Can You Really Hear the Difference Between Lossless, Lossy Audio? 749
CWmike writes "Lossless audio formats that retain the sound quality of original recordings while also offering some compression for data storage are being championed by musicians like Neil Young and Dave Grohl, who say compressed formats like the MP3s being sold on iTunes rob listeners of the artist's intent. By Young's estimation, CDs can only offer about 15% of the data that was in a master sound track, and when you compress that CD into a lossy MP3 or AAC file format, you lose even more of the depth and quality of a recording. Audiophiles, who have long remained loyal to vinyl albums, are also adopting the lossless formats, some of the most popular of which are FLAC and AIFF, and in some cases can build up terabyte-sized album collections as the formats are still about five times the size of compressed audio files. Even so, digital music sites like HDtracks claim about three hundred thousand people visit each month to purchase hi-def music. And for music purists, some of whom are convinced there's a significant difference in sound quality, listening to lossy file formats in place of lossless is like settling for a Volkswagen instead of a Ferrari."
Depends on the bitrate (Score:5, Informative)
Usually if the bitrate is above 256kb/s, i dont notice any difference.
Ofcourse it still effects some songs (especially the percussion parts).
44.1khz ought to be enough for anyone... (Score:5, Informative)
We recently discovered [arstechnica.com] that human hearing beats the linear response assumptions used in lossy codecs. So yes, their criticisms are scientifically founded.
Debunked (Score:5, Informative)
The concept of improving consumer listening experience using studio quality recording has been thoroughly debunked, right here on Slashdot...
Why Distributing Music As 24-bit/192kHz Downloads Is Pointless [slashdot.org]
Re:Depends on the source (Score:5, Informative)
Your preference for 24/96 audio as a listener is entirely due to the placebo effect.
Well, in all fairness, listeners may actually hear perceptible differences between 24/96 and 16/44.1 audio sources due to different mastering, but of course that says nothing about whether they can actually tell the difference between the two bitrates when everything else is equal.
This article [xiph.org] is a pretty good explanation of why 16/44.1 is as good as anyone needs for playback.
Re:Depends on the source (Score:5, Informative)
You sure can hear the difference if you stick a 44.1kHz DAQ in a 96kHz signal chain before filtering out ultrasonic high frequency components (if there are enough to make a difference). The advantage of 96kHz recording isn't that it can capture any more human-audible frequencies than 44kHz can, but that you have a lot more leeway to prevent aliasing of signals above the Nyquist limit down into the audible range (a 25kHz tone sampled at 44kHz results in a spurious, highly audible (25-44/2)=3kHz aliasing signal).
It's pretty much impossible to build analog frequency filters with a sharp cutoff (e.g. everything below 20kHz and below gets through, everything above 22kHz is -60dB attenuated), so recording at 44.1kHz sampling requires either being absolutely certain the original sound source has minimal high-frequency harmonics, or heavy analog filtering that cuts well into the audible high frequency range. With 96kHz sampling, it's much easier to build an analog filter that gradually rolls off high frequencies between 20kHz and 40kHz (...producing a >40kHz sound is tricky in the first place), preventing aliasing without the filter cutting into the audible range. Once digitized, it's trivial to make a *digital* filter with a perfect frequency cutoff to downsample the 96kHz to aliasing-free 44.1kHz.
Re:Depends on the source (Score:5, Informative)
44.1hkz 16bit audio is completely transparent to the human ear. No one has ever been able to detect when a 16bit DAC ADC pair has been placed in a 24/96 audio path.
Your preference for 24/96 audio as a listener is entirely due to the placebo effect. There are good reasons to master audio in high res, but for listening 16 bit 44.1khz audio is as good as anything.
As a former audio professional (specialized in location recording of choirs and orchestras) I must agree. But even my aging ears can hear the difference between 44.1 (or 48)kHz 16 bit uncompressed and a typical MP3. Side note: 24-bit has a few audible advantages for music with extremely wide dynamic range (from ppp to fff, say) where 16 bit will struggle a little at the very soft end.
Re:Better question (Score:5, Informative)
This is the real point: People are so used to listening to music with no dynamic range, on ear buds, in crappy acoustic environments that they wouldn't know where to start listening for a difference.
Re:Depends on the source (Score:5, Informative)
Speaking as someone who frequently does recording, your comment suggests that no one has done that test with classical music in a properly controlled listening environment using quality gear while giving the test subject the ability to control the volume arbitrarily. When you crank up the volume, the noise floor difference in soft passages alone should make the difference between 16-bit and 24-bit signal paths a dead giveaway, even for someone with moderate to severe hearing loss. It isn't even subtle. Of course, if the person doesn't turn it down for the loud passages, he/she will likely suffer hearing damage, but perhaps that's why he/she has moderate to severe hearing loss in the first place. :-D
The 44.1 vs. 96 kHz difference is more subtle, requiring someone with top-notch hearing (very rare), headphones that can accurately reproduce frequencies above 20 kHz, and 96 kHz DAC hardware that does not have a bandpass filter starting at 16 kHz. If you fail to verify even one of those requirements, you would expect no one to be able to hear the difference, because there won't be any difference.
Re:Depends on the source (Score:5, Informative)
No, not at all like 640K.
Re:Depends on the source (Score:5, Informative)
The point of the equipment is that you have quality in reserve as you go through the process of mastering the tracks. The more quality you have in reserve the more you're able to do before you start having to deal with artifacts and other nastiness. As with all such things, you have to think about the order in which you do things and the order in which you throw out data to get the best results.
The point of buying lossless music isn't so much that it's better for listening to, it's that you can compress it however you like later on without having to worry as much about the sound quality you get. Since you have more data to work with, you can get a better quality at a lower bitrate than if you were starting with an already compressed track.
Re:Depends on the source (Score:5, Informative)
If you have ever gone to a rock concert and been near the front or gone to most dance clubs and you will have sustained hearing damage. If you have ever left one of these venues with ringing ears, or been around loud machinery and noticed the same, then you have sustained hearing loss. Your hearing will recover mostly after the trauma and that will be indicated by the subsiding of the ringing of your ears.
If you want to find out how your good/bad hearing is, spend the money and see an audiologist. You will be surprised on to find out what your hearing is really like.
Re:Better question (Score:5, Informative)
Good point. Sadly, my $3k hearing aids don't seem to help either.
Bitrate doesn't matter much if your ears are the lossy part.
Re:Depends on the source (Score:4, Informative)
According to Wikipedia the audible range for human hearing is around 130dB. 16 bits can in best case offer a dynamic range of 96 dB, whereas 24 bits offer 144 dB.
So it should be pretty obvious that you can't fit the entire audible range into 16 bits. This might not be relevant to modern day music. But if you want to record what the ear is actually capable of hearing (not including sound levels above the pain threshold) you will need those 24 bits.
Re:Depends on the source (Score:3, Informative)
Two nits to pick:
1) You can get arbitrarily close but you can't get "perfect" frequency cutoff.
2) A 25 kHz tone sampled at 44 kHz gives you a 19 kHz tone. Remember the [-pi:0] (or [pi:2*pi]) frequency range comes first.A 41 kHz tone would get you a 3 kHz tone after sampling.
Otherwise all true, which is why most recording devices do exactly that, sample at a high rate and digitally filter before downsampling to 44.1. But none of that has much to do with whether or not, once you've gotten past the aliasing problem as you say, you can tell the difference between a 44.1 kHz playback and a 96 kHz playback.
Re:One word: YES. (Score:4, Informative)
Caveat: You have to have decent headphones (not Apple earbud BS), and/or good speakers, but that's about it. The difference is negligible once you hit ~320Kbps MP3, in my opinion, but anything under 256Kbps, regardless of lossy format, you can *clearly* hear cymbal hits turning to an underwater splooshy mess.
Highhats are even worse than cymbals. Even at 256 kbps, highhats tend to sound like they're being hit with a bag of broken glass, and is the easiest way to identify lossy compression I can think of. Except, perhaps, some of Mike Oldfield's earlier works.
Re:Depends on the bitrate (Score:0, Informative)
Generally, there is no such thing as lossless. I hear people keep using that word, but knows not what it means.
The great wisdom here is that this knowledge knows no bounds.
Captcha: localize
Re:Depends on the source (Score:5, Informative)
1) Digitally, yes you can. Take the DFT of the data; zero out all components above your frequency cutoff; reconstruct the signal as the sum of below-cutoff frequencies. Voila, a perfect sharp cutoff. The only subtlety is that you can only choose an exact cutoff corresponding to some integral number of cycles in your sampling window, so you can't cutoff at exactly sqrt(e*pi)kHz --- but you do have plenty of wave numbers from which to select a perfect cutoff (increasing with the size of your DFT window).
2) Untrue: a 44kHz *sampling rate* has a 44/2=22kHz Nyquist cutoff. Frequencies f>22kHz Nyquist limit "wrap around" to f-22kHz difference frequencies.
But yes, I agree, on the playback side there's no audible difference between a (sufficiently well made) 44.1kHz and 96kHz DAC.
Re:Depends on the source (Score:4, Informative)
When the music gets soft in 16b you have a lot of zeros in front of the number. So you effectively only have a three or four bit signal being fed into the DAC. This is fixed point math, not floating. With 24b you can put all of those zeros in the front and still have eight or more bits to feed into the DAC. This is even more beneficial when the amp implements power supply volume control. PSVC raises the effective noise floor the DAC has to deal with.
Re:Depends on the bitrate (Score:5, Informative)
I mean, if you're only listening to ear buds (even $$$ ones are limited in bass response, etc), or in a car (one of the worst listening environments conceived)....then sure it won't make a difference, and portability makes a lot of sense too.
However, in a nice listening environment, with good equipment...it is worth the effort IMHO.
For instance, I have a pair of Klipschorns [klipsch.com] ...paired with a couple of the much older models of the Decware SET amps [decware.com], running mono to each channel..plus an older 15" 800W Klipsch sub, etc......
Even with my older ears, I can hear differences in recordings and formats. Not as well as I used to be able to, but I figure, WHY would I want anything less than the best I can get for the given time/situation? When listening at home, I rip my music to flac, and have it play on my living room stereo.
And hey....kinda fun to watch the Flintstones in concert volume on tv too from time to time, or hell, once hooked the MAME machine to it....Robotron 2084 is fun with the room shaking around you.
God, my neighbors used to hate me when I live in a place where I had to share walls...
Re:Depends on the bitrate (Score:5, Informative)
You can actually practice listening to music, it's something you learn.
Sometimes the difference between two sets of speakers can be as little as one clarinet in the middle of an orchestral piece. On one set it sounds good, on another it doesn't (or it's hardly there at all).
It's not something you can pick out just by putting on a rap CD for ten seconds and turning the bass up to maximum in a store (which is how most "HiFi" systems are chosen these days and why the manufacturers produce so much garbage).
Re:Depends on the source (Score:4, Informative)
But yes, I agree, on the playback side there's no audible difference between a (sufficiently well made) 44.1kHz and 96kHz DAC.
No, but what makes a big difference is when you have a 48 kHz sound card that resamples everything to 48 kHz for an internal DSP stage that cannot be bypassed, and then back again. Yes, Soundblaster Audigy, I'm looking at you.
44.1 -> 48 kHz gives a lot more audible artifacts precisely because they're so close. Think of it as audible moire.
Also, for newer computer audio cards, if you have a choice, use 88.2 kHz for the internal rate instead of 96 kHz. The reason is that most high quality sound is in 44.1 which converts perfectly to 88.2. For 48 kHz, it's less of a problem in the first place, and likely also worse quality sound to start with.
Of course, unless the rest of the audio path is good, it doesn't matter much, but if you like to listen to FLACs with high end headphones, it sure won't hurt to use 88.2 instead of 96 kHz.
Re:Depends on the source (Score:4, Informative)
In a finite window, *any* signal can be represented as a sum of elements with frequencies corresponding to n=0 (DC offset), 1, 2, 3, ...., infinity integral cycles in the window. A signal corresponding to a non-integral number of cycles, e.g. 100.5, is indistinguishable over the window from some (infinite) combination of integral cycle waves. If you measured in a window twice as long, the 100.5-cycle signal would now be a unique, identifiable 201-cycle component. So, in an important sense, in a finite window the "intermediate" frequencies "don't exist" --- they can't do anything different from the (infinite series) of integral frequencies. Thus, you can create a cutoff that is as "perfect" as is meaningful in a finite window.
Re:Better question (Score:5, Informative)
I think the real point is that there are known limits to human hearing and many audiophiles fantasize about their hearing being superhuman. It just ain't so. Dynamic range compression is one thing, but perceptual compression, sample rate, and bit depth are a different matter. No audiophile has ever heard the difference between FLAC and 320Kbps mp3 audio in an ABX test at a statistical rate that is better than guessing.
Any time this argument starts, I refer people to this well written article [xiph.org] that lays out the limits of human hearing compared to the specifications of recording formats...
Re:Depends on the source (Score:4, Informative)
Actually, you've proven the GP's point. You can't tell the difference if you are listening to the program. Turning a program up in the "soft sections" is exactly what you should never, ever do when listening to a program. You may as well put on the IR headset with compression that came with your TV so you can watch late night TV without disturbing your wife.
Mastering is an entirely different ball of wax and, yes, you want all the headroom you can get. It's no different than photographers using RAW formats instead of JPGs (even lossless JPGs) out of the camera. You want all the bits you can get. But after your done mastering, dropping to 16bits isn't going to affect the outcome. That's the whole point of mastering - if we didn't want to be that soft, we would have engineered it to be louder.
Re:Depends on the bitrate (Score:5, Informative)
10 years ago, MP3 encoders couldn't encode decent cymbals and saxophones below 384kbps... it was just a stream of high pitched garbage.
These days they're both really good encoders. I still prefer AAC over MP3 just because the high freq nuances are better captured, but at AAC@256 and MP3@320, the differences are practically imperceptible to my ears.
The only time I'd look at lossless music is for Orchestral pieces. Compressed pieces still sound flattened and don't have the wideness because there's a lot more overtones, harmonics and variety of tones in live recordings. Microphones, recordings and engineering have adjusted in the past 5 years to compensate - so recent pieces are not too bad however.
Like anything, it's best to just try a few different methods and see what sounds best to you.
Re:Depends on the source (Score:5, Informative)
As a live sound engineer dealing with vocalists who do that regularly (sing at normal program levels and then BELT A PHRASE OUT)... let me say... ARGH.
I put a steep compressor on someone who's prone to doing that, and let me tell you, it makes my life much easier. I can't fix the clipping, but I can make sure they don't cause the audience to cover their ears.
Re:Depends on the bitrate (Score:5, Informative)
In my humble opinion, this old hoary debate will always remain a debate for several reasons. As you right mentioned, the reproduction environment in most cases is woeful at best. Most speakers are not even full-range to begin with, their cabinets resonate, their drivers cannot often keep up in complex multi-layered music, their passive crossovers do a half-assed job in distributing the sound to the various drivers, and so on. Then, the amps are weak so they start bottoming out and start clipping when the speaker impedance and phase dips sharply in certain frequency bands. Then the electronics, especially the capacitors and power supply cannot keep up. Then the cables are not fat enough or are not shielded enough so they load up the power amp even more. Then the pre-amp adds its own coloration to the already feeble signal coming from the source. Then the DAC does its own thing and further colors or degrades the source signal even more. Then the source adds its own share of noise and jitter to the audio signal that screws up not just the signal quality (bad enough) but even the timing of the music.
On top of it, the room comes into play. The room adds its own coloration and effect that is often a far bigger factor that the audio system itself - boosting certain frequencies while muddying and deadening others, and even adding echoes, reflections, etc.
Then there is the human being at the end of the chain. I personally can't even listen above 16KHz, and I have average ears. I suspect many people are like me too, at either end of our audible spectrum. On top of it, we humans hear music very differently - while our audio range may be fairly similar (20hz to 20khz by popular definition), our sensitivity to *variations* in tone and timing varies drastically - many often have off the charts sensitivity to even slightly off-key music (I do) or slightly off-beat music (I do not at all).
All in all, a decent headphone setup is far far more revealing than a decent audio setup. At a thousand dollars, you can probably assemble a decent headphone, but an audio system will sound atrocious, unless you are willing to spend a whole lot more effort and research in second hand discrete gear OR are willing to do serious DIY.
Anyway - I also wanted to say one thing - the thing that gets neglected the most in all this is actually the quality of the source recording - or what people call "mastering".
Most people who say something like "SACDs sound far better than redbook CD" or "vinyl sounds far better than CD" are most likely saying this because a whole lot more care went into recording the SACD or vinyl compared to the cheaper mass market CD or mp3.
If I look back at all the albums I have purchased or listened to (in whatever format), the one thing that stands out to me personally is that I have found less than 10% of them to be "recorded with care". And I'm not even being picky! Across the board, I can say that recording quality sucks when it comes to rock (which is what I listen to most often) - and I mean all kinds of rock.
If Neil Young's initiative (and even his Pono device) and Dave Grohl's initiatives are successful in improving the audio quality of music in general, I strongly suspect it will be because recording quality will be done with greater care, not because they decided to use a fancier digital format or use higher number of bits and samples to store their music. While everything becomes a factor by the time the music reaches your ears (heck, by the time it is processed by your brain, you even have to factor in psychoacoustics and gear bias and the "burn-in" syndrome) - the recording quality in general needs to improve (except for the jazz and classical pieces that audiophiles love to love, and are hence recorded with care), and this improvement will arguably make the biggest difference in audio quality.
Re:Depends on the source (Score:4, Informative)
Re:It doesn't matter (Score:3, Informative)
FLAC is about an order of magnitude simpler than MP3. I once implemented a decoder in about an hour over lunch just because I could. And because many lossless codecs feature error detection, they're much more likely to survive as a long-term archive than something like MP3 which doesn't even have a container or any reliable way to verify that the file's contents are correct.