A reverb is an audio effect that attempts to replicate natural reverberation – the multitude of delayed reflected sound repetitions when a sound is produced in a closed space.
We use the term "reverb" to distinguish between the natural reverberation and the artificial reverb effect. The artificial reverb is important, as it brings liveliness and natural qualities to the sound. This is what it is designed to do – to simulate the real life effects of a sound in a room, a club, or a concert hall. In addition, most contemporary music already has artificial reverb and that is what listeners are used to. A musical piece would sound "unnatural" without the artificial reverb. In short, reverb makes sound recording less dry and boring.
Properties of the natural reverberation
Natural reverberation is a complex phenomenon. Whenever a sound is produced in an enclosed space – a room, a club, or a concert hall – it is reflected by the walls, furniture, and other objects in the space. There is a multitude of reflections that interact with each other and that may be indistinguishable from each other. In a room of four walls, a single sound source will be reflected directly at least four times – one from each wall. That is, if the indirect reflections from the sound origin to a wall to a wall to the sound receptor and so on are ignored. Some of the original amplitude of the sound will be lost in each reflection, but each of the four reflections will result in at least four other reflections and so on.
Example: In fact, if only 70% of the sound amplitude is preserved in each reflection, there would be 67 million reflections before the amplitude of reflections falls to -20 dB, and 68 billion reflections before the amplitude falls to -60 dB, which is usually considered silence. In a room of 40 by 40 meters, this would happen in approximately one second. And that does not count the reflections from other objects in addition to walls, or the wall-to-object and object-to-wall reflections.
Since there are too many reflections in natural reverbs, it does not make sense to try and characterize the reverb by describing each reflection. Instead, one can describe the properties of the overall reverberation phenomenon.
- The total length of the reverberation is defined by how long it takes for the reflections to stop. Typically, this is defined as the time it takes for the reverberation reflections to reach -60 dB ("silence") and is denoted with T60. Reflections in larger rooms will take longer to bounce back than reflections in a smaller room. Reflections in rooms with very reflective surfaces will take longer to quiet down than reflections in rooms with surfaces that absorb more of the sound energy.
- Smoothness or diffusion occurs, because some reverberations have fewer spaced out reflections that are easier to distinguish, whereas other reverberations have more reflections that create one smooth sound blob. This can happen naturally, as some rooms are empty, while others contain a lot of objects and reflective surfaces.
- Irregularity happens when sound is produced close to an object. In this case, there may be more pronounced reflections at specific times, whereas in other cases those may be nonexistent.
- The absorption of the reverberation is defined by how much of the sound energy is reflected. Absorption is thus directly related to the total reverberation length in two different rooms of the same size. More reflective surfaces imply both less absorption and a longer total length, other things being equal.
- Natural reverberation has coloration as not all frequencies are reflected equally. There may be stronger reflections of higher frequencies than of lower ones.
These qualities may be different for different portions of the reverberated sound. Earlier portions of the reverberation, called early reflections, may contain reflections that are fewer in number and perhaps more distinct (usually during the first 100 ms, but depending on the size of the room). Later portions of the reverberated sound – late reverberation or reverberation tail – would contain more and indistinguishable reflections. The early reflections and the late reverberation may have different characteristics.
The following graph shows the reflections that result from the reverberation of an impulse (i.e., the impulse response of the reverberation).
Older recordings simulated reverberation in various ways – using chamber, spring, or plate reverb units. Some of these are still used today. A chamber reverb unit uses microphones placed around a room to pick up the natural reverberation of the room. A plate reverb unit directs the sound towards a metal plate. The metal plate vibrates with the sound. The metal plate vibrations, which include the original sound vibrations and some residual ones (the reverb) are then picked up. The spring reverb works similarly to the plate reverb, but uses a spring instead of a plate.
Digital reverb implementation take digitally recorded sound and use simple mathematical operations to create repetitions in the digital signal to mimic the reflections of the natural reverberation. The power of digital reverbs, in comparison with the reverb units above, comes from the ability of the user to control the reverb characteristics. The typical characteristics and controls of the artificial digital reverb are similar to the ones of the natural reverberation – coloration, length, etc. They may be named differently. Rather than absorption, we may see the wet / dry mix. Rather than coloration, we may see equalization, or a high pass cutoff. Rather than total length, we may see room dimension. Rather than irregularity, we may see left / right perception.
Digital reverb implementation
If the number of reflections in the natural reverbation is in the billions for a second of reverberated sound, then we cannot really model the natural reverberation. We can, however, create artificial reverb with fewer repetitions, but with the same characteristics as above. We will model early reflections separately from the late reverb. We will add multiple repetitions that decay due to absorption in some total amount of time in a way that creates diffusion and some irregularity.
Early reflections: Early reflections are usually modeled with a tapped delay line. The following is a tapped delay line of N taps.
Here x(k) is the digital input signal at sample k, y(k) is the output signal at sample k, M1 though MN are various delays in the signal in number of samples (N different delays in number) and g1 through gN are various gains applied to the delayed signal (usually g < 1 to signify that the repeated signal is of smaller amplitude and there is decay). In this formula, the output y(k) as the sum of the original input signal x(k) and N repetitions of the input signal with various delays M and decays g. In effect, a tapped delay line is very similar to an echo or a feedback comb filter, but it may be very different, as here there is no relationship between the gains gn or the delays Mn and there is only a finite number of repetitions N.
Late reverb: A typical implementation of the late artificial reverb is a collection of sequential all pass followed by parallel comb filters. This collection is called the Shroeder reverb.
In the Shroeder reverb, all pass filters create multiple successive repetitions of the original signal with decaying amplitude and use very short delays. These filters simulate the diffusion of the natural reverberation. The feedforward comb filters in the Shroeder reverb have longer delays and have two purposes. First, they provide some coloration through the characteristic comb filter magnitude response (see Comb filter). Second, they introduce irregularities in the reverb – peaks in the delays (see Shroeder reverb).
Alternative digital reverb implementations
An alternative to the Shroeder reverb is to use several Shroeder-Moore low pass feedback comb filters in parallel, followed by sequential all pass filters.
Another alternative is to use an impulse reverb. For an impulse reverb, one can record the natural reverberations of some physical space (a space that has nice reverberations). If we can also record the original sound (without the reverberations), the impulse response of the reverb can be computed from the two recordings – the recording with the reverberations and the recording without the reverberations. The impulse response can then be used as the reverb. The process is described in the topic Impulse reverb.
Unnatural properties of the artificial reverb
Artificial reverbs may also have "unnatural" controls.
- For example, we can have a very pronounced reverb, but one which cuts off quickly rather than decaying smoothly. This can be implemented by running the reverberated portion of the sound through a noise gate. A simple noise gate will allow the reverb portion of the signal (without taking the original signal through the gate) to pass through unchanged, if the amplitude of the signal is above a certain threshold, and will disallow that signal (zeroes it out), if the amplitude of the signal is below that threshold. A reverb such as this, with a gated decay threshold, can also be implemented with an expander. An expander, rather than completely disallowing the signal, will attenuate the signal when it falls below the threshold.A reverb with gated time, rather than disallowing the reverb based on an amplitude threshold, would cut it off based on the amount of decay time. For example, we may want a reverb that stops 250 ms after it starts, independently of whether the reverberations quieted down or not. This type of digital reverb is rare, but can be imitated by first compressing the original sound to make its dynamics more even, and then applying a reverb with a gated decay threshold.
- A preverb is a reverb that, rather than decaying after the original signal, builds up to the original signal. Sometimes a preverb is called a "reverse reverb". What is interesting about this type of reverb is that it is technically non-causal. A signal operation is causal, if its output depends only on the previous or current values of the input and not on future values (more precisely, this operation is non-causal in the time domain; in the frequency domain, this operation is causal, as it does not use negative frequencies). To make this operation causal and to be able to easily implement the preverb, one can create a copy of the input signal and shift that copy back in time.
- Reverberation is a spatial phenomenon. It is useful to be able to accentuate or modify how the reverb works in the physical space. One way to do that is to bounce the reverb between the two channels of a stereo recording – left and right, although this may sound too unnatural after some time.
- A better experiment is to differentiate between the reverbs applied to the left and right channels. We can, for example use a reverb that is has larger length but starts with lower amplitude in one of the channels. A reverb with 1250 ms total length on the right and 1100 ms total length on the left with minor differences in the initial amplitude is an interesting effect. It provides a better spatial position to the listener, rather than the sound origin, the position of which is defined with panning. In this case, however, to avoid an effect similar to the a slap back delay (see Delay) or a flanger, we must ensure that the digital reverb implementation does not mix the signal in the left and right channels, which some implementations do for a more natural reverb.
There are no strict guidelines on using reverbs in music recording and this is generally a matter of taste. Pronounced and long (over a second) reverbs were very prominent in the 1970s. Short and less obvious reverbs were used in rock music in the 1990s. Tastes will continue to change.
The following are a few notes that may help.
- Reverb should typically be applied after other effects and specifically after compression. A very pulsating and somewhat eerie song can be created with a pronounced reverb followed by a compressor with large compression ratio, but this is rare.
- Reverb can be applied to the whole mix or to individual tracks in the mix. It is normal, for example, to have pronounced and long reverb for solos, which are sparser than the rhythm and shorter and less pronounced reverb for rhythm sections. It is also normal to apply a second reverb to the overall mix after reverberated tracks have been mixed together.
- It is sometimes useful to apply a reverb that matches the tempo of the song. For example, if the song is at 80 beats per minute (80 quarter notes per minute, which means that one quarter note is equal to 750 milliseconds), then it may be good to use reverbs with total length of 750 ms or 1500 ms or early reflections that are some round portion of that.
- It is rare to have reverb on standard bass tracks as the reverb repetitions will likely make the bass muddy. It is possible to apply reverb to bass slap sections. It is also possible to extract the higher frequencies of the bass and add reverb to those, but that technique is more common of choruses and delay effects in general, rather than reverbs.
- Many artificial reverbs come with presets that add gain to higher frequencies (i.e., with a specific coloration). Thus, although the reverb itself creates a feeling of openness, some of this feeling is due to the equalizer that comes with the reverb and not the reverb itself.
- Plate reverbs (and digital reverbs designed to mimic plate reverbs) add gain to higher frequencies and are useful for vocals and instruments that need to be opened up. These higher frequencies are often the frequencies that correspond to sibilance and shimmer and so plate reverbs may not work on vocals that already have too much sibilance ("s") or for instruments that are too busy, such as rhythm guitars.