Digital Audio, Continued
Chapter 3: Digital Audio, Continued
In This Chapter:
Learn more about the basics of digital audio
Find out about quantization and bit depth
Understand the concept of a 16-bit word
Learn about dither
Quantization: One Step at a Time
Say you got an assignment from a photo editor: There's a new statue of Abraham Lincoln in the park, and she wants you to go photograph it. In fact, says your editor, we're doing a complete photo essay on this new statue, and we need you to photograph it from every possible angle.
This process of taking thousands of snapshots is like sampling. When you assemble all the photos (or samples) together, you'll know exactly what that statue of Abe (or your sound wave) looks like. That is, if the snapshots are high quality.
That's the question: How good is each snapshot? Did you shoot them with a cheap disposable camera, or with a high-end Nikon 35 millimeter? If each photo (or each sample) is low quality, it doesn't matter how many you have. You still won't get a faithful reproduction.
Hot Link: Webmonkey (http://hotwired.lycos.com/webmonkey/multimedia/audio_mp3/)
This audio tutorial offered by the tech-savvy folks at Hotwired is one of my favorites (see Figure 3.1). It's geared for desktop musicians, and it's clear and to the point.
Figure 3.1 The Webmonkey site is full of good audio tutorials.
The quality of each sample, its bit depth, determines how many volume variations it can reproduce. The bit depth of audio files is typically 16-bit, 20-bit, or 24-bit. (Occasionally you'll come across an 8-bit audio fileavoid them.) The greater a sound file's bit depth, also referred to as its resolution, the more accurately it reproduces a sound's changes in volume.
Sample Rate and Bit Depth: Related Terms
Sample rate is how many samples you take; bit depth is how good each sample is.
Let's look at bit depth. Sound can occur at a virtually limitless number of volume levels. From an autumn leaf falling, to a baby laughing in the next room, to a band practicing next door, not only are a vast variety of volumes possible, but most sounds are constantly and subtly changing volumethat's what makes them interesting and beautiful.
The process of digitally recording a sound's volume changes is called quantization. If you want your quantization to be faithful to the original sound, you'll need sufficient bit depth.
Hot Link: Audio Forums (http://www.audioforums.com/forums_frame.html)
We can't think of a topic related to desktop audio that's not discussedat lengthat Audio Forums (see Figure 3.2). Mac OS, Windows OS, audio hard drives, music software...the list seems never ending. It's also an online gathering spot for musicians, engineers, and producers.
Figure 3.2 The Audio Forums site is a great place to learn about music technology.
The decibel is named after Alexander Graham Bell, who invented the telephone in 1876. (What did people do before that?) While the decibel, or dB, is most commonly used as a measurement of sound pressure level (loudness), it's actually a ratio between two values. As a measurement of sound pressure level, its base reference is 0 dB, the lowest level at which the human ear can perceive sound. Sound is said to double in perceived volume with every 10 dB increase. Sound pressure level ranges from the threshold of hearing at 0 dB to the ear-splitting loudness of 130 dB, as you can see from the chart in Figure 3.3. But there are other types of decibels: dBv and dBm, for example. In dBv, based on voltage, the base reference is 1 volt, so 0 dBv = 1 volt. In dBm, which describes a signal's wattage, the base reference is 1 milliwatt, so 0 dBm = 1 milliwatt.
A related concept to sound pressure level is the Fletcher Munson equal-loudness curve. This is very important to understand if you want to create mixes that sound good at various volume levels. Fletcher and Munson were researchers at the Bell Labs who examined the human ear's perception of volume at different frequencies. They found that the mid-range frequencies sound the loudest to the ear, in particular the 2 to 5kHz range. To make low and high frequencies sound equally loud requires them to be played at higher dB (the lower frequencies need the most boost). What this means in practical terms is that when you mix at high volumes you'll hear those rich lows and bright highs, but when a listener plays it back at a lower volume the mid range is the most predominant. In short, it's always good to listen to your audio project at lower volumes before deciding you're finished.
Figure 3.3 Here are some examples of the sound pressure levels of various common sounds.
Hot Link: The Recording Web Site (http://www.recordingwebsite.com/)
This site is a good resource for the home recording enthusiast, with a library of product reviews and tips on recording (see Figure 3.4).
Figure 3.4 Check out all the home recording resources at the Recording Web Site.
But What's a Bit?
You may think your computer is devious (it seems mine knows just when to crash), but at root it's pretty simple. In a computer's world, everything is either a 1 or a 0. This two-digit system is called the binary number system, and binary digits are referred to as bits (Binary digITs).
A one-bit piece of data, a 1 or a 0, can only store two values: on or off. However, a two-bit piece of data, such as 01, can store four possible combinations: 00, 01, 10, 11.
A four-bit piece of data can represent 16 possible values, and an 8-bit piece of data can store 256 values.
Keep adding numbers and you'll arrive at a 16-bit piece of data, otherwise known as a 16-bit word. Here's an example of a 16-bit word:
It's not the kind of word you'd use in a conversation, is it? But it's a useful word, especially for audio. A 16-bit word can represent 65,536 volume variations.
Hot Link: Home Recording (http://www.homerecording.com/)
The Home Recording site is a well-known site, dedicated to audio education (see Figure 3.5). It's also part of a Webring, so it includes links to many similar sites.
Figure 3.5 You can learn about many aspects of do-it-yourself audio at Home Recording.
The 16-Bit Word and Quantization
Let's look at how this 16-bit word helps us record. In the diagram in Figure 3.6, each of the stair steps represents one of the 65,536 volume levels a 16-bit recorder assigns to a sound wave. (Our illustrator didn't want to draw all 65,536 stair steps, but you get the idea.)
As sound is recorded digitally, a sample and hold circuit inside the analog-to-digital converter measures that wave's amplitude thousands of times a second (this is the sampling process). The A-D converter then assigns that amplitude a binary number (in the case of a 16-bit recording, it could be any one of 65,536 values).
Figure 3.6 Notice how the rounded contours of a sound wave are represented by many small stair steps. This is the process of quantization.
The quantization process assigns a sound wave's volume level to one of these 65,536 stair steps. In the case of 16-bit recording, the resulting sound files can reproduce 65,536 different volume levels, from silence to rock-band loud.
Think about it. If a digital recorder is set to record at 44.1kHz, 16-bit, it's taking 44,100 samples a second, and each one of these samples is assigned to one of 65,536 possible quantization values. Whew. That's a lot of work for one second of audio.
Hot Link: ModernRecording (http://www.modernrecording.com)
The ModernRecording site is a labor of love by four experienced engineers/producers (see Figure 3.7). It covers the gamut of techniques and audio gear, and has an interesting Question and Answer page.
Figure 3.7 The ModernRecording site covers all manner of audio technology issues.