We converted the published sensitivity of 26 loudspeakers to a common basis. On 22 of 300 possible comparisons, the corrected order is the opposite of the printed one. Two of those reversals are between speakers made by the same company.
Last verified: 8 August 2026
The problem in one comparison
KEF publishes both of these:
| Published sensitivity | Nominal impedance | |
|---|---|---|
| KEF LS50 Meta | 85 dB | 8 Ω |
| KEF Q3 Meta | 87 dB | 4 Ω |
Same manufacturer, same spec sheet convention, same measurement voltage. Read the numbers and the Q3 Meta is 2 dB more sensitive, which would mean it plays 2 dB louder on the same amplifier.
It does not. Corrected to a common basis, the LS50 Meta is 1 dB ahead. The printed gap is not merely wrong in size, it points the wrong way.
Neither number is a lie. KEF publishes what almost everyone publishes, correctly labelled. The problem is that the label is the part nobody reads.
Why it happens: 2.83 volts is not one watt
Sensitivity is meant to answer “how loud does this speaker play for a given amount of amplifier effort.” The old convention measured it at 1 watt at 1 metre. The modern convention measures at 2.83 volts at 1 metre.
Those are the same thing into 8 ohms, because 2.83 volts across 8 ohms is one watt. That is exactly why 2.83 was chosen.
Into 4 ohms it is two watts. Same voltage, half the resistance, twice the current, twice the power.
So a 4-ohm speaker measured at 2.83 V is being handed double the power of an 8-ohm speaker measured the same way, and the number it earns is about 3 dB higher for that reason alone. The specification is honest. The comparison is not.
The correction is one line of arithmetic, and you can check it:
correction (dB) = −10 × log₁₀(8.0089 ÷ nominal impedance)- 8 ohms → 0.00 dB, nothing changes
- 6 ohms → −1.25 dB
- 4 ohms → −3.02 dB
And a second convention nobody flags
Klipsch publishes the RP-8000F II at 98 dB, which looks extraordinary next to a typical 87. Their own footnote explains why:
SPL at 1M, half-space anechoic with 2.83V input
Half-space means the measurement was taken with the speaker on a boundary, so the sound radiates into a hemisphere instead of a full sphere. Concentrating the same energy into half the space is worth roughly 3 dB. Most brands publish full-space figures.
Again, Klipsch is not hiding it. It is printed on the spec sheet. It is simply never carried across into any comparison anybody makes.
Klipsch speakers really are efficient. Corrected to full space, the RP-8000F II lands at 95 dB, which is still among the most sensitive speakers in our database by a wide margin. The correction does not demolish the claim; it right-sizes it.
What we found across 26 speakers
We converted every speaker in our database to the same basis: one watt, one metre, full space.
Four different measurement conventions are in use among 26 speakers from 13 brands:
| Convention | Models | Correction |
|---|---|---|
| 2.83 V / 1 m, full space | 22 | 0 to −3.02 dB, depending on impedance |
| 2.83 V / 1 m, half space | 2 (both Klipsch) | a further −3 dB |
| 1 W / 1 m | 1 (Triangle) | none, this is already the common basis |
| Single 500 Hz tone | 1 (Magnepan) | no correction is possible |
22 of the 300 possible pairwise comparisons reverse. That is about 7% of comparisons where reading the spec sheets gives you the wrong answer about which speaker is more sensitive. We only count a reversal when the corrected gap is at least 0.25 dB; at a zero threshold the figure is 27, and we publish the stricter one.
The starkest case is the KEF Q3 Meta against the Wharfedale Diamond 12.2. Published, the KEF looks 0.5 dB ahead. Corrected, the Wharfedale is 2.5 dB ahead. A three-decibel swing, from two numbers printed half a point apart.
See every model and its correction on the database page → The chart there shows the published figure and the corrected one for all 26, sorted by how large the correction is.
The one we refuse to correct
Magnepan publishes the LRS+ at 86 dB, and specifies it as 86 dB / 500 Hz / 2.83 V: a single tone, not a broadband average.
There is no arithmetic that makes a single-tone measurement commensurate with a broadband one. We could apply the 4-ohm voltage correction and print a number that looks like the others, and it would be false precision dressed as diligence.
So the database shows “no comparable basis” for that row, and the chart draws it as a hollow marker with no correction arrow. A gap you can see is more useful than a number you cannot trust.
What this actually changes when you are shopping
It does not mean 4-ohm speakers are worse. It means they extract their sensitivity partly by drawing more current, which is a real engineering trade and not a free lunch. The speaker plays as loud as it plays; what changes is how hard your amplifier works to get there. That is why impedance and sensitivity have to be read together, and why 14 of the 18 speakers that publish a minimum impedance drop below 4 ohms is a fact worth knowing about your amplifier.
It does mean cross-brand sensitivity comparisons are close to meaningless as published. Any article ranking speakers by their printed sensitivity figure, without stating the impedance and the measurement condition, is ranking on a number that means different things in different rows.
And it means a 3 dB error is bigger than it sounds. Three decibels is a doubling of amplifier power. If you choose a speaker believing it is 3 dB more sensitive than it is, you have quietly halved the headroom you thought you were buying. Run your actual pairing through the System Headroom Calculator, which applies both corrections automatically and shows you the arithmetic line by line.
Why we are the ones telling you this
We do not have a listening room and we do not own these speakers, so we cannot tell you which sounds better. What we can do is transcribe published specifications carefully, notice that they are measured four different ways, and do the conversion.
That is not a clever insight. It is unglamorous arithmetic on numbers anyone can read. It just appears to be work nobody else has bothered to do, and it is the reason our calculators produce different answers than the free ones that treat every sensitivity figure as interchangeable.
If you find an error in any of it, tell us and we will fix it and say that we did.
Sources
Correction arithmetic derived from the definition of the two conventions. Every specification linked to its manufacturer source:
KEF LS50 Meta · KEF Q3 Meta · KEF Q7 Meta · KEF R7 Meta · Klipsch RP-600M II spec sheet · Klipsch RP-8000F II spec sheet · Wharfedale Diamond 12.2 · Triangle Borea BR03 · Magnepan LRS+ · SVS Ultra Evolution Bookshelf · Q Acoustics 5040 · Focal Vestia N°3 · Full source list on the database page
Two manufacturer inconsistencies we hit while doing this. Q Acoustics prints the row label and the unit in contradiction with each other on the 5020 and 5040, labelling it “Sensitivity (2.83V @ 1kHz)” while giving the unit as dB/W/m; we treat it as a 2.83 V figure and say so. Wharfedale publishes 86.5 dB with a 3.7 Ω minimum on its US site and 88 dB with a 4.0 Ω minimum on its UK site, for the same Diamond 12.2, with every other specification identical. One of the two is wrong and we could not determine which, so we use the US figures and flag the discrepancy rather than quietly picking the flattering one.
Related: The gear database · System Headroom Calculator · Which AV receivers publish a 4-ohm rating · The receiver spec that decides whether you can ever upgrade · How we work