How can something be audible... that isn't?

Discussion in 'Audio Science' started by GoldilocksEars, Dec 28, 2019.

?

I tried comparing the files and they

  1. sound different to me.

    100.0%
  2. sound the same to me.

    0 vote(s)
    0.0%
  1. GoldilocksEars

    GoldilocksEars Rando

    Executed
    Joined:
    Dec 18, 2019
    Likes Received:
    6
    Dislikes Received:
    0
    Trophy Points:
    3
    Location:
    In the solar system
    TL;DR: With the foobar ABX plugin I can hear differences between files that should be inaudible - even below the threshold of hearing at times. Why isn't the difference masked by the much louder sounds or how is it possible to even perceive a difference between two files when the difference file itself is completely inaudible?


    I've been wondering about this since Marvey wrote on changstar that he could tell the difference between 20 and 16 bit files. I later did my own tests with a vinyl rip and at a normal volume level I was able to perceive a difference up to 20 bit aswell, maybe even 22 bit. This is with a very dynamic file at a level where the quantization noise should be just at or below the threshold of hearing.
    At another point in time I was trying to determine how much 2nd order distortion I could hear. When I still heard a difference at -100dB (despite not listening to sines at a level anywhere close to 100dB), I stopped the test thinking that I had gone mad. Still, it was repeatable, so I thought that this was somehow normal. Now I tried it again...

    With the volume level set so 0dBFS is 67dB at 500Hz with my HD800*, I think I can still tell the difference between the two files I posted in the 'How much distortion can you hear?' thread. I haven't tried lowering the volume level further, but it doesn't make any sense to me that they would sound different in the first place. At this level any distortion should be well below the threshold of hearing (even the distortion from the HD800).

    One file is 440Hz at -3dBFS, the other is the same 440Hz at -3dB with an added 880Hz tone at -103dBFS. I attached shortened versions of the same files so you don't have to create your own, but feel free to create your own files if you don't trust me.

    I measured the output from my HD800 and my 2i2 (without HD800 load) and all the harmonics should be well below the threshold of hearing, let alone ambient here now that it has started to drizzle. How does it work that I can hear a difference?
    One sounds higher pitched to me and I don't have absolute pitch by any means. At the moment I really do find it hard to believe that other people can't hear a difference between the two files, but I still don't get why they would sound different.

    I used the foobar ABX comparator plugin to try to eliminate placebo, but even then it's possible it's all placebo. I tried to measure the distortion at that level, but couldn't.**

    *As measured with a microphone at my ear canal opening. This was simply the level I had it set to for threshold tests earlier.
    **Distortion at 64dB/440Hz should be much less than 96dB/1000Hz, which I measured at -83dB. So even assuming it's -83dB relative to 64dB it should still be at -19dB - way below even the threshold of hearing.

    I uploaded the ouput of the Scarlett 2i2 G3 measured by the UR22C to imgur, since it was too big to attach directly. With the HD800 as a load the distortion will likely be a bit higher, but it shouldn't make a big difference here. File here: https://i.imgur.com/X6AmvZU.jpg

    I can't seem to attach the files, so I uploaded them to my dropbox. At higher levels (70-80 dB) it's really not hard to tell them apart. I got a 10/12 in a quick blind test I did. Just takes some concentration and I get distracted easily, that's all. I added my blind test results FWIW.
    Same link as before: https://www.dropbox.com/sh/j5lo6dkbhlc1xgg/AACFK-j25yGufJMUdcSNV7sHa?dl=0
    _________________________________________________________

    Some of my ideas on why this is possible:
    • I've read that the ear is sensitive to absolute polarity. The 2nd order distortion will add some asymmetry here, maybe this has an effect? Still, the difference is 100dB down.
    • It's easiest for me to hear the difference when trying to ignore the louder tone, focusing on what's past it. I can hear/feel the acoustic reflex kicking in and it seems to do so in a slightly different way for the two tones. But why would it be different?
    • Maybe the 440Hz tone gets amplitude modulated by the 880Hz tone in a certain way. But again, the difference is 100dB down. A fly farting two rooms over will be louder than the 880Hz tone.
    • Maybe it's not related to the auditory system at all? I really have no idea.
     

Share This Page