Resources: The Aging Anatomy of MP3 (why size matters)
But What Is An MP3?
The first standards for MPEG-1 came out of the MPEG group in 1991 and were finalized in 1992. MP3, MPEG-1 Audio Layer III's nickname, is a perceptual audio codec (coder-decoder). This process involves compressing the music file by applying a psycho-acoustic model which determines what information to keep and what to discard, its ensuing encoding according to a specified bit rate, the application of a lossless compression algorithm (some variation of Huffman Coding) that essentially squishes the file size further and the eventual decoding during playback. What you end up with is a file much smaller and ideally one that doesn't sound too much worse than the original.
How much smaller? That depends on the psycho-acoustic model employed as well as the bit rate chosen and here’s an important point to keep in mind when we talk about MP3s—the MP3 codec allows for bit rates from 32 kilobits up to 320 kilobits per second (kbps) according to the ISO standard and up to 640kbps for non-standard bit rates. Generally, the lower the bit rate, the worse the resulting audio quality. The most popular MP3 encoders include LAME, BladeEnc, Fraunhofer Encoders, and Xing.
For comparison purposes, a CD’s bit rate is 1,411.2 kbps and a high definition file with a resolution of 24-bit/96kHz (this 24-bit number refers to bit depth) has a bit rate of 4,608 kbps (bit depth x sample rate x 2 for stereo). Generally, the higher the resolution the greater the audio quality. But this is obviously “in theory” since the quality of the original recording is vastly more important than bit depth and sample rates. Which is vastly less important than how you feel about the music being played.
And while we can all march out the lower resolution copy that trumps the higher resolution copy, we have to decide whether we want to argue in theory or about a handful of really good sounding 320k MP3s. For the sake of moving forward as opposed to arguing backwards, I’m going to suggest that 320k MP3s can sound good, CD-quality can sound better and higher resolutions can sound even better. And lets accept the fact that all recordings are not created equal but we are here to talk about processing musical data not issues concerning the recording and mastering process. Let’s save dynamic compression, the loudness wars and recording engineer witch hunts for their very own post(s).
To answer our question, how much smaller?, with an example from the excellent Guide to the MP3 File Conversion Process on the h2g2 website, "20 seconds of stereo sound recorded in a WAV audio file will take up about 3,750KB of space" as compared to a 320kbps MP3 of the same 20 second clip which will take up 835KB, a 4:1 compression ratio.
Let Us Now Give Praise Where It Is Most Certainly Due
If we step back for a minute before the formation of the MPEG group, many very intelligent people were involved in developing the conceptual and theoretical framework that the MP3 format arose from including Karlheinz Brandenburg while working for the Fraunhofer Institute (he is also credited as a co-developer of the MP3 standard) and James D. (JJ) Johnston of AT&T who overlapped with Brandenburg for a time at AT&T-Bell Labs in Murray Hill, NJ.
We should not downplay the ingeniousness of the artful science behind MP3. We’re talking about true ingenuity, critical and creative thinking across disciplines and a practical and robust solution to a real-world problem that arguably ushered in the golden era of computer audio we enjoy basking in today. I like to imagine AT&T-Bell Labs as an amazing place with a great hi-fi history (reaching all the way back to Western Electric) filled with creative and fertile minds let loose by Corporate cash reserves to explore the limits of their imaginations unfettered. AT&T-Bell Labs makes most of those pre-busted Internet bubble incubators look like infertile chicken coops unless we value burning through others people’s money as an end in and of itself over real-world results.
Why was so much effort put into compressing audio? The shortest answer is space or lack thereof and poor plumbing. By the time the first MP3 player, WinPlay3, hit the market in 1995, computer hard drives were still measured in Megabytes (MB). For reference, a single 1-hour CD stored as an uncompressed WAV file takes up approximately 700MB of storage space so your average PC back in the day could at best store one CDs worth of uncompressed music. While this might suffice for some audiophiles, it doesn’t work for the average music lover (of course I’m joking—there’s no way a computer could play back a single file that consumed the majority of its hard disc's storage space). So the MPEG group squeezed out the MP3 standard for compressing audio data down to a more accommodating size.
Coupled with this paucity of storage were really thin tubes for transmitting data, to stick with our illustrious Senator from Alaska’s lasting Internet analogy. The average PC modem at that same time transmitted at 28.8kbps as compared to home DSL connection speeds of up to a theoretical 1.5Mbps today (i.e. much wider tubes).
MP3 Should Be Free—An Opinion
As we already know, once MP3 compression does its thing of throwing out those least significant bits, they are gone forever. We can also all agree that in an ideal world we’d like to enjoy exact copies of the original. In other words, if there is no good reason to throw away perfectly good audio bits, given the choice we’d keep them. Some argue that higher quality low quality MP3s at 320kbps are good enough. While even the staunchest objectivists agree there is an audible difference between an MP3 encoded at 320kbps and a non-compressed version of the same music, the difference is so slight as to not really matter. In other words, what's not really as good should be good enough for everyone.
With one terabyte hard drives available that can store about 2,500 CD-quality records in FLAC format for as low as $60, I would suggest that storage is no longer a concern. In terms of downloading compressed files versus uncompressed files, the Free Lossless Audio Codec (FLAC) has grown up to be the most commonly used lossless compressed format for CD-quality or better downloads. While a FLAC file is not as small as an MP3, it is about 50% the size of an uncompressed file, when coupled with our increased bandwidth makes for workable download performance. More importantly FLAC is a lossless codec meaning we do not throw out any bits of the baby or the bathwater. It’s all just packed up for tube travel and unpacked bit perfectly for playback. We’ll talk more about this process when we talk about FLAC in a future installment.
It's also important to note that once you have an MP3 or any other lossy compressed file, you are stuck with it. You can never recover the musical information that was discarded. And unlike lossless formats which can be transposed bit-perfectly to your format of choice, lossy formatted files degrade (further) when transposed. For example you can convert a FLAC file to ALAC so it will play in iTunes with no loss in sound quality—the two files will be bit identical. If you convert an MP3 to Ogg Vorbis, the latter will get rid of even more of the music than was slashed away by the MP3 compression process resulting in an even worse sounding copy of the already degraded MP3. I wonder if you converted a file through every lossy encoder available if you'd eventually end up with silence?
While there was a day and age when MP3s lossy compression made sense for delivering a music file, it no longer remains as sensible mainly due to advances in storage, processing and bandwidth. There’s no need to settle for anything less than an exact copy of the original which should be great news for anyone interested in hearing everything that was originally recorded and intended to be heard. For those who want and expect less, have at it. The ultimate caveat that applies not only to MP3s of all stripes but to any music storage media is—if the music you want only exists as an MP3, MP3 becomes the best choice.
Where MP3’s genius still shines brightly is its ability to deliver streaming music services where bandwidth limitations prevent higher resolution streaming and portable devices with limited storage although these days are numbered. While free streaming in lossy quality is OK, when paying for music, even streaming, go lossless.
I'd like to close with a quote from Karlheinz Brandenburg that remains painfully relevant where he addresses the concern that MP3 may signify the end of the music industry:
I think that will not happen, but we have changed the industry... The industry needs to know how to harness the new digital mediums and opportunities. They need to explore the positives rather than the negatives.