[ses1984]

I've been trying to A/B two pairs of headphones and someone suggested that I get a sound level meter, since that can help you get a better comparison. They're only $15-$20 on amazon so I figured why not?

I got one, now I'm not sure what to do with it. I removed the windscreen, and poked it through a piece of foam, and pressed the foam against the ear cup. Now what? I played pink noise, adjusted the volume until I got to 80dB, and went from there. But this is just where the confusion begins.

I can't just use any pink noise file, shouldn't I look at the levels of the file?

What about when comparing to one of my recordings? The meter I got only samples every 0.5 seconds, so if I use it to analyze a song, I think there's a chance I could miss the peaks. In that case, would I look at one of my songs, find the peaks, and cut that into a "noise file" and use that as the input? Then when I adjust the volume to 85dB I know that the peaks of that song go up to 85dB and no higher.

Then I suppose I need to Replay-Gain my whole collection (is replay gain still a thing? I used it about 10 years ago in foobar. Is there something else now?) Then pick a few songs at random, cut a "noise file" of the peaks with replay gain applied, and then adjust levels until sound pressure meter reads 85dB -- this should be my max safe volume?

 

[castleofargh]

getting relative variations between 2 gears, and getting a precise loudness value out of a headphone are 2 very different things(although you'll probably get none exactly right).
warning the last few times I tried to explain how I did things, it was critically unclear for the reader. so if what I post seems complicated and kind of overwhelming, keep in mind that most of the difficulty is caused by me and my mastery of heavy redundant English. ^_^ sorry.

to get 2 headphones at the same loudness is impossible because different signatures mean different levels depending on the frequency. so you can use some noises of whatever color and try to get a subjective equivalence by ear. or use a specific test tone and decide to match the headphones for that one frequency, measuring the results with the sound level meter. you might want to pick between 2 and 5khz where our ear is most sensitive, or 1khz because measurement graph usually align there, or anything you like as the measurements won't mean they will fell like the same loudness to you anyway. to objectively match 2 headphones would mean matching the frequency response first, which defeats the purpose of using listening tests to pick a headphone ^_^.

as for getting an accurate loudness level when using headphones, there is no particular trick. you need a reliable reference and some mean to measure the relative difference between your gear to that reference. the sound level meter could be a good approximation, use a test tone as signal, not music!!!!!(I suggest using a test tone at -3 or -6db just to avoid possible clipping, but you have to remember to add that value to your final measurement to get the full scale loudness). this should be enough to get a ballpark value which is really all you probably need.
as a double check, you could try to get a voltmeter and see the voltage getting out when playing a 1khz test tone(again if the test tone isn't full scale signal, remember to add up the difference at the end). you can then use the manufacturer's sensitivity provided in the specs of the headphone if you trust them. or use Tyll's measurements https://www.innerfidelity.com/headphone-measurements . on each PDF in the bottom right he measured the voltage to get 90dB at 1khz.
the main issue with that method is to get a voltmeter that is relatively accurate at 1khz(mine isn't and those which are known to be tend to be expensive). so if you don't know that you can rely on 1khz, use a low frequency test tone instead(the device is probably made to deal with 50 or 60hz AC signals in the first place). and using a raw frequency graph for the headphone, you try to estimate the sensitivity at 60hz(sensitivity in specs + the variation between 1khz and 60hz in dB on the graph). it's also not very accurate, but once we use non professional equipment, we have to accept that the results won't be. and mixing up all the methods can help improve the reliability of your measurements and give you tips about how to do them better.

and about how loud is your music specifically, once you have the full scale measured, you know that your music will never be louder than that (at the same volume setting on that headphone at 1khz or whatever you measured). you can also get the peaks, some average, or other readings for the music played on your computer, but it might not be very relevant when each song could give a different result. unless you use EQ or replay gain, relying on the full scale levels from a test tone feels to me like a relevant worst case scenario at that frequency.

 

[ses1984]

>getting relative variations between 2 gears, and getting a precise loudness value out of a headphone are 2 very different things(although you'll probably get none exactly right).

I think my main goal is figuring out what is a safe upper limit of listening levels, so I know what I should not exceed during listening sessions. I realized I actually have no idea what that sounds like. I *think* I've been overly conservative and generally listen very quietly. Maybe I can safely turn up the volume a lot, maybe I can't, who knows?

The thing that bothered me about using a test tone is what if the tone was what if that happened to be a valley for that particular set of headphones?

If I have a process for finding out what this is reliably, then I could get an approximate match between two headphones. I'm not that concerned about getting an exact match. I care much more about my subjective opinion than matching volume levels exactly.

Regarding voltmeters--I have a 10 year old $20 multimeter from radio shack. I could try to find a hacker space or meetup where I could borrow one, I guess.

 

[castleofargh]

the valley thing is a legitimate concern, but you can probably check your headphone FR in the link I posted, or some other website.

 

[ses1984]

How consistent are those fr measurements, from sample to sample?

I'm curious why a test tone is superior to something like white or pink noise.

 

[pinnahertz]

A multimeter is the wrong tool for measuring audio levels, always has been, but it's not a bad calibration reference. Yet most of us are sitting right next to the right tool: a computer, and sound card. All that's needed is the right software to create a low level digital AC voltmeter that remains accurate over the full audio spectrum, and means to calibrate it. Google a bit. TrueRTA (free) gets you the basics, but there are many others. Calibration can be done using your sound card and multimeter by generating a voltage and frequency at which the multimeter is accurate.

 

[pinnahertz]

I'm curious why a test tone is superior to something like white or pink noise.

For level matching / gain setting the best signal would be a one octave band of noise centered at 1kHz. You can start with pink and filter down to something like 700-1400Hz. A 3rd order HPF and LPF would be sufficient. Using pure tones for precision level measurement of any transducer would probably end up with erratic results.

 

[castleofargh]

A multimeter is the wrong tool for measuring audio levels, always has been, but it's not a bad calibration reference. Yet most of us are sitting right next to the right tool: a computer, and sound card. All that's needed is the right software to create a low level digital AC voltmeter that remains accurate over the full audio spectrum, and means to calibrate it. Google a bit. TrueRTA (free) gets you the basics, but there are many others. Calibration can be done using your sound card and multimeter by generating a voltage and frequency at which the multimeter is accurate.

I agree that once you have a mean to calibrate, a soundcard ends up more precise than my voltmeter at most frequencies, but where do you find a reference to first calibrate trueRTA or any other? the starting point is still that we need a reference from somewhere. I mention voltmeter because it is supposed to be already calibrated and often the manual gives some ideas about how and where it's reliable.

 

[pinnahertz]

The problem is people get a volt meter and then discover it doesn't have enough low voltage sensitivity or adequate frequency response, so they're frustrated and it's useless for the actual task they wish to accomplish. Yes, you do need it as a reference, and you can use it that way, even a real cheap one, to calibrate the sound card. Once that's done, you have both the range and FR to do pretty much anything audio related. Calibration would be done with the volt meter at 100Hz, or so, around .5 to 1Vrms, where they all work and are calibrated well enough. Some applications then can calibrate out any non-flat soundcard response. One meter application I found even has a way to characterize Windows volume control step sizes and work around that bugaboo too. Then there's the actual USB oscilloscope modules which software has RMS metering, 'scope functions, even spectrum, not quite as cheap as a multimeter, but not bad considering what you can do with them. And they are pre-calibrated, usually come with probes even.

I just hate to see people goofing around and being frustrated when the tools are actually within reach.

 

[spruce music]

I think some of the Harman research suggests pink noise over 500 hz to 2 khz as being a good method for matching. So that is basically two octaves of pink noise. Still no easy method for headphones.