385)
- Methods for the subjective assessment of small impairments in audio systems
- Efficient Adaptive Speech Reception Threshold Measurements Using Stochastic Approximation Algorithms
- Human observer sensitivity to temporal noise during B-mode ultrasound scanning: characterization and imaging implications
- Psychoacoustics
- A Study of Human Perception of Temporal and Spectral Distortion Caused by Subwoofer Arrays.
- THE EFFECTS OF DISTORTION ON THE PERCEPTION OF LOUDNESS IN LIVE SOUND
- The Correlation Between Distortion Audibility and Listener Preference in eadphones
- Question about which SNR and THD+N measurements to trust
- Low frequency harmonic distortion is almost inaudible. So what’s the point of low distortion drivers?