Why Distributing Music As 24-bit/192kHz Downloads Is Pointless 841
An anonymous reader writes "A recent post at Xiph.org provides a long and incredibly detailed explanation of why 24-bit/192kHz music downloads — touted as being of 'uncompromised studio quality' — don't make any sense. The post walks us through some of the basics of ear anatomy, sampling rates, and listening tests, finally concluding that lossless formats and a decent pair of headphones will do a lot more for your audio enjoyment than 24/192 recordings. 'Why push back against 24/192? Because it's a solution to a problem that doesn't exist, a business model based on willful ignorance and scamming people. The more that pseudoscience goes unchecked in the world at large, the harder it is for truth to overcome truthiness... even if this is a small and relatively insignificant example.'"
Pro recording (Score:2, Insightful)
I record my performances at 96 kHz sample rate, I have to say that the music sounds much better at 96 kHz than 48 kHz I think (feel?) because the higher sample rate gives audio effects like reverb a lush, deeper sound.
The more sample units per second give the effects more to work with, in addition, even though you can't hear above and below certain frequencies recording those inaudible frequencies has an effect on the final product.
You may be able to find some scientific proof of this but for me it's an ear thing, higher sample rates sound better.
Re:The article writer is a deaf idiot (Score:5, Insightful)
>There is a huge problem with file sizes
Not any more, pumpkin.
We hit the terabyte size in drives a couple of years ago. There's no reason to be buying this format vs "archive quality" cd-audio or other lossless.
Buy/rip lossless. Transcode to lossy as needed. Anything else and you're being ripped off.
I listen to real music with real instruments. The "swish" you get in high-frequency percussion with lossy algorithms is annoying as fuck.
--
BMO
Re:The article writer is a deaf idiot (Score:5, Insightful)
The bit depth does matter (Score:5, Insightful)
However, there is an increase in quality using 24 bit. Most people just assume increasing the bit depth is the same as increasing the sample rate, but this is incorrect and short-sided. With higher bit depths, you can get your analog components operating a little further away from the noise floor. This also makes dithering much less noticeable (the noise you hear when you crank the volume up as a song fades out). Why? There are more "levels" for each sample to be recorded into. It's like going from 16 to 24 bit color. You would notice this.
For the 192 KHz fans out there, there is direct and proven mathematical reasoning for why 44 KHz audio is plenty. That, and your equipment probably can't produce it. Your converters probably suck at this frequency, and your ears definitely can't vibrate that quickly. More samples doesn't "smooth out" the waveform.
Re:The article writer is a deaf idiot (Score:5, Insightful)
When you can tell the difference between 44.1/16 and 192/24 in a double blind trial, come back and we'll talk.
Subjective opinions about audio quality, particularly those accompanied by words like "deaf" or "idiot", are worse than useless. Subjective listening is deeply suggestible and unreliable. Claimed differences among any acceptably well designed audio electronics virtually always disappears under rigorous and controlled testing.
To give just one example, listeners reliably prefer the louder source in subjective testing, even if the difference is not consciously perceptible. If a 192/24 D/A is just 0.1db louder than a 44.1/16 source, listeners will tend to describe it in all sorts of subjective terms... "edgier," "richer," "more forward," "cleaner impact," "deeper soundstage" etc when in fact it is simply a little louder.
Re:The article writer is a deaf idiot (Score:3, Insightful)
If you're sure you can hear a difference, why don't you ABX and prove it (or give strong evidence for it)? It's easy to hear a difference if you think you're supposed to, or if you paid a lot of money for speakers, etc. But its a lot harder to hear differences if you're doing a double blind test.
It's certainly OK to allow your emotions to take over if it makes you feel better to know you're listening to 24/192, but that's different than there actually being a perceived difference. You feeling better listening to 24/192 is an opinion, but whether you can actually perceive a difference is fact; lots of people confuse the two, so don't feel too bad.
Re:Pfft. (Score:4, Insightful)
Doood ... just, dood. You originally posted this, word for word, elsewhere (http://www.investorvillage.com/smbd.asp?mb=1911&mid=10609989&pt=msg). Either you are a bug-eyed alien, a prankster, or a combination of the two.
For those who aren't in on the secret, you can look up "rotational velocidensity" -- on the Urban Dictionary. It is the supposed loss of bits in a file over a time, which is absolutely ludicrous. Digital is digital. It's ones and zeroes. Files stored digitally don't degrade, unless you're talking about media degradation (ex., CDs and DVDs can possibly suffer from loss of data over time).
Dood also talks about files "repairing themselves," which is somewhere south of ridiculous.
But enough of this. I fell for it and actually answered it.
("Digital dust." Heh.)
Re:The article writer is a deaf idiot (Score:5, Insightful)
Re:Pro recording (Score:4, Insightful)
44.1kHz will be able to capture the basic information of the signal, as the human ear can hear to 20kHz in some cases, and Nyquist's theorem says that to recover the information you need to sample at least double the highest frequency. Oversampling (i.e. 192kHz) allows much more room to develop a good anti-aliasing filter. It may be that the reverb is phase-shifted somewhat with standard AA-filters, but ones designed for the higher sampling rate can have more linear phase. Also, higher sampling rates allow for better reconstruction of the actual wave form, if you're interested in music rather than just information. So yes, sampling a telephone call at 192kHz would be stupid, but if you're an audiophile, doing it for music is quite reasonable.
Re:Audiophiles (Score:5, Insightful)
For the rest of us on /. haven't we had all of our music in FLAC for a decade now? I don't even listen to music much and mine is.
My music is mostly stored in whatever the default is for YouTube videos that I've saved locally. I'm apparently even less of a music fan than you are.
Fun fact: I'm also an audio technician. Yes, I can hear the occasional damaged sound, but I'm not enough of an asshole to care.
Re:The article writer is a deaf idiot (Score:5, Insightful)
Indeed. One of the overlooked but highly important issues with sampling rates is that although you can represent up to Nyquist in a periodically sampled signal, that is the limit for infinite length recordings. For finite-length recordings, it isn't all or nothing, represented perfectly or not at all -- instead the uncertainty (read: representation error) increases as you approach Nyquist.
Too bad Shannon and Nyquist are dead. It seems they've completely misunderstood the math. How embarrassing they passed on before you could correct their mistake. Now they'll never know.
Re:The bit depth does matter (Score:4, Insightful)
I would say that theoretically, 44 kHz is enough, but in practice the filtering is a bit of a PITA. WIth 48 kHz, you can use shorter filters and it's much easier to convert to-from other widely used sampling rates (e.g. 8 kHz and 16 kHz for telephony/VoIP). Otherwise, I fully agree that 192 kHz is totally stupid.
Re:The article writer is a deaf idiot (Score:5, Insightful)
It really doesn't matter (Score:4, Insightful)
Re:Can we stop using the word "truthiness," please (Score:3, Insightful)
I think Truthiness covers half truths too. A half truth is that 24-bit/192kHz audio is higher quality than 24-bit/96KHz audio.
The whole truth is that only your house cat would be annoyed at 96KHz, or an audiophile dog.
Re:The article writer is a deaf idiot (Score:2, Insightful)
The point is that the 44KHz 16bit track has already been compressed from the original recording. However you rip that track, lossless or lossy, it doesn't matter; you're still not getting the original track.
Actually - it wasn't compressed - it was the limits of the recording equipment at the time. 192KHz/24 bit wasn't common in the 80s.
Knowing this, it doesn't mean that the tracks some sites are selling as 192KHz 24bit are from the original sources, or will even sound better, either. The original track could have been recorded with bad equipment or settings. In other cases, when doing comparisons on CD tracks vs high resolution tracks from sites like HDTracks, you can sometimes find that the HDTracks track is just the CD track with increased reported resolution/file size - possibly due to the inability to acquire the original material, though it could also be as simple as pure greed and laziness. Not that all of the albums on those sites are fakes, but a few of them have been found to be ripoffs.
Unless the track are genuine 192KHz/24 bit tracks, that is true. CD tracks can sound as good or better than 192KHz/24 bit tracks, it all depends upon settings. CD tracks can also sound worse than 92KHz encoded MP3s, again, it depends upon settings.
There's also the fact that it's extremely unlikely anyone can tell the difference between an encode at 96KHz vs 192KHz. If they are both properly encoded from the same source, it's unlikely there will be any audible difference between them.
This, however, is patently false. Given appropriate equipment and a person with a reasonable ear (mine aren't even that great and they suffice) and you can definitely tell the difference between 92KHz and 192Khz, and even straight CD tracks they were encoded from. It does require that the original source have enough depth that something is lost, however. Simple electronica, or other music that samples heavily from trivial sources will not provide enough depth to tell.
At this point my entire collection is lossless (CD quality at a minimum), and yes, it even makes a difference in my car, which has a halfway decent audio system. The other vehicle needs new speakers and an amplifier, the former sound blown and the latter was never clean to begin with, enough so that I pretty much haven't listened to music in it in years, just haven't gotten around to replacing it as it was only short trips anyways.
Re:The article writer is a deaf idiot (Score:5, Insightful)
Re:The bit depth does matter (Score:4, Insightful)
as a current audio engineer (doing dac's and spdif circuits), let me inform you that 88.2, 96, 176.4 and 192 are well alive and working well and showing some really impressive test measurements.
I can't hear any better than cd redbook (even then its better than my aging hearing) but I sure can see it on the test gear I use to design my own gear with.
its cheap, too. wolfson dac chips are $10 or so, give or take. that's a current high-end pick and it tests very very well. so well that most analog buffers can't keep up and many power supplies are not low noise enough.
I do agree tha 192k is overkill for final delivery. I also shoot photos and I downmix to 8bit jpg but I insist on getting 24bit raw images, doing all my processing at 24bit color, then finally going down to 8 again for jpg saving. audio is exactly like that, too.
but in photo, you are either slim (8bit jpg) or really a pig and taking up far too much room. in audio, there are many grades. you can be 88.2 (relative of 44.1 cd) or 96k (never intending to go down to cd silver disc format). you can record at a multiple of 96k (first multiple is 192k) and then downconvert to 96k for user distribution.
dacs at 24/96 are a GOOD break point for performance (chip and circuit) and cost. files are not that big at 2496 either, really. 192 is nuts for end users but 2496 is quite good.
Re:The article writer is a deaf idiot (Score:5, Insightful)
Fair point. The people who go on about 24/192 probably don't really listen to the kind of music which is affected by the loudness war. Most audiophiles I know are heavily into jazz or classical music, the recordings of those usually try to be quite faithful to the original.
Re:The article writer is a deaf idiot (Score:4, Insightful)
1. Find post asking for results of a properly conducted double blind test.
2. Ramble on about your various stereo equipment for a couple paragraphs, show a complete ignorance of confirmation bias.
3. Completely fail to provide the requested evidence, wasting every ones time.
4. ???
5. Profit!
Re:It really doesn't matter (Score:4, Insightful)
16 bits isn't enough dyanamic range, sort of. (Score:3, Insightful)
If it weren't for the fact that all popular music has its dynamic range compressed to provide maximum loudness for the entire song, dynamic range would be be a problem.
The problem is that, on soft passages, where the high 8 or 10 bits are zero, you're listening to 8 or 6 bit audio. That quantization can be heard. This is a problem for classical recordings made without any dynamic range compression. Of which there are very few.
This is an issue only if you listen to classical music in a very quiet environment. It doesn't matter for car audio. It doesn't matter for Apple's trendy crap earbuds. So almost nobody cares.
Re:The bit depth does matter (Score:2, Insightful)
I'll grant to you that for most people AND for most kinds of recordings, what you say is absolutely true.
"For the 192 KHz fans out there, there is direct and proven mathematical reasoning for why 44 KHz audio is plenty."
Both you and the article say this, but my understanding of sampling theorem differs from the conclusions you both draw. The main issue is that Nyquist sampling theory is based around the idea that you are moving from a continuous to a discrete path in one dimension only. The theory that you can reconstruct frequencies perfectly is based around the y-axis being continuous. In digital audio, it isn't. So both of the graphs he has in the section "Sampling fallacies and misconceptions" are actually incorrect; a proper graph would show the "stair steps" being slightly off-center where the line goes (and off-center by different amounts). In fact, that he equates bit depth with dynamic range shows he really doesn't understand the mathematics of PCM audio very well at all.
What's more, despite the article author's excellent description of how we hear, he never really connects the FFT the ear performs to how it limits the effectiveness of anti-aliasing, and assigns to anti-aliasing algorithms magical properties that they don't have; heavy metal and string orchestra music in particular represent worst-case scenarios for anti-aliasing algorithms.
So the math IS clear, but it doesn't show what either you or the article author think it shows. It may not justify 192kHz, but it definitely justifies a sample rate greater than 44.1kHz for certain kinds of music.
Re:The bit depth does matter (Score:4, Insightful)
First, most studio masters are 48kHz. Finding 96kHz or even 192kHz mastered audio is HARD. The range and selection of media capable of those sample rates is extremely low. Maybe under 100 Blu-Rays have 96kHz audio tracks, and far fewer have 192kHz tracks. And 96kHz has been around since the DVD days, and we still get audio mixed at 48kHz.
They do, however use 24bit sampling.
As for why go 96kHz or 192kHz, it's quite minor. For this, we need to explore sampling theory.
First, you have an analog signal. Then you MUST pass it through a low-pass filter (called an anti-aliasing filter) that bandlimits the input signal so it doesn't exceed the Nyquist limit (which will cause aliasing in the sampled waveform).
The trouble spot is the analog filter. If we assume that human hearing stops precisely at 20kHz, at 44.1kHz, we have to have a filter that basically has a stop band from 20kHz to 22.05kHz. It takes a lot of work to do this and the filters tend to be pretty big if you want to achieve filters that have flat passbands and low phase-distortion.
At 48kHz, you have a stop band of 20kHz through 24kHz, which makes for a much easier design. At 96kHz, you have a LOT of stop band. Enough so that you can perhaps set the passband higher (you have to block frequencies above 48kHz, so you can start your stop band somewhere between 24-25kHz which should cover the majority of people's hearing. And you'll have a whopping 24kHz or so for the stop band, making for a very clean filter with gentle rolloff (which generally gives you better passband performance - flatter response on the pass band, and very low phase distortion).
At 192kHz, that's really getting excessive - even if you set the pass band at a ridiculous 48kHz to cover every possible human and dog, there's a pile of bandwidth available for the stop band.
96kHz audio may sound better if you're young (or a dog), but a good chunk of the older population has hearing that rolls off starting around 16kHz or so.
Hence why the vast majority of works are sampled at 48kHz - it really is good enough and those that can hear ultrasonic will lose the ability in a few years.
Re:The article writer is a deaf idiot (Score:5, Insightful)
Not wanting to go deaf, I use high quality devices with low THD percentages so I can listen at lower volume with maximum impact. Most people don't realize that high volumes are much less necessary as noise is removed and SNR goes up. With a very low noise level, you can play music at relatively low volumes that sounds incredibly good, whereas the high THD injection from a pair of crappy headphones or terrible stereo will cause you to turn up the volume repeatedly to counteract the noise.
Re:Pro recording (Score:5, Insightful)
When I listen to music, its not for the data -- its for the feeling. You should try listening to music for the feeling too ;-)
My opinion.
Re:44KHz (Score:2, Insightful)
Except some of us have tested our hearing well up to 24kHz. Trust me, its annoying more often than pleasant. Things that are thought to be inaudible simply aren't (excuse the double-negative).
I'm reminded of the science of vision ... and how eventually it was discovered that in fact various colour cones could detect light from neighbouring colours, making RGB reproduction of cyan for example impossible. Then the fact that some people (almost always women) have twice the red sensitivity, and see colours quite differently than others.
The science of hearing hasn't even come close to our understanding of vision, and neither is perfect.
Sampling rate (Score:3, Insightful)
96KHz isn't the audio frequency. It doesn't mean that the audio contains a 90Khz tone. It's the sampling rate. The higher the sampling rate smoother the signal.
Human perception wise a audio signal recorded at 96KHz sampling rate might well be indistinguishable from one sampled at 192Khz, but so is the file size between these files for practical purposes. I don't deceive my self thinking that I'm hearing better sound from a 192Khz file, specially considering that I'm using a basic pair of headphones on a my basic phone to listen to them. But my thinking is that future technologies might let you do interesting things with the extra bit of data which is useless to us right now. So given the choice I opt to get the higher sampled versions. Kind of like with digital pictures which are too noisy or blurred, but which might be cleaned up with future algorithms to give us a slightly more useful picture.
Re:The bit depth does matter (Score:4, Insightful)
However I've always doubted it as, it's can be defeated with a pen and paper.
Basically, what you're saying is that you have no background in maths, but you can disprove a very well known and thoroughly proven theorem by sketching lines on a peice of paper.
You can also draw a triangle with bent edges to disprove Pythagoras too, if you like.
You can also disprove Fermat's Last theorem by showing 1782^12 + 1841^12 = 1922^12 on many common calculators, too.
It is well known that the Nyquist frequency (and that frequency only) cannot have the phase or amplitude reconstructed correctly. *every* *single* frequency below that can, no matter what you think your bar graphs look like.
22kHZ will not be reconstructed correctly. 21.9999kHz will be, and that's still above the threshold of hearing.
Since you've gone to the effort of drawing them, now draw them after they've run through an analog filter. That's a little bit harder...
The sampling errors you refer to add noise over the entire frequency spectrum. This is well known and the article even addresses it very obliquely (noise floor).
Re:Pro recording (Score:5, Insightful)
My favourite audiophile rebuttal quote:
"If your hifi costs more than your music collection you have missed the point." - Unknown Source
Re:Pro recording (Score:5, Insightful)
The problem with low-pass filtering was resolved eons ago with a concept called "oversampling."
Only the earliest and ruddiest of CD players (and a lot of computer sound cards) had a brick-wall filter at ~22.5 KHz. The rest of them resampled the input by 4x or 8x, or converted the original signal to PWM, and then applied the anti-aliasing filter at a frequency several octaves above the range of human hearing.
This hypothetically pushed the nastiness inherent of a steep filter to a realm well outside such that humans could hear, and at least far beyond the limited confines of a CD.
Welcome to 1985, where your stated concerns are both accurate and already solved.