Reading a Frequency-Response Graph: Targets, Ear Gain and Why Measurements Disagree

The part on sound signatures said to read graphs, not adjectives. This post is how. A frequency-response graph looks precise, but it only means something if you know what it was measured on, what it is being compared against, and which parts of it are reliable.

What the axes mean

Axis	Shows	Notes
Horizontal	Frequency, 20 Hz to 20 kHz	Logarithmic: each octave gets the same width
Vertical	Loudness in dB	Usually normalised so curves line up near 500 Hz or 1 kHz

A few dB matters. A 3 dB lift across a wide band is clearly audible. A narrow 3 dB spike is often less obvious. Broad trends matter more than small wiggles.

Most graphs are normalised to the same level at one frequency. That makes two IEMs comparable, but it also means “more bass” on a graph is always relative to the midrange, not absolute.

Why an IEM graph is never flat

Sound in a room reaches your eardrum after passing your head, outer ear and ear canal. Those shapes boost the upper midrange. The open ear canal resonates, and the result is a broad lift centred roughly in the 2–5 kHz region. Reviewers call this ear gain.

You hear that boost all the time and your brain treats it as normal. An IEM sits inside the canal and bypasses part of it, so the IEM has to put some of that lift back. Measured on an ear simulator, a good IEM therefore shows a hump around 2–5 kHz. A flat line on that graph would sound thin and recessed.

This is why raw graphs are judged against a target, not against a flat line.

Targets

A target curve is the response a reviewer or researcher thinks an IEM should have on a given measurement rig.

Target	Idea	Shape on the graph
Diffuse field (DF)	Sound arriving equally from all directions in a reverberant room	Ear-gain hump, flat bass, no bass shelf
Free field	Sound from one speaker straight ahead in an anechoic room	Similar idea, different upper-mid detail
Harman in-ear	Built from blind listening tests run by Harman’s research group	Ear-gain hump, a sub-bass shelf, gentle treble
Reviewer variants	Personal or community tweaks of DF or Harman	Usually less bass than Harman, or less upper midrange

Harman’s in-ear work, led by Sean Olive with Todd Welti and Omid Khonsaripour, measured how listeners rated many IEMs and modelled their preferences. Later analysis split listeners into groups. As reported, the largest group liked the target as it is, a smaller group preferred less bass, and another preferred more. A target is a statement about the average listener, not about you.

The sub-bass shelf

The Harman in-ear target raises the low end below roughly 200 Hz, flattening out in the sub-bass. That shelf is the “weight” people describe. A purist who sometimes wants more bass should look closely here: an IEM that tracks diffuse field below 200 Hz will sound lean next to one that follows Harman. Neither is wrong. They are different answers to the same question.

Which parts of the graph to trust

Region	Trust level	Why
Bass, 20–500 Hz	High, if the seal is good	Easy to measure; mostly depends on seal
Mids, 500 Hz–5 kHz	High	Where couplers are designed to be accurate
Lower treble, 5–8 kHz	Medium	Peaks depend on insertion depth and ear shape
Upper treble, above ~8 kHz	Low	Standing-wave effects and coupler limits dominate

Above about 8 kHz, small changes in how deep the IEM sits move peaks and dips around. Your own ear canal is a different length from the coupler’s, so a sharp dip at 8 kHz on a graph may be a peak, or nothing, in your ear. Read treble as “more or less energy overall”, not as the exact shape of each spike.

The bass region has its own trap: a measurement with a leaky seal shows less bass than the IEM really makes. The same is true in your ear, which is the subject of the fit post later in the series.

Why rigs disagree

Most hobbyist measurements use a coupler built to IEC 60318-4, often called a 711 coupler after the original design. It models the ear canal and eardrum impedance of an average adult. It is standardised for accuracy up to around 8–10 kHz, and it has a known resonance in the upper treble, around 13.5 kHz, that is part of the rig rather than the IEM.

Newer and much more expensive rigs, such as Brüel & Kjær’s 5128 head and torso simulator, use a more human-like ear canal and pinna. They are designed to be more accurate in the bass and further up the treble, and they produce visibly different curves for the same IEM, especially in the upper midrange and treble.

The practical rules:

• Never compare graphs from different rigs. A 711 graph and a 5128 graph of the same IEM can look like two different products.
• Never compare a graph against a target made for another rig. Harman’s 2019 in-ear target was defined for a 711-style setup; there are separate translations for the 5128.
• Even two 711 couplers differ. Clones, ear-canal extensions and insertion depth all shift results, so trust comparisons made by one reviewer on one rig.

Using comparison sites

Sites built on the open-source “squig.link” graph tool let you overlay many IEMs measured by the same reviewer. They are the best free tool in the hobby if you use them carefully.

1. Pick one database and stay in it. Each site is one reviewer and one rig.
2. Load a target first. Choose the target the site was built around, then add IEMs.
3. Add something you own and know. Compare candidates against an IEM you have already heard. That calibrates the graph to your own ears better than any target.
4. Normalise the same way. Most sites let you align at a frequency or by average level. Keep it constant.
5. Look at the broad shape. Bass shelf height, ear-gain height, overall treble energy. Ignore single spikes above 8 kHz.
6. Check for a “compensated” view. Some sites show the IEM minus the target. A flat line there means “matches the target”, not “flat sound”.

Reading a graph in practice

A quick routine:

1. Find the bass shelf. Is it level with the target, above or below? That is your bass quantity.
2. Look at 1–3 kHz against the target. Too high sounds shouty; too low makes vocals distant.
3. Look at 4–8 kHz. A big peak here often means sibilance or harshness.
4. Glance at the region above 10 kHz for overall air, and do not overthink it.
5. Remember the graph assumes a perfect seal and average ears. Yours are not average.

That last point is why the next parts turn to the IEMs themselves and then to fit and eartips, which can change the bass more than any graph difference between two good IEMs.

Sources

• Testing Headphones and Earphones, audioXpress
• The Shape of IEMs To Come, headphones.com
• GRAS: are the RA0401/RA0402 high-frequency ear simulators fulfilling IEC 60318-4 (PDF)
• A Deep Dive Into Harman Curves, Part 3, Headphonesty
• Harman research targets the B&K 5128, SoundGuys
• Talking With Dr. Sean Olive, PS Audio Copper