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

24bit vs 16bit, the myth exploded!

Discussion in 'Sound Science' started by gregorio, Mar 19, 2009.
338 339 340 341 342 343 344 345 346 347
349 350 351 352 353 354 355 356 357 358
  1. castleofargh Contributor
    as an effort to avoid pointless discussions(at least some), nobody is saying that one cannot or should not use hires files if he wants to. we all have some amount of freedom and can use it for such decisions.
    this topic is about bit depth and while in practice sample rate and bit depth go together in a number of ways, bringing sample rate into the conversation is only complicating things. so does comparing typical hires files lower resolution as whatever comes out is not specific to bit depth. we have other thread to go crazy over hires files in general.
    the first post is pretty clear about the intent of this topic, and the conversation should be about bit depth and the benefits(increasing fidelity of the encoding, pushing the quantization noise down, etc). and then wonder when those changes are audible for us or when they're expected to be audible for non mutated humans listening at sensible levels to correctly created audio albums.

    now about the subjective benefits, because we're in this section, we do expect statements of audibility to be backed up with supporting evidence. even more so when we happen to have tested this for ourselves under controlled conditions and have consistently failed to pass a blind test between 16 and 24bit with our favorite tracks at normal to loud listening levels. that makes us even more anxious to see evidence from those who say that they do notice a clear difference. maybe it's about hearing abilities, maybe it's about listening skills, maybe it's about the equipment. but maybe it's made up stuff in the mind of a listener who never bothered to test his hearing ability properly. we could really stop wasting so much time and efforts if that last possibility was cleared by the people themselves before coming here to spam their overconfident claims based on garbage testing methods.

    it is a fact that 16bit is more than necessary under most circumstances. the debate only concerns niche cases and those who say otherwise are wrong. that much has been well established by decades of trials and I'm still waiting to see a legitimate research suggesting otherwise. if I take my own listening habits and listening environments, 12bit dithered is all I seem to need, and most people seem to have a hard time hearing differences beyond 13 or 14bit while listening to music at non stupidly high levels. I would not say the same of 8bit or 6bits where I could clearly notice at least the background hiss when testing music at those values. it's clearly a matter of magnitude, and of course if the same track was adjusted to peak at -20dB instead of say 0dB, and I adjusted my listening level by +20dB to get the same typical listening level, that difference would have to be reported on the lowest bit depth I need for transparency.
    so the question becomes, how often does that happen? and for me the answer is never! it's not the case for everybody, but it is for me. I do have classical music with quiet passages that are stupidly quiet compared to the rest of the symphony. but the rest of it is way too loud for me to just crank up the volume of the entire piece based on that quietest part. which leaves me with 2 options:
    -I leave the volume knob where it usually is and I won't hear crap on the super quiet passage. <= usually what happens and why I stop listening to those particular albums.
    -I become a human compressor and keep increasing the gain on quiet passages, then rush to lower back down when it's loud again. <= I hate that because it means I will have to endure overly loud music for at least a moment. so again in the long run I would stop listening to those musics.

    conclusion my listening habits have no circumstances where I could audibly benefit from more than 16bit. my DAC measures better if I send 24bit signal to it for some reason, so I send 24bit padding to it from 16bit albums and everybody's happy. I pay for the cheaper files, I hear the same, my DAC measures pretty well. I'm objectively and subjectively satisfied.

    a different listener with different habits and priorities, may encounter moment when 12bit isn't transparent on a regular basis despite how it's ultra rare for me. but I would argue that very few people on very rare occasions end up with musical content sounding audibly different because it has more than 16bit. and I would argue that among those, probably more than half get sound differences that have nothing to do with having higher fidelity. instead it's often about the master being different or the playback gear doing some crap when fed with some particular resolution. the legitimate cases remaining, where audibility correlates with the quantization noise going down so bit depth is the relevant factor, I would be surprised if we can find a dozen on the entire forum. and I'm confident that all of them listen to music too loud, or created the circumstances to achieve audible difference(purposefully or by malpractice, like having the digital volume on the computer at -80dB and compensating with the amp or whatever). I'm very confident about that and after all those years hanging around audiophiles, I have yet to see one solid counter example.
    the hundreds or thousands of people who "know what they're hearing" under sighted conditions might contain such counter examples. I can't know that when those never demonstrate their abilities. to me they're no different from guys saying they have seen flying saucers from mars. some could be correct. but in the absence of proper demonstration, we all save time treating the all group as making stuff up. it's just the most pragmatic conclusion. if we consider that this is the "sound science" section, no scientific research would draw conclusions based on knowing a guy who claims he can do it. facts are demonstrated, they're not acts of faith. and as we happen to be on the web, taking random statements as facts without supporting evidence, that's just gullibility. of course we don't want to that, it's internet!
    old tech likes this.
  2. chef8489
    The loudness wars
  3. Pro-Jules
    I wrote to a mastering engineer friend of mine for some assistance, one career audio professional to another, on the merits of hi-res formats

    Here is what he advised:

    1) For really dynamic material (such as Classical) where brief silences often happen then the extra 8 bits, that you get from 24bits compared to 16bits, helps. The dither noise applied during mastering can become audible far sooner for 16bits (-96dB) than 24bits (-144dB) as we can generally hear to around -120dB we're told. This being said the noise floor of the mic preamps, amp hiss etc. would be audible far sooner than either (maybe -50dB) so can negate this point.

    2) The anti-aliasing filter that is basically the cut off point/slope that is used to determine sample rate affects the audio we hear. So for 44.1kHz the slope is at 22.05kHz and cuts off (quite abruptly) anything over this. No problem cos we're told we only hear to 20kHz right, and the Nyquist Theory suggests we need twice the frequency (which would be 40kHz) to capture what we can humanly hear correctly. Therefore 44.1kHz gives us this plus a couple more frequencies (22.05kHz technically).

    But what you get is reflected back harmonics into the audible range that don't sound right in 44.1kHz, compared to 96kHz where the reflected back harmonics are above our hearing range. So with 44.1kHz audio, a 20kHz harmonic gets reflected back at 10kHz due to the 22.05kHz slope. We can hear 10kHz so this is not favourable. If the sample rate is 96kHz the slope is at 48kHz, so the lowest reflected back harmonic ends up as 24kHz which is above our hearing range - enough for it not the be a problem. So when you compare the two 44.1kHz vs 96kHz it is hard to determine exactly why we prefer it, but you get something 'clearer' or 'truer' in the upper range, bringing transient clarity that excites, to my ears.

    Finally digital converters just sound better at higher frequencies despite not hearing the information captured above 20kHz or so. I've conducted the test countless times. I don't care to argue if someone believes I'm wrong - it's obvious with the right system and ears in my opinion.

    So there's my view on it Jules. Carry on enjoying your Hi-Res audio mate, it's worth it.

    He added...

    Now if only the world embraced 32bit float 96kHz audio (negating the need for dither) we truly would have an exact replica of the mastering studio file.
    Last edited: Nov 29, 2019
    Peter Hyatt likes this.
  4. castleofargh Contributor
    1) is as he said, although -120dB is very optimistic with musical content and a normal adult. but yeah, noises of all kinds can and will negate that point.
    2) is not on topic.

    so, here we are.
    and again, having the desire for higher resolution, or the desire to get the sound exactly like whoever intended it to be, those are fine. misrepresenting why you desire higher bit depth with empty claims and arguments about sample rate is not.
  5. sonitus mirus
    Now you are simply repeating the same conflated issues that always prop up in this forum with no evidence to support these ideas. Yes, if you crank up the volume to ear-splitting levels during the silent sections in music, this noise may be measurable at an audible level. If you are listening to music at such loud volume levels that you might technically be able to notice any dithering noise during the silent parts, your ears/brain more than likely have already adjusted to the louder sounds due to a temporary auditory threshold shift, making it impossible to hear quiet noise from dithering. Same thing if you put your ear inches from the transducer.

    These are mostly pathological examples which would also depend upon the type of dither being utilized and anti-aliasing filter choice. Bit depth at 16-bit is not something to worry about with music. It is more than enough.

    This is why audiophiles are their own worst enemy. If someone is playing a Hi-Res file with content in the ultrasound frequency range over speakers that can play ultrasound frequencies at a high enough dB with an ADC that is using a.45/.55 anti-aliasing filter and playing music at significantly high volume levels due to their hearing loss from age/abuse, and voila, detectable noise.
  6. old tech

    I have several classical CDs of direct live recordings which are dead silent in quiet passages, even at loud listening levels - some are non-dithered (being original 16 bit recordings and others are dithered but no audible difference. Some of my early CDs which are known to be flat transfers from analog tape. The only noise I hear is tape hiss from the master. I think there could be an issue with incompetent mastering though.

    Ian Shepperd, a well known mastering engineer, has compared the sound of competently dithered 8 bits vs 24 bits in the video below. Note there is no loss of music information, just low level white noise. It would be humanly impossible outside contrived circumstances to hear this white noise with 16 bits.

    I have heard this claim before, I think it was in 1983 largely solved by 1985 - it it was ever noticeable to most people. But is this an issue in the 21st century with oversampling? Why over the past 20 years or so there are no controlled tests supporting this claim? In fact, this issue is more likely to present itself with higher sample rates which result in ultrasonic noise messing with the playback gear and creating distortion in the audible range. Monty gives a good explanation here. https://people.xiph.org/~xiphmont/demo/neil-young.html

    So there's my view on it Jules. Carry on enjoying your Hi-Res audio mate, it's worth it.

    He added...

    Now if only the world embraced 32bit float 96kHz audio (negating the need for dither) we truly would have an exact replica of the mastering studio file.[/QUOTE]

    Yet all the credible controlled test conducted over the past 30 years do not support this conclusion, like the decade old test below. I don't doubt your engineer friend hears a difference - it is either different masterings, volume levels or most likely, placebo. There are no other credible explanations given what we know about human physiology and psychology, digital audio and its measurements and decades of controlled tests.
  7. gregorio
    A. Why did you have to ask a friend for assistance, I thought you were supposed to have 25+ years of professional knowledge yourself? Let's look at the actual responses your friend gave though:

    1. Yes, the extra 8 bits "helps" when recording, not on playback though because:
    1a. This statement is INCORRECT. Standard triangular dither would technically become audible at about -92dB for 16bit, not -96dB. However, for really dynamic material (such as Classical), no competent mastering engineer would use triangular dither! It's been standard practice for nearly 3 decades to apply noise-shaped dither, which indeed provides a dynamic range down to about -120dB with 16bit. And in practice, -120dB is about the limit of any 24bit DAC anyway, due to thermal noise. Furthermore, science does NOT tell us that human hearing has a dynamic range of 120dB, it tells us that we have a dynamic range of about 60dB or less, that's employed in a moving "window" much like human vision with brightness. In other words, given certain conditions (say an anechoic chamber) we can hear down to 0dB and given the condition of a high noise floor environment (say a live gig) we can hear up to 120dB but of course, those two conditions do NOT occur concurrently! So, a dynamic range of 120dB covers all eventualities/conditions but is not appropriate for any single eventuality/condition. Plus, in practice, a 120dB dynamic range above the typical consumer listening environment noise floor (30 - 50dB) would mean a peak level of 150dB - 180dB, which is both impractical and pretty much guaranteed to cause severe hearing damage (or worse). Clearly you HAVE NOT read the OP!
    1b. Yes EXACTLY, it's negated anyway and his figure of -50dB (about 3,000 times less than -120dB) is even less optimistic than what I stated (about 1,000 times)!! How many commercial music recordings can you name that have a dynamic range greater than about 60dB?

    2. As stated this is off-topic but is not true of modern anti-alias filters. And, it is NOT hard to determine why some people prefer 96kHz vs 44.1kHz; numerous controlled tests have determined that it's the result of an expectation bias (placebo effect).

    3. Now that one is nonsense! Again, current professional mix environments are 64bit float, so you still have to dither or truncate to get to a 32bit float and therefore you would NOT have an exact replica of the mastering studio mix/master. Although it's completely irrelevant because it's nowhere even vaguely near audible anyway (at any reasonable playback level), either at 32bit float, 24bit fixed or 16bit (with noise-shaped dither)! In fact, his statement above (a recording noise floor of -50dB) would only require about 10 bits (with noise-shaped dither) for the dither noise to be below the recording noise floor and inaudible!!

    You stated a few pages ago that "I am out", to which I responded; "Probably a wise move. Your "appeal to authority", against some basic engineering facts and against a body of reliable scientific evidence, won't work here in the sound Science subforum and is unlikely to end well!" - However you decided to continue, dig yourself an even deeper hole and indeed, it is not ending well for you!

    sonitus mirus likes this.
  8. TheSonicTruth
    Yet all the credible controlled test conducted over the past 30 years do not support this conclusion, like the decade old test below. I don't doubt your engineer friend hears a difference - it is either different masterings, volume levels or most likely, placebo. There are no other credible explanations given what we know about human physiology and psychology, digital audio and its measurements and decades of controlled tests.

    Ian's technical series is exquisite and educational - but - I find him to be a bit hypocritical in his crusade to both fight the loudness war and promote dynamic range. He supports amd uses tools, as I do, such as the TT Dynamic Range Meter and Foobar 2000. You know, the ones that return a figure of "DR6"(squashed) or DR14(dynamic and open). His threshold for promoting dynamics in the popular genres was, at least a few years ago, DR8 by the scale of those plugins. I even argued that he should set the bar higher, DR10 maybe, but he insisted, baby steps.

    And yes, I do remember exactly what portion of the signal those tools are measuring. I'm just talking about Shep's tendancy on this subject to talk one way but not carry the stick.
  9. board
    Please just provide the log of the properly conducted ABX test (level-matched down to 0.1 dB, time-aligned, enough trials, etc.) that you convincingly passed between 24 and 16 bit (Foobar's ABX plugin is perfect for this). Then we won't pester you any more.
    By convincingly I mean something like at least 75-80 % correct. I usually use 16 trials for my blind tests, and when I don't pass with 16 out of 16 correct, I usually pass with 15 out of 16 (93.75 % correct).
    I've given three girlfriends blind tests as well, and some they did no better than flipping a coin, although I could easily pass the same test (inexperienced vs. experienced listener), even in some cases from my kitchen (seriously), and others they passed with at least 14 out of 16 correct.
  10. board
    If I understand the point he's trying to make correctly, then what he's trying to say is that there are no anti-aliasing filters in DACs. This is only true if you buy one of those NOS DACs that purposely leave out filters. These DACs/CD players are preferred by some audiophiles, but have by some people subjectively been described as having a "dirty" sound.
    It should be noted, however, that some DACs do have anti-aliasing filters but let a lot of ultrasonic content through, but this is simply due to faulty engineering. Often these DACS are the ones costing extraordinary amounts of money. Maybe this faulty technology is what makes audiophiles rave about how it sounds more "analogue" (meaning imperfect).

    Your friend's final point about how converters sound better at higher frequencies is again just talk, unless he's willing to submit a log from a passed blind test (or take one, since he probably hasn't taken one).
    As one of the other participants in this discussion has said several times: What is asserted without proof can be dismissed without proof.
    Submit your logs of your passed blind tests, both of you (you and your friend), please, because no one in this discussion who claims to hear the audible superiority of 24 bit over 16 bit has offered any proof other than stubborn claims to know what they heard.
  11. board
    I once read a mastering engineer say that if he had an artist asking for a squashed and loud mastering he would then make two masterings: One the way he wanted it to sound; and then the same but squashed. Then after compressing it, he would lower the volume of the loud master to match the non-compressed one and burn those two CDs for the artist to compare. He then said that the artist always chose the non-squashed mastering.
    I know that procedure would waste time, but it's an attitude that I wish more mastering engineers would have instead of just assuming that the artist want something loud and aggressive.
    I do remasters for fun, and if I would play my remasters for the artists, then I think many of them would like them (but this is of course speculation), but I also think that they would realize that they had an idea of a certain aggressive sound that would sound good, but when they hear my mastering (or any good mastering by a professional) they would say "Aaaaah! This sounds much better than before!" and be thrilled. Again, it's of course speculation, but I'm hopeful - we sometimes think something sounds great until we hear something that sounds better :).
  12. board

    Maybe the rest of you have heard about the experiement where they played people the same song two times in a row and a whopping 76 % preferred one over the other.
    Every time I have had similar arguments with audiophiles, whether it's about bit depth, sample rate, jitter, mp3 vs. wave, etc. it's the same story EVERY SINGLE TIME: The person did not take a blind test and refuses to do it and prefers to yell and scream at me instead.
    You might know the Bertrand Russel quote:
    "The whole problem with the world is that fools and fanatics are always so certain of themselves, and wiser people so full of doubts."

    I can add to that that wise people who are confident that they're right, or at least on to something, are too polite, permissive, agreeable and open to debate, whereas the fools and fanatics refuse to meet halfway (which in this discussion would be: "Yes, you can make up your mind by listening, as long as it's under controlled conditions, meaning blind"). The wise, and right, people, need to learn to make the fools shut up and then educate them instead of saying "you're welcome to think whatever you want - I'm not stopping you", which is an attitude I've come to despise more and more as time has passed by. I use to hold that position as well, and it's good in many ways, 'cause I don't advocate other countries or anything for being "wrong". But there are limits. If we don't educate people properly and we permit them to believe whatever they want, they will spread this misinformation to other uneducated people, who will spread it to other uneducated people who will spread it to ...
  13. board
    I live in Spain, so I'm not coming.
    If you're so intent on doing it in person you can come here (seriously).
    But we've all proposed an easier solution for you: Convert a 24 bit file to 16 bit, do an ABX test and post the log here.
    And you didn't answer my question. I assume you probably won't this time either, but I'll repeat it for you:

    What, if anything, can make you change your mind about your claim that you can hear a difference between 16 and 24 bit?
    Or to be more specific, what, if anything, can make you change your mind about your claim that you actually heard audible differences, and can demonstrate it in a blind test, between those particular 16 and 24 bit files that you claim to have listened to?
    Here is my point set up a google share point or any way we can share some files
    Give me the files to comment on
    I promise to do the test as given. I think many here should participate in this. set my rules to follow I would love this really would. I’m Human as such not perfect period. Please do this. Of you wish I’ll set up the share file location even if we email it.
    Lets do this please
    castleofargh likes this.
  15. board
    We will do this on one condition, and only on this one condition: That you use Foobar's ABX plugin to do an ABX test with 16 trials (or more if you wish) and that you post your log and provide other kinds of evidence that we might ask for in case we suspect foul play (although by using Foobar's ABX plugin there most likely won't be anything to ask for).
338 339 340 341 342 343 344 345 346 347
349 350 351 352 353 354 355 356 357 358

Share This Page