1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Test 24bit/96kHz vs. CD resolution audio

Discussion in 'High resolution audio' started by tigre, Sep 10, 2003.

  1. tigre

    tigre Moderator Staff Member

    Joined:
    Feb 18, 2003
    Messages:
    790
    Likes Received:
    0
    Trophy Points:
    26
    About Prisoner's concerns (?):
    I think everyone involved has the chance to learn from this. On points where everyone agrees no new knowledge will be discovered. So for me everything is fine :)
    My only problem related to this: Sometimes I'm frustrated because my English skills are not good enough to express myself as clearly as I want to.

    To me Wilkes appears knowledgable, helpful and open-minded which I appreciate.
     
  2. listen

    listen Member

    Joined:
    Sep 11, 2003
    Messages:
    30
    Likes Received:
    0
    Trophy Points:
    16
    Surely there was no dissension here... I don't see any reason why anybody might have been offended, it's just a chance to share ideas and knowledge. And tigre, your english is perfectly fine, it's just that often the most simple concepts are the hardest to explain. I wonder why people make such a big deal over things...


    Anyway, back to some actual test results.. you guys are going to like this.. :)

    This morning I ABXed 24/96 against 24/44.1

    This means that either the resampler is at fault, or 44.1KHz is an insufficient sampling rate.
    Yes, all the files I'm feeding to the ABX program are 24/96 (superficially at least), and I'm resampling with foobar2000 in slow mode. I'm using waveOut.. 24 bit output.

    The foobar resampler is a good one isn't it? ssrc?

    My result was only 11/12 this time, but it's just concentration really. I spent much less time on it than I did on my 16/44.1 test, and I'm not exactly 100% after new years eve...

    Today, I was just concentrating on the tambourine at 6s. I decided that the good one was more detailed. If it helps anyone else listening, the difference is less than the power of my imagination, so it becomes progressively more difficult. I just about always get the first 3 or 4 right if I work through them quickly, but then they start to sound different from when I started (boredom of repetition, time perception, imagination etc.) so I have to take more care. If I imagine that one of them is better and compare it, and then imagine that the other one is better and compare it, the true 24/96 file seems to be able to get a better boost from my imagination.
    Having said that, there does seem to be a consciously audible difference before I go mad from repetition. At this moment I would say that the shake of the tambourine seems to have more attacks and decays in 96KHz.. the 44.1KHz version seems more smoothed over or cohesive. I might also say that the 96KHz version has more crunch.

    Is anyone else having any luck?

    maxg, other than what tigre said, don't forget that blind testing is very important... this test is really not that complex to set up at all, don't let the pages of step-by-step instructions put you off.
     
  3. tigre

    tigre Moderator Staff Member

    Joined:
    Feb 18, 2003
    Messages:
    790
    Likes Received:
    0
    Trophy Points:
    26
    Listen, thanks for your answer. Unfortunately there are two more issues:

    1.
    There are chances that when using waveout or directsound fb2k's output is somewhat changed before it's passed to the soundcard. This seems to depend on Windows version and only Microsoft knows for sure.

    Best bet to avoid this issues is using Kernel Streaming output with fb2k. It's included in normal or special installer or as separate plugin on fb2k components page. So could you please try to ABX it again with Kernel Streaming?

    2.
    If I understand correctly you ABX the resampled 24/44.1kHz file vs. the 24/96kHz original. Besides the capabilities of the formats (and the resampler's quality which I doubt) there can be on more thing that causes the differences: Your soundcard's DAC could have noticable performance differences between 44.1kHz and 96kHz sampling rate. To find out if the difference lies here resampling the 44.1kHz file back to 24/96 before ABXing would be a good idea.

    I've tried to ABX the positions where you heard differences with this setup: original vs. (original -> 16/44.1 -> 24/96) using Kernel Streaming, 24bit output padded to 32bit, but the differences I heard always disappeared when I didn't know which sample I heard (= immagination).
     
  4. listen

    listen Member

    Joined:
    Sep 11, 2003
    Messages:
    30
    Likes Received:
    0
    Trophy Points:
    16
    no no..
    they are resampled back to 24/96 before I test.

    Yeah I was wondering about kernel streaming... I will try that... probably tonight..

    Another result (might be irrelevant without kernel streaming):
    Last night I couldn't sleep so I loaded up [lovely_1->16/96->24/96] against [original 24/96] and listened to a random new spot, between 9 and 10 seconds. To my great astonishment I had 8/8 within about two minutes, then I lost it so I went back to bed. I will try this one as well with kernel streaming.

    I was wondering about something else too... do you think that my not so fantastic headphones could make the difference more obvious than if I was using a really good set? I don't really like listening to music with them, but I'm not sure if they are exposing shortcomings of the recordings, or causing problems themselves...
     
  5. tigre

    tigre Moderator Staff Member

    Joined:
    Feb 18, 2003
    Messages:
    790
    Likes Received:
    0
    Trophy Points:
    26
    Listen, thanks for your efforts. About your question: I can't tell for sure. In theory it's possible that equipment (e.g. speakers) that is not designed for reproduction of > 20kHz equipment reproduces these frequencies distorted. This distortion might be audible in < 20kHz range and cause differences.

    I suggest to post your results at hydrogenaudio - there are some knowledgable people who can answer questions like this. (There's an "Resend validation email" button if necessary ;) ).
     
  6. listen

    listen Member

    Joined:
    Sep 11, 2003
    Messages:
    30
    Likes Received:
    0
    Trophy Points:
    16
    Hmm, that's an interesting thought. So... even though I can tell them apart, it [bold]could[/bold] be that the 24/96 file is distorted, and downsampling [96->44.1->96] fixes the problem and brings the sound closer to how it should be. Although, that wouldn't apply to word length..

    Anyway.. Kernel Streaming.
    Thanks for pointing it out.. my music doesn't sound like crap any more! I was starting to think there was something wrong with my headphones.. Although I can't be 100% sure, because there's no way I could try to ABX waveout vs. kernel streaming.
    I retested all of them, and yes I still hear the difference. I'm not sure if it's easier. It took me much less time (I only spent a few minutes on one of them), but I'm probably just getting used to it.

    So, I can tell apart:
    [24/96] vs. [24/96->16/44.1->24/96] (slow resampling, dither)
    [24/96] vs. [24/96->24/44.1->24/96] (slow resampling)
    [24/96] vs. [24/96->16/96->24/96] (dither)

    All were less than 1% chance of guessing, and I was pretty confident of the difference most of the time.

    I tried 16/96 at the spot I tested the other night, (between 9 and 10s) and I think the difference that I hear is somewhere in the attack of the guitar chord. They both have equal power, but the original 24 bit one seems more simultaneous.. there's some delay or offset in the 16 bit one. This doesn't really align with the theory of word length (does it?), so I guess it's a subjective impression caused by something else that I can't identify.

    Yeah, I'll post at HA.. hopefully will encourage some more ears too...
     

Share This Page