AES-2id: Common Impulse Response Measurement Errors and Solutions
TL;DR
AES-2id documents the most frequent measurement errors that degrade impulse response quality. Time variance during MLS measurement (people moving, HVAC cycling) raises the noise floor. Clipping the recording saturates the waveform and introduces harmonic contamination that deconvolution cannot remove. Insufficient post-silence truncates the reverberant tail. Sample rate mismatch between playback and recording causes frequency scaling errors. Incorrect deconvolution (using the wrong reference signal) produces garbage results. Missing calibration makes absolute level measurements meaningless.
Time Variance
The room impulse response is assumed to be linear and time-invariant (LTI) during measurement. Violations of time invariance include:
- People moving in the room during the sweep
- Doors opening or closing
- HVAC cycling on/off, changing the air flow pattern
- Temperature changes (air absorption varies with temperature)
MLS measurements are particularly sensitive to time variance because the entire sequence must be coherent for circular cross-correlation to work. Swept sine is more robust — time variance primarily affects frequencies near the instantaneous sweep frequency at the moment of the disturbance.
Recording Clipping
Digital clipping occurs when the signal exceeds 0 dBFS. Unlike analog saturation, digital clipping creates harsh harmonic distortion that cannot be removed by deconvolution. The distortion products appear at all frequencies and times in the impulse response, raising the apparent noise floor.
Prevention: Set recording level to peak at -6 to -3 dBFS. Monitor the level during a test sweep before the actual measurement. Use 32-bit float recording if available (virtually eliminates clipping risk).
Insufficient Post-Silence
If the recording stops before the room's reverberation has fully decayed, the impulse response is truncated. This causes:
Prevention: Include post-silence equal to at least 1.5× the expected RT60. For unknown rooms, start with 5 seconds of post-silence.
Sample Rate Mismatch
If the playback device runs at a different sample rate than the recording device (e.g., 44.1 kHz playback, 48 kHz recording), the deconvolution produces a frequency-scaled impulse response. All parameters will be wrong. Always verify both devices use the same sample rate.
Wrong Reference Signal
Deconvolution requires the exact excitation signal as reference. Using a different sweep duration, frequency range, or generation method than what was actually played produces meaningless results. Some systems regenerate the sweep internally rather than using the played-back version — ensure the reference matches the actual transmission.
Electrical Feedback Loop
When using a PA system as the source, monitor outputs or recording buses can create electrical feedback paths that bypass the acoustic channel. This adds a direct electrical impulse to the measured response. Disconnect all monitoring during measurement.
Missing Calibration
Without calibration, absolute levels (SPL, sound strength G) cannot be determined. Relative parameters (RT60, C80) do not require calibration. If absolute levels are needed, calibrate the measurement microphone before and after the session.
SonaVyx Error Prevention
The IR tool includes automatic checks: level monitoring during sweep to detect clipping, post-capture INR verification, and sample rate consistency checks. The averaging feature helps overcome marginal SNR conditions.
Try It Now
Open this measurement tool in your browser — free, no download required.
Last updated: March 19, 2026