The Quick 1. . .2 of Digital Recording

Editor's Note: The Windows Sound System v2.0 software user's guide is mostly concerned with the setup and control of audio into and out of a computer running Microsoft Windows. Within this rather dry manual we found several especially useful chapters that describe, in simple terms, audio compression, sampling frequencies and related subjects.
The material is so straightforward and helpful that we wanted to include them as part of the EUonline Continuing Education page.
Please keep in mind that this material is from 1993, and that specific technology has advanced such that 32kHz sampling is really a minimum for broadcast quality recording, storage and transmission---but the underlying concepts are still valid.

Understanding Sound | Audio Compression with Quick Recorder
Return to Continuing Education

Sound can be recorded in either analog format (as an electrical replica of the waveform) or digital format (as a series of numbers, which you can then manipulate with your computer). Your choice depends on both the capabilities of the audio recording equipment and software you have, and on what you plan to use the recording for when it's ready. The quality of sound you expect in playback, as well as the amount of space required to store the recording will also factor into your decision.

What's So Cool About Digital Audio

"Digital audio" is the application of computer-based technology to the recording, processing, and reproduction of sound. An audio signal is reduced to a series of numbers which then can be easily manipulated by the computer to effect time delays, reverberation, equalization, editing, and storage.

And What's The Catch?

While digital audio reproduction enables you to make copies of the master recording with virtually no degradation in sound quality and very little distortion, it requires much more storage space. If you are making a digital audio recording that will be copied, stored, and distributed, you may find yourself struggling to make a compromise between sound quality and storage requirements.

Analog vs. Digital Audio, An Efficiency Contest

The difference between analog and digital audio recording systems lies in how they store acoustic pressure waveforms, or sound.

Analog audio systems use electrical signals to create a continuous replica of the waveform. Only one set of infinitely variable information is provided, as shown in the Analog waveform diagram.

Digital audio systems, instead, periodically check, or sample, the waveform and record an approximate value of the signal at each checkpoint. This produces many discrete sets of information, as shown in the Digital waveform diagram.

Both analog and digital audio systems can store and replay acoustic pressure waveforms. With digital audio, once audio information is converted to numbers, it is processed, stored, and reproduced as numeric data. Managing and manipulating numbers is very easy. When you play a digital recording, the numeric data is converted back to an approximation of the original acoustic pressure waveforms and then played.

Analog waveform representations are stored, processed, and reproduced by the recreation of the waveform each time. The magnetic tapes used in analog recording and playback devices have inherent distortion and noise that contaminate the waveform, so the quality of the recording is degraded each time it is played or reproduced. By contrast, digital audio is converted back into its original acoustic waveform only when it is played back, so the quality of the sound is not affected by repeated processing.

Also, when magnetic tapes wear out or are degaussed, the quality of the recording deteriorates. With digital recordings, repeated use does not affect the quality because the information is stored as numbers, and numbers do not change with wear (a 1 is always a 1). It takes a significant loss of magnetic particles in a digital recording to lose a part of the recording. Note that while both analog and digital media will degrade with playback, the degradation is less significant with digital audio.

Digital Recording

The process of collecting sound information has two variables: how often information is collected and how much information is taken when it is collected. As explained earlier, when you make a digital recording, the waveform is checked periodically, and an approximation of the value of the signal is made at each checkpoint, about 8,000 to 44,000 times a second. The readings taken during this periodic checking and recording are called samples. Each digitally recorded sample is made up of bits—the actual digital wave data, stored as numbers, which represent the signal.

When you make a digital recording, you choose the number of samples to take (the sampling rate) and the number of bits taken (the bits per sample) to produce the sound quality you need.

Note While sampling rate and bits per sample are described separately, these factors work together to determine the quality of your recordings.

Sampling Rate, or How Many Samples Do You Take?

The sampling rate (measured in samples per second) determines how many samples are taken each second while recording.

The higher the sampling rate, the more data that describes the waveform is collected, and the higher frequencies your results will reproduce. The highest frequency a digital waveform may reproduce is half of its sampling rate.

There are four common sampling rates: 8 kHz, 11 kHz, 22 kHz, and 44 kHz. What you will be recording, the playback quality you expect, and how much storage space you have will determine the sampling rate you choose.

8 kHz sampling rate

An 8 kHz. recording sounds like a voice speaking over a public announcement system. There is some distortion of human voice, but recordings sampled at this frequency are small. This sampling rate is suitable for embedded voice messages, annotations in your documents or when the quality of sound is not important and you have a lot to say.

11 kHz Sampling Rate

An 11 kHz recording sounds like a voice speaking over a phone line. A recording made at this sampling rate is adequate for recording the human voice, although sibilant 'S' sounds are sometimes cut off. Recordings sampled at this rate are relatively small. This sampling rate is suitable for embedded voice messages or annotations in your documents, when the quality of sound is not important and the file size is.

22 kHz Sampling Rate

A 22 kHz recording sounds like a recording played through a cassette recorder. This is adequate for almost all sounds, except those with very high frequencies, such as the clash of cymbals or the ringing of bells. These recordings require more disk space than an 11 kHz recording, but have better sound quality and may be suitable for voice presentations or musical recordings.

44 kHz Sampling Rate

A 44 kHz digital recording sounds like music played from an audio compact disk (CD). A recording of this quality requires a lot of storage space, and is best used when the quality of sound is critical and file size is of little concern; for example, when you're recording a voice presentation to the executive management that will be stored on a large hard disk.

Bits per Sample, or A Little Bit More

Digital waveform data is stored as a number, or bit. The quality of a recording can be measured by how many bits of data are contained in a sample (the bits per sample). Recordings made with a higher bits-persample rate require more data to be collected and more disk space for storage, but they also contain less noise and reproduce quieter sounds because there are more samples of the original sound and less noise is generated in between samples. Two standard values are used, 8 bits per sample and 16 bits per sample.

The more digital wave data stored (in other words, the denser your data collection), the less space you will have in between the data recorded, and the better the playback quality of your recording.

This is because, in the absence of an input signal—that is, in the silent time between data bits the playback device generates noise, usually a soft hiss. This is known as the noise floor. For each additional data bit you add to your sample, the noise floor is reduced by 6 dB. Thus, the background noise you hear during playback is reduced.

8 Bits per Sample
Since each bit reduces the noise floor by about 6 dB, an 8 bits-per-sample recording reduces device-generated noise by about 48 dB, producing about the same sound quality as that of a cassette tape recording.

16 Bits per Sample
A 16 bits-per-sample recording reduces device-generated noise by about 96 dB. producing about the same sound quality as that of an audio CD.

The Trade-Offs

You've seen how the choices you make creating a digital audio recording can affect the sound quality you get when you play the recording. You've also seen how storage space can be a factor in your decision. The advantage of digital audio is the ability to transmit or make copies of your recording with virtually no degradation in sound quality and very little distortion, but digital recordings also require a larger amount of storage space. This may be a problem if you plan to store or transmit the file. If this is the case, consider settling for diminished sound quality or using a compression scheme. (See Audio Compression with Quick Recorder, for information on using Quick Recorder's compression schemes).

Audio Compression with Quick Recorder | Understanding Sound
Return to Continuing Education