I have a question about amplifiers and how the output voltage relates to the dB level produced in headphones. If I understand correctly, the sensitivity of a headphone is expressed as dB per Volt (or dB per mW) and that headphone manufacturers usually state the sensitivity in reference to a 1khz sine wave signal. So for example, if the sensitivity rating of my headphones is 83dB/V and I play a 1khz sine wav test tone from REW (at 0dB rel) and my voltmeter says the amplifier is outputting 1V, I know that I’m getting 83dB - more or less considering unit variation, optimistic manufacturer specs, seal, etc. Now if I change the signal to something with more bandwidth than a single frequency like music, will the db/V remain constant? Or does the amplifier need to output more voltage to achieve the same dB?
If the db/V remains constant, then I must be misinformed as to manufacturer specs using a 1khz sine wave since it would seem to not matter. If it requires more voltage then aren’t headphone amplifier power requirement calculators a bit optimistic?
Bit of a necro on my part here, but in case this is of interest to folks…
Your headphone’s sensitivity (expressed in either a voltage-to-SPL or power-to-SPL relationship, as you say) will be measured either at a specific frequency (1khz and 500hz are common), or as an average within a band (say 500-2000hz). Single-frequency sensitivity is more common, and since it’s pretty atypical for a headphone to have a really sharp spike or dip at precisely 1khz, it’s not an issue.
If you change the signal to something wider band, you would (assuming that you kept the RMS input value at the same level) get the average of the sensitivity in that band, which is just the sensitivity at a given frequency plus the frequency response. If a headphone is -5dB at 2khz vs. 1khz, and is 83dB/V sensitive at 1khz, then with a 1Vrms 2khz sine, it’d produce 78dBSPL at 2khz.
All of this, of course, assumes that there’s no weighting to the sound pressure level, which would further confuse things, but if you’re only interested in the un/Z-weighted SPL your headphones are outputting, it does just boil down to voltage vs sensitivity+FR.
tangential but related: How does one go about measuring an impedance and sensitivity curve of a headphone? I’m assuming impedance can just be done with an oscilloscope but sensitivity is more complicated?
Impedance is measured with a sense resistor and a frequency response sweep - you determine Z vs. frequency by looking at voltage drop as a function of frequency over a known resistance.
Sensitivity in dBSPL vs. voltage is just frequency response with a known input voltage and a microphone/ear simulator with known sensitivity - ARTA’s default “frequency response” view actually gives you dBSPL for a 1V input. If you want dB/mW, it’s just the standard ohm’s law math with the impedance and voltage sensitivity.
Well that so obvious I should have just figured it out.
Thats actualy super cool. I havent ever thought about FR that way. Do you have any idea if there is anyone that has played with doing a 3D plot of like FR vs V? I listen extremely quietly to a lot of my cans and I often find the sound of cans seems to change pretty significantly below 60dB or so
Esentialy a standard FR graph done at a bunch of different gains. XY would be dB vs Freq like normal and Z would be the gain level (would be easier to do by voltage but may be more useful if you did like 5dB decreases normalized at 1k like normal)
Thanks for your answer. I’d pretty much figured it out, but it’s good to see confirmation of my understanding.
To confirm if a magical headphone had a perfectly flat FR then at a given set volume setting on the amplifier, the SPL would be the same regardless of whether I was using a 1khz sine wave test signal, or several frequencies of sine wave, or white noise.
I’m still not entirely understanding the purpose here - is it like, a deviation from linearity plot, showing (if extant) change in frequency response as a function of input/output level?
Well, the RMS SPL would be the same as the RMS value of the test signal, yes - bear in mind that complex waves will typically not have the same crest factor as sine waves, so like, going from 0dBFS peak 1khz sine to 0dBFS peak white noise isn’t going to be the same level. But if the signal level is the same, then the unweighted SPL would be the same in that case, yes.
Thats exactly what I was going for. Esentialy I find well over half of cans just utterly fall apart at my usualy listening levels (mid to high 50dB range) so im currious if we may see whats causing that if such a graph were to become a common part of measurement reports
You will see some variation at 50dB, but it’s gonna be mostly environmental noise intruding on the measurements - if your headphones were non-linear at low levels like that, you’d have some pretty significant nonlinear distortion.
Got it. So the failures of the headphones likely come from other issues at playing that low? Ignoring loss of dynamics (cause that obv makes sense) I find a lot of stuff (especially planers) gets pretty harsh
I mean, all else equal, lowering volume is almost uniformly going to make things easier on your transducer. The only exception I can think of is this small section from Olive & Temme 2015
