VNandor
500+ Head-Fier
- Joined
- Oct 17, 2014
- Posts
- 813
- Likes
- 432
The actually expensive part of the measurement would be to figure out how sensitive the headphone/earphone is once you actually put it in/on your own ears. Generic sensitivity measurements are already published by the manufacturers as well as by some third party reviewers (such as Tyll) and generally they are 1-2dB off compared to each other. I don't remember coming across a 3dB difference but I also only ever check on headphones, not IEMs. If you take the published sensitivity as "granted" (ie not off by a wide margin) then the rest of the measurement can be done quite cheaply because all that is left is to measure/calculate how much voltage is sent to the headphones.
I never had the qudelix but as far as I know, it can already measure where the "peak" output voltage is set to. The sensitvity of the headphone can be entered manually I think.
I have no reason to think that this would be inaccurate but if you don't trust it, you could borrow a multimeter that can measure AC voltages to confirm the output voltage yourself. In my experience, they are fairly common household equipment. You could also buy them new or second hand if you want to stay away from unknown brands for less than 100$. Even if you ended up with a truly, truly horrible (but still properly working) multimeter that was +/- 10% that would still translate to less than +/-1dB inaccuracy. In terms of accuracy, this is really the least of your problems and the part that is the easiest to get right.
There's this blog post that points out the qudelix ignores the volume level of the digital signal. Essentially, the qudelix tells you what the output level of the headphone would be if you played back a full scale sine wave. If you did just that, it would be quite accurate. However, you will have to compensate yourself for the volume if you play back something else. The way to do that is to figure out the RMS level of the digital file you are listening to. This can be done with free programs, foobar+dynamic range plugin can do it(yuck), audacity can do it right out of the box. A full scale sine wave is -3dBFS rms. If your music comes out to let's say -13dBFS rms, then your music is 10dB quieter than the full scale sine wave. That means you have to subtract 10dB from whatever the qudelix told you to get the volume level for your music instead of the full scale sine wave. If you used foobar with the dynamic range meter you would have to subtract 13dB because the meter uses a different, "unique" definition of RMS compared to audacity (and the rest of the world basically...). If the qudelix works well, this could already get you in the correct ballpark with some caveats and there's little reason to worry if you got a something that's way below safe listening levels. I used this technique (except with a multimeter) to measure my listening levels and it came out to around 70-80dBSPL rms with 80dBSPL being almost uncomfortably loud.
If you don't trust the qudelix you'll have to make the voltage measurement yourself, do the bit of the math that the qudelix does to convert the voltage measurement from the multimeter to dBSPL but that part is fairly easy as well, it's like a weekend project for people who know what a multimeter is but never measured the output of a headphone amp.
Something else to be aware of is that the sensitivity is typically given at only 1 frequency. This wouldn't be a problem if headpones had a flat frequency response. Since they don't, you have to look up a frequency response graph to confirm that the frequency point is not placed in a dip or a peak. Even if it is placed at a representative point of the graph, this is still something that also causes some inaccuracy and the amount of inaccuracy depends on both how flat the headphones' frequency response is and the music's exact spectrum. Technically, it's possible to make a software that could import a frequency response and analyze the spectrum of digital file being played back and spit out how much of the overall level is being amplified/reduced but that's far more complicated than just reading the RMS and I don't think such a tool exists yet. If the frequency response don't have peaks or dips that are both big and relatively wideband it's still fine though and you can totally get in the correct ballpark. Some of the differences can be eyeballed as well, for example if a headphone is known for rolling off some low frequencies, then the measurement would report slightly higher than the real SPL numbers with bass heavy music, probably not so much with music with little bass. One could go as far as manually defining how much of the low frequencies are being rolled off by looking up a frequency graph, then using a spectrum analyzer to find how much of the music's spectrum is down that low and use that to calculate how much the overall SPL is being attenuated by the low frequency cutoff.
Last edited:
