″Nobody knows where you are,
how near or how far.″
— ″SHINE ON YOU CRAZY DIAMOND,″ PINK FLOYD, WISH YOU WERE HERE
(EMI RECORDS, 1975)
In the grand opera houses and symphony halls of the world, reverberation is an integral part of the music listening experience. The reverberant sound of a space is intimately bound to the sound of the music being performed within. The acoustic design of any music performance venue is directly influenced by the intended programming for the space. Musical theater, opera, classical music, chamber music, worship music, jazz — all types of music expect a specific sonic contribution from the hall.
Recorded music also places its own unique demands on reverberation. This form of music, importantly, is enjoyed through loudspeaker playback. This music is typically recorded in a sound studio, using natural and synthesized reverberation that flatters, alters, enhances, or otherwise refines the aesthetic value of the loudspeaker playback performance.
A wide variety of pop-music listening environments exist: living rooms from high-rise condos to country farm houses, automobiles from two-seater sports cars to family-of-five minivans. Unlike symphony halls, the acoustics of these listening spaces is rarely influenced by music because these rooms serve other more utilitarian priorities. Unable to rely on help from the acoustics of the loudspeaker playback venue, recorded music seeks to sound pleasing in any space. For recorded music, the most important reverberation exists within the recording, not the playback space. In the creation of music recordings, reverberation is selectively added, avoided, and/or manipulated to suit the creative needs of the music. That reverb — a fixed part of the recording — then follows the recording to every playback venue. The reverberation in the recording is thus even more intimately bound to the music than the natural reverberation of a symphonic hall is to the sound of the orchestra.
This chapter tours the technologies used to create the reverb used in recording studios and illustrates the broad range of studio effects it generates.
Imagine listening to the sound of someone′s voice as they sing in the peaceful outdoors. Now imagine the sound of her voice as she sing in Paris′ Notre Dame Cathedral. What changes? It is well known, even to those who have not been lucky enough to hear music performed at Notre Dame, that the cathedral adds a rich, immersive, heavenly decay to every note sung. That augmentation of the music by the resonance of the space is reverberation.
When one sings outside, there are no walls to reflect the sound back to the audience, no roof to hold the sound energy in. Listeners hear the voice directly. When the singing stops, the sound stops immediately. In a cathedral (or symphony hall, or shower, or any mostly-enclosed, sound-reflective space), listeners hear the voice directly plus the sound of the multitudinous reflections of that original sound bouncing off all the surfaces of the room.
Ignore for a moment the direct sound from the singer and focus only on the reflected sound energy from the building enclosure. Each sound reflection in that reverberant decay is both lower in level (due to all that traveling and bouncing) and later in time (it takes time for sound to travel, about 1 millisecond (ms) for every foot, as discussed in Chapter 1) than the original sound that was sung. The size and shape of the room and the geometric complexity and material make-up of the walls, floor, ceiling, and furnishings, drive the number, amplitude, and timing of the reflections. There are generally so many reflections arriving so close together that none of them can be heard distinctly. The countless reflections fuse into a single continuous wash of energy arriving steadily after the original sound. The particular blur of reflections that is associated with every sound in a specific space provides the sound signature of that space, its reverberant character.
Fans of classical music are well aware of the contributions made by a hall to the sound of the orchestra′s performance. Performances are sought out based on the repertoire, the orchestra, the conductor, and the hall.
Experienced recording engineers know that even slight changes to a studio reverb setting can have a significant effect on the overall recording. It is perilous to try to reduce reverb to a few numerical quantities. Art defies such a distillation. Imagine trying to describe Jimi Hendrix′ ″Little Wing″ with just a few numbers.
Something as complicated as reverberation cannot be reduced to a handful of numbers. The field of architectural acoustics employs a vast range of measurements, tests, and analyses in an attempt to predictably improve the sound within a space. The broad field of room acoustics is understood by studying many dozens of books and participating in many years of experience and experiments. Nevertheless, engineers focus on a short list of measurable quantities to summarize the quality and quantity of any kind of reverb. The most important such parameters available to recording engineers working on popular recorded music are reverb time, bass ratio, and predelay.
Reverb Time (RT60)
The perceived liveness of a hall is measured objectively by reverb time, RT60. Perhaps the most noticeable quality of a room′s acoustics, reverb time describes the duration of the reverberant wash of energy. More specifically, it is the length of time it takes the sound to decay by exactly 60 decibels (see ″Decibel″ in Chapter 1).
Allow sound to play in the hall. It could be music or pink noise. Abruptly cut off this sound. The hall does not instantly fall silent. It takes a finite amount of time for the hall to return to silence again. RT60 is the standard measure of this length of time.
Bass Ratio (BR)
The reverb time measurement procedure just described is very much dependent on the spectral content of the signal used to initiate the reverb. Pink noise is a good choice because it contains sound energy distributed evenly across all the audible of octaves. If the playback system can handle it (and this is often a challenge, particularly at the lower frequency ranges), the hall is energized at all frequencies of interest to an engineer, nominally 20–20,000 Hz (see Figure 3.1, Chapter 3). The decay observed in this way represents the decay across the entire range of frequencies relevant to music.
Imagine the reverb is caused by a music signal instead of pink noise. This requires a choice to be made as to the type of music. Is the music selection bass heavy? Is it overly bright, having a strong high-frequency emphasis? Then one must decide the moment when the music must be stopped and the duration of the resulting decay measured. When is the best time to stop the music? During a chorus? After a snare hit? After a kick drum? While the vocal is singing?
Reverb time is better understood across a range of frequencies. Rather than relying on a single number to describe reverberation, it is helpful to find a low-frequency, mid-frequency, and high-frequency reverb time. In fact, measuring reverb time as a function of octave bands is the preferred approach.
The generation of reverb remains the same: use full bandwidth music or pink noise, abruptly stopped. The resulting decay is then band-pass filtered into the octave bands of interest. Finally, the time it takes each individual band to decay by 60 dB is found. Measuring reverb time as a function of frequency makes moot the issue of the spectral content of the test signal. The frequency-dependent decay of a space is a function of the size, shape, and materials used in making the space. The test signal only governs which frequency ranges are energized, and which are not.
RT1000 describes the length of time it takes sound energy in the one-octave band centered on 1,000 Hz to decay by 60 decibels. RT500 describes the 60-dB decay time one octave below, centered on 500 Hz. As the audio window spans some 10 octaves, this more refined method of calculation suggests that a space is better described by 10 reverb time measurements! In truth, even 10 frequency-dependent decay times do not come close to fully describing reverb — more and different measurements are needed to more meaningfully describe reverb. On the other hand, a set of 10 numbers is too much information to keep up with in the recording studio, with the engineer responsible for so many other aspects of the recording besides reverb. Further simplification is called for.
Bass ratio offers a single number comparison of lower octave reverb times to middle frequency reverb times. specifically:
where BR = bass ratio, and RTX = 60-dB decay time in the octave band centered on the frequency, X.
If the lower octave reverb times are longer than the middle frequency reverb times, the bass ratio will be greater than one. The perceived overall warmth and low-frequency richness of a performance space is very much influenced by its bass ratio, and ratios slightly greater than unity are often the design goal for a hall expecting to play romantic orchestral music.
Due to the importance of reverb time as a function of frequency, RT60 stated alone is generally understood to be a middle frequency reverb time (usually 1,000, maybe 500 Hz). Bass ratio adds required extra context.
Bass ratio is a term borrowed from the field of architectural acoustics. Measuring real symphony halls is a challenge, even with the advanced measurement equipment available today. Low-frequency measurements are the most difficult of all. Getting a portable transducer to reliably create significant sound pressure level in the bottom octaves remains impractical. Making reliable measurements in the bottom two octaves (with center frequencies of 31.25 and 62.5 Hz) is difficult today, and was essentially impossible until recently. Low-frequency behavior was best viewed at these, the third and fourth octave bands in human perception, centered at 125 Hz and 250 Hz, respectively. There are certainly exceptions, but in large, well-behaved spaces, it might reasonably be assumed that the behavior at frequencies below 125 Hz and 250 Hz is a reasonable extension of what is observed at these more accessible frequencies.
Predelay
In addition to the duration of the reverberant decay and the relative duration along the frequency axis (low versus mid), every recording engineer must understand a third reverb parameter, predelay. A gap in time exists between the arrival of the direct sound straight from the sound source to the listener and the arrival of any sound reflections or the reverberant wash of energy that follows. Predelay is the difference in time of arrival between the direct sound and the subsequent first associated reflection.
The size and shape of the performance space is the key determinant of predelay time. In orchestra halls, it is often the sidewalls that create the first reflection. Reflections off of the ceiling or the rear walls arrive much later. Therefore, a narrow hall is likely to have a shorter predelay time than a wide hall. In smaller spaces, the ceiling may be the closest reflecting room partition. With reverb-generating signal processors, of course, predelay is simply an adjustable parameter almost without limits.
The very idea of reverb for music comes from real spaces, such as symphony halls and houses of worship. The reverb used in recording studios is typically generated by signal-processing devices. These user-adjustable pieces of equipment are wonderfully — and sometimes frustratingly — independent of the physics of sound constrained by architecture. The total freedom to synthesize any kind of reverberant sound is at times paralyzing for the novice engineer, and has been known to bog down even veteran engineers. It is useful to bracket the range of studio reverb parameters based on the architectural acoustics of classical performance venues. An engineer is welcome to venture beyond physically-realizable reverb properties, but clever engineers knows when they have done so.
The symphony halls most adored by conductors, orchestras, critics, and enthusiastic music fans represent perhaps the highest form of achievement in reverb, specifically reverb for romantic orchestral music. Three halls are consistently rated among the best halls in existence today and serve as our reference point in the recording studio:
- Boston Symphony Hall, Boston, MA, United States
- Concertgebouw, Amsterdam, The Netherlands
- Musikvereinssaal, Vienna, Austria
Detailed analysis of these halls and other halls approaching their quality leads to a useful set of representative reverb values (Table 11.1). Set-ting up a studio reverb so that it′s parameters fall within these preferred values does not guarantee success. This quality of reverb may not sound appropriate for the music production at hand. These values have proven themselves appealing for the live performance of romantic orchestral music, not all forms of music. Also, a hall that falls within the desirable ranges for these three values might fail on other fronts. As mentioned above, something as complex as the reverb within a hall is not fully defined by so few numbers. These particular values could be met, yet for reasons not captured in these quantities, it remains a disappointing hall. Even within the great halls, more contemporary classical music may not be flattered by this reverb. Opera, jazz, and rock each sound best with at least slightly different reverberant qualities. There is no universally right, best, or perfect reverb. Use these values as a reference. Step out of these defined ranges deliberately, armed with an understanding of why the current multitrack production needs a shorter reverb, a lower bass ratio, yet a longer predelay than what might sound good for a Mahler symphony.
11.2 The Need for Reverb Devices
In a reverberant room, the ambience cues that perceptually connect a sound to a specific space coexist with the sound itself. The sonic contribution from the room in which the music is performed is inseparable from the music itself. A recording approach that captures an appropriate balance of the direct sound from the instruments and reverberant sound from the performance space should be an effective means for conveying ambience through loudspeaker playback of music.
This is precisely the approach for those styles of music that are closely connected to a space. Most classical music is recorded on-location, so that the music is created in, and recorded with, the sonic signature of the hall, church, or other space where it is expected to occur. Similarly, many jazz performances are recorded with their own, albeit smaller, space.
However, the vast majority of recorded music is created in recording studios. For technical, practical, and creative reasons, recording studio ambience information is rarely a significant part of the recorded audio signal. Instead, reverberation is generated from the reverbless recording. Temporary, artificial reverb will feed headphones during tracking and overdubbing. More carefully tailored reverb will be created and combined with the recorded tracks during the mixdown session.
The very process of multitrack production inspires and all but requires that the multitrack arrangement be built up over time, not in a single live performance (see ″Types of Sessions″ in Chapter 2). Unable to accurately predict the reverberant quality each track will ultimately need, the recording engineer very often wishes to record tracks with little to no reverberation within the signal. Ambience and related sound qualities are to be added later, at mixdown, when the interaction of all tracks is before the engineer.
As a result, recording studios often provide highly absorptive rooms that approach nearly anechoic conditions at mid to high frequencies. In these spaces, microphones are placed around instruments and performers primarily to capture their direct sound, rejecting as much as possible the reflected sound energy from the room. In many circumstances, it is an explicit goal of the recording practice to minimize the sound that the recording room adds to the instrument being recorded.
To further reduce the recording of reverberation that may later prove to be inappropriate when the multitrack music is ultimately mixed down into the final stereo or surround release format, recording engineers often use close microphone techniques. Microphones, frequently with unidirectional/cardioid pick-up patterns, are placed very near the instrument (within a foot of the singer, within inches of a drum), very much in the near field of the instrument. This ensures the direct sound will be much higher in level than any reflected sound from the room; recorded reverberation slips toward inaudible.
This tendency toward ″dry″ (without recorded reverberation) tracks is reinforced by other priorities in the multitrack recording process. Consider the situation in which more than one music performer is to be recorded. Maximum creative engineering flexibility is maintained at the mixdown session if individual performances are well-isolated from each other. The recording of a horn section is likely accomplished by placing a microphone near each and every individual horn in the section. Particularly in pop music, the multitrack production process motivates the recording engineer to utilize a separate audio track for each instrument so that adjustment of the relative levels of each horn in the section, and the adjustment of the sound quality of each individual horn in the section, can be made later, at mixdown. As discussed above, it is desired that these individual horn tracks be recorded with little to no recorded ambience. It follows also that each individual track should not have much, if any, acoustic leakage from a neighboring horn player into the track. That is, the flute track should have as little tenor saxophone signal in it as practical. The tenor saxophone track should be recorded with as little flute sound as possible.
It is often a recording goal to minimize this leakage. This influences the audio engineer′s placement of musicians in the room, and the selection and orientation of directional microphones around the instruments. Clever use of high-transmission-loss sound barriers between players within highly absorptive rooms is also part of this approach. The result is a further reduction in the recording of room ambience in every session.
The desire for isolation between and among the musicians contributes additional motivation to use close-microphone techniques as described above. The sound engineer places microphones closer still to the instrument being recorded, shifting the relative levels of the target instrument versus the ″leaking″ instruments decidedly in favor of the target.
11.2.3 CREATIVE MICROPHONE TECHNIQUE
When microphones zoom in close to the instruments they record, they find an unusual ″view″ of the instrument. Good microphone practice requires the engineer to be able to capture a natural sound of an instrument despite such proximate microphone placement. Moreover, the varied, unusual, and at times unnatural sounds that exist in various locations so close to an instrument provide the creative recordist with the chance to record sounds that have a distinct sonic flavor, exaggerated sound feature, extreme intimacy, or heightened timbral detail. Microphones are brought in close to the instruments ultimately in search of ways to benefit the recorded end product. Recording engineers take advantage of this, moving beyond a goal of accuracy in pursuit of more creative sounds. For all recorded instruments, any characteristic sound qualities to be enhanced are identified and emphasized. Unwanted portions of the sound are strategically avoided. The engineer is, therefore, further motivated to place the microphone(s) quite close to the instruments being recorded.
With so many forces conspiring to deemphasize recorded reverberation, the multitrack production process often requires reverberation to be added back into the loudspeaker music at a later time. At the mixdown session, all individual tracks have been recorded and the technical and creative process of combining them into the single loudspeaker listening experience occurs. To add reverb, the engineer has two choices: record natural reverberation separately (i.e., onto separate tracks of the multitrack recorder, often called room tracks) at the time of the original music performance, or employ reverberation devices that create the desired spatial, ambient, or other qualities at mixdown via signal-processing effects.
While recording overdubs with close-microphone techniques, the engineer may also place distant microphones in the studio in order to capture the room′s liveness. These ambient signals are recorded as separate audio tracks so that they later appear as fully adjustable elements of the multitrack arrangement. While a single mono room track has value, stereo productions will record room tracks in stereo pairs. Surround productions will record room tracks in four or five track sets. During mixdown, these room tracks can themselves be adjusted and processed in any way the engineer chooses. Some high-end recording studios are prized for the sound of their live rooms that have proven particularly effective in recorded music. Architecturally much smaller than opera houses and symphony halls, these live rooms offer particularly supportive early reflections with a dose of relatively short reverberant decay.
The history of recorded music documents the value of recording even popular music in reverberant spaces. Many of the most important works of recorded art were recorded in studios much larger than is common today. Studios existed in converted churches and giant loft spaces. The talented engineer was able to capture recorded ambience at the tracking stage that worked for the production when finally mixed and enjoyed over loudspeakers. Few engineers today get the chance to record in such large, live spaces.
While it does not happen very often, multitrack music productions will sometimes go to the trouble to record elements in very reverberant spaces. The session will leave the recording studio and track drums in a church or string section in a hall, and background vocals in a solid concrete basement. Room tracks are a critical part of such sessions.
11.3.2 ACOUSTIC REVERBERATION CHAMBERS
Clearly it is impractical to utilize an orchestra hall as an effects device. Such buildings are rented at great expense and cannot be proximate to every recording studio. As a result, the natural reverberation of a large hall is rarely part of a pop or rock multitrack music production. Knowledge of room acoustics — beginning with the work of Wallace Clement Sabine in the 1890s — can still be employed directly in the recording studio, how-ever. Sabine′s well-known equation illustrates the concept of the reverb chamber.
RT60 is the length of time in seconds required for the reverberant sound level to decay by 60 decibels, V is the volume of the room (cubic feet), and A is the total sound absorption of the room (square feet).
The reverb time is directly proportional to the volume of the room, and inversely proportional to the room′s total sound absorption. For a long reverb time, seek out a large room volume. As large halls are impractical for recording studio work, multitrack productions must generally make do with a much smaller space. If the room volume cannot be large, use a room with little sound absorptivity. A small space is utilized and equipped with loudspeaker(s) and microphones (Figure 11.1). To achieve a long reverb time, the lack of cubic volume in a reverb chamber (Figure 11.2) is overcome through high sound reflectivity. Plaster, concrete, stone, and tile are typical materials for such a space.
Sabine′s equation (Equation 11.1) assumes a fully diffuse sound field. It is unlikely that a space as small as a reverb chamber generates a truly diffuse reverberant field within the reverb time window of interest, but the equation still provides informative guidance about using reflectivity to overcome small volume. It follows that the surface treatment within a chamber may benefit from being highly irregular to maximize diffusion, increase modal density, and decrease audible coloration within the relatively small space.
Figure11.1 Reverb from a small, highly sound-reflective room volume.
Figure11.2 Chambers must generate reverberation within a significantly smaller room volume than symphony halls.
There is no such thing as a ″typical″ reverb chamber. Recording stud affordable space. Custom-built rooms just for reverb are an expensive endeavor for recording studios, which are often located in urban areas (e.g., New York, Los Angeles, London, Tokyo) where real estate costs are substantial. Practicality has motivated recording studios to award the lofty title of ″reverb chamber″ to such unlikely spaces as multistory stairwells (CBS on 7th Avenue, Avatar/Power Station); attics (Motown, Capital Studios on Melrose Avenue); bathrooms (project studios around the world); and basements (CBS 30th Street, Capital Studios on Hollywood and Vine, and Abbey Road).
Some trends in the architecture of the proven chambers may be observed. Highly sound-reflective materials finish theses spaces. Concrete, stone, brick, and tile are common. Frequently, thick layers of paint cover the naturally-occurring pours in the concrete and masonry for minimum sound absorption. For reasons of tradition and superstition as much as science, irregular shapes and sound-diffusing elements are frequently sought out. Stairs, columns, pipes, highly articulated surfaces, and nonparallel walls are the norm.
Selection and placement of the loudspeakers and microphones are critical to the sound of the chamber. The loudspeakers are often selected for their dynamic range capabilities. To energize the chamber and overcome any extraneous noises within, the ability to create very loud sounds is desired. Horn-loaded, sound-reinforcement speakers make a good choice, with efficient ability to handle high power. Practicality influences the loudspeaker selection too. Last year′s control room monitors are frequently repurposed as this year′s chamber speakers. As coloration is so prevalent in chamber reverbs, often being deliberately sought out, loudspeakers with a flat frequency response are not strictly required. Nonflat frequency response becomes a creative variable. Clever engineers turn the spectral imperfections of the loudspeakers in combination with the frequency biases of the space into a chamber reverb with a distinct and hopefully pleasing flavor.
Microphones are placed very much in the same way room microphones are placed in a recording session. Engineers develop intuition about where in a room a microphone might sound best. Experimentation follows, and the microphone placement is revised as desired. Condenser microphones are most common, ranging from large diaphragm tube condensers to small diaphragm electret condensers. While omnidirectional microphones may seem best for fully picking up the sound of the chamber, directional microphones are common too as they reward the engineer for even slight changes in location and orientation.
Generally, a line of sight between the loudspeakers and microphones is avoided. The loudspeakers are placed so as to energize the space generally. The microphones are placed to capture the subsequent reverberation. The direct sound from speaker to microphone amounts to an acoustic delay only, whose level is likely to be much higher than the reverberation that follows. Avoiding that direct sound maximizes the reverb that is captured by the chamber microphones and returned to the mixing console. Reverb is the intended effect, after all. Directional loudspeakers (horns) and microphones, though not required, help the engineer achieve this.
Because reverb chambers are so small, a certain amount of modal coloration is unavoidable. It is hoped that the coloration this causes will sonically flatter any ″dry″ sounds sent to it. Equalization (EQ; see Chapter 5) is a signal processor purpose-built to introduce coloration to the sound. Engineers reach for reverb generated in a chamber when they feel the chamber′s coloration will lead to a desirable change in sound quality. Otherwise, the strong coloration must be avoided (no chamber reverb is added); ignored (other elements of the multitrack arrangement mask the coloration problem); or perhaps deemphasized with some EQ on the send and/or return.
The typically small dimensions of a chamber lead to a predelay that is on the order of just a few milliseconds, well short of the 20 ms associated with large halls. Typically a delay line (see Chapter 9) is inserted on the send to the chamber so that all reverb is appropriately delayed when it is combined with the other elements of the multitrack production.
A simple spring can be used to create a kind of reverberation. A torsional wave applied to a spring will travel the length of the spring. Upon reaching the end of the spring or encountering any change in impedance along its length, some of that twisting wave reflects back down the length of the spring from which it came. In this way, the wave bounces back and forth within the spring until the energy of the wave is converted into heat and dissipated through friction. Analogous to reflections between just two walls, a spring mechanically emulates the sound of a reverberant space— a one-dimensional space. Combine several springs of different lengths, thicknesses, and spring constants into a network of coupled springs and the pattern of reflections can be made more complicated.
The spring reverb has a unique sound but falls well short of simulating a real, physical, reverberant space. Its most important shortcomings in this regard are:
- The frequency response of the driver-spring pick-up system has limitations in the audio band.
- The number of reflections in even a multispring system provides only a finite imitation of the nearly infinite collection of reflections in a real room.
- The modal density of so simple a system is insufficient to prevent strong coloration.
- The build-up of reflection density does not grow exponentially with time as happens in a real room.
The spring reverb is simply not a room simulator. Yet it remains in (limited) use today. Affordable, analog, and portable, it is a common reverb in many electric guitar amps. Artists and engineers have locked onto it for its unique, if unreal, sound. Measuring one example of this device in the same way that acousticians measure a space offers a better understanding of its reverberant behavior.
The impulse response of Figure 11.3 shows an extraordinarily long reverb time. Reverb times vary from about 2 seconds in the upper octaves to more than 5 seconds in the 250-Hz octave band (Figure 11.4).
Figure 11.3 Impulse response of a spring reverb.
Figure 11.4 Reverb time by octave band of a spring reverb.
The frequency response of this mechanical system very much favors the upper octaves, with an approximately 6 dB per octave tilt up to the 4-Hz octave band (Figure 11.5). This helps give the spring reverberator its characteristic metallic sound. Yet the late decay becomes distinctly modal, leading to further coloration of any sound sent to the spring reverb (Figure 11.6).
While the spring does not closely resemble the sound of a hall, it is informative to compare its performance to the three referenced orchestra halls.
Figure 11.5 Frequency response of a spring reverb.
Figure 11.6 Spectral decay for a spring reverb.
The spring reverb is compared to Boston Symphony Hall, the Concertgebouw, and the Musikvereinssaal in Figures 11.7 and 11.8. The reverb time of the spring is clearly much longer than that of any desirable hall. The bass ratio of the spring reverb might seem to suggest that the spring could to do an adequate job of creating the warmth associated with good low-frequency reverb. This is not at all the case. The bass ratio metric alone is misleading.
Figure 11.7 Reverb time, spring reverb versus benchmark halls.
Figure 11.8 Bass ratio, spring reverb versus benchmark halls.
Refer again to the reverb time plot of Figure 11.4 and note the extreme difference in reverb times in the two low octave measurements that are a part of the bass ratio calculation, 125 Hz and 250 Hz. While these two octave bands share the same average reverb times as the two mid-band octaves, there is a nearly two-second difference in reverb times between the two lower octaves. Such extreme reverb time variation in the lower octaves does not lead to the same quality of warmth that a more consistent low-frequency device or space would possess.
Though differences certainly exist, it might be surprising to any reader who has used spring reverb how similar the spring reverb measurements are to benchmark room measurements. However, these metrics provide an in-complete picture. The sonic signature of a spring reverb, apparent to experienced sound engineers, is not fully explained by this analysis. Sound engineers are aware (sometimes only intuitively) that, among other things, a spring′s initial density of echoes as well as the rate of change of the echo density over time are different from that of an actual reverberant space. Resonance in a space builds and decays as the sound propagates in three dimensions. A spring offers a single dimension of propagation for the torsional wave within. This contributes to a unique and distinctly unnatural-sounding decay. In the right musical and sonic context, engineers use spring reverb precisely because of this different sound quality.
The concept of the spring reverb is improved upon through the use of a metal plate instead of a spring. This upgrades the mechanical reverb to a two-dimensional design. A thin plate of metal has attached to it a driver that initiates a bending wave. This bending wave propagates through the plate, reflecting back at the edges, leading to an accumulation of reverblike energy. A pick-up transducer captures this reverberated signal. Reverberation on this plate behaves in a way very much analogous to a room with two pairs of opposing walls (but without a floor or ceiling).
Because the plate is excited with bending waves, not compression waves, the speed of propagation is determined by the elasticity and mass distribution of the plate and by the plate′s thickness and suspended tension. Plate reverb designs can, therefore, slow the speed of propagation down to one thousandth the speed of sound in air. A mechanical device much smaller than a hall can be used to generate reverberation similar in duration to that of a large hall.
The plate offers further advantages over the spring. By applying any of various means of damping (e.g., placing liquids or porous materials against the plate), the reverb time is adjustable across a useful range, offering the sound engineer highly desired production flexibility.
The plate reverb, like the spring reverb, also fails to accurately simulate a real, physical, reverberant space. Its most important shortcomings versus acoustic reverberation are:
- The frequency response of the driver-plate pick-up system has limitations. There is a lack of response in the lower portion of the audio band (measurements are provided in the discussion that follows). In addition, even with damping, there is too much high-frequency resonance, leading to an unnaturally bright sound.
- The modal density of even a large plate is still insufficient to prevent strong coloration.
- While the build-up of reflection density does grow exponentially with time, the two-dimensional, rectangular, mechanical reverberator does not grow at the same rate as a real, three-dimensional space with complex shape.
As with the spring reverb, the plate reverb is simply not a room simulator. However, plates are extremely popular in multitrack production even today. Artists and engineers have found the quality of its sound to be useful for many applications (discussed below). A plate reverb, damped to a commonly used short reverb time, is measured as if it were a reverberant performance space.
The impulse response (Figure 11.9) shows a very steep slope. Reverb times (Figure 11.10) measured in octave bands all fall below about 0.5 seconds.
Though it varies a great deal from unit to unit, and depends very much on tuning and maintenance, the frequency response of a well-tuned plate reverb can be made remarkably flat (Figure 11.11), at least above about 100 Hz. These devices are tuned very much like a musical instrument, with the tension of the plate adjusted until a pleasing sound is created. For the most part, the plate is made as tight as practical for maximum undamped sustain, with uniform tension around the entire perimeter for a less clouded sound.
Figure 11.9 Impulse response of a plate reverb.
Figure 11.10 Reverb time by octave band of a plate reverb.
Figure 11.11 Frequency response of a plate reverb.
The relatively flat frequency response, from about 100 Hz up through the rest of the audio band, is preserved throughout the decay of the reverb (Figure 11.12).
Figure 11.12 Spectral decay for a plate reverb.
Figure 11.13 Reverb time, plate reverb versus benchmark halls.
Comparison of this plate reverb to the benchmark halls reveals noticeable differences. The plate′s reverb times (Figure 11.13) are markedly lower than these halls at all octave bands measured.
The bass ratio comparison in Figure 11.14 highlights the lack of warmth in the reverberation of this mechanical device. However, creative recording engineers have used this to advantage, finding successful applications for the thinner sounding reverb. Popular music has little in common with the orchestral music typically played in these benchmark halls after all.
Figure 11.14 Bass ratio, plate reverb versus benchmark halls.
As with the spring reverb, the full sound quality of the plate reverb is not fully captured by these measurements. Sound engineers rely on intuition, experience, and critical listening when choosing to work with this effect.
While the synthesis of reverberation through math requires an intense amount of calculation horsepower, the resources are certainly available to do so. In fact, digital reverberators have been the most common tools for adding reverb in popular music recordings for some 20 years.
Countless algorithms exist for generating reverb digitally. Below is a discussion of the basic building blocks that are a part of the digital reverb.
Infinite Impulse Response Reverberators
Digital signal-processing algorithms, based on recirculating delay systems forming a combination of comb filters and all-pass filters, form a class of infinite impulse response (IIR) reverberators, which are very much in use in recording studios today.
Figure 11.15 Reverb is a very large collection of chosen delays.
Comb Filter
Reverberation is a wash of sound reflections. It follows that each reflection could be modeled using a well-chosen digital delay. Figure 11.15 simplifies this approach, beginning to trace out all available reflected sound paths propagating from a sound source in one location to a sound receiver in another. Map out the time of arrival and amplitude of each reflection, and the concept for a machine generating reverb through delay begins to emerge. A key question arises. How many delays are needed?
In a time of significantly less digital signal-processing horsepower, Manfred Schroeder famously explored simple ways to fabricate reverberation from a ″dry″ signal. Back in 1962, when the Beatles were singing, ″She loves you, yeah, yeah, yeah,″ Schroeder first proposed creating reverberation through the use of a single delay with regeneration, the comb filter (Figure 11.16a). The time response to an impulse shows the reverblike quality of the filter, with its exponentially decaying reflections. The frequency response of a comb filter (see Chapter 9) reveals its spectral shortcoming as a source of reverb.
When the delay time is set to a high value, flutter echo is unmistakable. When the delay time of the comb filter is set to a low value, a dense set of echoes is satisfyingly produced. However, the frequency response will exhibit a wide spacing of peaks and notches. Strong coloration results. It is only these frequency peaks that are reverberated, and the spectrum in-between decays quickly.
Figure 11.16 Reverberation from (a) a comb filter, (b) an all-pass filter.
All-Pass Filter
Schroeder immediately improved on his reverberator. Through feed-forward and scaling, the comb filter can be made into an all-pass filter (Figure 11.16b) creating a series of exponentially decaying reflections with flat frequency response. This flat frequency response is based, mathematically, on an infinite time integration. Of course, human hearing does not wait that long. The result, unfortunately, is that in the short term, the frequency response is not necessarily flat, some coloration remains, and the sound is dependent on the delay time selected and can sound quite similar to the comb filter discussed above.
Hybrid Combs and All-Pass Filters
Combining multiple comb filters and all-pass filters is productive. Perceived coloration can be reduced if the peaks of a comb filter are made more numerous (increasing modal density). Flutter echo is diminished if the number of delayed signals is increased (increased echo density). Comb filters combined in series are ineffective because only the frequencies in the peaks of one filter are passed on to the next. For a signal passing through several comb filters in series, only frequencies within the peaks common to all comb filters will appear at the output. Multiple comb filters, when used, are placed parallel to each other.
All-pass filters, on the other hand, may be connected in series freely, as the net output will still be all-pass. Using multiple all-pass filters will, therefore, increase echo density while preserving the (long-term) frequency response. Use of multiple comb filters in parallel, with strategically chosen delay values that have no common factors, can create a frequency response with peaks distributed throughout the frequency range of interest, increasing modal density and decreasing perceived coloration. Parallel application of multiple comb filters also increases the echo density as the output is simply the sum of the individual comb filters, with a corresponding increase in echo density.
Schroeder founded an industry of digital reverberation when he offered his hybrid combination of parallel comb filters feeding a series of all-pass filters back in 1962, as shown in Figure 11.17. Countless permutations exist for choosing any number of parallel comb filters and serial all-pass filters. Designers are also free to adjust in complex ways the variables, such as delay times and gain factors, for each of the individual comb and all-pass filters. Furthermore, the introduction of low-pass filters within the comb filter simulates air absorption and makes for a more realistic reverberation. Limitless options are available to the creative digital signal-processing designer.
A common, high-performance digital reverb based on this approach, with parameters set to emulate a large orchestral hall, is measured. The measured impulse response (Figure 11.18) is analyzed to reveal a mid-frequency reverb time of about 2.5 seconds. Octave band calculations of reverb time (Figure 11.19) demonstrate behavior consistent with an idealized real room. No single octave band dominates, minimizing unpleasant coloration. Additionally, there is a clear trend for high frequencies to decay more quickly, mimicking the air absorption of a real hall.
Figure 11.17 Reverberation from parallel comb and serial all-pass filters.
Figure 11.18 Impulse response of a digital reverb (large hall).
Figure 11.19 Reverb time by octave band of a digital reverb (large hall).
The frequency response (Figure 11.20) shows a natural roll-off of high frequencies consistent with the performance expected in a highly diffused, air-filled, real hall. The digital algorithm decays naturally, without pronounced modes, but with lingering low-frequency reverberation adding desirable warmth (Figure 11.21).
Figure 11.20 Frequency response of a digital reverb (large hall).
Figure 11.21 Spectral decay for a digital reverb (large hall).
Of course, digital reverbs like this one have many adjustable parameters that enable the recording engineer to modify this particular measured response. The reverb can be significantly altered. Nevertheless, it is informative to make comparison to the three benchmark halls.
Reverb time comparison (Figure 11.22) shows this popular reverb algorithm is consistently longer than the most desirable halls across all octave bands. Keep in mind that the recording engineer has many tools to control the perceptual impact of this reverb (discussed below), with the result that a very long reverb can be integrated into a music recording without the expected clutter and obscuring that a similarly performing room would have; the digital reverb in this common configuration creates a longer reverb than might be expected to be pleasing.
Figure 11.22 Reverb time, digital reverb (large hall) versus benchmark halls.
Figure 11.23 Bass ratio, digital reverb (large hall) versus benchmark halls.
The trend, octave to octave, of the digital reverb parallels that of the benchmark halls. The highly sought-after warmth that comes from a bass ratio slightly above unity is clearly built in to this reverb algorithm (Figure 11.23).
Convolution Reverberators
If there were a way to capture some sort of blueprint that describes the detail (both the time of arrival and the amplitude) of all the reflections that make up reverberation, and apply it to other sounds, reverb could be synthesized. Convolution reverbs do exactly that.
The impulse response that drives the convolution might be measurement data from an existing space, calculated data from room modeling analysis, or wholly synthesized data fabricated by the creative sonic artist.
Convolution Defined
Earlier in this chapter, comparison was made of a voice singing outside, and then singing inside the Notre Dame Cathedral. Consider a handclap instead of singing: ″Pop!″ The hands colliding make a sound of very short duration. Outdoors the single handclap is over as soon as it began. Inside the cathedral, the handclap is followed by a burst of reverberant energy that is made up of a very specific set of reflections — handclaps that have been attenuated and delayed according to the size, geometry, and material properties of the space in which the sound bounced around. The way the space converts a single spike of a sound into a more complicated wash of decaying sound energy is a defining characteristic of the space. Impulsive sounds are sounds of very short duration, like the handclap, a gun shot (blanks only, please), a balloon pop, a clave, a snare drum, etc. The way the room responds to an impulsive sound is called its impulse response. The impulse response contains the signature of the space as it augments every bit of every sound with its own specific pattern of reflections (see again Figure 11.15).
Convolution is the process of applying that impulse response to other sounds, like a vocal, an orchestra, a guitar, a didgeridoo, to create the wash of energy the space would have added to such audio events if they had occurred there.
It helps to simplify the complicated world of a real space to one that is easier to understand: a space that impossibly offers only two reflections (Figure 11.24). A single impulse within this space would always be followed by this exact pattern of two reflections, arriving exactly at the time and amplitude described by the impulse response. A louder impulse would trigger a correspondingly louder pair of reflections.
Figure 11.24 A simplified impulse response
If two impulses are played one after another in the space, each would trigger this impulse response and they would combine into a more complicated sound (Figure 11.25).
Convolution is the word that physicists and mathematicians use to describe this math. As a concept, it is fairly straightforward. As a calculation, though, it gets complicated fast.
Consider a more complicated sound — a snare drum, for example (Figure 11.26). Zooming in on the digital waveform to the individual sample level reveals the sound to be nothing more than a stream of pulses. For a sample rate of 44,100 Hz, the audio is made up of 44,100 such pulses of varying amplitudes each second. Let each one of these trigger the impulse response individually. Combine new sounds with those sounds still lingering. Keep track of the decay associated with each sound so that it can be combined with future sounds. Add them all up. The overwhelming data crunching rewards us with convolution reverb.
Figure 11.25 Accumulation of impulse responses.
If one convolves the complicated impulse response of a real space with the close-microphone recording of a snare drum, the result is a sound very much as it might have occurred if the snare had been played in the real space. Knowing what a space does to a simple impulse is enough to predict what it would have done to almost any signal, as long as there is a machine willing to keep up with all the associated calculations.
Figure 11.26 A snare hit, up close.
This is the basis for the family of reverb devices variously known as sampling, room modeling, and convolution reverbs. The math is so daunting that while theory suggested it many decades ago, engineers could not take advantage of it until computers got fast enough. In this, the twenty-first century, home computers are more than capable of implementing convolution in the recording studio.
To better appreciate the math, consider a symphonic hall with a two-second reverb. At a sample rate of 44,100 Hz, the two-second impulse would be made up of 2 × 44,100, or 88,200 samples. For stereo, two impulses are needed: that seen by the left ear and that received at the right. The hall is then characterized by a stereo impulse response that has 2 × 88,200, or 176,400 pulses. When this is convolved with an audio waveform like that snare drum that is also digitized at a 44,100-Hz sample rate, each one of those samples triggers 176,400 numbers that must be stored and added to the variously decaying elements of all the other pulses of the audio. That is, 44,100 times per second, the system gets another wash of 176,400 numbers to keep track of. Each second of audio requires 3,889,620,000 samples of reverb to be stored and included in the calculations. Raise the sample rates in those calculations from Redbook CD to the middle-of-the-road high-resolution sample rate of 96 kHz, and some 18,432,000,000 samples are part of the calculation for every second of source audio that generates stereo reverb. It takes a fair amount of CPU horsepower to do this in anything close to real time. This horsepower was not readily available in off-the-shelf computing until about the year 2000. These are good times to be creating audio.
Summing up convolution:
- Pulse code modulation converts audio into a series of pulses.
- Each audio pulse triggers an associated impulse response.
- Each impulse response is scaled by the height and polarity of the audio pulse that triggers it.
- Keep track of the amplitude, polarity, and time of arrival of all associated pulses and add them up.
Convolution Parameters
Engineers who have been using those legacy digital reverbs not based on convolution have grown accustomed to the process of selecting a reverb patch and then tweaking it into submission. That is, they select a ″hall″ or a ″medium room″ program as a starting point and then adjust parameters until the quality of the reverb fills the needs of the multitrack production.
Convolution reverbs do not offer quite as much flexibility. Their strong basis in reality (it is a theoretical representation of what actually happens in a real room after all) makes them difficult to modify. To select the type of space, simply choose an appropriate impulse response — one from a hall or a medium room, for example. So far, so good.
With nonconvolution reverbs, adjustable parameters beyond reverb time, predelay, bass ratio, and such are freely available for modification. Such parameters are not always available in a sampling reverb. In a real room, the reverb time gets longer when the space gets larger or more acoustically reflective. This implies that to lengthen the reverb, one has to move the walls or change the sound absorptivity of the ceiling, remeasure the resulting impulse response, and then execute the convolution. In other words, to lengthen the reverb time in a convolution engine, one must select a different impulse response entirely, one with a longer reverb time (Figure 11.27).
This causes headaches. The reverb time of the Concertgebouw in Amsterdam, a much-loved hall, is fixed. There is simply not a knob to adjust reverb time in the space. Therefore, there is not an easy way to get that Concertgebouw sound at 1.5 seconds, 2.0 seconds, and 2.5 seconds. The Concertgebouw sound is what it is. If it sounds gorgeous on the lead vocal, then the engineer is satisfied and gets on to the next task. If it sounds good, but the reverb time is a bit long, the engineer is going to have trouble shortening it without altering the many subtle qualities that lead to the selection of this impulse response to begin with.
Figure 11.27 Changing the reverb quality generally means changing the impulse response.
The impulse response contains not just the reverb time of a space, it contains the many other properties of a space that give it its character. While any two impulse responses may look quite similar, the ears can extract quite subtle differences between them. Human hearing systems perform a miraculous amount of analysis on the sound quality of a space that is only partially explained by parameters such as reverb time, predelay, and bass ratio.
Consider two different halls with the exact same reverb time. Even though they both might take exactly 2 seconds to decay by 60 dB, they will likely sound different. One hall might be a bit brighter, adding airiness and shimmer as the sound decays within. The other hall might be warmer, adding low-frequency richness to the sounds within. The frequency response, not just the reverb time of the hall, is very much captured in the complex detail of a room′s impulse response.
Achieving reverb through convolution is terrific for emulating a known space. However, working in the recording studio, production demands may ask the engineer to modify the sound quality of that known space. It is common to find, for example, the sound of the convolution reverb on the vocal is directionally correct, but needs to be a little shorter or a little brighter. With limited adjustability, convolution can back the engineer into a signal-processing corner. The sound is perfect, almost. No small tweak is available that can bring it in line with what the engineer really needs. On a nonconvolution reverb, engineers simply adjust the appropriate parameter. On a convolution reverb, engineers most commonly hunt around for another impulse response and hope it sounds similar, only shorter. That is a different process. Hunting for different impulse responses and auditioning them is clumsy enough to interfere with the creative process.
Increasingly, convolution reverbs are adding adjustable parameters. Three signal-processing targets present themselves. The quality of the convolution reverb can be manipulated by processing the audio signal before it is sent into the convolution engine, the reverberation coming out of the convolution engine, and, perhaps most intriguing of all, the impulse response used by the convolution engine (Figure 11.28). Reverb time can be made shorter if the impulse can be shortened. The trick is shortening the actual impulse response of the Concertgebouw without robbing the reverb of other sonic features that make the Concertgebouw so desirable. If the engineer wants a shorter reverb overall, and a bass ratio that favors low frequencies a little more, it is difficult for the convolution reverb to respond and still preserve the original sound quality of the defining impulse response. Engineers must listen very carefully when modifying convolution reverbs in this way. The history of digital signal processing makes clear, however, that audio professionals can look forward to continued dramatic improvements. Tomorrow′s reverbs will likely be better than today′s as the software engineers create better algorithms, and the computer develops more signal-processing power.
Figure 11.28 Effects processing can be implemented on the input, the output, and the impulse response.
Convolution Caveats
All of this is well and good as long as the convolution theory discussed above is true. Skeptical readers may be wondering: Is the impulse response really a perfect picture of a space that can be applied to any signal? It is worth pointing out the limitations so that engineers can make better use of this kind of technology in their productions.
Reverb Quality = Impulse Response Quality
The convolution reverb is only as good as the impulse response measurement. That is, a bad measurement (noise, distortion, poor frequency response, undesirable measurement positions, etc.) is going to flow through the convolution calculation to create similarly bad-sounding reverberation.
When a performance hall is measured, high-quality gear is required. The measurement system requires both a loudspeaker and associated electronics to produce the test signal (which might be an impulse, noise burst, swept sine, pseudorandom noise, or other) and a microphone/microphone preamplifier system to capture the measurement data. This entire measurement system has a number of demands placed upon it. The system needs to have a flat frequency response so that the equipment itself does not color the measured spectral content of the hall. The test equipment needs to have a full-range frequency response — reaching as low and high as practical — to engage and capture the hall′s full bandwidth, low- to high-frequency characteristics. The impulse response measurement system needs to be omnidirectional to trigger and collect the contributions from all the room′s boundaries. The measurement gear must have great dynamic range — low noise and high level before distortion — so that the resulting impulse response is itself low in noise and not clipped.
In order to measure spaces and add to the collection of available impulse responses, one needs very high-quality gear. Impulse response measurements are approached like an important location-recording gig. The same mindset that leads to great classical recordings captured at the hall leads to great impulse responses for convolution reverbs. Discipline, practice, a certain amount of paranoia, and the highest quality equipment can lead to clean and beautiful impulses.
Engineers do not just shop for the convolution reverb product that meets their needs. Impulse responses for many of the famous spaces are available for purchase too. Acquiring impulse responses is a different kind of gear acquisition, perhaps without precedent in the professional audio industry. Behind every impulse response considered, someone else gained access to the space, chose the measurement gear, placed the speakers and microphones, and captured the measurement. The art of measuring a hall is not unlike the art of recording drums. Spaces, like drum kits, are big, vague instruments that respond well to good microphone selection and placement. They are also unforgiving of bad recording craft. The simple fact is some measurements are better than others. The quality of the reverb coming out of the convolution engine depends on their skills in capturing the impulse response. When a recording engineer uses a convolution reverb in their multitrack production, he is also using the recording skills of another recording engineer, the one who orchestrated the impulse response measurement.
One Space = Countless Impulse Responses
No space is completely defined by a single measurement. It depends on the location of both the sound source and the listener (Figure 11.29). A measured impulse response changes whenever the measurement position changes. After all, the hall sounds different in the front row than it does on the balcony. Moreover, it sounds different at the left ear compared to the right ear. For good sounding convolution reverberation, one seeks to measure in locations that ″sound good.″ Recording experience, technical know-how, creativity, and intuition influence these decisions.
Figure 11.29 The impulse response changes when the source and/or reciever position changes.
But this is not just about microphone placement. The impulse response changes whenever the position of the sound source is changed. That is, a singer standing center stage has a certain reverberant sound. But a singer standing at the rear right of the stage creates a different impulse response. Any single impulse response is only representative of a single location for the source and a single location for the listener. A 2,000-seat auditorium needs 2,000 stereo impulse response measurements for each stage location to even begin to tell the whole story of that performance space. A large stage capable of holding an orchestra of 100 plus a chorus of 100 leads to a sobering realization: One must choose which of the 800,000 choices of impulse response will be used.
The engineer first envisions the type of reverb needed for the multitrack project — perhaps a little Carnegie Hall on the snare drum. This is not enough. Decisions must be made about the location of the snare drum in Carnegie Hall. Is it on the stage or not? Sure, use the stage. But exactly where on stage will sound best? In addition, the engineer must consider the seat in which the listener is to be virtually placed. When these decisions are made, one then finds the correct impulse response — where the measurement engineer puts the measurement loudspeaker at the desired snare drum location and the measurement microphones at the preferred seat in the house.
In the course of the mix session, an engineer may be inspired to try the snare drum on the balcony, with the listener on stage. Good idea. Tough luck. They will need an impulse response with that exact configuration. Such ideas cannot be explored with the push of a button or turn of a knob.
Impulse ResponseNOW ≠ Impulse ResponseTHEN≠
The impulse response is really only representative of the reverberation of a space if the hall′s performance doesn′t change over time. Just as the impulse response is specific to a given source and receiver location, it is also somewhat specific to that instant in time.
Does the sound of a hall change over time? Surprisingly yes — ever so slightly. We look here, not at how the symphony hall sounds different now than it did in 1917, but at how the hall sounds different at the end of the song than it did at the beginning of the song. That distinct pattern of reflections that the impulse response documents does change slowly with time. If the walls move, the patterns of reflections in the impulse response must change. But even if the room partitions are not moving (which is generally the case), the impulse response can still change. These changes are caused primarily by the typical air drafts that occur whenever a group of people occupies a space.
We warm-blooded humans generate heat. Hot air rises, and does so inconsistently. Air conditioning fights back with an occasional breeze of cool air. As the air speeds up, slows down, and changes direction within these thermal drafts, it shifts ever so slightly the time of arrival of each reflection. The speed of sound is about 1 ft/ms plus the speed of the medium. If the air is moving with the sound, it propagates a bit more quickly. If the air is moving into the sound, it slows its progress ever so slightly.
This problem is most apparent at shorter wavelengths where a slight breeze might push the sound wave by a larger portion of a cycle. That is, a large wavelength (low frequency) pushed by this small convection current might vary a degree or two of phase sooner or later when it arrives at the listener during the song — likely imperceptible.
The same breeze would create a relatively larger phase shift for the shorter wavelengths (high frequencies) — possibly more perceptible. A single impulse response, therefore, might be thought of as being reliably representative of a hall at low frequencies. But it provides only a snapshot in time of the high-frequency portion of the hall′s reverberant character. An instant later, the actual impulse response of a room would drift into a slightly different pattern.
The slight time variations that occur to the reverberation of a real space are not captured by convolution of a static impulse response.
Sampled Gear ≠ Actual Gear
The time variability of the impulse response turns out to be a problem in a surprising way. With the convolution reverb′s ability to measure and reproduce the sound of a real space, it did not take recording engineers long to decide they could also sample the expensive racks of artificial reverb gear that the fancy studios have and that most engineers can′t afford.
Need an actual plate reverb? Sample one. Missing that top-of-the-line $15,000 digital MegaSweetVerb? Measure one. No access to unobtainable, vintage, first generation, weird sounding, early digital reverbs? Capture one.
Not so fast. Unfortunately, many of these other reverbs exhibit time variation. In fact, the best sounding digital reverbs that do not rely on convolution (e.g., those terrific sounding high-end reverbs by folks like Lexicon and TC Electronic) have gone to great lengths to make the reverb change over time. It is built into the algorithms of these devices. The designers of this equipment figured out pretty quickly that the reverb they synthesized sounded in finitely more natural and beautiful if the algorithm had some dynamic processing within, not unlike the modulation of a typical delay line set to sweep ever so gently.
Sample a top-of-the-line digital reverb to drive a convolution algorithm and the resulting reverb may have noticeably inferior sound quality. More than natural spaces, infinite impulse response digital reverbs strategically shift things in time. They cannot be convincingly defined by a single impulse response.
Convolution is not mathematically valid as a way to describe a system that changes over time. That is why simple convolution will not let us model other devices like compressors, wah wah pedals, distortion boxes, autopanners, chorus effects, etc. These devices cannot be fully described by any impulse response. Their response to an impulse changes over time. With no overarching impulse response, convolution has nothing useful to convolve.
On the other hand, those devices that do not change over time can be sampled, such as equalizers. Measure the impulse response of a reputable EQ and the convolver just became an equalizer. Trouble is, the impulse response is different for each and every knob position on the equalizer. So while the vintage Pultec tube EQ is tempting to capture with a convolution reverb, it will need several hundred impulse responses to even start to emulate it. Oh, and by the way, some of the appeal of that old EQ is its tubes — the way they distort so sweetly. Trouble is, since distortion depends on how hot the signal is going through the tube-based gain stage, and that amplitude changes over time, convolution does not apply.
In the final analysis, all of these reverb technologies — chamber, spring, plate, IIR, and convolution — are valid today. Each possesses unique advantages that the informed engineer knows can be strategically leveraged as needed on any production.
This single device, reverb, offers not just one but a vast range of effects, helping engineers solve technical problems and pursue creative goals. It takes many years to master the wealth of reverb-based studio effects, but the work pays big dividends. Exquisite use of reverb sets apart the truly great productions from the merely average ones.
Reverb effects, almost without exception, are parallel processes using effects sends, not serial processes using inserts (see ″Outboard Signal Flow″ in Chapter 2). Patch up aux sends into the reverbs so that a single effect can be accessed by any and all parts of the multitrack production. The reverb outputs are patched to effects returns or spare monitor paths feeding the mix bus.
For many applications, evoking the sound of a space (and all the feelings, memories, and social importance invested in that space) is the goal of artificial reverberation. Three-dimensional space, as it exists in a hall, church, club, or canyon, can be a useful analogy for the sound engineer who seeks to create a sonic landscape between or among loudspeakers. Among all of the reverb effects discussed here, simulating the sound of a real space is the primary motivation for adding reverb to a multitrack production.
specific Real Space
Discrete elements of a multitrack production, or possibly the entire mix, might be processed to convert close-microphone studio recordings into a sonic illusion that the instruments were played in a real space. Reverb devices synthesize a pattern of early reflections and a reverberant wash evocative of an actual space.
specific spaces, such as Boston Symphony Hall or Carnegie Hall are some-times the goal. Engineers, faced with a need to add reverberation to a recording, may wish to stay true to the original recording venue. A recording made in Carnegie Hall is often required to have the reverberant signature of Carnegie Hall. The end of a movement often reveals noise within the hall: people coughing, traffic just outside, air conditioning, etc. Some-times the recording is edited immediately at the end of the movement, and artificial reverb replaces the acoustic decay at that critical moment. The artifical reverb provides a clean, realistic musical ending free of extraneous noises. Any additional reverb added in the recording studio must, in this case, be perfectly consistent with the sound of — or our memories of the sound of — Carnegie Hall. Reverb devices are selected and the parameters are adjusted to achieve this.
In video postproduction, it is often necessary to replace the dialog. The dialog is recorded at the time the film is shot. Problems arise: some words are difficult to understand, the acting is not quite what was desired, the script changes, a distracting and unwanted noise happens off camera, the field-gathered sound quality is poor, etc. A single word or the entire scene may need new speech recorded. The actor wrestles with getting a great per-formance while syncing their timing with the original take. The engineer must worry about matching the ambience information in the replaced dialog. If the scene takes place in a church, taxi, or on another planet, the reverberation surrounding the dialog must sound exactly like that church, that taxi, or that particular planet.
Matching studio reverb to natural reverb elsewhere in the recording is a fine skill that the great engineers possess. An IIR reverb patch is selected, and its various parameters adjusted, until a satisfactory match is obtained. A convolution reverb is the ideal reverb for matching an existing space, if a quiet impulse response consistent with the original recording approach can be obtained.
Creative opportunities specific to convolution reverb await the adventurous engineer. If it can be measured, it can be convolved. When one wants the sound of the snare in the car, all that is needed the impulse response of the car. For the engineer who cannot help but notice how awesomely thunderous it is when someone slams a door in the concrete stairwell at the parking garage at work, who likes the sound of an empty swimming pool, who sings in the shower — all of these spaces can become reverb patches as soon as an impulse response is captured. Many convolution reverbs come with this very desirable ability. Take loudspeakers, microphones, and the convolution reverb anywhere, capture its impulse response, and then use the signature sound of that space — any space — in a future recording.
Moreover, thoughtful and generous people are sharing and swapping their own impulse responses with other engineers. Reverb transactions occur: ″I′ll trade you my parking garage stairwell for that church in your neighborhood.″ Every engineer has access to these opportunities.
Generalized Real Space
A realistic space might be evoked through loudspeakers for purely creative reasons. Relieved of any burden to simulate a specific place, the goal of the reverb signal processing might simply be to create a believable sound of a likely, surprising, or otherwise appealing space. Here the engineer seeks a reverberant character that sounds appropriate for the music and evokes an architecturally-grounded image of a realistic space. It is not necessarily Boston Symphony Hall. It is some kind of hall, convincing and realistic. Maybe it is a little smaller than the real hall. Maybe it is a little darker (sonically or visually or both). The creative engineer seeks to stimulate the imagination and sonic memory of the listeners and let them fill in some of the details themselves.
Evoking realistic spatial qualities is not limited to symphony halls and opera houses. Production goals might lead the engineer toward the sound of a canyon with its long echoes, a gymnasium with its shorter echoes, or a shower with its distinct resonance. Many of the same digital reverberators that evoke real halls can also be adjusted into configurations evocative of other real-sounding spaces. In these applications, the engineer has more freedom to adjust parameters. The space must be realistic, but not specific, allowing the engineer to be quite creative.
Spatial Adjustments
Less obvious spatial applications are common in popular music recordings. The lushness of a reverberant symphonic hall is easily heard, but more subtle approaches have value.
Close Microphone Compensation
Much of the audio in popular music is recorded using close-microphone techniques. Placing the microphone very near the musical instrument leads to tracks full of intimate detail. Done well, this can certainly be a pleasure to listen to; it is a proven technique for many important and successful recording artists. It is very much part of what makes the recorded music sound ″better than real.″ However, a multitrack project consisting entirely of close-microphone tracks may be too intense, too intimate, too exaggerated, or too unnatural for some styles of music. On any given individual track of the multitrack arrangement, the close-microphone intimacy may not suit the creative goals of the art.
In this case, audio engineers use a reverb device with a good dose of early reflections (generally a simple cluster of strategically chosen delays) and very short decay time (digital or plate reverb), mixed in at a just-noticeable level, with the goal of diminishing this close-microphone immediacy. The sonic result is that the recording sounds as if the microphone were further away from the musician at the time the recording was created. To the less-critical listener, the sonic result is a believable and pleasing reduction in the intimate detail of the particular track. It sounds more realistic, less exaggerated, and less intense.
This type of effect might use any of a number of reverb technologies, such as a plate or a digital reverb. Even if a ″large hall″ type of reverb setting is employed, the large sound of a large space is very much suppressed here. The reverb is used only to diminish and blur the spectral detail of a music track to better assemble a loudspeaker performance that is consistent with the artist′s goals.
Liveness
Employing reverb signal processing to simulate a collection of early reflections only, without the wash of late energy typical of a real space, can perceptually diminish the close-microphone signature and contribute to the illusion that the audio track was performed in a real space. A direct sound followed by a volley of reflections indicative of a real room geometry can make for a more compelling stereophonic illusion of location (left to right). In addition it can help evoke in the listener′s mind a real location such as a room or stage.
Springs and plates do not create discrete early reflections. These mechanical devices initiate dense reverberation very quickly. For sparse early energy, digital devices are needed. The spacing, in time, of these early reflections is typically adjustable. The pattern of the early reflections — their relative amplitude and time of arrival — is determined in a real space by the room geometry and acoustic properties of the materials used to construct the space. The overall timing of these delays is determined by the scale of the space. Larger rooms will have these reflections spread out in time more than smaller rooms. Many digital reverbs emulate and make this property adjustable (Figure 11.30).
Application of reverb in this way adds the pattern of early reflections of any space, small to large, to a recorded signal with an adjustable amount of the associated late decay of a reverberant room. With digital reverb, it is possible to have early reflections evocative of a space with certain size, without any of the reverberant energy such a space would physically be required to create. The audio track achieves a more precise placement within a believable, real space without the reduction in clarity and intelligibility that the late reverberant energy would typically cause.
Note that the creation of the sonic illusion of a large or small space without the associated reduction in clarity and intelligibility caused by reverb benefits not only the individual track being processed but also the many other elements of the mix seeking to be heard clearly. A long, reverberant wash on the snare drum can mask some speech frequencies (see Chapter 3) and diminish the intelligibility of the lead vocal. The decaying energy of a real space can mask other important elements in the multitrack production. A reverb consisting of early reflections without this late energy does not contribute as much to this masking.
Multitrack projects are often crowded with 24, 48, and often more tracks filling the arrangement simultaneously, fighting to be heard. It requires great care on the part of the mixing engineer and the arranger to prevent the mix from sounding cluttered when played back over loudspeakers. Long reverb times can be the bane of clarity. Unburdened by room acoustics, engineers regularly remove the late energy and use only the early energy, creating a live, but intelligible sound. In this manner, recorded music is made to sound as if it were occurring in an actual space, while the perceptual detail of the tracks remains unnaturally, unbelievably vivid.
Figure 11.30 An IIR reverb with and without added early reflections.
Width
Through reverb, any single track of the multitrack production can be made to sound as wide as the loudspeakers are placed, perhaps a little wider. Augmenting a sound with reverberant energy that is clearly driven by the source sound (but provides each ear with even slightly different signals) can create a thrilling effect. Subtle differences at the two ears can lead to a perceived widening of the apparent width of the sound source, and can create a more immersive feeling for the listener. Surround sound offers the engineer the chance to make this effect more pronounced and more robust.
An additional hazard of the close-microphone craft is that even sections of musicians (a string section, a horn section, a choir) might be tracked in sonic isolation. That is, each member of the section has an individual, independent track on the multitrack recorder and thereby its own, isolated sound. It is not blended with the other members of the section by the acoustics of a stage or a room. The close microphone picks up the individual member with very little acoustic energy from the other players in the section. Valid, practical motivations push pop music recording in this direction. The result is that sometimes reverb is added back later, when desired, to reassociate the players into a single ensemble.
Multitrack production gives artists the freedom to have possibly every single element of the multitrack arrangement fall into its own, unique ambient environment. This freedom is not always exercised. Using a single kind of reverb effect on many instruments helps to reattach these isolated tracks of audio back into a single section again. A collection of isolated point sources is merged into a single, broad section of players coming out of the loudspeakers with the unifying sonic signature of a single ensemble. Any reverb type is appropriate to this application, but digital reverbs set to ″hall″ or ″medium room″ are a good starting point.
Unreal Space
Beyond creating a space that is at least conceptually born in architecture, reverb devices are also used for synthesizing a reverberant character that may not exist in nature. Wishing to enhance the music with lushness or some other form of beauty and held only to a standard that it ″sound good,″ an engineer might dial in settings on a reverb device that violate the physics of room acoustics. For example, an engineer might collect a pattern of early reflections typical of a small room, combined with a late decay typical of a large hall. In addition, the decay might be brightened (high frequencies emphasized) for a pleasing timbre. Digital reverbs make these parameters adjustable. Such an amalgamation of acoustic elements is gathered despite the implications that there now exists a small room geometry (early re-flections) inside a large hall (long decay) in which air absorption has been miraculously overcome (the bright quality).
Such a reverb can be made to sound glorious coming out of loudspeakers. When reverberation is synthesized in a device, it is not burdened by the physics of room acoustics. It is a signal processing creation. Sound engineers regularly abandon any connection to room acoustics and stretch the capabilities of a reverb device to whatever limits necessary in pursuit of improved sound supporting the art of music, not the physics of acoustics.
Reverberation as described above to evoke spatial qualities is part of the signal processing — along with level settings, panning, EQ, and other effects — used to assemble an auditory scene. Within the sound field created by stereo or surround monitoring, this processing begins to make adjustable the subjective spatial parameters of depth and distance, the immersive attributes of discrete sound sources, groups of sound sources, and the size and quality of the performance and playback environment. All the spatial variables are accessible and changeable for the sound engineer.
11.4.2 NONSPATIAL APPLICATIONS
Reverb devices have their roots in room acoustics. The human experience in reverberant spaces motivated scientists and engineers to invent reverberant devices to ″?x″ tracks recorded in the highly absorptive rooms of recording studios, using close-microphone recording techniques. This historical cause (room acoustics) and effect (reverb devices) does not limit reverb devices to this single use, however. Engineers will apply reverb signal processing for other reasons: for some other practical benefits or simply for the aesthetic change it brings to the production. In this way, reverb devices take on many of the attributes of musical instruments. A given reverb might be sought out for its own subtle, subjectively beautiful sonic character. Another reverb might be utilized for its ease of use, its playability. Engineers select and adjust reverbs to create timbres and textures that support the music, to influence the audibility of the various elements of the multitrack arrangement, and to synthesize completely new sounds; all of these results are not spatial in nature. Reverb signal processing finds other functions.
Timbre Through Reverb
In some situations, reverb is used specifically because of the coloration it brings to a sound. Their frequency response, density of reflections, and resonant nature make springs, plates, and chambers useful devices for coloration. Productions that use these devices (and the digital devices that seek to simulate them) in this way do not seek to mislead anyone into believing that a real space exists around the instruments. There is a reverberant wash of energy. The character of reverb is certainly there. However, evoking an illusion of a symphony hall or an opera house is beyond the capability (or, indeed, the modern day intent) of these devices. Yet, despite the existence of many powerfully effective digital signal processors that can reliably evoke illusions of real spaces, preceding technologies with all their shortcomings are still used in popular music.
State-of-the-art digital signal processors provide the ability to simulate real spaces with presets labeled ″hall,″ ″medium room,″ and ″cathedral.″ It is interesting to also note that they simulate other reverb processors, with patches called ″chamber″ and ″plate.″ While springs, plates, and chambers were invented out of necessity to add reverberant character back to dry recordings, pop music engineers (and less consciously, pop music listeners) came to like their sound qualities even as more realistic and natural reverberation technologies were developed.
The coloration of a chamber comes in part from its small size. The highly reflective surfaces of the chamber have indeed lengthened the reverb time per Sabine′s equation. The modal density is not sufficient, however, especially at the lower end of the frequency range, to prevent obvious resonances. The room is simply so small that the existence of resonances may be audibly obvious. In a hall, this is a well-known problem to be avoided. In popular multitrack production, the engineer creatively matches this resonant behavior with tracks of music that are flattered by this coloration.
In this way, reverb devices are used as an alternative to equalizers and filters (see Chapter 5), which directly alter the frequency content of the signal by design. The resonance of a chamber, plate, or spring might add some sort of glow to the track that sounds pleasing.
Beyond resonance, reverbs can influence the perceived timbre of a signal through frequency-dependent reverb time differences. A bass ratio target value of 1.1 to 1.5 is known to contribute to the perceived warmth of the instruments that play in the space. Likewise, employing a digital reverb that creates reverberation without modeled air absorption, high-frequency reverb times are often stretched longer than mid-frequency reverb times. This adds high-frequency energy to the loudspeaker music, leading to a perceived airiness, sparkle, shimmer, or other pleasing high-frequency quality.
The frequency response of a spring reverb (see Figure 11.5), sloping steeply upward with frequency through about 4 kHz, influences the perceived timbre of the audio track being processed. The mixture of an audio track with reverberation like this colors the sound, ideally in a flattering way. A steel string acoustic guitar might be made to sound a bit brighter still through judicious use of spring reverb.
The plate reverb offers a different spectral adjustment (see frequency response in Figure 11.11), offering more uniform output from just above 100 Hz to just above 10,000 Hz. Brightness with midrange complexity can be created by adding some plate reverberation to an audio signal.
In this way, reverbs are used to subtly shape the perceived timbre of the audio tracks being processed. Chambers, springs, plates, and radically manipulated digital reverbs are well suited to this approach.
Texture Through Reverb
The unique sonic character of some reverbs, especially springs and plates, leads to their use for elements of texture. Reverb is added not for the spatial attributes it evokes, but for the quality of the sound energy it offers. This approach enables a sound engineer or sound designer to add to a signal perhaps a pillowlike softness, a sandpaperlike roughness, a metallic buzziness, a liquid stickiness, etc.
Musical judgment motivates this application, and it requires experience and a good understanding of reverb and synthesis because there is no reverb patch labeled ″pillowlike softness.″ There is, however, the capability to make a track take on a soft, pillowlike texture if the sound engineer makes clever use of the reverb devices in the recording studio. Much as the disciplines of architecture and acoustics must collaborate to create a great sounding space, the fields of sound engineering and music intersect to find musical, aesthetically appropriate reverb applications.
Examples abound. A reverb might be used to slightly obscure a particular part of the multitrack arrangement giving those instruments an ethereal, veiled texture. The metallic sound of a spring reverb might be used to help a track take on a steely, industrial texture. That pillowlike soft texture can be created through use of a long (reverb time exceeding 1.25 seconds) plate reverb with a long (about 120 ms) predelay, and a gentle roll-off of the high frequencies (starting at about 3–4 kHz).
When a composition is orchestrated, a horn chart arranged, or a film scored, elements of texture are a part of what motivates the creative thinking. A similar approach can be applied to reverb to help create textures and their associated feelings in support of the music.
Overcoming Masking Through Reverb
The multitrack arrangement contains musical elements of style, melody, harmony, counterpoint, and rhythm plus the sonic elements of loudness, frequency, timbre, texture, and space, generally with the overarching intellectual element found in the lyrics. All of these are explored, created, composed, and adjusted to values that are appropriate to the music. Such musical and audio complexity must somehow be communicated through loudspeaker playback. The difficulty of fitting so much detail into a pair of loudspeakers is captured by the common saying in audio circles that, ″Mixing is the process of making each track louder than all the others.″ Of course, this circular reasoning makes no logical sense. But it resonates with every pop music recording engineer who has tried to mix a complicated multitrack production. A given guitar performance might sound terrific alone. Add in the horn parts, and the guitar performance likely becomes more difficult to hear and enjoy; its musical impact is diminished. A talented mixing engineer figures out ways to introduce the horns to the mix and yet maintain the important musical qualities of the guitar. Reverb is one of the tools used by the engineer to extract selective clarity, size, richness, etc. from the various elements of the multitrack production.
Contrast Through Reverb
Taste and style might dictate that the entire multitrack production have the same reverb signal processing, putting the entire mix in a single space. The accumulated overdubs, instrument by instrument, are processed with the goal of placing each of them in the same, single space spread out between and among the loudspeakers. The lead vocal, drums, piano, and hand percussion are all treated with reverb so as to create the illusion that all these players are in the same room together. That the players played separately, at different times, possibly in different studios, does not necessarily diminish this illusion of a single space. Signal processing is applied with the goal of uniting these instruments together in a single fabricated sonic space. While less likely, a similar global approach may be used with any of the nonspatial applications of reverb as well.
However, as a long list of rebellious rock-and-roll musicians will testify, there are no rules in music. While the creation of a single unifying space may be a goal for many pieces of music, it is perfectly reasonable to pursue contrasting spatial qualities among various musical tracks in a multitrack production instead. The vocal might be made to sound as if it is in a warm symphony hall with one reverb, while the drums may appear, sonically, to be in a much smaller and brighter room courtesy of another reverb, and the tambourine gets some high-frequency shimmer from yet another reverb. Applying different processing to different elements of a multitrack production reduces masking, making each sound or group of sounds treated with a unique reverb easier to hear (see Chapter 3). This enables a broad range of spatialities and effects to coexist in a single, possibly crowded multitrack recording. In popular and rock music, this approach is the norm. Even simple productions commonly run three or four different reverb units at once, each creating very different reverb qualities. A globally applied, single space or effect is the exception.
As so many pop productions have shown, the sound of a voice in a hall coexisting with the drums in a small room does not lead to any mental dissonance. Listeners are not troubled by the fact that such sounds could never happen naturally in a live performance. Listeners are motivated by what sounds beautiful, exciting, intense, and so on. Listeners are drawn to each of the different reverberant effects, making the variety of musical tracks within the multitrack production easier to hear, and therefore easier to enjoy.
Scene Change
Similarly, the reverb effect used can change during the course of the song. Typical approaches include a scene change from the intro to the first verse. Perhaps the introduction occurs in a lush, highly reverberant soundscape. When the first verse begins, the reverb vanishes, leaving a more intimate setting.
This approach knows no limits. The chorus might be accompanied by a transition from intimacy to a live concert feel, and so on. Active manipulation of the reverb choices and parameters along with advances in the song form is a cliché pop music effect.
Reverb Extrema
Applying a given type of reverb to isolated elements of the multitrack arrangement enables extreme reverb to be employed. Every instrument in the orchestra is treated to the same reverb because they play together in a single hall. Multitrack productions are able to add extraordinarily long reverberation in part because that reverb can be applied to selected tracks only, not the entire mix. Generally, a rock-and-roll drum kit would not be very satisfying to listen to with a reverb time in excess of 2.5 seconds. A noisy wash of reverberant energy would make it difficult to hear any details in the drum performance. But the vocal of a ballad — the vocal alone — may soar toward the heavens with such a long reverb. The drums, meantime, might be sent to a different, much shorter reverb to better reveal their impulsive character.
The multitrack production process makes it possible not only to apply a given reverb to any single track, but also to apply reverb to a single phrase or single note of a performance. The reverb effect can be turned on and off, instant by instant, during the song. It is all or nothing for an orchestra in a reverberant hall. Any degree of fine control is allowed in a multitrack mix using reverb devices.
In this way, it is not unusual for a reverb — when applied to isolated multitrack elements — to climb to reverb times in excess of 10 seconds. It is not uncommon for predelay to range from some 60 ms to beyond 100 ms, or even more. Bass ratio moves freely from a value 0.5 to beyond 2.0. High-frequency reverb times are routinely allowed to rival mid-frequency reverb times even though this could never happen in a symphony hall. The recording engineer reaches for whatever reverb settings support the sonic art they seek to create. Radical, physically impossible reverb settings are applied at any time, to any part of any track in search a better sound.
To be clear, many styles of music are not allowed so liberal an approach to reverb-based treatment. Some pop, most folk and jazz, and nearly all forms of classical symphonic music are presented through realistic recordings. This realistic approach looks for a space — a real, believable space — recorded in an actual room and/or synthesized by signal processors that creates a single convincing performance location. In the world of popular music, however, there is typically great freedom to apply spatial attributes and special effects to any element of the multitrack production, unburdened by the realities of room acoustics. Limits are pushed to the extreme, often with thrilling artistic results.
Unmasking the Reverb
While some reverb techniques strive to make the audio tracks easier to hear, other reverb signal-processing approaches seek to make the reverb easier to hear. The contrasting reverbs and the reverbs with extreme settings of parameters can serve to unmask the reverb itself. Predelay, for example, pushes the reverberant wash of energy further back in time, separating it from the audio track that causes the reverb to be generated. Classify the unprocessed audio track as the masker, and the output from the reverberator the signal to be detected (see Chapter 3). This point of view reveals predelay to be a key tool for overcoming temporal masking. A loud individual word within a vocal performance, as might occur in the chorus of a pop song, might easily mask the reverb effect that has been so carefully added to it. Predelay separates the vocal from the reverb making it easier for listeners to hear both.
Unmasking the reverb has an additional benefit. As reverb (or any other effect) is unmasked, its level within the mix may be reduced without removing the perceptual impact of the effect. Attenuating the reverberant energy within the mix without diminishing its aesthetic impact will leave more room sonically for the other elements of the multitrack production. The task of mixing becomes a bit easier. The clarity and precision of even a complicated mix improves.
Synthesis Through Reverb
Reverb signal processing is often the basis for the synthesis of wholly new sounds, acting very much like an electronic music synthesizer, played very much like a musical instrument. As with sound design and sound synthesis, the options are limited only by the imagination and creative force of the musician and the technical capability of the equipment. Some examples among limitless options of using a reverb device as a synthesizer are discussed below.
Gated Reverb
Signal processing aggressively violates laws of room acoustics in this application. A reverberant wash of energy is sent through a compressor (see Chapter 6), followed by a noise gate (see Chapter 7) that radically alters the decay of the reverb, resulting in an abrupt, double-slope decay.
A gated reverb system is connected (Figure 11.31). The outputs from the reverb device are sent to a stereo (two-channel) compressor, whose outputs in turn are sent to a stereo noise gate, whose outputs are sent to the mix via the mixing console. The gate is keyed open reliably by the close-microphone snare track.
An illustrative example begins with a snare drum recording. Reverb is added to this sound (Figure 11.32a). The resulting reverb is sent to the compressor. The compressor, by attenuating the louder portion of the reverberant wash, flattens out the initial decay of the reverb, giving it a more gradual slope implying a longer reverb time (Figure 11.32b). The gate then cuts off the decay with a slope indicative of a much shorter reverb time (Figure 11.32c).
The result is a concentrated burst of uncorrelated energy associated with each strike of the drum. This reshapes the sound of the snare drum into a more intense and more exciting sound. The effective duration of each snare hit is now stretched longer, making it easier to hear (see Chapter 3).
This effect can be made prevalent in the mix, for all to hear. It is unmistakable. It is unnatural. Recordings from the 1980s elevated this effect to a cliché. If one uses it today, a bit of 1980s nostalgia is attached to the production.
Figure 11.31 signal flow for gated reverb.
However, gated reverb is also used in more subtle ways. A valid philosophy is to make the gated effect so subtle that it is barely noticeable, if at all, to the untrained listener. The goal is to add a bit of sustain to the sound so that the track becomes easier to hear without having to turn it up. In the case of snare drum, the additional harmonic complexity and stereo width are also welcome (see Chapter 13).
Reverse Reverb
Another sound synthesis technique based on reverb requires the temporary reversal of time in the recording studio. The goal is to have the reverb occur before the sound that causes it.
Consider a snare drum back beat within a pop song, falling on beats two and four of a measure (Figure 11.33). The typical addition of reverb to such a track is shown in the lower portion of the same figure. Note the lower waveform is the reverb from a snare drum, not an impulse response. This is the typical use of reverb for such a track. Reverse reverb takes a different approach.
First the snare track itself is played backwards in time (Figure 11.34). This can be done on an open reel analog tape machine by turning the tape over and playing it upside-down. Alternatively, within a digital audio workstation, the audio track is selected and the computer is instructed to calculate the time-reversed waveform. Reverb is then added to this time-reversed snare, creating the signal shown in the lower part of Figure 11.34 and recorded to an available track on the multitrack recorder. Finally, the original snare track and the reverse reverb track are reversed in time (by flipping the tape back, or executing another time-reversal command). This restores the snare track to its original place in time. The reverb derived from the time-reversed snare is found to occur before the snare drum sounds (Figure 11.35).
Figure 11.33 Adding reverb to a snare back beat.
Figure 11.34 Adding reverb to a reversed snare back beat.
The result is a kind of preverb. Sound energy swells up and into each snare hit, creating an unnatural, but musically effective, new snare sound. A short plate reverb was used for these illustrations, but there are no constraints on the type of reverb in this application. Musical anticipation, rhythmic syncopation, and the desire to fabricate an otherworldly sound motivate the sound engineer designing this effect.
Figure 11.35 Creating reverse reverb on a snare back beat.
This effect is somewhat simulated by reverb devices without resorting to backwards playback. Labeled ″reverse reverb″ or, more accurately, ″non-linear reverb,″ the amplitude envelope of a reverberant decay is aggressively altered so that the reverb tail seems to go backwards in time. That is, the reverb starts off relatively quietly, becoming increasingly louder before abruptly cutting off. Unlike true reverse reverb, nonlinear reverb happens after the stimulating sound, not before. The unnatural, antidecaying shape to the reverb tail alludes to reverse reverb. This is a powerful way of lengthening the sustain of short sounds such as percussion, lifting them a bit up out of the mix so that they are easier to hear. It also provides dramatic ear candy without risk of mix-muddying, sustained reverberation.
Regenerative Reverb
Reverberation in a hall is initiated by the sounds coming from the musicians within. Reverberation in the recording studio is not so constrained. Figure 11.36 shows a reverb being fed by two sources. This might be any type of reverb: spring, plate, or digital. First, the typical signal routing using an aux send to a reverb is used. The track for which reverb is desired is sent to the input of the reverb (using effects send 1 in Figure 11.36). The reverb outputs feed other inputs on the mixer, sending the reverberant signal into the stereo (or surround where applicable) mix.
Figure 11.36 Signal flow for regenerative reverb.
For a regenerative reverb, this standard approach is augmented by a delay. The track for which this kind of reverb is desired is sent not only to the reverb, but also to a delay unit (shown using effects send 2 in Figure 11.36). The delay time is likely set to a musically-relevant time interval, such as a quarter note, a dotted eighth note, a quarter-note triplet. The output of this delay in turn feeds the same reverb used by the original audio track.
The delayed signal is likely lower in amplitude as it reaches the input of the reverb than the original, undelayed signal. The sonic result of this second, delayed feed to the reverb is a subtle extra pulse of reverberant energy, in time with the music. In addition, regeneration on the delay unit, routing its own output to its own input, further delays the already delayed signal. This feedback of the delay to itself creates gently decaying, musically-timed repetitions of the signal. This is combined with the direct, dry signal and fed to the same reverb device. The result does not sound like any physical space in existence. This elaborate reverb system creates an ethereal, swirling, enveloping wash of energy resonating underneath the track.
While any reverb type is effective for creating regenerative reverb, large hall programs from a digital reverb are most common. Clearly, such a dramatic effect would generally be applied only sparingly to a single element or two of a multitrack arrangement — not the entire mix. While it is a matter of taste, it is often the case that this pulsing, regenerative reverb is placed in the mix at a very low level. Acting almost subliminally, it offers a rich, ear-tingling sound that only exists in the music that comes from loudspeakers.
Dynamic Reverberant Systems
As the reverb-based synthesis approaches described above make clear, signals may be processed both before and after the reverberation device to alter and enhance the quality of the sound. Effects devices are placed before the reverb, as was done with the delay unit to create the regenerative reverb shown in Figure 11.36. Effects devices are placed after the reverb, as was done with the compressor and noise gate used to create the gated reverb shown in Figure 11.31. This figure also shows that multiple effects devices may be connected in series to further develop the effect.
This approach can grow still more complicated by introducing additional processors in parallel, both before and after the reverb device. A single such system is discussed here as an illustration (Figure 11.37), but an infinite number of options exist for the sound engineer to fabricate a reverb-based sound.
Consider having two parallel feeds into a reverb, each processed differently. For example, each of the two input options might have different EQ curves applied to the signal. Each equalizer is tuned to a different resonant frequency, or perhaps one is bright (high-frequency emphasis) while the other is warm (low-frequency emphasis). Preceding these two filters is an autopanner.
An autopanner (see Chapter 8) is a machine-controlled pan pot. It automatically adjusts the amplitude of a signal between two outputs so that, as the level is raised on one output, it is lowered by an equivalent amount in the other. It is two synchronized amplitude modulators moving with opposite polarity. Its typical application, used in a mixer, is to create the illusion of an audio track moving left to right and back again as the perceived localization follows the louder signal. In this dynamic reverberation system, the autopanner is applied in a different way.
Figure 11.37 Signal flow for dynamic reverb.
The effects send from the mixer feeds the autopanner. Each autopanner output feeds a different equalizer. Both equalizer outputs are combined to feed the input to the reverb. The reverb system is shown in Figure 11.37. The resulting reverb is a constantly changing sound, perhaps subtle, perhaps obvious. This dynamic reverberation takes on a life of its own, a soft pad of ever-changing, uncorrelated energy adding life and mystery to an audio track.
Convolution Creativity
No one requires that engineers must convolve their audio tracks with actual impulse responses. Why not convolve the vocal track with a snare drum sound? Or convolve the vocal with the piano track? This has no physical meaning. There is no space with an impulse response like a snare hit or a piano performance.
Just because it is physically meaningless does not mean it will not have creative value.
You are free to convolve anything with anything: background vocals with an insect buzzing, kick drum with glass breaking, a guitar solo with a dog barking, etc.
So the sound of a singer ″in a snare″ is rather puzzling to think about. But how does it sound? While such questions may bother those obsessed with reality, they are rather inspiring to those of us obsessed with unique and beautiful sounds.
When a recordist starts to use convolution as a signal-processing technique applied freely to any sound, convolved with any sound, they have begun to use convolution reverb as a synthesis device. In this way, engineers fabricate new sounds and textures using the convolution hardware/software as a sophisticated sound design tool.
Artist: Roxy Music
Song: ″Avalon″
Album: Avalon
Label: Warner Brothers Records
Year: 1982
Notes: This mix makes significant use of Chamber One at Avatar in New York City, a multistory stairwell. Note in particular the lush vocals.
Artist: Bryan Adams
Song: ″Summer of ′69″
Album: Reckless
Label: A&M Records
Year: 1984
Notes: More ear candy from Chamber One at Avatar. The snare hit at the top of the tune offers an exaggerated example, but the rhythm guitar has plenty too.
Artist: Michael Penn
Song: ″Figment″
Album: Resigned
Label: Epic Records
Year: 1997
Notes: Dramatic change of scene driven by reverb shift between the A section (″Leave it for a while …″) and the B section (″Before the day is done …″).
Artist: Michael Penn
Song: ″Out of My Hands″
Album: Resigned
Label: Epic Records
Notes: Multiple simultaneous reverbs: Electric Bass is close-microphone track and dry. Drums enter with liveness and early reflections of a medium room. Acoustic guitar is also a close-microphone track and dry. Lead vocal is extremely wet, with reverb resembling a very large, dark hall program.
Artist: Paul Simon
Song: ″Spirit Voices″
Album: Rhythm of the Saints
Label: Warner Brothers Records
Year: 1990
Notes: Reverb on a single note: The conga accent at 0:47 in the intro gets emphasis through a nonlinear reverb.