Does Portable Music Have to be Compressed?

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."

Comments

I would pay a few cents less

[CastrTroy]

I would personally pay a few cents less to get CD Quality music. Often when I buy CDs they are priced anywhere from 7.99 to 13.99. I think that if you average it out, the CD ends up being about the same price as iTunes, possibly a dollar or two more. But for that extra dollar, you get a physical copy, that's lossless, and doesn't contain any DRM. I try not to buy CDs with copy protection, and even for the few I do, I can still easily rip them, by disabling autorun. The only advantages of iTunes and other music services are, the ability to buy one track, and the ability to have it right away. I don't usually buy music from artists who can't fill up a whole CD with good music, and I'm not that impatient that I can't wait for the CD to arrive from Amazon, or wait until the next time I happen to be in the mall. Sometimes, if I know I won't be in the mall for a while, I'll download the cd in MP3 format and then buy it later. So, I could buy off iTunes, but i'd get music that was of inferior quality, and locked by Apple, which means that I couldn't play it on another MP3 player without degrading the quality even further.

Re:more for non-DRM

[CRCulver]

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.

Not all sound engineers are as dedicated to the art as you suggest. Okay, sure, if one wants to listen to something recorded in a state-of-the-art lab by consummate lovers of both the music itself and clean audio in general, then one should invest in the right conditions.

From my own collection, I'll take the world premiere recording of Boulez's Repons as an example. It was recorded in the projection space at IRCAM, one of the world's foremost music and acoustics research laboratories, and I only listen to it from the CD on my home stereo system, which isn't the most whizbang, but the best I can afford.

Contrast this with Rush's 2002 album Vapor Trails , a musically strong release which was recorded in poor circumstances and remastered in worse. The clipping that plagues every track in the album has long been criticized by fans (see the Amazon reviews for further info). So, since the guys who engineered the album didn't aim for clear audio, I feel no shame in putting this in 160 kbps Ogg Vorbis and listening to it with merely average headphones on my portable MP3 player.

As has already been said in many places in the discussion, lossless is probably going to be a draw mostly for classical (or, in my case, modern-classical) fans.

Re:FFS shut up already

[denoir]

Audiophiles have consistently been failing double blind tests when it comes to lossy vs lossless audio compression.

Now, if you wish to sell stuff to audiophiles, then players supporting lossless compression are excellent - they will buy it (along with anything you claim, on whatever grounds, will improve the playback quality).

If you however want to bring better music quality to the general population - make them get better headphones.

Re:Lossless is compressed

[Yvan256]

Just that there's no law of nature establishing CDDA as the gold standard in the first place.

There is, however, a rule of market that establish the CD as the only source users can encode their music from.

Main problem with lossless is battery life.

[DdJ]

I put lossless content on my iPod sometimes. The main problem is battery life.

Yeah, lossless content can be compressed, but it's not compressed as well as it would be with lossy compression. So, on my iPod, the hard drive spends a lot more time working when I listen to lossless content. The result is a significantly lowered battery life. Go ahead and test this yourself if you have an iPod, or other drive-based MP3 player.

It's not as bad as it is with completely uncompressed content, but it's a good deal worse than it is with AAC and MP3 content.

IMO, lossless is the right choice for media centers and other applications that are able to draw power externally, and lossy is the right choice for battery-powered playback.

Re:FFS shut up already

[hankwang]

It does mask (cut off) but it's based on the psychoacoustic modelling

Yes, they are based on psychoacoustic modelling. But I believe that it is mostly a few curves that define the hearing threshold for certain frequencies in the presence of a loud masking tone. The rest is trial and error, with lots of fine-tuning of a zillion parameters in the algorithm while listening to compressed music and asking the golden ears at hydrogenaudio to compare different versions of a codec (at least for the OSS ones). There is no algorithm that will give you the degree of transparency of an encoding as a number that realiably matches the results of double-blind trials.

Regarding generative losses of enoding: masking can for example be done by using the fact that a listener doesn't hear pre-echoes before sharp attacks as long as they don't come earlier than X milliseconds before. The encoder uses this fact to get the bitrate of sharp attacks down. But on the second encoding, the pre-echo might become 2X milliseconds rather than X milliseconds, and be audible.

I've seen reports on hydrogenaudio that codecs such as LAME that use complex psya modelling extensively do a worse job as a source for transcoding than fast high-bitrate codecs that have much simpler algorithms for throwing away information.

So it's entirely possible to make a lossy codec which loses NO information through further encodings (though what would be the purpose?).

I suppose you are talking about transcoding to the same bitrate MP3 with the same psychoacoustic model. That could be useful if you want to get rid of DRM by burning to CD and then re-ripping. But the question is whether transcoding to a lower bitrate or even different codec will give audibly different results from encoding directly from the source.

Re:Lossless is compressed

[jd]

I'd start by massively oversampling the data at initial recording time. 100 MHz, 26 bits per sample, 8 channels, etc. Something totally outrageous and utterly unusable for anything resembling sane. The next step is to split the sound up NOT according to source or frequency, but according to what groups together the best. Here's where the oversampling comes in - you don't NEED the actual data points to get lossless encoding, you only need to be able to recreate them. Thus, once we have grouped the compressable information, anything that is left over that can be reconstructed is of no further interest and we can ignore it. The same goes for any complete grouping that we have formed - if the complete group can be synthesized directly from one or more other groups, we don't need it. You then compress the groups - in isolation or as a simultaneous set of systems - either losslessly or using a lossy method. When you downsample, you eliminate the guesses that are wrong first and then eliminate duplicate guesses that are right between the groups.

What you will end up with is some set on N systems, which will be large amounts of noise with small amounts of useful sound in them, which when superimposed with each other AND a filter function produce the original sound and which when taken individually are highly compressable. (The noise is simply there to create fake patterns that we can compress. It won't be random noise, because that doesn't compress, but is noise in the sense that it has no meaning or purpose other than to produce nice mathematical functions. The filter is simply something that's used to extract this deliberately injected deluge, so that the output is valid.)

Is this a valid technique? Well, yes - it's not that unusual to add noise to simplify compression, then subtract the noise afterwards. That's fairly standard. Splitting the data up to simplify the noise is merely a variant on the idea, and is used in plenty of compression methods. Compressing individually seems to be the customary method, but computing power is more than adequate these days to use fancier techniques IF justified. (Since you can encode the decoding method at the start of any track, it should be wholly irrelevant as to what method is used, provided the computing power is there to run it in real-time.)