Previous Table of Contents Next

PC-Based Sound

Some exotic work in digital signal processing has been going on for years, but it usually involved expensive special-purpose peripherals far out of reach of the average computer installation.

Early desktop computers did not really push the state of the art in sound reproduction. Original IBM and Apple computers both had built-in speakers as standard equipment, but they gave the programmer only a single bit with which to control the speaker. This meant the speaker could generally be used only to emit beeps and buzzes, not true digitized sound.

In the early 1980s, however, many computer manufacturers saw that a true digitized sound capability could be added to their computers at a relatively low cost. Apple was the most prominent manufacturer, adding an eight-bit DAC to the Macintosh, which opened the door to the use of true digitized audio.

Most desktop computers today are IBM compatible ISA computers based on Intel’s 80x86 CPU chips. Unfortunately for sound enthusiasts, IBM has not yet elected to add sound capability to the PC, but third-party solutions are relatively inexpensive. The sound samples used in this book have been created and manipulated using the Sound Blaster card, manufactured by Creative Labs. But several other cards are on the market that can play digitized sound samples, and any of these can be used, provided file-format conversion utilities exist.

The next generation of digitized sound on the desktop is now here. Many of today’s consumer machines can digitize and playback 44 KHz sixteen-bit CD-quality sound data. Only a few years ago, this capability seemed a bit unusual. The exotic black cube from NeXT Computer seemed to presage the future when it was first introduced, incorporating a digital signal-processor (DSP) chip as a co-processor; the intent was to offload work, such as manipulating digitized audio, from the main CPU. For a while, other manufacturers followed this design, for example, Apple with its AV line of Macintosh computers. Today, the vast majority of PC-compatible machines sold in retail consumer outlets come equipped with sound cards and CD-ROM drives—by one count, over 75%. More recently, it seems that the pendulum may shift back in the other direction, as a new largesse of processing power in the CPU will allow its deployment for audio and video processing, in addition to handling its regular duties. Intel is promulgating such a configuration with its new P6 processor, which has cycles to spare, that can be used for compressing and decompressing audio and video on the fly, even while crunching numbers in a spreadsheet. Regardless of how it’s done, the multimedia capabilities of today’s machines only highlight the need for data compression, since they fill up a hard disk faster than ever before. The explosive growth of the Internet and the World-Wide Web, which allows multimedia-enriched distributed documents, also increases the need for compression, because the bandwidth of communications links is not increasing as fast as the processing power of the host computers.

The files distributed with this book will be “raw” sound files. These will be pure binary recordings of eight-bit input data. Virtually all sound software on desktop machines today expects more than that for a sound file, but many software packages have utilities to convert raw sound files to a particular format. The Sound Blaster, for example, includes an executable program called VOC-HDR.EXE that prepends a header file to a raw sound file. The sound samples here were all sampled at 11KHz, a commonly used rate for medium-fidelity digital recording.

By supplying sound data only, the code here can concentrate on compression, without worries about additional superfluous data in the file. A full-fledged sound-file compression package by necessity needs to support the dozens of different file formats in existence, but that mostly consists of implementation details.

Some sound capability resources are available for a relatively small investment. Many on-line services, such as Compuserve, America Online, GEnie, and BIX, have active forums for audio manipulation. There are also active forums on the Internet, such as Web sites and Usenet newsgroups, focusing on digital audio. Freeware and shareware utility programs available in these forums do a passable job of playing sound out of the PC speaker. Other programs convert sound files between various formats. It wouldn’t be feasible to try to list specific examples here, but it should be relatively simple to find this type of software. In addition, third-party sound cards are available for a relatively low investment.

Lossless Compression of Sound

The original applications for sound compression could not take advantage of lossless data-compression techniques. One characteristic of all the compression techniques discussed so far in this book is that the amount of compression they achieve on a given data set is not known in advance. In some cases, the compression program can actually cause the data to expand, taking up more space than it occupied before.

In the 1960s, telecommunications researchers were trying to find ways to put more conversions on digital trunk lines, particularly on “expensive” lines, such as undersea cables or satellite links. Unlike disk space, which is somewhat flexible, these links have a fixed total bandwidth. A single telephone conversion might be allocated a 64Kbps slot on one of these channels. If it suddenly needed 100Kbps because the compression code hit a rough spot, there would be a major problem.

These early researchers were attempting to divide a 64Kbps channel into two 32Kbps channels to get two for the price of one. This required compression techniques that would consistently compress data by 50 percent, even if it meant losing some resolution.

Today, when trying to compress sound on disk for multimedia applications, we are in a slightly better position. We store and retrieve data from fixed disks, a more flexible medium for our work. If our files are compressed by 95 percent in some cases, and -10 percent in others, it will not really cause any trouble.

Previous Table of Contents Next