Technical The Internal Components Sound Cards

Sound Cards

Many personal computers come with a sound card or sound support built into the motherboard. Digital audio has relatively low performance requirements and is commonly used to support games and multimedia applications. Pentium CPUs are powerful enough to perform much of the sound processing directly.

Sound in the analog world is represented by a continuous current or voltage varying in amplitude and frequency. The amplitude of a signal is thought of as the volume and the rapid variations in amplitude (frequency) thought of as the tone. An analog audio signal is amplified and applied to an electromagnet, which moves the cone of a speaker and produces an audible sound wave in the air that your ear picks out. For the computer to handle audio it must convert the analog electrical signal to digital data using an ADC (Analog to Digital Converter) device. The process of converting a digital sound back to an analog signal is done by a device called a DAC (Digital to Analog Converter).

The ADC converts an analog signal to a digital one by sampling the amplitude of the signal at a periodic rate. The size and rate at which the samples are taken determines how well the analog signal is represented by the digital data. The other element of digital quality is the sample size. If an analog signal is sampled and assigned 8 bit value, 256 signal levels can be used to represent the signal's amplitude. The large steps between the 8 bit samples does not reproduce a smooth analog signal and introduces high frequency noise in the output of the DAC. Increasing the sampling size to 16 bits, 65535 different signal levels can be used which produces a smoother digital pattern, more closely representing the original analog signal.

As an idea of the effects of sample size and rate on audio quality, a telephone system uses an 8000 Hz sample rate with a sample size of 8 bits. A CD quality audio recording uses a 44100 Hz sample rate with 16 bits per sample. The amount of memory required to store one minute of CD quality stereo audio requires 10.584 MB.

When the ADC and DAC functions are combined they are referred to as a CODEC, which stands for COder and DECoder. CODECs can compress and decompress audio data and perform oversampling of the data. Oversampling at faster rates allows the sampled analog signal to recreate a smoother sound. On DAC output it is possible to create additional sample points between two real points by interpolating new intermediate samples and feeding them to the DAC. Since oversampled data is data created from the original samples the basic data rate and size of the playback files are not changed. Oversampling on the ADC conversion is actually running at much higher sampling rates. A weight average sample is actually passed on by the ADC. Rapid signal changes are more detectable than with a slower sample rate.

Audio compression and decompression is one solution to the problem of large file sizes of sampled digital audio. Two basic algorithms are employed. Redundant information is first removed and replaced with a code indicating that the next samples have the same value. This scheme tends to be lossless and can completely reconstruct the exact input data from the compressed information. A second algorithm actually removes information during the compression and are said to be lossy. There are hundreds of compression, decompression schemes for audio.
Companding is used by telephone companies to give a higher range in the mid volume levels. 8 bits are used to store 256 steps that are not equal. Assigning a different value to each step code in the sample, from a hand picked value increases or decreases the importance of a specific volume.
ADPCM (Adaptive Differential Pulse Code Modulation) compression is a lossy technique that works by outputting the difference between two samples. ADPCM works best with audio signals that have small, slowly changing amplitudes or volumes such as speech.
MPEG audio compression exists in two standards for MPEG-1 and MPEG-2. The MPEG-1 standard supports a nearly CD audio quality and MPEG-2 supports multiple levels of audio compression.

Once analog audio signals have been converted to digital audio data it can be processed to create all types of effects such as reverberation, surround sound, chorus and other sound characteristics. A dedicated processor called a DSP (Digital Signal Processor) has an architecture and instruction set designed to handle analog information that has been converted to digital. DSPs are capable of sound and music synthesis, decompression and compression and special effects such as Simulated Surround Sound.

Audio synthesis creates digital audio as it is required in real time. FM synthesis produces a audio signal by summing up different frequencies and amplitudes. FM synthesis chips contain functions called Operators that combine two frequency sources and add other special effects such as envelope control and noise. Functions can be combined to produce more complex sounds. A Yamaha chip called an OPL3 is used on modern sound cards that supports 15 melodies and 5 percussion effects and 10 stereo instruments and special effects. FM synthesis requires very little processing power or data.

Sound Blaster has become a defacto standard board for the PC. Creative Labs created and manufactured the sound blaster board and many others produced Sound Blaster compatible products. Two other basic functions were added to the sound card. An embedded audio processor that accepts commands and controls the basic functions of the sound card such as routing commands and data, compression and decompression and control of the audio mixer function. The audio mixer can accept analog signals from several sources and individually control their volume and mix them to create a single stereo output with master volume control for the left and right channel. Additional features can be added by an open chip socket that accepts an ASP (Advanced Signal Processor) used to add new capabilities to the sound card or through a connection that supports a MIDI Wave Table synthesiser feature.

Wave table synthesis can produce very high quality natural sounding musical instrument audio. Each synthesised instrument to be played is taken from a small sampled data segment which is stored in a memory chip. The sampled data segment is of actual instrument converted to digital data and possibly compressed. A DSP processes the sampled data and produces realistic audio simulating the actual instrument. FM synthesis and wave table synthesis can be mixed so that the music comes from wave tables and the special effects from the FM synthesis chip.

MIDI (Musical Instrument Digital Interface) allows for remote access to sound boards and a standard way for controlling synthesis. The MIDI interface supports 16 channels along which commands can be sent to control a number of voices. The commands are sent as codes that invoke a specific synthesiser instrument or effect. A standard set of 128 program codes defined as General MIDI is used to overcome different sounds being produced by different manufacturers boards.

Microsoft adopted a basic business audio system and developed software and hardware to support a set of basic audio functions and became known as the Windows Sound System. WSS consists of a 16 bit CODEC supporting programmable sample sizes and rates up to 16 bits and 48 kHz with basic mixer functions with programmable volume controls. FIFO memory and DMA support provided increased performance and low CPU assistance. The Windows Sound System has no synthesis or compression, decompression hardware.

NSP (Native Signal Processing) introduced by Intel on their processors moved signal processing functions such as audio compression, decompression, sound synthesis, speech synthesis and recognition from special DSPs and hardware to the PCs CPU.

Microsoft introduced a Windows API into Windows 95 called DirectX, designed to provide software with direct access to low level functions. Software simply makes calls to DirectX which today's sound cards and graphics cards understand. DirectSound is the audio portion of DirectX that offers stereo left and right panning effects. DirectSound acts as a mixing engine, adds software based effects, supports multiple sample rates and uses system memory to hold multiple audio streams.

DS3D (DirectSound3D) places a sound anywhere in 3D space and is known as positional audio which manipulates the characteristics of sound so it seems to come from a specific direction. Software can use the set of API commands to position audio elements behind or to the left and right. DirectSound3D provides support for effects using the CPU.

EAX (Environmental Audio Extensions) adds reverb which helps to identify different environments. EAX provides an API to predefined reverb effects for a variety of environments. EAX 1.0 provided 26 reverb pre-sets as an open extension to Microsoft DS3D. EAX 2.0 enables simulation of muffling effects, echoes, a better perception of environmental size, sound source location and air absorption. EAX 3.0 adds many more effects and synthesising positional audio on a single pair of speakers.

Other Co-processors

Graphics Cards