FORMAT WARS
Apr 1, 2006 12:00 PM, By Doug Eisengrein
Preparing recordings for specific delivery media is nothing new; first there were radio and 78 rpm records, then came reel-to-reel, cassette, CD, DVD… and the list goes on. Optimizing audio for posting on the Web, however, presents a unique set of challenges. Though many of us are now spoiled by the availability of high-speed Internet, bandwidth is still a core issue when dealing with media on the Web, especially since the majority of global Web surfers are still connected by slower 56K modems. Therefore, although you should be concerned with fidelity, you must also keep file sizes in check. There is a plethora of digital formats that can be used for Web audio, including old-school file formats like WAV and AIFF; the current standard, MP3; and modern formats such as FLAC and Ogg Vorbis. For this column, I'll explore which formats sound the best and which ones are the smallest in size or are the most compatible with delivery methods.
MAIN SQUEEZE
Due to the enormous number of file types, this article considers the most popular formats. Of the basic file formats, WAV (.wav) and AIFF (.aif) are not the most suitable for the Web. WAV and AIFF files are native to Windows and Mac platforms, respectively, and while they typically sound excellent, they are uncompressed and enormous in size, rendering them unsuitable for the Web. The same applies to CDA, or CD Audio. Probably the most common audio format on the Web is MP3 (.mp3) because of the popularity of portable MP3 players and MP3 CDs. MP3s indeed have many advantages, such as variable bit rates, a high (12:1) compression ratio and, often, very good sound quality. Compression is an important point here. There are generally two classes of compression: “lossy” and “lossless.” As the names imply, lossy compression schemes discard data to achieve smaller file sizes, while lossless schemes do not. MP3s use lossy compression. That means that more and more data is thrown out as the chosen bit rates (sometimes selectable in the conversion process, depending upon the tool used) are lowered. Commonly labeled as “CD quality,” 128 Kbps is a loose standard for MP3 conversion rates. When creating MP3s at a 192 Kbps, using variable bit rate (VBR) encoding often yields a good balance between file size and quality. A younger cousin of MP3 is MPEG-4 (.m4a), a common multimedia format that can include audio and/or video (Apple's iTunes Music Store uses that format). Though comparable in file size to MP3s, MPEG-4 files sound better in side-by-side comparisons. When used in a nonstreaming environment, however, MP3s by comparison still enjoy far more integration in Web browsers. Similar to MPEG are RealAudio and RealMedia (.ra and .rm) files. The Real formats are still fairly common, but much less so than MP3. Huge advances have been made in RealAudio's sound quality, so that it now sounds comparable to MP3. Also, RealAudio supports a huge range of quality: from 12 Kbps to 800 Kbps.
TO LOSE OR NOT TO LOSE
Ogg Vorbis, an open-source audio format (meaning it is not proprietary — programming code is available to the general public) is similar to MP3 in many ways: It uses lossy compression, it can handle source files of various bit depths and sampling rates and it converts to various qualities. Ogg Vorbis files typically sound better than MP3s of similar bit rates. The advantages to limited bandwidth are obvious, and you can often convert to higher-quality Ogg Vorbis while retaining small file sizes. Ogg Vorbis uses quality settings on a scale of 1 to 10 (rather than using bit rates). The lowest setting compares roughly to 48 Kbps, while the highest setting is equivalent to approximately 500 Kbps. They are typically encoded at quality 5 or 6, which yield acceptable file sizes and excellent sound quality. The Free Lossless Audio Codec (FLAC) is another popular, newer format similar to other compressed audio only in that it creates smaller files. FLAC is fundamentally different because it incorporates a lossless compression scheme; virtually all essential data in an audio file (and therefore, the file's full fidelity) is completely preserved. That also means FLAC files are substantially larger than lossy-compression files such as MP3, though they are still much smaller than their original source files. FLAC encoders use compression settings on a scale of 0 to 8, with 5 being a popular norm. With FLAC files, I prefer to compress at the maximum setting of 8; because we're dealing with lossless compression, there isn't much audible difference — if any — from the lowest to the highest compressed files. The results sound excellent in all cases.
A SOUND SELECTION
A typical CD-quality (16-bit/44.1kHz) song of 2:45 in length saved as a WAV takes up approximately 28 MB of drive space. The same file converted to 128 Kbps MP3, MP4 or RealAudio becomes approximately 2.6 MB. Conversion to an Ogg Vorbis file of quality 5 is approximately 3 MB. The lowest-quality FLAC file is 16 MB, while the highest-quality FLAC is 28 MB. All lossy files end up being small and roughly the same size (with similar settings), while the lossless files are much larger. The format you ultimately use should depend on your goals. If you want to reach the most people in the simplest way, MP3 is still the way to go because almost all browsers and media players support it natively. If sound quality is of utmost importance, FLAC (or another lossless compression scheme) clearly wins. My suggestion is to create your files in various formats with different bit rates, choose the best sounding ones that won't hog bandwidth and then offer a variety to your Website visitors.
| Want to use this article? Click here for options! |  |
