Final Cut Pro HD | H•O•T Hands-On Training [Electronic resources] نسخه متنی

The Math of Audio

But, first, you need to understand some of the math behind the audio. Why? Because, when it comes to audio, what you don't know can make your life really unpleasant.

You've probably tossed around terms like "48K," "44.1K," "32K," or "8-bit," "16-bit," even "24-bit." But, what do those terms mean? When I ask my students, I either get "quality," or some seriously blank looks.

But understanding what these terms mean has a direct impact on your audioespecially when, as a marketing ploy, many DV cameras ship from the factory with the wrong audio settings for digital editing.

So, sharpen your pencil and grab an extra cup of coffee. This stuff is important and won't take long to read.

Frequency 101

In order to understand digital audio, you first need to understand "normal" audio.

All audio is analogvariations in air pressure that cause your eardrum, or a microphone diaphragm, to vibrate. Those vibrations are translated by our brain, or the mic, into electrical signals, which can then be heard or recorded.

The "normal" range of human hearing is 20 to 20,000 cycles per second. This is called the frequency response of human hearing. However, human hearing changes with age. As people get older, they tend to lose the ability to hear higher frequencies. A middle-aged man, for instance, can only hear frequencies up to 12,000 to 14,000 cycles. Women tend to lose high-frequency hearing more slowly than men.

Chapter 11, "Filters and Keying."

Sample Rate

The first set of terms, "48K, 44.1K, and 32K," all have to do with the frequency response of a digital audio clip.

By its very nature, analog signals are infinitely variable, like the smooth curve of the sine wave in this illustration. And, although we can deal with analog signals perfectly well, a computer can't. It needs to convert these infinite changes into discrete numbers that it can store and process.

This conversion is called quantizing, and it works like this: The computer "listens" to an incoming audio signal and, every so often, records the average level of the sound it "hears." If those digital samples are far apart, as in the second illustration, only a very rough digital image of that sound is recorded. However, if those samples are very close together, as in the third example above, then a much more accurate picture of the sound is recorded.

The sample rate is the measure of how many digital samples are recorded by the computer. The higher the number of samples per second, the more accurate the picture of the sound is captured by the computer.

According to the Nyquist Theorem, the frequency response (or range of frequencies a clip contains) of a digital audio signal is equal to half its sample rate. Thus, a 32K sample rate contains a range of frequencies from 20 to 16,000 cycles. A 48K sample rate ranges from 20 to 24,000.

Final Cut does not like editing 32K audio files. It works much better on 44.1K and 48K sample rates. However, many inexpensive digital cameras are set to 32K sample rates for weird marketing reasons. So, make sure the production crew sets the audio sample rate on the camera to 48K (the digital standard). This is often listed in camera menus as 16-bit. Converting files from 32K to 48K after shooting never improves quality and should be avoided.

Dynamic Range

The second set of numbers, "8-bit, 16-bit, and 24-bit," all have to do with dynamic range, the amount of difference between the softest and loudest sound in a clip.

But where did the word "bit" come from? Not surprisingly, it came from the computer. A bit is the smallest amount of information the computer can record. Since all computer numbers are recorded in binary ones and zeros, an 8-bit audio file means that all its different volumes are represented by a series of numbers that range from 0 to 255or 256 different values. Why 256 values? Because 256 is equal to 2 to the 8^th power, where 8 represents the bit-depth of the audio file. The standard for CDs, 16-bit audio, has 65,536 discrete volume settings (2 to the 16^th power).

This means that an 8-bit sound with a level of 1 is very, very, very soft. A sound with a level of 255 is as loud as it can get; because in an 8-bit environment no loudness number can be greater than 255. If your audio track gets too loud, so that the combined signal is greater than 255 (shown as 0 dB on the Audio Meter), your audio gets clipped and parts of the signal are thrown away.

As a warning, Final Cut lights these two small clips lights to tell you your audio is too "hot," or loud.

Why is this important? Because audio mixing is additive. As you add more tracks, the additional signals increase the overall volume, which makes it very easy for those levels to get too high. Then, because the computer does not have any more numbers to use to represent this extra loudness, it throws that excess audio away, leaving flat-top waveforms where audio should be.

This missing signal creates digital distortion. Little computer bits, flying out of your computer, landing higgledy-piggledy on the floor. Homeless. (Well, OK, maybe they don't really fly out of your computer. But digital distortion sounds awful, and if you output it to tape, there is no technology on the planet that can fix your audio to make it sound good.)

So, watch your Audio Meters! Be very, very careful those clip lights don't light. Not once. Not even flicker.

Final Cut can handle either .WAV or .AIFF files, 8- or 16-bit depth, with sample rates from 8 to 48 kHz.