AES-2id Excitation Signals: Swept Sine and Alternatives

|

TL;DR

AES-2id recommends logarithmic (exponential) swept sine as the primary excitation signal. Key specifications: frequency range should cover at least 20 Hz to 20 kHz, sweep duration 3-30 seconds (longer = higher SNR), sample rate ≥44.1 kHz (48 kHz preferred), fade-in/out of 10-50 ms to avoid transients, pre-silence ≥0.5 s, post-silence ≥ expected RT60. The sweep rate is constant octaves per second: f(t) = f1·exp(t·ln(f2/f1)/T). Logarithmic sweep provides time-separated harmonic distortion products during deconvolution, unlike linear sweep or MLS.

Logarithmic Swept Sine

The logarithmic (exponential) sweep is generated by:

x(t) = sin[ 2π·f₁·T/ln(f₂/f₁) · (exp(t·ln(f₂/f₁)/T) - 1) ]

Where f₁ is the start frequency, f₂ the end frequency, and T the sweep duration. The instantaneous frequency increases exponentially, spending equal time per octave. This matches the logarithmic frequency resolution of human hearing and acoustic analysis.

Duration Selection

Sweep duration affects measurement SNR directly:

DurationSNR improvement vs 1sTypical use case
3 seconds+5 dBQuick survey, quiet rooms
10 seconds+10 dBGeneral-purpose measurement
20 seconds+13 dBLarge venues, high precision
30 seconds+15 dBMaximum quality, research

Doubling the sweep duration adds approximately 3 dB of SNR. For meeting ISO 3382-1 INR requirements, 10-20 second sweeps are typically sufficient with a decent loudspeaker.

Bandwidth and Sample Rate

The sweep should span at least 20 Hz to 20 kHz for full audio-band characterization. AES-2id recommends starting slightly below 20 Hz (e.g., 10 Hz) and ending above 20 kHz (e.g., 22 kHz) to ensure the useful band has full energy. Sample rate should be 48 kHz for professional audio work (matching common PA system rates).

Fade-In/Fade-Out

Abrupt signal onset creates broadband transients that contaminate the impulse response. AES-2id recommends cosine-squared fade-in of 10-50 ms at the sweep start and fade-out at the sweep end. The fade duration should be short relative to the sweep duration but long enough to suppress spectral leakage.

Pre/Post Silence

Include at least 0.5 seconds of silence before the sweep to capture the noise floor for SNR assessment. Post-sweep silence must exceed the expected reverberation time — if the room has RT60 = 3 s, include at least 4 s of post-silence to capture the full decay. Truncating the reverberant tail corrupts the deconvolution.

MLS Alternative

Maximum Length Sequences offer constant sound level (no crest factor issues) and are computed via circular cross-correlation. However, MLS is more sensitive to time variance and does not separate harmonic distortion. AES-2id recommends swept sine over MLS for most applications.

SonaVyx Implementation

The impulse response tool generates logarithmic sweeps with configurable duration (2-20 s), automatic fade-in/out, and appropriate pre/post silence. MLS generation (orders 2-18) is also available. See signal processing for the deconvolution step and common errors to avoid.

Try It Now

Open this measurement tool in your browser — free, no download required.

Open Tool

Last updated: March 19, 2026