1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.

    Dismiss Notice

Lossless vs mp3 ABX results. (Among other ABX's)

Discussion in 'Sound Science' started by deadlylover, Sep 30, 2011.
1 2 3 4 5 6
  1. kiteki
    Oh well, this test is much harder with the replaygain OFF... there's no way I can replicate what I did above now... 100 A/B's in under 20 minutes, that's just not possible in what I'm listening to right now.
    I'm still getting results in my favour, but the difference in listening is so subtle right now I'm just not sure if 192kHz can justify the stupendous file space required for it and lack of current playback support (no portable players support native 192kHz at the moment).
    and for a moment I was excited I could differentiate scarletbook and redbook 100 times in under 20 minutes on $20 Sony earbuds!!
    It turns out reality is much more, real.
  2. deadlylover
    Heh well, I never said going against the world would be easy.
    -0.50 and +0.98 is a pretty big difference on terms of volume, that's almost 1.5db all up, which should be a piece of cake to distinguish in an ABX.
    Don't beat yourself up if you fail to ABX 192khz vs 44.1khz, I don't think it's supposed to be theoretically possible anyway (frequencies above audible range), compared to lossy vs lossless. Then again, you can't help but think that it might be possible that the ultra high frequencies would have some effect on sound/feeling.
  3. kiteki
    Then why did you fail?
  4. deadlylover
    On the two high res tracks I tried with replaygain, the track gain difference was 0.02db. I don't think that's audible.
  5. kiteki

    Oh, why didn't you say so! I'll just try a different track then... your listened to "Come Away, Death" and the other piano one, right? What did you downsample with?
    The thing is, it seems like upsampling to 192kHz can sound very good at times and is not a matter of ultrasonic frequencies:
    In the example below, the piano intro to "Rihanna - Unfaithful" (I just chose some random pop music, mp3 192kbps, 44.1kHz) the piano notes have more weight, articulation and air, almost like a filter is removed on the music, it actually sounds good and distinct enough with the SoX upsampler version (which I downloaded last night) for me to consider upsampling all of my music to 192kHz and feeding that signal from foobar into my DAC into my headphones/speakers - all the time! :wink:
    This is not the case in comparing native 192kHz... versus 192kHz downsampled to 44.1kHz... which is more like trying to listen to a butterfly landing on top of a church... in rome.
    ...at a concert.
  6. deadlylover
    Yep, I used that foobar pluggy in thing, Resample (SoX) DSP applied for the conversion process.
    If it sounds better to you, then that's all that matters. [​IMG]
    I'll try some upsampling later so see what effect it has on my tunes.
    I gave it a try with Bad Apple!!, upsampled to 96khz and I couldn't tell a difference. I also failed the ABX. [​IMG]
  7. kiteki
    Yeah, you need to pull the SoX up on your music and listen to more dignified schoolgirls, for real.
    I don't get how you downsampled the high-res stuff with SoX, I'm kinda doubting you did an ABX comparator test for the high-res stuff at all.
  8. deadlylover

    I used foobar's converter to do the resampling with SoX, I just converted FLAC to FLAC with the resampler applied. You can see the option there in foobar's converter under processing.
    You can see the converted tracks up the top, the bitrate should be enough proof. As you can see, I've been a busy boy doing all those comparisons from the playlist.
  9. kiteki
    Michael Jackson 24/192 vinyl rip.
    Downsampled to 16/44.1 with Audacity.
    Replaygain ticked ON: (this test has become fairly quick and simple now)
    Replaygain ticked OFF: (this is pretty difficult, it could be simply luck at times... It's hard to tell)
    Replaygain data:
    I tried exporting a Rihanna MP3 from foobar with SoX upsampling into Flac/192kHz as per your post #98 and then I ABX'd it with original and couldn't tell the difference, something is glitchy there, ABX'ing is a bit too difficult and annoying, the upsampled Rihanna should have been the easiest one, and the replaygain glitchiness and everything, I just don't get it...
    blah... that's enough kilohertz for my head today.
    Thanks for the cool thread anyway.
  10. deadlylover
    Aahah it's okay, thanks for helping to contribute. ^^
    If anything, we've worked out that upsampling/downsampling fudges with the files a fair bit depending on how you go about it, and care must be taken in doing comparisons. That's very helpful to know for anyone else out there trying to ABX high-res and wot not.
  11. Head Injury
    There shouldn't be such a huge difference in volume from a simple downsample [​IMG]
    kiteki, can you try a few things? You don't have to do any more blind testing, not yet anyway.
    1. Do you have any native 96kHz files you can downsample? I want to see if the volume difference is anywhere near that high. Don't downsample 192kHz to 96kHz, just 96kHz to 44.1kHz (you can do the 192 -> 96 too if you want, to see about volume differences there).
    2. Could you try to downsample to 48kHz instead? 192kHz is a multiple of it, so it would make for a simpler and potentially less error-prone downsample.
    Hold on, I think I have a 96kHz album somewhere. It's all it was available in [​IMG]
    Dang, nevermind. It's just 24/44.1kHz.
  12. Tegmentum
    I've run ABX tests in the past that yielded I pretty much can't reliably pick out the lossless source from the lossless until it dips down to around ~v4 mp3. However, I still primarily rip and listen to FLAC for peace of mind and because hard drive space no longer runs at a premium these days.
  13. Willakan
    keteki, are you using your HifiMan DAC for the tests?
  14. Tetsuma
    I never bothered doing ABX for compressed vs lossless, didn't want to give people the satisfaction. I've always found V0/320 to either:
    Lose power in vocals
    Fail at percussive transients (they sound flat)
    Exclude little details
    I had Gentle Giant's The Power and The Glory in V0, and decided to re-rip it when I got my K701s.
    The V0 was made from the .wavs I previously had in 2006 using the then current build LAME encoder via dBpoweramp. But I had HDD limitations then.
    re-ripping showed detail I hadn't heard before, most evidently is the clapping in the opening song at the beginning.
    Compressed songs are very noticable with my current setup (Benchmark DAC1, K701, Canamp), even when I was using a Xonar DX as my source.
    That being said, my brother can't hear a difference between:
    Xonar>Heed Canamp>K701
    DAC1>Heed Canamp>K701
    And struggles to see the merits of my K701s vs some shoddy $20 Sennnheiser headphones. In fact, the thing he likes so much about my headphones over his ones are the ear cushions.
    I honestly remember back when I started playing bass guitar in primary school that I couldn't hear the bass line of a song unless it was mixed to be forward (see primus/incubus/RHCP etc). So being critical and listening analytically has been a passive practice ever since.
    I realize this is anecdotal, and probably won't carry much weight. I think i'm posting this because its nice to see others hear exactly what I hear too. (I hope cable <3ers don't misconstrue this).
  15. anetode

    It seems like you're still having problems with the level matching. Previous tests have shown that even .1 db difference can cause and audible difference. Deadlylover's matching to an order of magnitude less (.01/.02) than the minimum perceivable difference makes sense, yours do not. Unfortunately it often takes more than preset algorithms included in programs to achieve the desired result. Audacity is awesome because it's free, but pro mixing software can produce tighter tolerances in level matching.
    I'm gonna be gone for a while and won't be able to keep up with the thread, but on return I'm gonna spend some time to furnish test samples of various processing of the original.
1 2 3 4 5 6

Share This Page