Almost no headphone maker publishes a frequency response

Nearly every headphone is sold with a frequency response printed on the box. “5 Hz – 50 kHz”. “12 Hz – 40,500 Hz”. It looks like a specification, it sits in the specification table, and it is the first number most people compare. It is also, in almost every case, unfalsifiable.

A frequency range means nothing without a tolerance. “5 Hz to 50 kHz” does not say the headphone plays 5 Hz at a useful level; it says a microphone picked something up at 5 Hz. It could be thirty decibels below the rest of the response and the sentence on the box would still be true. The number that turns a range into a claim is the one beside it: ±3 dB. Without it there is nothing to be wrong about.

So this page counts. 350 models have been checked against their own maker's document. 315 have a frequency range printed somewhere. 25 print a tolerance with it — 8 per cent. 25 makers print no range at all.

The widest claims in the database

Sorted by the top of the range, exactly as the maker prints it.

265 of the 315 models with a printed range claim a top end above 20 kHz, which is roughly where human hearing ends in a young adult and well above where it ends for anyone reading this. What your ears can still hear puts your own number on that. None of these claims comes with a level, so none of them can be checked — and none of them tells you anything about how the headphone sounds in the range you can actually hear.

They do not even agree what to call it

Across 350 models the same row appears under 35 different labels. AKG calls it audio frequency bandwidth. Sennheiser calls it frequency response (speaker). Shure calls it frequency range. DITA calls it FR. One maker prints it only as a marketing headline, with no specification table on the product page at all.

The makers who print the number that matters

25 of 315. They are split between in-ear and full-size models. 14 of them name the coupler standard as well as the tolerance, which is the part that makes the figure reproducible.

And they do not print the same tolerance. The figures here run from ±2 dB to -10 dB, and those are not the same promise: a range measured to -10 dB may be far further down at its edges than one measured to ±2 dB. Two ranges are comparable only when the decibel figure beside them matches too.

One headphone, two tolerances, two different answers

Sennheiser prints the Sennheiser HD 700 twice in the same table. At -3 dB it runs 15–40,000 Hz (-3 dB). At -10 dB the very same headphone runs 8–44,000 Hz (-10 dB) — 7 Hz lower at the bottom and 4,000 Hz higher at the top.

Nothing was measured differently. The driver did not change. Only the question changed: “where is it still within three decibels” versus “where is it still within ten”. That gap is what every range printed without a tolerance is hiding, and it is why two such ranges from two makers cannot be compared at all.

Moondrop prints both numbers for the Moondrop Aria 2, one under the other. The headline is 16Hz - 22kHz (IEC61094, Free Field), with no conditions attached. Below it sits 20Hz - 20kHz (IEC60318-4, -3dB), labelled “Effective Frequency Response Range” — the point where the response is 3 dB down, measured on IEC60318-4. Same headphone, same page, and the honest number is the narrow one.

And one that forgot the unit

Grado prints the frequency response of the Grado GS1000 (Classic) as “8-35”. Not hertz, not kilohertz — two numbers and a dash. Read as hertz at one end and kilohertz at the other they make sense, and that is almost certainly what is meant, but the maker never says so, so this database does not say so either. It is the same failure as a sensitivity printed without dB/mW or dB/V, and it is why the spec sheet audit exists.

The makers who print nothing

25 of the 350 models checked so far carry no frequency range at all: AFUL Explorer, Audeze LCD-3, Denon AH-MM400, EPZ Q5 Pro, Final Audio A4000, Final Audio A5000, Final Audio E3000, Final Audio E5000, Grado GT220, Hidizs MK10 ArcSonics, Juzear 41T, Kiwi Ears Orchestra Lite, Letshuoer S12 Pro, Letshuoer S15, Softears Twilight, Thieaudio Ghost, Thieaudio Monarch MK3, Tin HiFi T4 Plus, TRN MT1, ZMF Aeolus, ZMF Atrium, ZMF Auteur, ZMF Auteur Black, ZMF Caldera Closed, ZMF Verite Open. That is not laziness. A maker who prints nothing is telling you less than one who prints “5 Hz – 50 kHz”, but it is also claiming less, and nothing they have claimed can be misread.

One maker answers the question directly. Dan Clark Audio keeps a frequency-response row in the specification table for Dan Clark Audio AEON 2 Closed, Dan Clark Audio AEON 2 Noire, Dan Clark Audio AEON 2 Open, Dan Clark Audio AEON RT, Dan Clark Audio E3, Dan Clark Audio Expanse, Dan Clark Audio Noire X, Dan Clark Audio NOIRE XO, Dan Clark Audio ÆON Core, Dan Clark Audio Stealth, and what it prints in it is “Yes, it has one”. It is a joke, and it is the shortest summary of this page that exists.

What to do with the number

Treat a bare range as marketing copy, because that is what it is. If it carries a tolerance and a coupler standard, it is a specification, and you may compare it with another figure measured the same way — but only with another figure measured the same way. Two ranges from two makers, both without tolerances, cannot be compared at all; there is no shared definition underneath them.

This is the same failure as the sensitivity problem documented in the spec sheet audit: the number is printed, it looks authoritative, and the unit or the condition that would make it mean something is missing. It is why every figure on this site says where it came from.

Method

Each entry was read from the manufacturer's own product page, product sheet or manual, and recorded verbatim — the range, the label above it, the tolerance if printed and the measurement standard if named. Where a maker publishes two ranges, both are recorded. Nothing here is measured, converted or estimated, and no third-party measurement is reproduced. The count grows as models are checked; it currently stands at 350 of 578. How this data was built covers the rest of the database.