Auditory Filters, A/B/C Weighting, and Loudness: How We Measure What We Hear

When measuring sound, raw pressure levels are not enough. Two sounds with the same physical intensity can feel very different to our ears depending on frequency and context. To deal with this, audio engineers and researchers use models that imitate how the ear processes sound.

This includes:

Auditory filter banks
A, B, and C weighting curves
Loudness measured in phons
Critical band (CPB) analysis

Together, these tools bridge the gap between physics and perception.

Auditory Filters: How the Ear Splits Frequencies

Inside the cochlea, different parts respond to different frequencies. This behavior is often modeled using auditory filters.

Each filter:

Covers a small frequency region
Acts like a band-pass filter
Represents one critical band of hearing

Instead of thinking in single frequencies, the ear processes sound in bands. Energy inside a band interacts strongly, while energy in distant bands interacts much less.

This is why masking mostly happens within nearby frequency regions.

Filter Banks

To simulate this behavior digitally, we use filter banks:

Many overlapping band-pass filters
Each representing a section of the audible spectrum
Often spaced using Bark or ERB scales

This allows systems to analyze sound the way the ear does: not line-by-line in frequency, but band-by-band.

These filter banks are widely used in:

Audio compression
Speech recognition
Hearing research
Loudness modeling

A, B, and C Weighting: Measuring Loudness the Human Way

Sound level meters often report values like:

These letters refer to frequency weighting curves that adjust measurements to reflect human sensitivity.

Why Weighting Is Needed

Humans are not equally sensitive to all frequencies:

We hear mid-frequencies best
Very low and very high frequencies must be louder to feel equally loud

So instead of measuring pure physical energy, weighting curves shape the spectrum before measuring level.

A-Weighting

Models hearing sensitivity at low sound levels
Strongly reduces bass and extreme treble
Commonly used in environmental noise measurements

Used for:

Workplace noise
City noise regulations
General exposure limits

B-Weighting

Intended for medium loudness levels
Rarely used today

It was largely replaced by more accurate loudness models.

C-Weighting

Flatter frequency response
Better for loud sounds
Preserves more low-frequency energy

Used for:

Measuring loud music
Industrial noise
Peak sound levels

So A, B, and C weighting do not change the sound — they change how we measure it.

Loudness and Phons

Decibels measure physical intensity, but loudness is psychological. To represent perceived loudness, we use phons.

Definition:

A sound has X phons if it sounds as loud as a 1 kHz tone at X dB SPL

This links physical measurements to human perception using reference tones.

For example:

A low-frequency tone may need much higher dB to match the loudness of a 1 kHz tone
When they sound equally loud, they have the same phon value

This leads to equal-loudness contours, showing how sensitivity changes across frequency and loudness levels.

From Phons to Sones

Phons are still relative. To model perceived loudness growth more directly, engineers use sones:

1 sone = reference loudness
2 sones = twice as loud
4 sones = twice as loud again

This gives a scale closer to how people naturally describe loudness changes.

Modern loudness standards (like broadcast loudness meters) are based on more advanced versions of these concepts.

Critical Band (CPB) Analysis

Critical Band analysis groups frequencies into perceptually meaningful bands instead of narrow FFT bins.

In CPB analysis:

Each band roughly matches a region of the cochlea
Energy is summed within each band
Masking and loudness are evaluated per band

This is much closer to how hearing works than simple spectrum plots.

CPB analysis is used in:

Audio codecs
Noise assessment
Hearing aid design
Perceptual audio models

It also forms the basis of Bark-scale and ERB-scale processing used in psychoacoustics.

When Filters Become “Bad Filters”

Not all filters are perceptually useful.

Problems with poorly designed filters:

Sharp edges that create ringing
Poor overlap between bands
Artificial spectral gaps or spikes

These artifacts may look fine mathematically but do not reflect how the ear behaves.

Good auditory filters should:

Overlap smoothly
Have asymmetric shapes like real cochlear filters
Follow perceptual frequency spacing

That is why many psychoacoustic models avoid simple equal-width bands and use biologically inspired filter shapes.

Why All of This Matters

These models are not just academic — they are used everywhere:

Audio compression decides which data can be removed
Loudness normalization controls broadcast levels
Hearing protection standards depend on weighted measurements
Sound design tools simulate perceptual balance

By modeling how the ear filters and groups sound, engineers can build systems that match what people actually hear, not just what microphones measure.

Measuring Sound the Human Way

Raw decibels describe energy, but perception depends on:

Frequency sensitivity
Band-based processing
Masking interactions
Loudness growth

Auditory filter banks, weighting curves, phon scales, and critical band analysis all exist to translate physical sound into perceptual meaning.

They allow machines to approximate what the ear and brain do naturally — turning waves in air into structured, meaningful auditory experiences.