Does Portable Music Have to be Compressed? 540
FunkeyMonk writes "The Christian Science monitor has an article discussing the gap between music fans and audiophiles when it comes to portable music. Would you pay a few cents more to have lossless downloads from iTunes and other online music retailers? As a classical musician myself, I choose not to download most of my music, but rather rip it myself in lossless format."
more for non-DRM (Score:5, Informative)
Actually I'd like to be able to get an "original" image a la the CDs you buy, but allow single CD tracks. Would I pay more for that? I don't know. I've never bought any of the DRM'ed crap because it's DRM'ed, so I don't know how badly (or well) compressed they are.
If there are audible compression artifacts anywhere in today's downloadable DRM'ed music I'd probably insist the compression be less or not at all, after all I'm paying for music, and a compression artifact (to me) is analogous to stuck pixels in a monitor or camera... my threshold of tolerance is zero for that.
(I had one of the very original SONY Mini-disk recorders, and remember a passage of a Doobie Brothers track where some high pitched bells instead of sounding like high pitched bells sounded like someone sneezing... unacceptable... completely altered my experience of MD (along with numerous other things about SONY).)
So, bottom line, DRM aside, I consider it the responsibility of the music industry to deliver what they claim they are delivering... music (usually). I'm willing to bet what they are delivering has artifacts... I wouldn't pay more to get rid of that, I'd demand they replace the defective product.
The nice thing about my CDs and my derivative mp3 collection (recorded at 320 VBR) is if I hear an artifact in my track, I have the unedited original, I rip it at higher quality until the artifact isn't there.
(As an aside, I think the article makes an exceptionally great point not directly related to the users:
So, in addition to short-shrifting consumers with less-than-perfect (to the ear) product, the movers of downloadable music thumb their noses at the collective profession of sound engineers and engineering... pretty rude.
Granted, a lot of the music out there is crap -- it's no justification for compromise on the medium.
Oh, and re the subject line of my post... I'd pay a little more for non-DRMed music, not uncompressed music.
Re:GIGO -- Garbage In, Garbage Out (Score:3, Informative)
What you really should get all in a knot about is the continously low quality of shite music being promotoed. Payola's a bitch.
Tom
Re:Lossless audio file downloads? Hell no. (Score:2, Informative)
Double blind test (Score:5, Informative)
"Audiophiles" like to make all sorts or ridiculous claims that lead to things like $2000 speaker cables, gold CDs and just a general proliferation of nonsensical technobabble.
Psychology simply has too strong of an effect on questions like this to get an actual answer from a forum like this.
What you'd really find is that as the bitrate of an mp3 goes up, the number of people who can tell the difference goes down. At some point the number of people who can tell the difference becomes a statistically insignificant sample. This would be a good project for some grad student.
Re:What's the point? (Score:1, Informative)
Mr. Goddard need to get with the program (Score:4, Informative)
Lossless is coming soon to most of us. With the 5.5g iPod at 80GB and the Zune hackable to 80GB as well, all but the top 3-4% of all consumers can fit their entire (legal) collection on a single portable device in lossless compression. I've got about 6500 tracks, most as FLAC rips, and I'm right about 81GB (plus about 40GB in books, but those are all low-bitrate). If I jettisoned the extra downloded stuff I have that I didn't like (but didn't get around to deleting), I'd probably drop to 75GB or so. I suspect that my entire family (three of us) buys less than 5GB worth of content each year. There's no reason to expect that the size of the players, in capacity, will not continue to decrease. As for those with bigger collections...well, just get more portables, or learn to live with a smaller subset on your player (or a higher compression).
As long as the high-qualtiy masters are available, portables can become a calculated compromise. Since my threshhold for accuracy happens to be at about 256kb/s LAME, that's where I transcode my FLAC library for my portable. If I had a car player, it would probably be more like 160kb. Heck, it's practically impossible to hear artifacts at 128kb in my Pilot at 70mph at a normal volume. My wife's 8GB flash player will be encoded in the 160-192 range, becuase I know she doesn't have the gear to hear much more, and she's just not that picky. With good music managers, you can automagically sync and transcode at the same time (I use mediamonkey). Transodeing is a bit slow right now, but as PCs get faster, the sync/transcode process will get better and better.
I do agree that it is a travesty that the online services will not offer home-archival-quality tracks, but I'm probably a top-10% listening geek. I buy all my music on CD, and rip to FLAC. Okay, okay - I've bought some at AllOfMp3.com, too, but I can get lossless there. The key is that the studios will continue to have qualtiy masters - but will they be willing to sell that quality to the public?
Re:GIGO -- Garbage In, Garbage Out (Score:2, Informative)
Who the hell uses 128 kbps MP3 anymore? If you use iTunes, like a sizeable group of mainstream consumers, then you're getting 128 kbps AAC, which is indistinguishable from the source when it comes to loud, over-compressed pop music. When it comes to something like classical, that's when you probably need to move up to 160 or 192 (which iTunes doesn't offer, unfortunately). I don't have a clear idea of wma's quality, which is the other mainstream consumer digital music format. My point is that you probably have nothing to worry about concerning MP3 becoming the standard format, at least through official means of distribution. After all, it's too hard to DRM it and lock your customers into one unshiftable format and player.
That said, I really like Bleep [bleep.com], which distributes music in non-DRM, high-quality VBR MP3 and sometimes FLAC, both of which create sample-perfect representations of whatever's encoded with it.
Re:Lossless is compressed (Score:3, Informative)
Re:Lossless is compressed (Score:5, Informative)
I'm not arguing that a lossy encoding of CDDA is as good as CDDA; it isn't. Just that there's no law of nature establishing CDDA as the gold standard in the first place.
Lossless vs. Good Lossy -- We've Tested It (Score:5, Informative)
I became curious about just how the various compressions stacked up against each other. I knew Vorbis was better than "normal" MP3 by a long shot, but newer MP3 variations have definitely gotten better. Here are the formats tested: WAV (straight from the CD), FLAC, Vorbis, and about 15 different MP3 variations (VBR, CBR/ABR, 32k to 320K). I tried both down-convert from FLAC and ripped-direct-from-CD (there should be no difference, and I certainly couldn't hear any). This was done on a variety of material, choosing particularly demanding/revealing passages from acoustic guitar, cafe jazz trios, brass ensembles, Beethoven's 6th, piano (jazz and classical), rock and vocalists (Streisand, Baez, Queen - Bohemian Rhapsody).
I did a few tests and verified that I could not distinguish between WAV and FLAC -- no surprise there -- so for convenience the other formats were compared to FLAC as the baseline.
I did extensive A-B, B-C, A-C, etc., etc. comparisons using my main system (Marantz A/V amp with Magneplanar MG-IIIa speakers) and also with Sennheiser HD595 headphones. Below 128k, MP3 is complete crap. Starting at 128-CBR, it got more difficult to hear the difference. At CBR/192 or VBR/medium, I could rarely distinguish MP3 from FLAC, although sometimes the high-hat cymbals sounded like they lost a little bit of brilliance.
Although I'm a fairly discerning listener, I do have high-frequency hearing damage in my right ear. So I brought in a friend who is a serious audiophile. We did a lot of listening and comparing (many hours over several days because your ears get "tired"), both on my system and back at his house.
The Verdict: Vorbis is good, really good. But MP3's produced by Lame at VBR/Medium to VBR/High are also really, really good, maybe even better. MP3/VBR/Medium is approximately the same size as Vorbis/Normal (-q 4.99) at about 1MB/minute -- 1/5 the size of the FLAC files. Although there are players out there that can handle Vorbis, there are many more that don't.
Ps. We're not going to throw out the FLACs, because something better *will* come along. By that I mean 'smaller than' MP3/VBR/HIGH.
Re:Double blind test (Score:4, Informative)
Re:Lossless is compressed (Score:5, Informative)
Ahem, http://flac.sf.net/ [sf.net]
A used for Magnatune downloads (among others), and supported by decent media player software and a handful of MP3 players
Re:FFS shut up already (Score:5, Informative)
Every encoder will generate ringing and other artifacts. Every good encoder tries to put those artifacts just a bit below the hearing threshold according to an algorithm that has been tested extensively with normal music. However, encoders are generally not fine-tuned to deal with the unnatural type of noise that results from another encoding process, resulting in the noise ending up above the hearing threshold after the second time.
You might wish to check some double-blind test results on HydrogenAudio [hydrogenaudio.org]. Short version: reencoding 256 kbps MP3 to 128 kbps MP3 sounds horrible compared to 128 kbps MP3 straight from the lossless source.
Re:FFS shut up already (Score:2, Informative)
Re:Lossless is compressed (Score:3, Informative)
My original post did say "show me non-lossy, non-PCM".
Re:Double blind test (Score:5, Informative)
Using up to date encoders, for the vast majority of people, for the vast majority of tracks, 128 kbps is indistinguishable from source.
Link. [maresweb.de]
Everyone should try to ABX at least once. You'll be shocked how much worse your ears are that you'd believe them to be... ABX Just Destroyed My Ego [hydrogenaudio.org] is a very informative read for any would be audiophiles:
Re:Lossless is compressed (Score:3, Informative)
But last time I checked, all sampled music was PCM, and that's lossy by definition. You're limited in the sampling rate and the bit resolution, which makes is lossy when comparing with the original (i.e. "real-life") source.
Then again, like my original post says, audio CDs are what most of us have to use as the "original lossless" source.
So no, FLAC isn't "lossy" in the MP3/AAC/VQF/WMA sense, but it is PCM, which my original post clearly pointed out (I asked for "non-lossy non-PCM". FLAC is non-lossy but is PCM.
Re:Double blind test (Score:3, Informative)
http://mtsu.edu/~record/ [mtsu.edu]
Speaking without detail is useless. (Score:3, Informative)
Taking an analog signal and representing it digitally is an application of Nyquist-Shannon sampling [wikipedia.org]. The important bit to understand (for those of you who've never heard of it), is that the Nyquist rate [wikipedia.org] is twice that of the sampling rate you want to record.
A 44.1Khz sampling rate perfectly records a 22.05Khz signal, 48 Khz does 24Khz, etc. Human hearing peaks out at 20Khz for most people, and many people spend a good chunk of their life destroying their upper hearing range with various tools (rock concerts, overly loud headphones, etc) anyway. 48Khz is marginally better, but 44Khz is more than enough to sample anything most people can hear perfectly.
"Let's not perpetrate the myth that music can be recorded losslessly in the first place. All sampling is lossy." -- so, since we're directly sampling (sector-by-sector) the raw bit values, or sampling a perfect reconstruction of a 22Khz signal, there is no loss either way (although the 2nd one has to deal with cables and other noise in the electrical system, since you pass through DAC -- analog -- ADC). At least, not loss humans can hear.
Re:Lossless is compressed (Score:2, Informative)
You are correct that it is impossible (even theoretically) to record music perfectly accurately...but this doesn't have anything to do with "lossless". CDDA encoding uses lossless compression. That means that it is a perfectly accurate representation of what was recorded, though obviously the recording is not a perfectly accurate representation of the sound wave.
This is an important distinction in that you can perfectly accurately convert between anything compressed in a lossless manner, but you lose accuracy every time you convert between anything compressed in a lossy manner. That is, I can convert CDDA->FLAC->Apple Lossless->CDDA over and over ad infinitum and still get exactly the same CDDA file. On the other hand, if I were to try this CDDA->MP3->WMA->CDDA, I'd end up with crappier and crappier reproductions.
Re:Speaking without detail is useless. (Score:5, Informative)
It's funny, I have an audiophile acquaintance who swears that records are superior in every way to "digital," and for the same reasons described above. The funny thing is, because of the large number of quantization levels used in a CD, the CD's dynamic range far surpasses that of any record player. More info here [georgegraham.com]
Theoretically, yes, analog would always be superior. But in reality, physical limitations of the stylus on a record player limit that medium far more than quantization limits the CD. Those same physical limits exist in the human ear, too.
So, while digital might not be "perfect" theoretically, it's "perfect enough" allowing for the limitations of the human ear.
Observation on music quality (Score:5, Informative)
common Nyquist fallacy (Score:2, Informative)
While I don't have any reference to give you, I find it a matter of common sense. If you sample a 1hz signal @ 2hz, you'll see consistent peaks & valleys, and the signal can be assumed almost immediately, after 3 samples (ignoring issues of quantized amplitude sensitivity over time). If you sample a 0.9hz signal @ 2hz, you'll see peaks & valleys alternating as before, but their amplitudes are both approaching zero, then cross zero, approach peak, and repeat. After analyzing this signal for a duration, you could assume it was a 0.9hz signal because of the relationship between the rate of amplitude change and the rate at which those amplitudes cross zero.. although this also assumes that you'd never see a 1hz signal simply increasing and decreasing amplitude at that same rate - considering this condition places stipulations on both frequency AND amplitude over time, whereas a 0.9hz signal only stipulates the frequency over time, we can only make a definitive assumption if we know the frequency doesn't change over time.
Hence, considering the frequencies are changing over time, we can't possibly accurately reconstruct an audio signal using a sample rate at twice the highest frequency, unless you get very lucky. As we consider a lower and lower highest frequency, our chosen sampling rate becomes more and more accurate, though I don't believe you ever reach perfect 100% reconstruction because of the irrational nature of true time-varying frequencies. One could, theoretically, calculate the accuracy of a given sampling rate for a given maximum frequency - I'm sure someone has at some point.
In fact you could analyze the typical audio signals that are digitized today, and develop some rough statistical analysis of how often a given frequency changes at a rate that could be interpreted as another frequency. This would likely vary depending on the individual frequency, the relative location within a song, and the musical genre. You could use these numbers to select an appropriate sampling rate to achieve N% accuracy of frequencies up to a X-hz maximum.
Re:common Nyquist fallacy (Score:1, Informative)
For example an AM radio signal is a signal that has its base frequency, the one you see on your radio, modulated by an audio signal. This means its amplitude changes with the audio signal, the effect on the frequency content of the signal is to spread it out around the original unmodulated frequency. The amount of spreading is related to the frequency content in the modulating signal.
Re:Speaking without detail is useless. (Score:3, Informative)
He's not talking about formats, he's talking about the way samples are recorded. Each sample is a number from 0 to 2^16-1. He's saying that human ears can't hear the difference between 2^16-1 and 2^16-2 (and so on, down to 0). This means that there's no point in adding more bits to each sample, since you can't hear the difference anyway. (The only reason to add more bits is if you have a really small signal and you're going to amplify it. Try listening to music through an amp fed by a digital amp, but with your music player digitally reducing the volume. Sounds really weird, because you're reducing the number of bits.)
> Red Book audio, the standard for CDs, is not the highest quality humans can hear
Any proof here? I don't have anything to test with personally, but considering that CDDA can sample any sound that your ears can, and that each level is represents is indistinguishable from other levels by your ears, it's probably pretty close to perfect. The output of a CD and a live performance will look different on an oscilloscope, but they'll probably sound the same through your ears.
Re:Speaking without detail is useless. (Score:2, Informative)
One step of CD mastering involves quantizing a signal to linear PCM at 16-bit. This process introduces quantization error [wikipedia.org], which shows up in the reconstructed signal as a noise floor at roughly 93 decibels below full scale (-93 dBFS). This means if a recording is played with volume set such that full scale = 100 decibels sound pressure (100 dB SPL), quantization noise will be about 7 dB SPL. The human ear cannot hear sounds below the absolute threshold of hearing [wikipedia.org], and this threshold is much higher above 8000 Hz than it is in the region of peak sensitivity (1000 to 6000 Hz). Noise shaping [wikipedia.org] algorithms have been developed that move most of the quantization noise above 10000 Hz. Therefore, at comfortable listening levels, a properly mastered CD moves all quantization noise out of range of the human auditory system.
Re:Intervieww is an ***HAT? (-1 Pedantic) (Score:4, Informative)
No it isn't, but perhaps he is SPECIFICALLY talking about Apple's implimentation.
Completely wrong. ASF is the container used by WMA and WMV files.
WMA is indeed the name of the audio codec, and WMV is a video codec.
He didn't say these were codecs. Included in your own quotation, he said: "audio file formats."
Yes it is. You'll get exactly the bits out that you put in. Your complaints are about DIGITAL SAMPLING OF ANALOG AUDIO AND HAVE NO SPECIFIC RELEVANCE TO WAV.
FLAC is not a lossless format. It is limited by in it's dynamic range (bits per sample) and sample rate. Compared to analog or a raw sound source, FLAC loses a lot of the sound.
Re:Speaking without detail is useless. (Score:2, Informative)
Analog recordings have a soft window so there isn't hard clipping like there is in the hard window of digtial.
Re:Observation on music quality (Score:1, Informative)
(posting anym cause I've already modded)
a good friend of mine is the leading contrabass in one of Germany's most respected symphony orchestra's (NDR.de). When it comes to (classical) music, he knows what he's talking about. He uses an ipod for pop AND classic and repeatedly tells me, when properly ripped (depending on what it is, starting from 128kbit up 196kbit or VBR), he can hear his cello colleague's fart in Bruckner's d-minor symphony. He generally doesn't give a shit about so called overexpensive "audiophile" equipment. Maybe he turned deaf from performing too many Mahler synphonies