Question 1

Why is my coherence low below 200 Hz?

Accepted Answer

Low coherence below 200 Hz is the most common measurement challenge and has several causes. Room modes create standing waves where the pressure at the microphone position fluctuates unpredictably. Background noise (HVAC, traffic) is often strongest at low frequencies, contaminating the measurement. The microphone may be near a node of a standing wave where the direct sound and room mode partially cancel. Insufficient averaging means the statistical estimate has not converged. To improve: use more FFT averages (16-32), increase FFT size to 16384 or 32768 for better frequency resolution, move the microphone away from walls and corners, increase the source level, and ensure the source speaker actually reproduces low frequencies. SonaVyx color-codes coherence so you can immediately see which bands are reliable.

Question 2

Why does my transfer function look noisy even with pink noise?

Accepted Answer

A noisy transfer function usually indicates insufficient averaging. Each FFT frame is an instantaneous snapshot contaminated by ambient noise and room mode fluctuations. Averaging multiple frames reduces this variance. With 1 average, the transfer function is very noisy. With 4 averages, it becomes usable. With 16-32 averages, it is smooth and reliable. Other causes: the reference signal level may be too low relative to ambient noise, the microphone may be picking up mechanical vibration, or there may be a feedback loop in the signal chain. Also check that the reference signal is actually being received (verify the input level meter shows signal). SonaVyx supports linear, exponential, and peak-hold averaging. For system tuning, exponential averaging is recommended because it tracks slow changes.

Question 3

What causes periodic dips in the transfer function magnitude?

Accepted Answer

Periodic, evenly-spaced dips in the magnitude response are the signature of comb filtering, caused by a delayed copy of the signal summing with the direct signal. The dip spacing in frequency equals 1/delay_time. For example, dips every 1 kHz indicate a 1 ms delay (34 cm path difference). Common sources: a nearby reflective surface (floor, desk, lectern) creating a short reflection, two speakers at slightly different distances to the microphone, or a misaligned driver in a multi-way speaker. Identify the delay by measuring the dip spacing: if dips are at 500 Hz, 1500 Hz, 2500 Hz (spacing of 1000 Hz), the delay is 1 ms. SonaVyx comb filter detector in the problem detection suite automatically identifies these patterns and reports the implied delay.

Question 4

Why does my phase plot wrap around at ±180 degrees?

Accepted Answer

Phase wrapping occurs because mathematical phase is defined within ±180 degrees (or ±pi radians). When the actual phase exceeds 180 degrees, it wraps to -180 degrees and continues upward, creating vertical jumps in the plot. This is a display artifact, not a measurement problem. Unwrapped phase shows the continuous accumulation of phase shift, which is more intuitive for understanding system behavior. A speaker with a crossover at 2 kHz might have 360 degrees of phase rotation at that frequency, which appears as two wraps. SonaVyx can display either wrapped or unwrapped phase. Group delay, which is the derivative of unwrapped phase, avoids the wrapping issue entirely and is often more useful for alignment decisions.

Question 5

My coherence drops to zero at certain frequencies. What is happening?

Accepted Answer

Coherence dropping to exactly zero at specific frequencies means there is no correlation between the input and output at those frequencies. Common causes: a notch filter in the signal chain completely removing that frequency, a deep null from destructive interference (two paths canceling perfectly), the source not producing energy at that frequency (e.g., a speaker with a sharp low-frequency rolloff), or a feedback suppressor actively notching those frequencies. If the zeros are at harmonically related frequencies (100, 200, 300 Hz), suspect a comb filter. If at a single frequency, check for a notch filter in the DSP. If coherence is zero across a wide band, the reference signal may not be connected or the wrong channel is selected. SonaVyx flags zero-coherence bands for investigation.

Question 6

Why does my measurement change when I move the microphone slightly?

Accepted Answer

Sound in a room creates a complex interference pattern where pressure varies significantly over short distances, especially at higher frequencies. At 1 kHz, a half-wavelength is 17 cm, so moving the mic 17 cm can shift from a constructive to destructive interference point. At 10 kHz, half-wavelength is 1.7 cm. This position sensitivity is a fundamental property of room acoustics, not a measurement error. To get a representative measurement: use spatial averaging (average multiple nearby positions), apply smoothing (1/6 or 1/3 octave), focus on the broader trends rather than individual narrow-band peaks and dips, and avoid over-EQing narrow features that are position-dependent. SonaVyx trace overlay lets you compare multiple positions to identify which features are consistent.

Question 7

How do I deal with latency between my reference signal and measurement?

Accepted Answer

Latency (propagation delay) between the reference and measurement signals causes the transfer function phase to accumulate rapidly, making it hard to read. Most analyzers use a delay finder to detect the propagation delay and compensate for it, aligning the reference and measurement in time. Once delay is compensated, the phase plot shows only the system phase response (crossover behavior, reflections) without the linear phase from propagation. SonaVyx delay finder uses the impulse response peak to automatically detect and compensate propagation delay. If auto-detection fails (low SNR, multiple arrivals), you can set the delay manually. For a speaker 5 meters away at 20 degrees Celsius, expect approximately 14.6 ms of propagation delay (5 m / 343 m/s).

Question 8

Why does my transfer function look different from the RTA measurement?

Accepted Answer

The transfer function and RTA show different things. The RTA shows the absolute spectrum at the microphone, which is the product of the source spectrum, the system response, and the room response. The transfer function shows only the system-plus-room response, with the source spectrum divided out. If your source is pink noise (which has equal energy per octave, rolling off at -3 dB/octave on a linear frequency scale), the RTA will show a downward slope while the transfer function will be flat for a perfect system. Additionally, the transfer function averages multiple FFT frames coherently (preserving phase), while the RTA typically averages power spectra. This means the transfer function can reject uncorrelated noise through coherent averaging, while the RTA includes all energy regardless of its source.

Transfer Function Troubleshooting: Low Coherence, Noise, and Measurement Issues

Frequently Asked Questions