Hear What You’re Doing

The only way to avoid straying from the direction you wanted to go with your mix is to thoroughly compare if your manipulations actually make things better. Based on this comparison, you decide if the pros outweigh the cons (every manipulation has its drawbacks!) and if you need to refine anything. But making a good comparison is easier said than done. Just conduct the following experiment: tell an unsuspecting listener about a new processor you’ve used to give your latest mix more impact. To demonstrate its quality, you prepare the same mix on two different tracks, titled ‘A’ and ‘B.’ You don’t say that the only difference between the two is that you’ve made mix B 0.3 dB louder than A. Then you proudly switch back and forth between the two mixes, while you ask the listener to describe the differences. Chances are that the test subject will hear non-existent differences in dynamics, frequency spectrum, depth and stereo image. Due to the slight loudness difference between the two mixes, you can’t objectively compare them anymore.

Because of this, the most important thing when setting a compressor is using the make-up gain parameter. This will allow you to compensate for the loss in loudness caused by the compressor, so you’ll keep the audible loudness difference to a minimum when you switch the compressor in and out of the signal path. Still, this is easier said than done, because a compressor—unlike an equalizer, for example—doesn’t cause a static difference in loudness, but changes it with time. Since it’s not possible to constantly compensate for the varying level change—which would ultimately undo the compression, should you succeed—it’s best to set the make-up gain by gut feeling. Try to give the sound the same place in the mix, whether the compressor is on or off. If you have doubts about the amount of make-up gain, it’s better to use a bit less. This way, you’re less likely to think the compressor has a positive effect, and it makes you work harder to really find the optimal setting.

The final hurdle for the objectivity of our before/after comparison is time-related as well. Because the effect of a compressor is so variable, it’s hard to assess its impact by switching it in and out during playback. The effect of the compressor on the line right after you turn it on isn’t necessarily the same as it would have been on the preceding line. You can’t predict the effect of a compressor on the entire song based on an instant comparison, like you can with the effect of an equalizer. Therefore, it’s important to listen to a section of the song that’s representative of its overall dynamics, both with and without compression. Only then you can assess how the compressor affects your perception of the dynamics, and whether this is an improvement.

There are compressors that automatically adjust their make-up gain to the amount of compression, which would facilitate equal-loudness comparisons. However, I’ve never heard this work well in practice, and although it can provide a good starting point, you can’t trust it completely. Your own ears are a much more reliable tool for estimating loudness differences. Besides this, the meters that display the amount of compression (gain reduction) aren’t the best guides to base the amount of make-up gain on. These meters can give you a rough indication of what’s going on in terms of compression, but they hardly tell you anything about the sound. In fast compressors that use VU meters (or a digital simulation of these) to display gain reduction, peaks can already be significantly compressed before there’s any deflection on the meter. This is because the meter reacts too slowly to display the compression. Conversely, the gain reduction meters in digital limiters (which only block peaks) can react very strongly, while the peaks hardly mean anything in terms of perceived loudness. So remember to only trust your ears!

Zooming In

Compression is often used to solve a specific problem. For example, a vocal recording contains loud bursts that suddenly stick out sharply, which makes it hard to position the vocals in the mix. You can solve this by setting the compressor in such a way that it mainly blocks these annoying peaks and leaves the rest alone. This can be tricky, for example when the vocalist sings a low line in the verse that contains a relatively large amount of energy, but that doesn’t go off the map in terms of dynamics. The compressor doesn’t see any difference between a low note that feels balanced, and an equally intense high note that you perceive as piercing and too loud. Because of this, you sometimes need to lend the compressor a hand, by exaggerating the problem it has to solve.

Many compressors have a built-in sidechain filter, or the option of patching a processor of choice in the sidechain. This way, you can filter the signal used by the compressor for its detection (not the signal it’s processing). In the example of the vocal recording, you could boost the 2 kHz range in the sidechain of the compressor. This will make the compressor perceive the piercing high notes in the vocals as much louder than they really are, so it will apply more compression to these bursts. With the filter, you increase the level difference between the peaks and the rest of the signal, which makes it easier to only apply compression to the bursts and not to the low vocal line in the verse. This method can also be used to turn a compressor into a de-esser. If there’s no need for very extreme corrections, such a (broadband) de-esser can sound more transparent than a de-esser that uses dynamic filters.

Sidechain filters can also be used to make a compressor respond to loudness more like our ears do. Not a bad idea, since you usually need a compressor to balance sounds based on human auditory perception, not on how they look on the meter. At an average listening level, our ears are less sensitive to low frequencies than to midrange frequencies. With a sidechain filter, you can give a compressor the same frequency dependence. Of course, it won’t work exactly the same way—as the hearing curve depends on the sound pressure—but its overall shape is still a usable reference. Many compressors have a high-pass filter in their sidechain, which can be very useful here. Thanks to this filter, the low frequencies that sound comparatively weaker to our ears will also have a relatively smaller share of the overall sound level detected by the compressor (see Figure 5.4).

Serial Compression

Compressors come in many flavors. In section 17.6, I’ll discuss many different compressor types and their strengths and weaknesses. In practice, you’ll often need to solve complex problems that require more than just the strengths of one specific compressor. What you want to avoid at all times is using a compressor outside its ‘comfort zone.’ Outside that zone, the compressor will have too many negative effects on the sound, the most common of which are ‘pumping,’ ‘clogging’ and distorting.

Figure 5.4 From the ear’s response you can derive a filter curve for the sidechain of a compressor, which will make it respond to loudness in a way similar to how we perceive it. Sometimes it can be useful to not manipulate the sidechain signal with EQ, but the audio signal itself. For example, if a vocal recording contains loud popping sounds, it’s better to filter off some lows before you run it through a compressor. If you do it the other way around, every single pop will push the compressor into maximum gain reduction, because it detects a high peak. In other cases, it’s recommended to first compress and then EQ, especially if you’ve already found the right dynamics for a sound with the compressor, but you still want to refine its place in the mix with EQ. In that case you want to avoid changing the dynamics again with every single EQ adjustment you make.

Because of this, it’s sometimes better to break complex problems down into sub-problems and start tackling them one by one. An example: a recording of a plucked double bass contains large differences in dynamics between notes, plus there are extreme peaks on some of the plucks. Instead of using one compressor to solve both problems, you let two different compressors work in their own comfort zone. First, one compressor takes care of the peaks with a fast attack, fast release and high ratio. Now the peaks won’t affect the second compressor anymore, which balances the average energy between the notes with a lower threshold and a much longer release time.

Figure 5.5 A soft knee response made up of three serial compressors with increasing ratios and timing.

Mastering compressors by Weiss and Massenburg use this principle to tackle complex signals transparently. They consist of three compression sections in series, of which the first one has the lowest threshold, lowest ratio and slowest timing. The next two both take it a step further with a higher ratio and faster timing, but with increasingly high thresholds. This system can be seen as soft knee with a complex response (see Figure 5.5).

Parallel Compression

Another way to combine the advantages of different compressors is to connect them in parallel. For example, you can use aux-sends to run a vocal recording through three compressors: a fast compressor to make the voice sound clear and up-front, an optical compressor to make it warm and smooth, and an aggressive compressor as a special effect. The three compressed sounds return to the mix on three different channels, which you can then balance any way you want. A bit more to the front? More of compressor number one. Too flat? More of number two. Too polite? Boost the third one. This is a very intuitive process, and you can even use automation to adjust the compressor balance to the different sections of a song! Keep in mind that if you work digitally, the latency of the compressors must be properly corrected. If not, a parallel setup like this can cause some pretty annoying phase problems (see sections 7.1 and 13.3).

Of course, besides the compressed signals, you can also add the original signal to the mix. Some compressors come with a wet/dry control, but you can easily use an aux-send for this purpose as well. This is what is usually meant by parallel compression, and it’s the ideal way to make certain details of an instrument come out without making it sound flat. For example, some compressors won’t really sound interesting unless you ‘hit’ them pretty hard, and to prevent the sound from becoming too flat, you just mix in a decent amount of dry signal again. The effect of this is that the peaks in the original signal remain largely intact, while the signal unaffected by the compressor is boosted in terms of intensity (see Figure 5.6).

Figure 5.6 The effect of parallel compression: the peaks of the processed signal (orange) are still virtually identical to those of the original (yellow), but the information in between is boosted.

Making Connections

As a rule, individual instruments or samples will sound more connected if they have a joint influence on the sound image. Shared acoustics can act as a binding agent, but bus compression can contribute to this as well. Bus compression means sending a number of instruments to the same bus and putting a compressor on it. This can be the entire mix, or a select group of instruments that have to blend together more. Think of all the individual drum microphones, the drums and the bass, all legato parts, all parts that play on the beat, or any other combination.

If an instrument is relatively loud, it affects the overall compression, and therefore your perception of the softer elements as well. This makes for a much more exciting and consistent mix, in which everything affects everything else. Because of this, compression on groups of instruments is just as useful as compression on individual elements. The rule of thumb that usually works is that problems should be fixed by focusing on individual elements, but if you want to create consistency and compactness, it’s often better to do this in groups. Chapter 12 will elaborate on the use of bus compression.

Dividing the Beat

Drums in particular are often played at the same time as other instruments, potentially causing them to mask a lot of each other’s detail. Simultaneous masking like this can be reduced by making these sounds less simultaneous. Of course, you don’t do this by shifting the drums and the other instruments away from each other, so they don’t play at the same time anymore. But what you can do is divide a beat in several time zones, each dominated by a different instrument. For example, the first zone can be dominated by the attack of the drum sound, after which the guitars and then the bass take over, followed by the sustain of the drums, until finally the sustain of the guitar reappears: all of this within the 50 to 100 ms of a beat.

The nice thing about this concept is that you’re less dependent on instruments that continuously dominate certain frequency zones, which happens if you use a lot of EQ to give sounds a particular place in the mix. Thanks to the different time zones, instruments that occupy the same frequency range can become more interlocked. This will make your mix much more robust when it’s translated to other speaker systems, because in the midrange that’s so important for this translation, all the instruments can now have a dominating presence, just not at the exact same time.

A Compressor per Place in the Beat

A system of time zones per musical beat is easy to explain, but how do you get something like this to work? The trick is to have compressors attenuate the parts of a sound that you want to reserve for other sounds. Take the drums, for instance. You want to emphasize the initial attack, and a bit later some of the sustain. By using a compressor with a slow attack and an average release, the part of the drums right after the first hit is attenuated. This leaves some room in the mix, mainly for the electric guitar. Usually you don’t even have to manipulate the electric guitar for this, because its transients are already a lot slower (and less sharp) than those of the drums. But if you compress the guitar with a slow attack and release and a very low ratio, this will emphasize the guitar’s sustain, so it can fill the gap after the drums have faded away. Next, you can run the bass through the same compressor as the drums, so it will be most strongly audible when the kick drum doesn’t play. It’s thanks to these mechanisms that compression is such a major factor in your perception of rhythm and such a useful tool for giving sounds a place in the whole.

You might think this type of compression affects the drums too much or makes them sound unnatural, but surprisingly this is not the case. Especially if you use the compressed signal parallel to the original, it can sound very natural and work like a zoom-in function. In the mix, the drums will sound more like they do when you listen to them on solo without compression. Why is this? Due to the Haas effect you don’t perceive every reflection as an individual sound source (you can read more about this in section 7.2), and as a result the time range just after the first hit of the drum is not so important for your perception of the drum’s sound. In a way, your ear combines it with the first hit. It’s not until much later that you perceive the acoustics around the drum again, which determine a great deal of the sound. So the middle time range of a drum hit might not contribute much to your perception of the drum, but it still takes up space in your mix. It’s a relatively loud and long-lasting part of the drum hit, so it will get in the way of other instruments playing at the same time. You may not notice anything about the drum sound itself, but your mix isn’t working out the way you wanted: all the energy and power seems to be blocked because you’re wasting too much space. Compression can help to attenuate the part of the drums that doesn’t contribute much to their size and power, but that does take up a lot of space. As a result, your perception of the drums in a busy mix can be more similar to how they sounded in the recording room.

Figure 5.8 Compression can help with the temporal redistribution (envelope shaping) of different parts playing the same beat (A). This way, they will give each other more space without the need for EQ (B).

Too Flat?

The drawback of using compression as an envelope shaper is that it affects the dynamics of the sound. A popular way to keep this effect within limits is by applying heavy compression to some drum microphones while leaving others unprocessed, so you still have some dynamics left. You could see this as a form of parallel compression. However, even this can be too much sometimes. When sounds already have a lot of density— like distorted guitars—the tiniest bit of compression can make them lose their definition and squash them into a tiring mush of sound. In that case, it’s better to manipulate the envelope in different ways. For example, recording the guitars on an analog recorder with a low tape speed will smooth the peaks in a way that won’t make the guitars as flat and swampy as compression does. And a transient designer (which can change the level of the attack and sustain part of a sound, independently of the signal level) will serve this purpose as well.

Too Good?

In music that was made on a grid, there’s a risk that the parts are so perfectly aligned that their individual character is barely audible anymore. Transients are the biggest giveaways of sound sources. You can see them as tiny attention grabbers that briefly announce a new instrument. If the transients of simultaneously played instruments are easy to distinguish from each other, the instruments themselves are likely to be distinguishable as well. You can create distinctions through the use of panning and timbre, but if the transients all occur at the exact same moment, it will be very hard to tell them apart.

In this case, the compression method I previously discussed can’t make enough of a difference. With music played by humans you won’t have this problem, because even though there are insanely great musicians out there, they’re definitely not robots. Their interpretation of the part they play—by delaying or advancing certain notes—can even give it a comfortable place in the overall groove. However, with loops or other quantized material this doesn’t happen automatically, and sometimes it can be worthwhile to move the timing of the parts around a bit (see Figure 5.9). Running an instrument through a distinctly sluggish compressor with a fast attack can also help to set it apart from the rest. You can take this concept pretty far, as long as you end up with one well-defined, strong transient (usually the drums) taking the lead. In arrangements that sound a bit messy, I often solve the problem by choosing the most stable instrument to lead the rhythm, and making the others less important by attenuating their transients with compression, EQ, reverb and tape saturation. Or by shifting the messy part a bit backward in time, because your perception of timing is usually based one the first transient you hear, so that’s the most important one.

Figure 5.9 Parts don’t need to line up exactly, as long as you give the listener a clear reference of where the center of gravity of the beat should be. ‘Unquantizing’—moving tight loops slightly out of sync with each other— like in example B, can bring stacked parts to life, because it will be easier for your ear to distinguish them afterwards.

Perspective Masking

Masking is an easily recognizable problem: you hear entire instruments disappear. Once you’ve detected it—and with a certain degree of skill— you can usually avoid it quite well through the use of balance, EQ and compression. It gets harder when the problem is less easy to recognize. Just imagine the concept you’re trying to achieve with your mix. This is the creation of a world of sound that you connect to as a listener. In this world, you’re mainly focused on the foreground, but all the surrounding elements together make it into a whole with a certain meaning. You maintain this illusion by having all the parts of the world relate to each other in certain proportions. Together, they form an image with perspective: their mutual proportions give the image meaning. A mix with a foregrounded bass drum and vocals lurking in the depths will have a very different meaning than a mix in which you make the vocals jump at you, while they’re being supported by the same bass drum. The way in which you design this image is completely up to you, but the pitfall is the same for every mix. If the foreground of the image moves around too much, the meaning of the rest of the world is constantly changing. As a result, you can’t connect to the whole as a listener. You need a consistent foreground to guide you through the world (see Figure 5.10).

This kind of perspective masking—in which one inconsistent element distorts the perspective of the whole—lasts a lot longer than the ‘normal’ masking that has been extensively studied scientifically. This time, it’s not about the inability of the ear to distinguish different sounds, but about the image of the music that you create in your head. If this image is disrupted, it can sometimes take minutes before you get used to it again. Especially long, sustained sounds with a lot of energy in the range your ears are most sensitive to (like vocals or loud cymbals) can disconnect you from the rest of the music.

Fortunately, compression is a perfect tool to prevent this. Because if there’s anything a compressor does well, it’s keeping things within limits. However, the compressor does have to ‘realize’ how the foreground is made up. For example, bass frequencies are less important there, but the range between 2 and 5 kHz all the more. Through sidechain filtering or even multiband compression, you can keep everything in this range within clearly defined limits, so you’ll never be pushed away too far as a listener.

Figure 5.10 Perspective masking in music. If the foreground isn’t consistently present, the meaning of the overall image distorts, just like it would in a photo.

Limiting

You might have noticed that in this chapter on dynamics—and even in this section on loudness—hardly a word is devoted to the subject of limiting. That’s because of all the equipment in a studio, limiters are the least interesting when it comes to manipulating musical proportions. Limiters curb the peaks of a signal, and they’re designed to do their job as inconspicuously as possible. Ideally, you can’t perceive the difference before and after limiting. Therefore, it’s usually not a very interesting process if you want to improve something audibly, and contrary to popular belief it’s not a good tool for generating loudness either. Limiting can increase the absolute loudness of a sound: by attenuating the peaks, you can make everything louder without exceeding the boundaries of the medium you use. But absolute loudness says nothing about the perceived musical loudness of a song (a drum-and-bass track will feel louder than the average ballad, for instance), only about the amount of energy that you fire at the listener.

Limiting Basics

A limiter works like a compressor with a ratio of ∞:1 and a very short attack time. As a result, it will never let a peak exceed the chosen ‘ceiling.’ But this sounds easier than it is. A very short attack time means that, at the moment the limiter is engaged, it cuts the signal so fast that it can create sharp edges (see Figure 5.11). These sharp edges are distortion, and on some signals (like vocals or other long sounds) this can sound very annoying. The solution is to set the timing a bit slower, and have the limiter ‘look into the future’ with a buffer (look-ahead), so it will still be in time to block the peaks. But this results in audible ‘pumping,’ plus the transients will feel much weaker.

Look-ahead can work well for sounds that distort easily, but short peaks like drum hits sometimes sound better when you simply cut them off completely (clipping). This process has its limits, because beyond a certain point you’ll clearly hear a crackling distortion that makes the entire mix tiresome to listen to. But if you use it in moderation, it can sound more transparent than limiting, because it doesn’t introduce any timing effects (pumping). Some limiters combine both methods (a clipper for short peaks and a look-ahead limiter with a slow timing for long peaks) to reach the best compromise. But often the very best compromise is simply not to use so much limiting.

Figure 5.11 The pros and cons of different types of peak limiting. In practice, the original signal largely determines what the ‘least bad’ method is. For example, if there are a lot of short peaks, clipping sometimes sounds better than look-ahead limiting.

In mixing, limiting is mainly interesting as a problem solver. For example, if you want to add a piano that’s recorded so close to the hammers that every note starts with a loud click, limiting can help to better balance the sound. The same goes for tambourines, congas and other sharp-sounding percussion. The limiter can attenuate the transients, which can almost sound as if you’re moving the microphone further back. Some limiters add a certain character to the sound when you crank them up: they can distort the sound or change the envelope in an interesting way. These kinds of side effects can be used as a creative tool, and in this role limiters can be musically useful. In section 14.4 you’ll find more about loudness, limiting, and the considerations that come with it.