Clarity and Definition (C50, C80, D50)
Clarity measures how much sound energy arrives early — soon enough to reinforce a note or syllable — versus late, where it blurs into reverberant wash.
Reverberation time tells you how long a room rings, but not whether you can understand it. A room can have a moderate RT60 and still be perfectly intelligible if the early energy dominates, or muddy if the late tail does. Clarity and definition put a number on that balance by splitting the impulse response at a time boundary and comparing the energy on each side.
Clarity needs an impulse response
C50, C80, and D50 are produced only by analyzeImpulseResponse(...). Blind analysis of ordinary music (detectAcoustic(...)) recovers a decay rate and nothing more: all three come back NaN, and the per-band c50Bands / c80Bands arrays come back empty. If you need clarity figures, record a clap or a balloon pop, or record a sine sweep and deconvolve it with the known sweep before passing the resulting IR to the analyzer. A raw recorded sweep is the excitation convolved with the room response, not an impulse response.
- early 0–50 ms
- +30 ms for C80
- late
C50 — speech clarity
C50 is the ratio, in decibels, of the energy in the first 50 milliseconds to all the energy after it:
C50 = 10·log₁₀( early energy (0–50 ms) / late energy (>50 ms) )
Fifty milliseconds is the rough window within which early reflections add to the direct sound and help a syllable be understood, rather than smearing it. Energy arriving later is heard as separate reverberation that masks the next syllable. This is related to the precedence (Haas) effect, but it is not the same thing: precedence is about the first-arriving sound deciding where a source seems to be, while the 50 ms boundary is about early energy supporting intelligibility.
- High C50 (positive, several dB) — consonants stay crisp; good for speech, dialogue, lectures.
- Low C50 (negative) — the tail overwhelms the direct sound; speech smears and intelligibility drops.
C80 — music clarity
C80 uses an 80 millisecond boundary instead. Music tolerates — and often wants — more reverberant blend than speech, so the early window is widened.
C80 = 10·log₁₀( early energy (0–80 ms) / late energy (>80 ms) )
C80 is the standard "clarity index" for concert halls. Hall designers balance it carefully: too high and the music sounds dry and analytical; too low and fast passages turn into mush.
| C80 | Musical feel |
|---|---|
| > +4 dB | Dry, articulate, close |
| 0 to +4 dB | Clear but supported by the room |
| −2 to 0 dB | Reverberant, blended |
| < −2 dB | Washy, indistinct |
D50 — definition
Definition (D50) — Deutlichkeit is simply the original German name for the same quantity — expresses the same 50 ms split as a fraction rather than a ratio:
D50 = early energy (0–50 ms) / total energy
D50 and C50 are two views of the same measurement and convert directly into each other: C50 = 10·log10(D50 / (1 − D50)), so a D50 of 0.72 is a C50 of about +4.1 dB. libsonare returns D50 as a fraction between 0 and 1 — 0.72 means 72% of the energy arrived inside the first 50 ms — and interfaces conventionally display it as that percentage. Higher means more of the energy is early, so the room is more "defined." Many people find the percentage more intuitive than a dB ratio for intelligibility.
Why all three
They answer subtly different questions: C50 for speech, C80 for music, D50 for an intuitive percentage of early energy. A space optimized for spoken word (high C50, high D50) is not the same as one optimized for an orchestra (moderate, balanced C80). Seeing all three lets you judge what a recording space is actually good for, not just how long it rings.
How libsonare computes clarity
libsonare integrates the squared impulse response on each side of the 50 ms and 80 ms boundaries to form the early and late energy sums, then reports C50 and C80 as 10·log₁₀ of the early/late ratio and D50 as the early/total fraction. The integration stops at the noise-floor truncation point (found by the Lundeby method) — where the decay meets the recording’s noise floor — so background noise after the reverberant tail does not count as late energy and drag the clarity values down. All of that runs on the impulse-response path only. Blind analysis of ordinary music never forms an early/late split at all — it fits a decay rate to the fragments of tail the programme happens to expose — so C50, C80, and D50 are returned as NaN there rather than as rough approximations. Because clarity depends on the exact arrival time of the direct sound, the analysis first locates the direct-sound peak and measures the windows relative to it; a mislocated direct sound is one reason a low confidence score should make you treat C50/C80/D50 as approximate.
Related: Reverberation Time (RT60 and EDT), Source Distance and DRR, Inverse Room Estimation, Room Acoustics