The transfer function had a beautiful, evenly-spaced pattern of peaks and nulls above 800 Hz. Beautiful in the way that a disease is beautiful under a microscope — technically elegant but deeply unwelcome. The nulls were 12-15 dB deep, spaced approximately 430 Hz apart.

Comb filter spacing of 430 Hz corresponds to a delay of 1/430 = 2.33 milliseconds, which means a path length difference of approximately 0.8 meters. Something was creating a reflection that arrived 2.33 ms after the direct sound. But what? The nearest wall was 4 meters away — too far. The ceiling was 6 meters up. The floor was carpeted and couldn't produce a specular reflection strong enough for 15 dB nulls.

We spent an hour hunting. Moving the mic confirmed the comb filter shifted with position, proving it was a spatial effect rather than an electrical one. The pattern was worst at the center seats and diminished toward the sides. Whatever the reflecting surface was, it was directly behind the seating area.

Then the lighting shifted and a glare caught my eye. Behind the last row of seats, running the full width of the room, was a floor-to-ceiling glass partition wall. Perfectly flat, perfectly smooth, and acoustically — at mid and high frequencies — essentially a mirror. The direct sound from the PA was passing over the audience, hitting the glass, and reflecting back with a path length difference of exactly 0.8 meters at the center seats.

Acoustic curtains in front of the glass wall eliminated the comb filter. The client's interior designer was unhappy. Physics was indifferent to her opinion.

The Moral: Comb filter spacing reveals the physical cause. SonaVyx's problem detector calculates the implied delay and distance from comb filter patterns — turning 'mysterious interference' into 'the glass wall 0.4 meters behind row twelve.'