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

Perceptual coding

1 2
  1. Music Alchemist
    You have submitted zero evidence that you can reliably distinguish between them under controlled conditions. All you have done is make claims without any objective data to back it up. If you want to prove that you can reliably distinguish between them, it's easy to attempt to do so: just take a lossless file and convert it to high bit rate lossy (or a hi-res file and convert it to lossless Red Book) with dBpoweramp, then do an ABX test and show your results. If you cannot pass 15 out of 20 trials, that means you cannot hear the difference. Period.
    I'm talking about playback of audio, not production.
    This thread really should have been posted in Sound Science, anyway...
  2. FFBookman

    Yeah I was banned from "sound science" which goes to show how they feel about things that are completely obvious to actual audio professionals. Strange bunch over there, in full denial of true science, which is discovering new details all the time.  Those guys are stuck in 1998, a very odd timewarp.
    We are just getting started learning about the details of the auditory system, the impact of sound on our psyche, and the power of musical interpretation and detection. They are in the first stage of understanding it, which is why there is little terminology agreement.
  3. FFBookman

    Well i've been in this same dance before. You want me to pass a test i've continually claimed, and am not alone in saying, is utter trash for mixed music quality detection purposes.
    It appeals to my ego though....I know I could pass it, but that's the trap.  I even have it built into my DAP, the revealer, to play different versions.
    But It's still not a fair test, regardless of my results, 25/25 or 10/25 or whatever. It completely takes us away from natural listening and enjoyment of music. If you aren't enjoying it there's no point in trying to be a scientist.
    If I passed 20 or 25 out of 25 what would that prove to you?   Math is already on my side, science is already on my side, logic is already on my side.  You are counting on the inability of human senses to detect something obvious to many.
  4. FFBookman

    Evidence?  3000k/sec of data vibrates more than 300k/sec of data when converted back to analog vibrations.
  5. Music Alchemist
    If you can get 15 (or more) out of 20 correct, then that would indicate that you can reliably distinguish between them. But you have to isolate the variables first by following my instructions.
    Side note: anecdotes (claims) do not count as evidence in this context.
    That relates to the data, not necessarily the audible sound. All the objective evidence indicates that there is no audible difference between them.
  6. VNandor
    There's no evidence behind that people can hear a difference between them. I bet someone who devoted their precious time to learn about how humans' hearing work then developing perceptual coding would love to have a word or two with you! Unfortunately, I haven't encountered such people in these forums so far.
    I can see why they did that since you refuse the idea of ABX tests. Like it or not, double blind testing is the accepted way of testing and not only in the audio word. As for what the audio professionals say, it doesn't make anything more or less true. It has zero effect on people's psychology, hearing, and on how perceptual encoding works.
  7. FFBookman

    Not true.  Go into a studio and sit down and listen to things at different resolutions.
    You can get all the evidence you need.
    Talk to the engineers at that studio. Talk to the people that make music, especially if it's not the modern pop/EDM/hip-hop. They'll give you all the evidence you need.
    Your adherence to a garbage test and it's useless results have given you a fatal flaw in your logic. Sorry it's just true.  If you can come at me with anything other than an ABX test for your evidence I'll address it.  If that's all you have I ain't interested.  
    It's totally flawed, i've given you several examples why. Get a better test or stay with that pile of crap for an argument.
    BTW -- neuroscience just had a major breakthrough this year and has released a far more detailed map of the brains regions than ever available before. most accepted neuroscience today is based on brain research from decades, if not a century ago.
    I have to admit I did start this thread to either see if your type were extinct yet, or if not, what are the reasons for staying lossy in 2016.  I've heard 1 good reason - I want my whole collection to fit in 128gb - but other than that it's this tired ABX stuff.
  8. FFBookman
    here's a decent mid-level interface for recording, circa 2012.  listen and learn, there's nothing even close to lossy going on in something like this.   lossy was invented was to get cd's through dial-up modems.  playback at 24/192, when mated to proper analog, is amazing.

    here's a documentary style video explaining the overall decline in sound quality as digital technology has taken over most areas of music.

    here's andrew schepps famous engineer giving a talk about lossy streaming at a google event a couple years back:

  9. Music Alchemist
    What you are talking about are subjective anecdotes (people telling stories and making claims), which is not objective evidence. Objective evidence would be a test under controlled conditions. These types of tests are well established in the scientific community. You can't say that they are flawed, because they are not. There is no other way I know of to prove or disprove that someone can reliably distinguish between two things. It is your own proposed tests that are flawed.
    I agree that it only makes sense to use lossy if you are trying to fit music onto a portable device. I listen at home and only use lossless.
  10. VNandor
    People who care about audio but refuse to to admit that lossless and lossy do not *sound* the same should be the least of your worries. There are a lot of people who don't care at all about quality audio, or really, quality music in general. Please try to "enlighten" them first, then you can worry about the rest.
  11. FFBookman

    The real debate I am interested in is 16/44 vs higher resolutions. I've studied that for 20 years now.
    I don't believe "lossy" was ever meant for the popularity it has achieved as the standard format for music listening these days. It was meant for a temporary bandwidth restriction that has gone away. If you went back to the 90's and tried to find anyone that thought an MP3 sounded exactly the same as a CD you'd still be looking.
    I think our society after 9/11 and other terrorist attrocities has become "lossy" as a mental state. Quality and longevity are not admired anymore. "Good enough" rules the day and we are used to accepting poor digital substitutions for analog things. Someday they'll get better, we tell ourselves, and I'm back in the mix of pure music love again.
    I know I just accepted bad science with MP3. By 2010 I was hating it even at 320k, kinda forgot about lossless, was just about to give up on music. I thought it was me. Age, whatever. Then I woke up, realized i could buy 24bit music or rip lossless, and got a real DAP that let me take pro audio mobile, and now here I am taking all these words, all these minutes, to try to get you to see past bad data and trust your instincts.  Of course 3200k gives your speakers more data than 320k does. It's counter-intuitive to think it doesn't. Where is all that extra data going?
  12. FFBookman

    I accept your logic of needing controlled conditions.  But those conditions cannot pollute the sample! Quick-switching unfamiliar mixed music in unfamiliar spaces (often on an unfamiliar rig) with the singular goal of detecting resolution is highly flawed.  It really is, but I don't know how I can get you to understand that.
    I AB different dithers when I give my 24bit masters to my mastering engineer and he downsamples them. But I mixed the song and usually tracked the instruments so I know how it's supposed to sound at master level. I know it's sound signature. I know where things are, how they blend.
    The dithers are all very different. I have heard 4-5 of them I think. Each sounds different, and each tries to recreate the 24bit signal, but none of them can. To a certain extent dither is perceptual coding also since it attempts to get the 16bit audio sounding like the 24bit version.
    That's why I might sound like I'm copping an attitude with pro-lossy people. "Yeah but can we hear the loss at 320k?" gets in the way of my overall argument that you are already being given degraded versions if they were down sampled from 24bit to 16bit, or if they were recorded to tape but transferred at 16/44 only.
    The grid for 24bit audio has 16 million possible values. The grid for 16bit audio has 64 thousand possible values. You just must not believe sound or music is very complex if you refuse to believe people telling you there's more to the sound than what your ABX test is "discovering".
    Your test is telling you that 320k/sec = 1200k/sec and that your body cannot detect any difference. There is clearly a flaw in your test, your gear, your material, or your ears. Since I believe you have good gear, materials, and ears, that leaves the test methodology itself.
  13. FFBookman
    OK, that's cool.  
    My mobile device holds 128gb cards so I don't need lossy anymore. Hopefully you will get that upgrade soon.
  14. Music Alchemist
    So use music you are familiar with, on whichever equipment you like. No one's forcing you to use random music you've never heard before for a listening test.
    Try using dBpoweramp to convert some files. (Since you hate ABX, don't worry about that for now.) Then you can open the original and converted file in a player like foobar2000 and switch between them at any part of the song you want. Do that simple AB comparison and let me know whether you still perceive a difference.
  15. VNandor
    I fear that I can't say (or anyone else for that matter) anything new about it. Go check the Nyquist-Shannon theorem if you haven't already. I think you already know that the only difference between 16 bit and 24 bit is the noisefloor. You could try doing null tests by downsampling one of your 192/24 to 44/16 to see what's left, how big the differences are. You could organize for yourself any kind of BLIND test, not necessarily ABX. Because sighted evaluation won't cut it. Noone says differences can't be perceived, we question where the differences come from.

    You could also change the title of the thread as well...
    I agree blind testing is not the end-all be all method to determine if we can hear a difference (as in difference because of different file formats and not other things that have effects on what we hear). The correct approach would be to research how the ear works what it's sensitive to or insensitive to, (like bandwidth or masking effects) I believe there are serious studies about that. After that perceptual coding should be analyzed to see if it only removes information that could not be heard in the first place. These would be cold hard facts, not only observations.
    I mean come on, do you really believe that your ears can detect every type of air movement at any amplitude under any given circumstances? Because if you do then you are probably not a hearing expert.
    Science also explains why someone can perceive differences (not only in audio) where there wasn't any at the first place. But it is closer to psychology than physics.
    Now I may appear as if I don't "trust my ears" but I do. That's why I like to use my ears only when I evaluate music. What I accept is that what I hear can be affected by price tags or the good-old more=better argument or by expectations in general. Have you tried to listen to 352kHz/32bit float recordings? Were they any better than 192/24?
1 2

Share This Page