How to Read a Frequency-Response Graph

A frequency-response graph plots level (dB) on the vertical axis against frequency (Hz) on the horizontal axis. That is the easy part. Everything else is interpretation.

The axes

Frequencies are almost always spaced logarithmically — each octave gets the same width — because hearing works in ratios, not differences. The vertical axis is in decibels; ±1 dB is roughly the smallest change most listeners can detect in steady tones, and ±3 dB is unmistakable.

What “flat” means (and doesn’t)

A speaker measured perfectly flat on one axis, in a reflection-free environment, one metre away, is one particular fact. Your room adds its own response on top. So when you compare two graphs, ask: same distance? same smoothing? same measurement environment? A graph without those footnotes is decorative.

Smoothing: the friendly lie

Raw speaker responses are jagged. Display smoothing (1/6 or 1/12 octave) averages the curve so you can see the trends. The trend is what matters for tonality — but smoothing can hide a narrow, deep notch that would be audible. Our archives always state the smoothing used; see Methodology.

Which deviations matter

• Broad tilts (everything above 2 kHz −2 dB): heard as warmth or dullness. Large, obvious differences.
• Narrow spikes at crossover frequencies: can read as hardness or edge.
• Bass wiggles: usually the room, not the speaker — ignore them unless two speakers show different wiggles in the same room.

Reading our archives

Every graph in our measurement archives lists distance, smoothing and evidence level in its header. Sample datasets carry a watermark. If a graph ever lacks those footnotes, treat it as decoration.