The magnitude of a sound wave is measured in decibel (dB). Theories aside, all you need to know right now are the followings:
- You can adjust the gain (or volume) of the incoming signal so long as it does not exceed 0 dB. Once the signal goes beyond 0 dB, it is clipped and is unpleasant to the ears.
- Stereo signal consists of 2 channels – left and right. Each with a waveform that undulate between -infinity to, ideally, 0 dB.
In real life, people seldom sing in a constant volume or that snare drum may sound a lot louder than the rest of the tracks. And since our recording signal should not exceed 0 dB and due to all these occasional and sudden burst of loudness, we set our recording level low in order to prevent clipping. In doing so, much of the track’s details may become too soft to the audience. So, what shall we do?
We compress the dynamic range of the music and raise the overall gain. This way, the details of the music are relatively magnified while the sudden burst of loudness is subdued to an acceptable level. The following diagram illustrates the mechanism of audio compression. I choose a symbol of a turbine (due to my engineering background) to represent an audio compressor.
One tiny piece of advice before going into the details. From my experience, there is no easy way to learn how compression works except to fully understand how each 5 key parameters work. I have tried experimenting each parameter and attempted to learn by ears. The combination is just too vast to handle.
The waveform as seen above largely undulates between -inf to -12 dB, which the details within this range can be too soft to the ears (as you will see later, a -6 to -3 dB would be ideal). Our aim is to compress the spikes and by reducing the dynamic range with a certain ratio, the overall sound can be more pleasant to the ears. If the compression ratio is set at 1:1, that means the signal will pass through the compressor unaltered. Some compressor may allow you to perform a 30:1 compression ratio. That means the incoming signal is reduced by 30 times. Of course, in some applications, you may wish to amplify those spikes instead. Hence, in another extreme, you may set the ratio to 0.4:1. That will amplify the signal by 2.5 times.
A threshold determines which portion of the signal will be unaltered (i.e. 1:1). Once the threshold is exceeded, the compressor kicks into action. But do you want the compressor to compress the signal immediately to the preset ratio once the threshold is exceeded? Or do you want to gradually apply the compression to those spikes? That is the setting of the knee, which is a range measured in dB. A hard knee (i.e. 0 dB) denotes an immediate compression when the threshold is exceeded. A soft knee (up to 30 dB in some tools) enables a more graduate application of compression. In some software tools, you may be able to specify the knee setting using a curve graph.
Knee setting refers to the magnitude dimension (in dB) while the attack and release setting controls the time dimension of when the compressor should activate and deactivate. Depending on the tools you are using, the range of the attack setting is usually lesser than the release setting (note: 0 to 400ms versus 1 to 4,000ms). You should consider if you want:
- Instant compression (i.e. faster attack)
- More variation in the signal (i.e. slower attack)
- A signal with more level and less distortion (i.e. longer release)
- Or maximize the overall compression (i.e. shorter release)
Some compression tools may offer more parameters for you to set and customize the sound of your music. However, if you could master all these 5 settings – threshold, ratio, knee, attack, and release – you are pretty much in good shape to make some real improvement to your work of art.
Some of you may prefer to control the input and output gain of the compressor as well. If it is only one track you are working with or you are not concerned over maintaining the same vocal volume throughout all the songs within the same album (part of mastering process), you could normalize the signal by scaling up the compressed signal to reach 0 dB. Normalization is a common function within sound processing and you should be able to find it within the tool you are using.
Next, we are going to see how the waveforms look like for the profession recording.
Do feel free to leave a comment at the main page.
| < Previous Page | Next Page > | |