---
title: "Frequency Response: How to Read the Spec"
description: "What a frequency response spec really says, why a range with no tolerance figure means nothing, and what to read on a headphone or speaker instead."
url: "https://theaudiostuff.com/glossary/frequency-response/"
type: "website"
author: "Jakub Charkiewicz"
---

> Markdown rendering of https://theaudiostuff.com/glossary/frequency-response/
> Canonical URL to cite: https://theaudiostuff.com/glossary/frequency-response/
> Generated from the published page. Text and links only; see the HTML for layout.

Breadcrumb: [Home](https://theaudiostuff.com/) > [Glossary](https://theaudiostuff.com/glossary/) > Frequency Response

# Frequency Response: What the Spec Does and Does Not Tell You

Frequency response describes how evenly a component reproduces the audio band: how much output it produces at each frequency relative to the rest. Quoted as a bare range like "5 Hz to 55 kHz" it is close to meaningless, because without a tolerance figure it says nothing about how far the output deviates inside that range.

## Work it out for your gear

[EQ Ear Training Game](https://theaudiostuff.com/tools/ear-training/) Train your ears to identify frequencies from 20Hz to 22kHz. Adaptive difficulty, five training modes, score tracking, and a shareable Golden Ear Rating.

## What the measurement actually is

Play a test signal that contains every frequency in the audio band at equal level, measure what comes out, and plot output against frequency. That plot is the [frequency response](https://theaudiostuff.com/glossary/#frequency-response), and it is the most fundamental description of a component's tonal character there is. Everything a reviewer calls warm, bright, dark, thin or full is, first of all, a description of this curve.

Two things make it the measurement to look at before any other. It correlates better with listener preference than any other single number, which is a well replicated result rather than an opinion. And it is the one property of a transducer that is large enough to hear easily: differences of several decibels between headphones are routine, while their distortion figures at normal listening levels usually sit far below audibility.

## The number on the box is not that measurement

Almost every headphone and speaker prints a range. "20 Hz to 20 kHz". "5 Hz to 55 kHz". Read on its own, that figure tells you nearly nothing, and here is why.

A range is only a specification when it carries a tolerance: the number of decibels the output is allowed to deviate across it. "20 Hz to 20 kHz, plus or minus 3 dB" is a real claim, because it says the output never strays more than 3 dB from the reference level anywhere in the band. "20 Hz to 20 kHz" alone makes no such promise. The manufacturer is free to have chosen the points where output had fallen by 10 dB, or 20, or to have picked the endpoints by eye.

This is why a headphone quoted to 55 kHz is not thereby better than one quoted to 20 kHz. Neither figure describes the part of the curve you will actually hear, and the wider one is often just a more generous reading of the same kind of driver.

## What the endpoints are worth

Human hearing runs to roughly 20 kHz in young ears and falls back through the teens with age, so almost all published ultrasonic extension is outside anyone's hearing. It is not a fraud, and there are engineering arguments that a driver with [headroom](https://theaudiostuff.com/glossary/#headroom) above the band behaves better inside it, but it is not a benefit you can hear directly.

The low endpoint is more meaningful, because [bass extension](https://theaudiostuff.com/glossary/#bass-extension) genuinely varies between designs and is audible. Even there the tolerance matters more than the number: a headphone that reaches 20 Hz at full level and one that reaches 20 Hz already 15 dB down can both print "20 Hz".

## The shape is the thing

What separates one well made transducer from another is not where the curve stops but what it does in the middle. A dip through the upper [midrange](https://theaudiostuff.com/glossary/#midrange) makes a presentation sound spacious and [recessed](https://theaudiostuff.com/glossary/#recessed). A peak between 2 and 4 kHz brings vocals [forward](https://theaudiostuff.com/glossary/#forward) and, pushed too far, makes them shouty. A rise from 6 to 10 kHz reads as detail at first and as [sibilance](https://theaudiostuff.com/glossary/sibilance/) by the end of an album. A broad lift below 200 Hz reads as [warmth](https://theaudiostuff.com/glossary/#warmth), and a narrow one as bloat.

None of that is visible in a range. All of it is visible in a curve, and audible in a minute of listening.

## Headphones and speakers are not measured the same way

A loudspeaker is measured in an anechoic space at a fixed distance on axis, so its published response is close to what the drivers do before a room touches it. What reaches your chair is that curve plus the room, and the room dominates below a couple of hundred hertz.

A headphone is measured on a coupler or a head simulator, and the result is not supposed to be flat. The outer ear and ear canal already impose a large boost centred somewhere around 3 kHz before sound reaches the eardrum, and a headphone that measured flat at the eardrum would sound severely dull. Published headphone curves therefore include a deliberate target shape, and different rigs and different targets produce curves that cannot be compared directly. This is the single most common mistake made with headphone graphs: two measurements of the same headphone from two sources can look quite different and both be correct.

## How this site treats it

Every review here that has a manufacturer figure publishes it verbatim in the specs table, unit and notation exactly as the maker states it, and never converts or tidies it. That is deliberate. The figure is the manufacturer's claim, not our measurement, and rewriting it would quietly turn someone else's marketing into something that looks like data of ours.

What the reviews do instead is describe the shape in words, region by region, and say what it sounds like on named material through a named chain. Where a correction is worth making, it is stated as a correction rather than implied.

## Learning to hear it

The fastest way to stop guessing about response is to learn what the bands sound like. The ear training game below plays a tone or a filtered band and asks you to name the frequency, which takes the vocabulary out of the abstract: once you can reliably pick out a boost at 3 kHz from one at 8 kHz, review prose about presence and sibilance turns into something you can check for yourself.

## Keep reading

- [Full audio glossary](https://theaudiostuff.com/glossary/)
- [How we test and score](https://theaudiostuff.com/about/#methodology-title)
