Streaming Audio Primer (Part 4):

Overview

The next major step is to compress that whopping 250 Mbyte WAV file into a 6.5 Mbyte file, or maybe even smaller 2 Mbyte file. How do you such wonderful things? Simple, use either a MP3 or VQF encoder. You may be asking, "That sounds nice, but which one should I use - MP3 or VQF, and what's the difference?". That's a good question. Here's a few thoughts to help you decide.

MP3 versus VQF

First you must decide which audio compression/encoding format you will use. Both have the same basic idea: analyze frequency content of audio and compress by encoding only audible content. Unbelievably this can reduce a file size by 10:1 with virtually no loss in quality. And, if you are willing to sacrifice quality, one can achieve compression of almost 100:1 at the extreme. So what's the difference:

• MP3: The MP3 format, which stands for MPEG 2 Layer 3, has been around for a while. It is a part of the second generation of the standard MPEG format that is used to compress both audio and video. Most video capture boards save files in this format. It is by far the most popular format. Many expect it to take over the music business and revolutionize how we listen to music. Already many companies sell handheld or wristwatch players that play MP3 files from Flash Roms instead of CD's. In general (except really low sampling rates), MP3 has slightly better quality than VQF. Also, there is an abundant source of players, encoders, and other related software for the MP3 format because of its immense popularity. Check out the web site of the guys who invented the MP3 format and are busy on MP4.
  
   
• VQF: The newcomer to the audio compression scene, VQF has yet to reach the popularity of MP3. However, it has advantages that may cause it to eventually topple MP3: it has slightly better compression than MP3 (maybe 5% to 10%). Also, the free encoder has an 11,000 Hz (11 kHz) / 8,000 bps (8 kbits per second) setting that produces a phenomenal file that is only 2 Mbytes. Although the quality is significantly lower than that of higher rates (such as 11 kHz or 12 kHz at 16 and 20 kbps), it may be an option for those who have limited web space or may worry about larger file sizes discouraging people from downloading the file. The MP3 encoders also support this low sampling rate, but surprisingly have much worse quality at this particular low rate.

So, which one? That's up to you. I prefer MP3 because of its widespread acceptance. Also, the popularity of MP3 and the abundance of related software makes it much more attractive to the majority of people. However, some prefer the slightly smaller file sizes that VQF has to offer. The bottom line is that both will work, and you will win no matter which choice you make.

Encoding the WAV file

First you must download an MP3 or VQF compressor. First check Fraunhofer Institute's web-page on its official MP3 encoders, which are licensed. I use AudioActive's Production Studio Lite which retails for about $35. They have been known to offer a free 30-day trial download, but it is currently discontinued. If you want a free VQF encoder, you can download a free Linux Version, or a free 90-day trial version from Yamaha.

Once you have downloaded an encoder program, you can use this program to encode the audio file into either the MP3/VQF format, depending on which encoder you selected. Just to get you started, try setting the compression preferences to the 16,000 Hz sampling rate, 20 kbps, the Mono setting, and the MP3 output format (using AudioActive's trial version). Compress away! This will take a while too. Depending on the quality of your compressor, the quality and size of your final MP3 file will vary. Using this sampling rate (20 kbps, 16000 Hz for MP3 encoding) produces a comparably small file (6.5 Mbytes for 45 minutes), and balances well between file size and audio quality. This is what I use for all of my recordings. For the VQF encoder, try the 11kHz/8kbps setting to produce the small 2 Mbyte files. For recording of speech, stereo is not required, since most speech are recorded in Mono anyway. Also, stereo would double the file size too.

If you are going provide streaming audio on your site, then your primary concern in selecting an encoding rate is the bits-per-second (BPS or "bit rate") parameter. You must make sure that the bit rate you select is below the modem connection speed you want to support. For example, any modem that connects at 28.8 kbps should easily be able to handle a streamed audio file at the 20 kbps rate. The 24 kbps rate should also be possible, but any momentary hiccup in the network may cause a slight pause in the playback, which is not uncommon. Because of this, I prefer the 20 kbps set of encoding rates. Typically the higher sampling frequency (22,050 Hz, 16,000 Hz, etc.) within a given bit rate is better. Also, it helps to plan ahead and choose the original recording rate (48,000 Hz, 44,100 Hz, etc.) such that the encoding rate is an integer factor of the original rate (48,000 / 16,000 = 3 - an integer, no decimal). This makes better encoding. Encoding rates that are not integer factors must be "resampled". That is why "Resample" is listed by some frequency choices, but not all, in the "Encoding Properties" window in AudioActive's Production Studio. Planning ahead so that resampling is not required produces superior results.

Once the encoder finishes crunching, try testing it to see if you are satisfied with the output. If not, try different compression levels to get the results you like - just don't forget to balance file size and bit rate against quality. Once you have completed the encoding step, then you are ready to update your web-site with your new audio files!