c.~. 5' DIGITAL SAMPLING & SYNTHESIS TEACHER UNIT GUIDE NOTES BACKGROUND INFORMATION Digital audio is considered to be the most important development in sound production in the last several decades. Experiments and research with digital synthesis began in the late 19508 with instruments like the RCA Mark 2 Electronic Music Synthesizer and other large digital computers. Itwas not until the 1980s that digital sampling and digital synthesis became a reality in instruments that most musicians could afford and use in live performance. Digital sampling first became widely available to musicians in 1981 when E-mu released the Emulator digital sampling keyboard. As for digital synthesis, the real revolution began in 1983 with the introduction of the Yamaha DX7 digital algorithmic FM synthesizer. THE DIGITAL RECORDING PROCESS Overview Electronic audio signals, exist in one of two worlds, analog or digital. An analog audio signal consists of a continuously rising and falling voltage that can be directly modified or amplified. The"shape" of the electrical signal directly corresponds to sound waves that travel through the air. In the digital world all information, including a recording of a sound, exists in the form of binary numbers. Digital audio recording has many advantages over analog tape recording including a virtually complete lack of wow and flutter, better signal-to-noise ratio (i.e., no tape hiss), no tape generation loss, and greatly enhanced editing capabilities. Digital sampling, digital synthesis, and digital signal processing have created sonic possibilities unthinkable without the power of digital computing. Digital audio technology has revolutionized the way we can create and listen to music. Digital Sampling The process of digitally recording a sound is called digital sampling. Simply put, digital sampling is the recording of a sound using a string of numbers to represent the sound. The string of numbers is created by measuring the amplitude (or height) of a waveform thousands of times per second. Each individual measurement or IIsnapshot" is referred as a sample. (The word "sample" is also used to refer to the entire recording as welL) In some ways the sampling process is similar the process of creating a motion picture. The camera's film captures the action as a series of snapshots or still photographs. The projector plays back the snapshots in rapid succession, creating the illusion of continuous motion. Fig. 5-1 When sampling, an analog signal from a microphone or other sound source is converted to numbers, or digital information, by an analog-to-digital converter (ADC). To playa sound back, the opposite process is carried out by means of a digital-to-analog converter (DAC); that is, the string of numbers is converted into an analog waveform and played through a speaker system. Some digital instruments will refer to the internal sounds or the recording process as PCM sampled. The term PCM, Pulse Code Modulation, is the digital coding technique used in virtually all digital instruments. The sound quality produced by the sampling process is determined by two factors: sampling rate and sample size (resolution). Sampling Rate The number of times per second that the measurements, or samples, are taken is called the sampling rate. Typical sampling rates are between 15,000 and 52,000 samples per second. For compact discs, a sampling rate of 44,100 samples per second (44.1 KHz) is used. Thus, for every second of sound you hear on a CD, 44,100 numbers must be stored! (Actually there are 88,200 numbers stored; since most CDs are recorded in stereo.) This sampling rate of 44.1 KHz has become the standard for digital synthesizers and sampling keyboards as well. Fig. 5 - 2 Figure 5 - 2 shows how sampling rate effects how accurately a sound is represented in the sampling process. Even though thousands of individual samples are taken each second, each measurement must be a discrete number. This results in a stair­ step representation of the sound rather than a continuous change of an analog signal. The higher the sampling rate, the closer these stair steps approximate the actual analog sound. Many CD players and DAT machines advertise oversampling as a feature. Oversampling is a process that attempts to smooth out the stair-step effect by interpolating points and adding more samples between each of the steps. For example, 16x oversampling places 15 estimated samples between each actual sample point. Sampling rate has a direct effect on the frequency response of the system. "_ , Fig. 5 - 3 Frequency response is a measure of the range of frequencies that the system is capable of producing (e.g., 30 - 22,000 Hz). As a rule, the highest frequency that can be reproduced in a sampling system is equal to one-half of the sampling rate. That number, one-half the sampling rate, is called the Nyquist frequency. Thus, for the CD standard rate of 44.1 KHz, the highest frequency that can be reproduced is just over 22 KHz, which is just above the range of human hearing. Any frequencies above the Nyquist frequency are filtered out b~fore and after the conversion process using specially designed filters. Many samplers allow for variable sampling rates. The reason is that the highter the sampling rate, the more memory is used up in the system, both for storage and for processing. Since some sounds (particularly in lower frequency ranges) don't require as high a frequency response for reasonable quality of reproduction, setting the sample rate lower can help minimize memory allocation. Sample Size (Resolution) Sample size or word size, refers to the number of bits used for each individual sample. The greater the number of bits used, the higher the resolution of the system. In an 8-bit system (where 8 bits are used for each individual sample), there are 256 possible numbers to represent the amplitude of the waveform at any given instant. In a 12-bit system, there are 4096 different possibilities; and with 16-bit resolution there are 65,536 possible numbers to represent each sample. Thus, with larger sample sizes, the grid of resolution is much tighter and there is greater accuracy in defining the waveform. Fig. 5 - 4 Sample size has an effect on the signal-to-noise ratio of the system. The signal-to­ noise ratio is a comparison between the level of desired signal and the amount of undesired noise present in the sound. It is measured in decibels; the higher the dB level, the better the signal-to-noise ratio. An 8-bit resolution provides a signal-to­ noise ratio of about 48 dB, which is roughly the quality of a low grade cassette deck. 8-bit digital audio is considered to be the lowest level of usable resolution and is used where memory conservation is critical. It generally has a grainy sound with noticeable hiss at low levels. A 12-bit resolution delivers about 72 dB signal­ to-noise ratio. A 16-bit system, the current digital audio standard, will produce about a 96 dB ratio. Compact discs, DAT, and most electronic music devices use the 16-bit word size for a very "clean" sound. Low end, economical devices sometimes use a smaller sample size for the same reason as a lower sample rate: since more memory is needed to store more bits for each sample, a smaller sample size allows for a longer recording with the same amount of memory. Memory Requirements As stated, sonic quality with digital audio comes at the price of increased memory requirements. As the sampling rate and bit resolution increase, the amount of memory to process and store the additional samples and bits also increases proportionally. The CD standard of 16 bit/44.1 kHz digital audio requires about 5 MB of memory per track per minute. That's 10 MB per minute for stereo audio and more for multitrack recording. High Definition Audio The 16-bit/44.1 KHz sampling standard set by the CD has remained for over a decade. Recently, however, the audio industry has begun to work with even higher fidelity digital sound. Referred to as high-definition audio, the new high-end standard is moving toward a 24-bit sample size and a 96 kHz sample rate. A number of devices already support 20- or 24-bit resolution and sample rates greater than 44.1 kHz. With the benefits of high-definition audio, however, comes the added cost of increased processing speed, faster disk access, and greater TYlPTYlOTV TPlllliTf'TYlf'nh:L At 24-hit/9t1 kH7.. thp mf'mOTV Tf'll111Tf'mf'nt<:: illTYln In 17 MB per track per minute. While the audiophile listener with high-end equipment may perceive a noticeable difference, the average end user with a typical stereo or sound system may not notice much difference in sound quality. DIGIT AL AUDIO APPLICATIONS Compact Disc (CD) The compact disc (CD) is the most common application of digital audio, and currently the standard medium for distribution of commercial recordings. Until recently it is has been a read-only format, but compact disc recorders (CD-R) are now affordable and becoming common place. The CD-R allows the average computer user to "burn" their own CDs which can contain digital audio tracks or computer data. Standard audio CDs use a sample rate of 44.1 KHz with a 16-bit sample size and can store about 74 minutes of stereo sound. While CDs have been the industry standard for over a decade, DVDs with a new high definition audio standard may eventually replace the current CD standard. Digital Versatile Disc (DVD) The DVD (Digital Versatile Disc) has become the new standard for distribution of consumer movies and will likely eventually replace video tapes.