Woo-hoo! You made it through the recording process, but your work’s not done yet. Now you have to take all those separate tracks and turn them into something other people can listen to—and, ideally, something they want to hear. This process is called mixing or mixing down. The mixdown stage can seem daunting at first, especially when you have many tracks to mix. However, with practice, it’s not that big a deal on a modern digital audio workstation (DAW). It’s easier if you’re also the person who did the recording—you’re already familiar with the tracks, and you were probably working on the mix during the recording. But that’s not essential, and for this column I’ll be assuming that someone else recorded the tracks. The same principles apply to both situations.
Your primary goal when mixing should be to make sure the main elements of the arrangement, such as the vocals or the melody, shine while everything else (e.g., the rhythm section) supports them. Some of the tools you will use affect the tone and dynamics of the sound—equalizers for tone and compressors for dynamics. In addition, there are time-based effects, such as delay (echo) and reverberation. Panning (left to right, or multiple directions in surround sound), volume adjustment, and phase manipulation round out your tools.
Firstly, you’ll want to get an idea of what you’re working with. I do this by playing the song through with all the faders at 0dB. It will be loud, because most of your faders will end up lower than the zero mark on the DAW’s mixer faders in your final mix, so you may have to turn down the main L/R output fader to avoid clipping. You’re listening to determine the main elements of the song—for most popular music, the focus will be the vocals—but also for any problems that need fixing.
The biggest problems you’re likely to find are distortion from poor gain structure, intonation problems, and timing issues. Gain structure, though hardly sexy, is fundamental. If the amount of signal (gain) within and between units—be they analog processors or digital plugins—isn’t in the optimal range, sound quality will suffer. Too little signal will reduce your signal-to-noise ratio. With digital, too little signal also reduces resolution. Too much signal results in distortion. In analog units, harmonic distortion is sometimes intentional and pleasing, but often it’s detrimental; in digital, it causes hard clipping, period. Making sure the signal level, or gain, coming out of one unit is within the optimal range for the next unit in line is what creates the gain structure.
You can set the basic gain structure of your system with a 1kHz tone from a tone generator. If you don’t have a tone generator, just go to YouTube and find a 1kHz tone; fidelity is not important to set the level. Play the 1kHz tone into your recording interface at -18dBFS (a commonly used zero level, equivalent to 0 on a VU meter) and make sure your DAW is picking it up at the same level with your fader set at 0. Since it’s a digital connection they should match; if they don’t, you’ve got a problem beyond the scope of this article. If you have any analog outboard gear connected, play back that 1kHz tone from your DAW and make sure that the levels in and out and back to the DAW are the same when the analog gear is inactive or bypassed (i.e., at unity gain). Adjust input and output levels as needed, and note the settings. In essence, you want your base gain structure to be unity gain throughout your system when no processing is being applied.
As I have mentioned in a previous column, you should continuously monitor your signal levels while you’re mixing, and leave at least 3dB of headroom on your final mix for the mastering engineer to make any final adjustments. A lot of what you’ll be doing is making different instruments fit together—even in mono. Making sure you have proper gain structure between every separate element is fundamental to a good mix, and when you’re done there shouldn’t be overs (where the signal exceeds 0dBFS) anywhere—not for individual tracks (even between or within plugins), not on any busses, and especially not on the mix bus. Busses are used for routing and summing signals within a console (“bussing a signal”). The terminology is a holdover from the olden days of analog mixing consoles. The mix bus is the main output of the mix—in other words, your final stereo mix of the tracks.
Harmonic distortion and clipping are both pretty difficult to undo. There are specialist plugins that can fix, or at least mask, a lot of technical problems, but they tend to be expensive and have steep learning curves. It’s best to get the levels right during the recording.
Intonation and timing issues are easier to fix. There are many DAW plugins available to address intonation problems: for example, Auto-Tune and Melodyne. Some of them are even user-friendly! Occasional timing issues can be addressed by simply cutting out the mistake and pasting it back in at the right spot. Some DAWs, like Pro Tools, can automatically fix consistent issues, like a drummer with unsteady time. Or you could get a better musician and re-record the drum track. . . . If the original recording was well done, there should be few, if any, technical issues.
Plugin tools, and some common uses
Equalization should be a familiar concept if you’ve ever used a tone control or room-correction software. Most equalizers (EQs) you’ll use will operate on multiple frequency bands, giving you control over the center frequency, the width of the frequency range over which it applies (known as the quality factor, or Q), the slope (e.g., 24dB/octave), and the type of curve (shelving or bell, usually) for each band. Most home stereo bass/treble controls are shelving EQs, which tilt the bass or treble up or down, and that ten-band graphic EQ in your dad’s sweet ’78 Trans Am had ten bell-curve EQs. Equalization is great for carving out sonic “space” for an instrument, but the more you use it, the more audible it becomes. Steep filter slopes and large amounts of boost and cut can make EQ sound unnatural, and generally what you’ll hear is a “phasey-ness” to the sound. A good song arrangement, where instruments don’t overlap tonally, and good recording technique are the best ways to avoid using a lot of EQ. Planning!
This is an example of an EQ plugin with several helpful features. Notice that it shows you the frequency response of the track(s) being equalized in gray, along with the change in response based on the current settings. This one is set to take a little boominess out of an acoustic guitar track. The center frequency is 257.06Hz, the Q is 0.475 (very broad, affecting the signal from around 20Hz all the way up to 3 or 4kHz), and the amount of gain is -2.15dB. There is no direct slope control on this EQ type (which is a bell curve), but the Q directly affects the slope. The lower the Q, the shallower the slope, and the wider the range of frequencies affected. There are also high-, low-, and band-pass filters, which are similar to what you see in speaker crossovers. A high-pass filter set around 75Hz is commonly used on guitars and drum overhead mikes. The lowest note on a standard-tuned guitar is around 80Hz, so this is an easy way to get rid of some bass-frequency bleed.
A stereo compressor, showing the basic controls
Dynamics processing, on the other hand, can be a little harder to wrap your head around, and includes several types of processing. Most involve various forms of compression.
A compressor controls dynamic range. The basic parameters of a compressor are threshold, ratio, attack, release, and makeup gain. Most compressors will give you control of these parameters, and some will provide additional controls. The threshold is the signal level at which the compressor starts to compress the signal at the set ratio. The ratio is simply the attenuation of the signal expressed as a ratio of the input level above the threshold to the gain applied to the output signal. For example, a ratio of 3:1 means that for every 3dB the input signal rises above the threshold, the output of the compressor rises by just 1dB. Ratios above 10:1 are considered limiting. Graphically, the effect on the signal can be visualized as a line moving upward at a 45-degree angle (i.e., a 1:1 ratio) until it encounters the threshold, at which point it will level off. At an infinite ratio, the line goes horizontal at the threshold. This is known as a brick-wall limiter, because the sound will clip like hitting a brick wall.
The threshold can be implemented as either hard knee, where it only acts on signals above the set point, or soft knee, where the compressor starts acting before the set point but with a lower ratio at first. This is similar to the way that analog tape compresses sound when recording at high signal levels. I personally prefer soft-knee compression, for the most part, as it sounds less obvious, but a hard knee is usually more effective for limiting the signal. Some compressors give you a choice of knee type, some don’t.
A compressor’s attack is how long it takes to react to the signal and start to modify it. Set this by ear, based on what sounds best for each instrument. Limiters tend to have the fastest attack times since any overshoot is undesirable, especially with a brick-wall limiter. The release is how long it takes for the signal to return to its unaltered level after the compressor stops compressing. Attack and release can be set for an effect called pumping that’s usually bad, but sometimes can sound interesting if done in tempo. Makeup gain is basically a volume knob. Compression lowers the average level of a signal, and you might want to turn it back up. Or not, because doing so may affect the gain structure.
Pumping is an interesting effect of compression. You can encounter a minor version with bad attack/release settings, but you can also use it creatively, by employing side-chain compression. This means that instead of determining the compressor’s behavior from the signal that’s passing through the compressor and being acted upon, some other signal, called the side-chain (or key) input, is used. As an example, I sent a pair of guitar tracks from a song I’m remixing for a friend to a separate stereo bus (return, or return bus) using an auxiliary/effects send (aux, aux send, or aux bus) and applied a stereo compressor. I then used another aux send to route the bass-drum mike to its own return bus. I assigned the key input to the bass-drum return bus and set the compressor to external side-chain input, the attack to its quickest setting, and the release to half the interval between bass-drum hits. Here is the compressor and its settings; sound samples can be found with the others at the end of this article.
There are many uses for side-chain compression, from subtle effects to my extreme example. Most of the time, the action of a dynamics unit will be based on the signal upon which it’s acting, but sometimes that’s not what you want. Let’s say you want something (like a screaming guitar solo) to come down in level when something else—the main vocal, perhaps—is playing. This is another job for the side-chain input, and is specifically an example of ducking—temporarily reducing the level of one or more tracks of a recording so that another track or tracks are easier to hear. You can hear this every day in radio broadcasts, where the level of the music is automatically reduced when the DJ starts talking. Ducking uses one track to control the dynamics of another track. The track that you want to preserve—usually the vocals—is fed into the compressor’s side-chain input. The compressor only acts when the side-chain input exceeds a set threshold. But it gets more complicated: low-frequency content, due to its greater energy, tends to trigger compressor detectors more quickly than higher-frequency content. A high-pass filter is commonly used as part of an internal side chain to reduce the low-frequency content of the side-chain signal so that the ducking effect is more consistent.
The salt to the compressor’s pepper is the expander or noise gate. Instead of reducing a signal’s gain above a set threshold, gain is reduced until the input signal reaches the threshold, when it’s let through unaltered. An expander reduces the signal level when it’s below the threshold and a noise gate essentially mutes the signal until the threshold is reached. Outside of certain effects that often use side-chaining, the noise gate has largely been superseded by editing—with a DAW it’s easy to just remove all the parts you don’t want to hear.
A de-esser, as the name implies, is used to reduce sibilance. It is a frequency-dependent compressor, applying compression only in the range where sibilance occurs (about 4–10kHz), without affecting other frequencies. Multiband compressors can apply different parameters to each frequency band. One common use is for taming snare-drum resonance in the overhead cymbal mikes, when EQ might negatively impact the sound of other parts of the drum set. Dynamic EQ, where the EQ parameters can be varied based on the signal level, can also be used. The combination of dynamics processing and EQ can be powerful, but is a bit fiddly to use.
The best way to learn how these controls work and their effect on the sound is to play with them on different types of instruments; and yes, vocals count as an instrument. You’ll find that each type of sound/instrument has its own optimal attack and release range. The more you listen, the better you’ll learn how compressors (or any processors) sound. This is the essence of critical listening: listening actively and with intent.
In a very real sense, a compressor is also a tone control device because it will change the tone of the sound being compressed. To illustrate this, I had myself recorded playing drums at a professional recording studio here in Austin, Texas, and applied a few different compressor plugins to show how they affect the sound they’re fed. I’ve included some audio clips as samples of a couple of basic concepts (compression and phase manipulation). Each sample is the same loop of approximately 15 seconds of me playing drums.
This is the whole drum kit, as it was recorded, but with no additional processing added. It sounds pretty good; the recording engineer used quality mikes, preamps, and other equipment for the multitrack recording. The hi-hat is a bit louder than I’d like and too far to the right in the soundstage: click to play.
I added some compression and EQ to improve the sound. Also, to tame the hi-hat and move it somewhat toward the left, I used a de-esser along with EQ: click to play.
You can achieve some tonal changes without using any effects by changing the phase of a track 180 degrees. This is the second clip, but with the “incorrect” phase applied to the track from the mike recording the bottom of the snare drum. Notice how the sound changes, and that there’s less low end: click to play.
Here is the same clip with both the top and bottom snare-drum mikes in the “incorrect” phase. Now the snare isn’t just a bit thinner, it starts to recede into the background: click to play.
When you use a compressor, you don’t just change the dynamics; you also change the tone of the sound. I processed the bass-drum track through three different compressors, each modeled on a different type of physical compressor. The first clip is the drum track without processing, followed by the three different compressor plugins (dbx 160, Purple Audio MC77, and a one-off custom compressor):
I ended up using the third compressor. Here it is in context: click to play.
The track without processing, and with pumping compressors applied:
Next time, I’ll talk about reverb and delay, and also about a song I’m remixing for a friend of mine. It needs a lot of fixin’ in the mix. I’ll include some before-and-after samples and explanations of the choices I will make to fix certain problems.
. . . Mark Phillips