Prominence Ratio (PR)

The Prominence Ratio quantifies how much a tonal component stands out above the surrounding broadband noise within its critical band. Defined in ECMA-74 and widely used in IT equipment noise standards, the PR provides an objective measure of tonal audibility based on the frequency masking properties of the human auditory system.

Theory

The critical band represents the frequency resolution of the human cochlea. Within each critical band, energy is summed by the auditory system. A tone is perceived as prominent when its energy significantly exceeds the broadband energy in the same critical band.

Critical Band Definition

The critical band around a center frequency f_c is defined using the Bark scale. The bandwidth increases with frequency:

CB(f_c) = 25 + 75 · [1 + 1.4 · (f_c/1000)²]^0.69 [Hz]

At 1 kHz, the critical bandwidth is approximately 160 Hz. Below 500 Hz, it narrows to roughly 100 Hz. Above 4 kHz, it widens considerably.

PR Calculation

The Prominence Ratio is computed by comparing the total power in the critical band to the estimated noise-only power in the same band:

PR = 10 · log₁₀(P_total / P_noise) [dB]

The noise power P_noise is estimated by interpolating the spectral levels from the edges of the critical band across the tonal region, effectively reconstructing what the spectrum would look like without the tone.

PR vs TNR Comparison

Both the Prominence Ratio and the Tone-to-Noise Ratio (TNR) assess tonal prominence, but they differ in methodology:

Aspect	Prominence Ratio (PR)	Tone-to-Noise Ratio (TNR)
Standard	ECMA-74, ISO 7779	ECMA-74, ISO 7779
Comparison Basis	Tone power vs. total power in critical band	Tone power vs. noise power in critical band
Noise Estimation	Interpolation across critical band	Subtraction of tone power from total band power
Sensitivity	Less sensitive to narrow tones in wide bands	More sensitive to narrow tones
When to Use	General-purpose tonal assessment	Strict tonal penalty evaluation

Classification Thresholds

ECMA-74 defines frequency-dependent prominence thresholds for classifying tones as "prominent" or "not prominent":

Frequency Range	PR Threshold (dB)	TNR Threshold (dB)
90 Hz – 1000 Hz	9.0	8.0
1000 Hz – 4000 Hz	Linearly decreasing 9.0 → 6.0	Linearly decreasing 8.0 → 5.0
4000 Hz – 11200 Hz	6.0	5.0

When the PR exceeds the threshold at a given frequency, the tone is classified as prominent. A tonal penalty is then applied to the declared noise emission value, typically adding 3–6 dB(A) to the broadband level.

Parameters

Parameter	Options	Description
Input Channel	Any signal channel	Signal to analyze for tonal prominence.
Spectrum Size	4096 – 65536	FFT size. Finer resolution better separates tones from noise.
Number of Averages	10 – unlimited	Spectral averages for reliable noise floor estimation.
Tone Detection	Automatic / Manual	Automatic mode scans for peaks; manual specifies frequencies.
Classification Standard	ECMA-74, ISO 7779	Threshold curve used for prominence classification.
Frequency Range	90 Hz – 11.2 kHz	Search range for tonal components.

Interpretation

PR < threshold: The tone is not considered prominent. No penalty is applied to the noise emission level.
PR ≥ threshold: The tone is classified as prominent. The product must declare a tonal-adjusted noise level, increasing the declared value.
Multiple tones: Each tonal component is evaluated independently. If several tones are prominent, the most penalizing case determines the adjustment.

Practical Tips

Use a minimum FFT size of 8192 (preferably 16384) to achieve sufficient frequency resolution for separating tones from adjacent noise, especially below 500 Hz.
Ensure adequate spectral averaging (at least 20 averages) for a stable noise floor estimate. Insufficient averaging causes fluctuations that mimic or mask tonal peaks.
When a tone falls near the edge of a critical band, the PR may underestimate its perceptual prominence. Cross-check with TNR for borderline cases.
Fan noise, transformer hum, and switched-mode power supply whine are common sources of prominent tones in IT equipment and appliances.

In V-Listen, the PR and TNR analyses can be run simultaneously. The results table shows both metrics side by side for each detected tone, with the prominence classification highlighted for frequencies that exceed the threshold.