The transfer function measurement was displayed on the screen. The magnitude trace jagged wildly between +20 dB and -30 dB with no discernible pattern. The phase trace was a blizzard of random angles. The coherence was below 0.2 across the entire frequency range. It looked like someone had given a seismograph to a caffeinated squirrel.

The client leaned in. 'So what does that tell you about the room?' It told me nothing about the room. It told me everything about the measurement setup.

Coherence below 0.2 means the output signal has almost no linear relationship to the input signal. There are three common causes: the noise floor is higher than the signal, the measurement system isn't connected to the right source, or something is fundamentally broken. In this case, it was option two. The measurement software was sending pink noise to output channel 1. The PA was connected to output channel 2. The measurement microphone was faithfully capturing ambient noise — HVAC, traffic, a bird outside the window — and the software was faithfully computing the transfer function between pink noise that the room never heard and ambient noise that had nothing to do with the pink noise.

We connected the PA to the correct output. The transfer function resolved into a clean curve with coherence above 0.9 from 100 Hz to 16 kHz. The room was actually quite good. We had been measuring the relationship between noise and nothing for fifteen minutes.

The Moral: If your transfer function looks like modern art, check your routing first. SonaVyx's coherence display immediately flags low-coherence measurements — coherence below 0.5 means your data is noise, not signal. Fix the measurement before interpreting the results.