A third of headphone spec sheets don't say what their numbers mean

Every headphone in this database was checked against the manufacturer's own documentation — 578 models, one spec sheet at a time. The point of the exercise was to record not just the sensitivity figure but the unit it was printed in. That turned out to be the more interesting number.

Sensitivity can be stated per milliwatt or per volt. The two differ by 10 × log₁₀(1000/Z): about 5 dB on a 300 ohm headphone, 18 dB on a 16 ohm in-ear. Read one as the other and you will conclude a headphone is easy to drive when it is not, or buy an amplifier you did not need.

Of the 567 non-electrostatic models here, 253 publish per milliwatt and 139 publish per volt. The remaining 175 — 31% — publish a number you cannot use: 142 print a bare decibel figure with no reference at all, 28 publish no sensitivity figure whatsoever, and for 5 the maker's own document could not be reached — the page is gone, or the site cannot be read — so those figures are untraced rather than unpublished.

Disagreement would be manageable. AKG, Campfire Audio, Elysian Acoustic Labs, Moondrop, Softears, Symphonium and Truthear publish per volt wherever they publish a sensitivity at all — not one model in a per-milliwatt unit between them — so a reader who knows the convention can convert and move on. Sennheiser is per volt on 23 of its 28 models here, and the gap it opens is real: at 300 ohms a per-volt figure sits 5.2 dB above the per-milliwatt one, so reading the wrong one makes a headphone look easier to drive than it is. That much is solvable arithmetic, and this site does it on every page.

What the arithmetic cannot solve is a printed number that is itself wrong, and a reader on Head-Fi was right to point out that the HD 600 is exactly that case. Sennheiser prints 97 dB (1 V) for it and 103 dB (1 V) for the HD 650 and HD 6XX, which are the same impedance and the same driver family; independent measurement puts the two within a decibel of each other. Convert the HD 600's printed figure honestly and you get 91.8 dB/mW, which is further from the measured value than the naive misreading would have been. Two errors, pointing opposite ways, and the careful one lands worse. It is the clearest argument on this page for what the whole site says in its footnotes: a converted number is only as good as the number that went in, and this database records what makers print rather than what is true.

The unsolvable ones are the makers who print a figure and stop. HiFiMAN, Hidizs, KZ, Sivga, TRN and Thieaudio state a sensitivity with no reference unit anywhere on the page, in the manual, or in the datasheet. The spec sheet says “94 dB” and that is the end of it. There is no honest way to convert a number whose unit nobody states — so on those pages this site shows both readings and the gap between them, rather than picking one and hoping.

There is a textbook distinction that would fix all of this if anyone used it: sensitivity belongs to a volt, efficiency belongs to a milliwatt. It is a real convention and it is worth knowing. It is also not what spec sheets do. The makers here who print a per-milliwatt figure almost all label that line “Sensitivity” anyway — Audio-Technica, Sony, Focal, Audeze, Meze, Shure and Etymotic among them. HiFiMAN is the one brand that reaches for the other word, and it prints “Sensitivity” on some models and “Efficiency” on others while stating a reference unit on neither. The clearest case is Sennheiser's HD 490 PRO, whose product-specification PDF prints “Sensitivity 105 dB SPL (1 kHz / 1 Vrms)” and “96 dB SPL (1 kHz / 1 mW)” side by side: one word, both references, one document. So this database goes by the reference the maker printed, never by the word in front of it. Where there is no reference, the word settles nothing.

Several makers are not even consistent with themselves. Beyerdynamic prints the reference on its newer models (“1 mW / 500 Hz”) and omits it on the DT 770, 880 and 990 — where it quotes one 96 dB spanning every impedance version, from 32 to 600 ohms, on a single line of a single table. That is what a per-milliwatt figure looks like, because rewinding a coil changes impedance while leaving efficiency roughly alone; read as per-volt it would imply a 9 dB efficiency spread across three versions of one headphone, which no driver does. So the number is probably per milliwatt — but that is an inference about Beyerdynamic's convention, not something Beyerdynamic has written down, and this site does not convert on an inference. FiiO publishes both units on one line for the FT1, FT1 Pro and FT5, and per-volt only for the FT3. Kiwi Ears prints a per-milliwatt unit on three models and a bare dB on the other three.

HiFiMAN states no reference unit on any model it sells. All 27 HiFiMAN headphones here were checked, and not one carries a reference unit — not on the product page, not in the owner's guide, not in the Chinese-language manuals, not anywhere in the download centre. The Susvara is the one that matters: its famous “83 dB” is the number every discussion of hard-to-drive headphones is built on, and HiFiMAN has never said what it is 83 dB of. Read per milliwatt it is a demanding headphone; read per volt it would be 71 dB/mW, which is a different object entirely.

And the pattern in HiFiMAN's numbers points one way without settling it. The Ananda at 26 ohms and the Ananda Unveiled at 22 ohms are both 93 dB. The Arya Stealth at 32 ohms and the Arya Organic at 16 ohms are both 94 dB. Edition XS at 18 ohms and Edition XV at 12 ohms are both 92 dB. Equal decibels across different impedances is what a per-milliwatt figure looks like and not what a per-volt figure looks like — but that is an inference about HiFiMAN's likely convention, not a statement HiFiMAN has made, and it is not enough to convert on. Two of the models are not even labelled the same way: the HE-6se and HE400se say “Efficiency” where the rest say “Sensitivity”. This site reports what is printed and says plainly that it has no published unit.

Superlux rates a headphone in watts it cannot take. The HD 668B datasheet gives 98 dB SPL referenced to one watt. The same datasheet caps input power at 300 milliwatts. Converted honestly, one watt puts the figure 30 dB below anything a headphone could plausibly measure, and the same sheet says the headphone cannot take one watt — so the 1 W reference cannot describe a measurement anyone actually made. Which half of the line is wrong is not determinable, and this database records the contradiction rather than guessing.

Campfire Audio publishes an SPL at 11.84 millivolts. Not per milliwatt, not per volt — at one specific drive voltage. It is perfectly convertible and this site converts it. 17 models here are specified against a stated drive voltage rather than per volt or per milliwatt: three Tin HiFi in-ears at 0.126 and 0.179 V, an OLLO Audio studio model at 0 dBu, and the Campfire Audio line at 11.84 millivolts. Campfire's is the most extreme of them. A reader comparing it to anything else without doing the arithmetic will be out by more than 16 dB.

Three products in wide circulation do not exist. A Limited Edition DT 1990 PRO, a Focal Utopia in-ear and a KZ ZA10 all had entries in this database, carried over from secondary sources. Beyerdynamic has never made the first, Focal has never made a Utopia in-ear (its in-ears were the Sphear and Spark lines), and KZ's nearest real models are the AZ10, AS10, ZS10 and ZAX. Specifications circulate for products that were never built.

Amplifiers have the same disease in a different form. Of the sources checked for this site, 208 were excluded outright. The three largest groups are a product whose page has simply gone (33), a power figure with no load attached to it (31) and a maker that publishes no headphone specification at all (30). A wattage with no stated load means nothing without the impedance it was measured into — it is not a specification, it is a number. The full breakdown is on the amplifier audit page.

If you are comparing two headphones, check what unit each figure is in before you compare them. If one is per volt and the other per milliwatt, the difference between them is mostly the unit, not the headphone. The converter does that arithmetic, and every model page here states which unit its maker printed.

If a spec sheet gives you a bare decibel number, treat it as an indication and nothing more. Ask what the impedance is, work out what the gap would be, and see whether the answer changes your decision. Often it does not — most headphones are easy to drive either way. When it does, it changes it completely.

The full method, including what was excluded and why, is set out in how this data was built. Every model page links the manufacturer document its figures came from.

Headphones an amplifier cannot tell apart

Reading 567 spec sheets one at a time turns up a pattern no single page shows. The two numbers that decide what an amplifier has to do — impedance and sensitivity — are not unique to a model. In 34 cases a group of headphones from different makers publishes exactly the same pair. That is 104 of the 567 models here, 18 per cent of the catalogue. (Another 57 groups share a pair within one maker, which is far less surprising.)

The largest group is 32 Ω at 96.0 dB/mW, where 10 models land on the same two figures. Here are the 18 largest:

This says nothing whatever about how any of them sound. It says that an amplifier cannot tell them apart: at the same impedance and the same sensitivity they ask for the same voltage and the same current, so every calculation on this site returns the same answer for all of them. Anyone reading a row as “these are the same headphone” has used this site against the thing it is for. Tuning, driver, distortion, soundstage, comfort and build are not in these two numbers and are most of what you are choosing between.

There is a second reading, and it is probably the right one for a good share of these rows: 22 of the 34 groups sit on round figures — a multiple of eight ohms and a whole number of decibels. Identical round numbers are weaker evidence of identical hardware than of rounded paperwork, which is the same complaint this page makes about reference units, measured one level up.