Question 1

Why does my coherence suddenly drop to zero at one frequency?

Accepted Answer

A sudden zero-coherence point at a specific frequency indicates that the signal at the microphone has no correlation with the reference at that frequency. The most common cause is a notch filter in the signal chain (feedback suppressor, graphic EQ set to maximum cut, or a crossover filter) that completely removes energy at that frequency. With no signal energy, only noise remains at the microphone, giving zero coherence. Other causes: perfect destructive interference from two equal-level paths with opposite phase at that exact frequency, or a null in the speaker polar pattern at the microphone angle. Check the DSP processing chain for any active notch filters. SonaVyx highlights zero-coherence points and suggests checking signal processing as the first diagnostic step.

Question 2

My coherence is uniformly low across all frequencies. What is wrong?

Accepted Answer

Uniformly low coherence (below 0.5 across all frequencies) usually indicates a fundamental measurement setup problem rather than an acoustic issue. Check: is the reference signal actually connected and receiving audio? Is the measurement microphone working and positioned correctly? Is the correct input channel selected? Is the signal level too low (check input meters)? Is there a sample rate mismatch between reference and measurement? Is the delay compensation set approximately correctly (a large uncompensated delay decorrelates the signals)? Is the system actually playing audio? Try a simple test: connect the reference signal directly to the measurement input (electrical loopback). If coherence is still low, the problem is in the setup. If it jumps to 1.0, the acoustic path is the issue. SonaVyx input level meters help verify that both channels are receiving signal.

Question 3

Why does coherence drop in the presence of air conditioning noise?

Accepted Answer

Air conditioning adds broadband noise (especially at low frequencies from fan and duct rumble and mid-frequencies from air turbulence) that is uncorrelated with the measurement signal. This noise appears at the measurement microphone but not in the reference, reducing the coherence at affected frequencies. HVAC noise is typically strongest at 63-500 Hz, which is exactly where room mode effects already challenge coherence. The combined effect can make low-frequency transfer function data unreliable. Mitigation: increase source level to improve SNR, use more averages, or turn off the HVAC during measurement if possible. If HVAC cannot be turned off, note the affected frequency bands and accept lower confidence in those regions. SonaVyx ambient noise indicator shows the noise floor so you can assess its impact.

Question 4

Can feedback in the sound system cause low coherence?

Accepted Answer

Feedback does not typically cause low coherence — in fact, it often causes high coherence at the feedback frequency because the feedback loop creates a strong, consistent signal correlated with the input. However, if the feedback causes the system to saturate or clip, the resulting nonlinear distortion generates harmonic content that is not linearly related to the input, reducing coherence at the harmonics of the feedback frequency. Also, if a feedback suppressor is active and constantly notching frequencies, the notched frequencies will show zero coherence because the signal is being removed from the output. If you see multiple narrow coherence dips at apparently random frequencies, check for an active feedback suppressor. SonaVyx feedback detector identifies ringing frequencies before they reach full feedback.

Question 5

Why is coherence worse when I measure at a distance?

Accepted Answer

At greater distances from the speaker, two things degrade coherence. First, the direct sound level decreases with distance (6 dB per doubling), while the reverberant field level is approximately constant. This shifts the balance from direct (correlated) to reverberant (partially decorrelated) energy. In the reverberant field, the sound has been reflected many times from many surfaces, decorrelating it from the original signal. Second, ambient noise remains constant while the signal level drops, reducing SNR. The critical distance (where direct and reverberant fields are equal) marks the transition: coherence is typically good within the critical distance and degrades beyond it. For most rooms, the critical distance is 2-5 meters. To measure beyond it, use more averages and a louder source.

Question 6

How do I tell if low coherence is caused by noise versus reflections?

Accepted Answer

Noise-caused low coherence improves when you: increase the source level, add more averages, or reduce the background noise. Reflection-caused low coherence does not improve much with more averages because the reflections are deterministic (same every measurement). Diagnostic approach: increase the source level by 10 dB. If coherence improves proportionally, noise was the dominant factor. If coherence barely changes, reflections or interference are the cause. Additionally, noise tends to affect coherence uniformly or at specific bands matching the noise spectrum. Reflection interference creates coherence dips at specific, regularly-spaced frequencies (comb filter pattern from a single reflection) or random-looking dips (multiple reflections). SonaVyx comb filter detector identifies periodic coherence dips indicative of a specific reflection.

Question 7

My coherence looks fine with pink noise but drops with music. Why?

Accepted Answer

Pink noise excites all frequencies continuously, providing the H1 estimator with energy at every bin for every FFT frame. Music has a time-varying spectrum: during a bass note, there is little energy above 2 kHz, so those bins have no signal and coherence drops. During a cymbal hit, low frequencies may be weak. The H1 estimator needs energy at each frequency to compute meaningful coherence. Remedy: use more averages (16-32) so that over time, the music excites all frequencies. Use broadband music (full-band mix) rather than sparse arrangements. Accept that some frequencies will have low coherence during quiet passages. The transfer function magnitude may still be usable despite low coherence if you average long enough. SonaVyx exponential averaging mode adapts to changing program content better than linear averaging.

Question 8

How do I verify that my reference signal is clean?

Accepted Answer

A clean reference signal should show coherence near 1.0 when connected directly to the measurement input (electrical loopback test, bypassing speakers and microphone). If loopback coherence is below 0.99, there is a problem in the signal path: ground loops adding hum, interference from nearby equipment, gain staging issues causing clipping or quantization noise, or a sample rate mismatch. Check: is the reference signal level appropriate (not clipping, not too quiet)? Are the cables shielded and balanced? Is the reference coming from before any time-varying processing (compressors, gates, auto-gain)? The reference should be a direct, unprocessed copy of the signal feeding the amplifier. Any processing between the reference tap and the amplifier will appear in the transfer function measurement. SonaVyx input meters show reference and measurement levels for verification.

Coherence Troubleshooting: Diagnosing and Fixing Low Coherence Problems

Frequently Asked Questions