...OR, HOW DOES TRENT REZNOR GET THAT DIRTY, STRAINED SOUND?
Signal processing devices can be technically defined as electronic devices that alter the audio signal in a nonlinear fashion. Given that definition, a fader, gain control, or amplifier is not a signal processor. Equalizers, effects processors, phase shifters, compressors, expanders, and noise reduction systems are all examples of signal processors; they all modify the input signal in some fashion. Equalizers modify sections of the frequency spectrum by boosting or attenuating certain frequency bands; effects processors are capable of introducing time delay, artificial reverberation through the use of strange phase-related algorithms, or a combination of both; phase shifters modify the phase characteristics of an input signal and add it back to the original, providing for "hollow" sci-fi effects; compressors essentially keep loud input signals from becoming too loud; expanders make soft signals softer; noise reduction units use a combination of compression and expansion to provide a lower noise floor; the list goes on and on and on. Most signal processing devices used in sound reinforcement come to us from the recording industry, in which artists and producers use signal processing to alter the sound of individual instruments, not to mention the overall mix. Theatrical sound designers generally use processing such as equalization and delay to properly align and "tune" their individual loudspeaker systems within a given acoustic space, and use other effects, such as reverberation and compression in more specific quantities on particular inputs.
ALL FREQUENCIES ARE CREATED EQUAL...
Equalizers are probably the most common example of a basic signal processor-- they are found on the input channel of most mixing desks and are also found as standalone units to modify the frequency response of a given input signal. Historically, equalizers were used by the telecommunications industry to make up for frequency-dependent signal loss when transmitting voice over long distances of cable. Special components were used to boost certain frequency bands that were attenuated over the long cables, thus making all frequencies equal in level (ampltitude)-- hence, "equalizer."
From electrical engineering, we recognize that a circuit component that changes a portion of the frequency spectrum is called a "filter." We also recognize that filters are designed to attenuate, or cut, certain frequencies. However, if a filter is set to cut most frequencies and allow certain frequencies to pass without being cut, the final result is similar to having those uncut frequencies boosted, especially when an amplifier component is added. In this fashion filters are used to boost frequencies. These days, it is common to find units that boost and cut specific frequency bands, although some equalizers, called passive equalizers, are designed only to attenuate frequency bands.
Equalizers are used extensively in sound recording and reinforcement. They are designed to be used to provide control of the frequency response of a specific portion of the sound system. Most mixing desk input channels provide at least a very basic equalization section, which may be as simple as a set of high and low filters (like your car stereo), or may be a four-band parametric equalizer which allows very precise control of the input signal. Equalizers are used further down the sound system chain as outboard processors which can affect the frequency response of groups of inputs, or even farther down the chain as system equalization designed to affect the frequency response of a given set of loudspeakers.
BASS-ICS AND TREBLE-ICS
The very basic form of equalizers, usually found as the "bass" and "treble" knobs on your car stereo, are simply two filter networks. The "bass" knob is a filter circuit centered around a particular frequency-- usually around 100 Hz, and has a given amount of attenuation and gain (+/- 10dB, usually). Turning up the "bass" knob will increase the amount of low-frequency information relative to the rest of the program material and will result in a richer sound, and in the extreme, boominess. When you turn down the bass knob, the amount of low-frequency information relative to the program material is reduced, resulting in a honky or thin sound. The tone controls are usually of the shelving type, which means that frequencies above the treble center frequency (or frequencies below the bass center frequency) will also be affected by the filter. If one were to plot a graph of a bass tone control turned fully counterclockwise (cut), the graph would show that the low-frequency response would slope downward such that at 100 Hz, there was 10dB of attenuation. Below 100 Hz, there would continue to be 10dB of attenuation. The point at which the frequency response starts to be affected by the equalizer circuit is called the hinge point. For instance, to provide for a 10dB reduction in amplitude at 100 Hz, the frequency response starts its downward slope at a higher frequency, say 500 Hz, continues to slope downward until it reaches 100 Hz, and then levels out again below 100 Hz. The treble control operates in exactly the same fashion, but mirrored. If the treble control is turned fully clockwise (boosted) and has a center frequency of 10 kHz, the frequency response will start being affected at its hinge point, probably somewhere around 3 kHz, and the amplitude will increase until it reaches 10 kHz, at which point there is 10dB of gain. Frequencies above 10 kHz will all have 10dB of gain.
The tone-control setup does not provide for much flexibility at all, what with a fixed center frequency and a shelving curve. What happens, for instance, when an offensive frequency falls between the two hinge points of the tone controls? The tone controls will not affect the band between the two hinge points, so what good is it? Mixing desk manufacturers thus provide somewhat more elaborate controls which allow the operator to choose the center frequency, the amount of boost/cut, and even more important quantity- the bandwidth.
In our tone-control example, frequencies above the "treble" center frequency are summarily affected in the same way- all frequencies above the 10 kHz point are boosted or cut. But this shelving arrangement isn't very practical, even if we're allowed to choose the center frequency. What happens if we want to center our equalizer at 500 Hz with a shelving equalizer? Either everything below 500 Hz or everything above 500 Hz will be similarly affected by our filters, which isn't a very helpful arrangement. Opposite to the shelving type of equalizer is the notch or peaking equalizer, which allows the operator to center around a particular frequency and leave frequencies outside of a given bandwidth largely unaffected-- thus, we can center on a 500 Hz frequency, apply attenuation or gain, and leave frequencies beyond, say, 600 Hz and below 400 Hz unaffected. If were to plot the frequency response of a 10dB boost at 500 Hz, we'd see that low frequencies remained at 0dB, started to slope upwards somewhere around 400 Hz, reach a 10dB boost at 500 Hz, then slope downwards around 600 Hz, leaving higher frequencies at 0dB. This is a very common type of equalizer; when multiple notch filters are combined into one unit, each affecting a different portion of the frequency spectrum, we call this a graphic equalizer.
A graphic equalizer is a multi-frequency bandpass/reject filter that functions as a standalone unit, usually used to equalize specific loudspeaker systems. While mixing desk input channel equalizers operate on three or four bands, a graphic equalizer can simultaneously operate on at least eight different frequency bands. The most common graphic equalizers in sound reinforcement align their center frequencies according to ISO standards which fall on one-third octave centers, and can be covered in thirty-one bands. There are more expensive and cumbersome one-sixth octave and one-twelfth octave devices, but these units are rarely used in sound reinforcement. It is generally accepted that one-third octave graphic equalizers are sufficient for most venue tuning and feedback reduction.
Further developments in equalizers provided a bandwidth control in addition to the boost/cut and center-frequency selections. Adjusting the bandwidth controls the range of frequencies affected by a given filter. In graphic equalizers, the bandwidth of each frequency band is preset by the manufacturer and cannot be changed. This configuration does not provide the flexibility that some designers and operators require-- in certain circumstances, one may wish to have a broad EQ curve, which begins gently and very gradually builds up to the maximum peak with regard to frequency. On the other hand, one may wish to have a very specific frequency cut or boosted without adversely affecting adjacent frequencies. The broadness or sharpness of the EQ curve is specified by a quantity called Q, or quality factor. Quality factor is the measure of the sharpness of a resonant peak, and many people often use Q interchangeably with bandwidth, which is not entirely correct. The quality factor of a filter can be said to actually determine the bandwidth of the filter; the higher the Q, the sharper the curve, and thus, the narrower the bandwidth.
Equalizers that provide control over Q, center frequency, and boost/cut, are known as parametric equalizers, so named because they allow control over all necessary parameters of equalization. Parametric equalizers are found on input sections of mixing desks and also as standalone units. Input channels can offer anywhere from one to three different bands of parametric equalization, which allows for precise manipulation of the input signal. Outboard units usually come in five-band versions; each band is set for a different frequency band, although there are overlaps. There are three potentiometers per band; one selects the frequency, the other selects the bandwidth, which is usually measured in octaves (.1 is a very tight bandwidth, and 1.1 is a wide bandwidth), and the last selects the amount of boost/cut produced by the filter.
Parametric equalizers are usually very daunting to the novice sound operator because they are mostly potentiometer-based. A one-third octave graphic equalizer has thirty-one center boost/cut sliders, which represent the thirty-one bands of one-third octave audible frequencies. They are comfortable to use because the frequency response of the equalized system can be quickly observed by looking at the resultant pattern of sliders. A variant of the parametric equalizer, called the paragraphic equalizer, is a fully parametric equalizer that uses a slider for boost/cut instead of a potentiometer. Big woop.
The debate over which type of equalizer to use rages on; rock-and-roll monitor engineers prefer graphic equalizers on their monitor mixes because of the instantaneous recognition of the equalization curve; theatrical sound designers prefer parametric equalizers on their multiple loudspeaker systems because of the flexibility of the unit. Some people prefer parametric equalizers because a minimal amount of equalization can be applied to correct a frequency response anomaly, and other people eschew parametric equalizers because so many parameters need to be adjusted, which makes it difficult to properly replicate the settings in a different venue.
TO EQ, OR NOT TO EQ
Any equalization creates phase shift in the audio signal, which can lead to distortion, reduced headroom, and comb filtering in multiple loudspeaker systems, if allowed to run rampant. Equalization should kept to a minimum, and errant frequency response should be corrected by proper loudspeaker positioning or acoustical treatment of the room. Most designers never use system equalizers for frequency boost; boosting particular frequency bands can result in reduced amplifier headroom, distortion, and more potential for feedback. Most designers prefer instead to attenuate only, and use equalizer gain only in very specific instances, such as at the input channel.
Equalization can reduce the resonant peaks and dips in loudspeaker systems as they interact in a given acoustic environment, which can be used to reduce the possibility of acoustic feedback. As the overall gain of the sound system is turned up, feedback occurs first at that frequency where the system of microphone + system + loudspeaker + room has a peak. It is heard as a slight ringing and can become a loud howl. If the offending resonant frequency appears due to the loudspeaker + room interaction, an equalizer set to attenuate that first peak can be used on that particular loudspeaker system in order to increase the available gain before feedback. Subsequent frequency peaks can be eliminated in the same way. Graphic equalizers provide a quick and dirty method of equalizing a system within a room, while parametric equalizers provide more control over the attenuation of frequencies.
Regardless of the type of equalizer used, they are a very powerful tool that should be used carefully. Inputs on a mixing desk may require some sort of equalization, be it radical frequency-specific tuning, or just a low-cut rollof filter to reduce boomy microphone noise. Most system designs require an equalizer on each signal feeding a loudspeaker system, which is usually installed just prior to the amplifiers or electronic crossover systems. In recording and broadcast situations, the equalization that is required is often radically different from that required in a reinforcement system; instead of equalizing frequency response peaks and dips in loudspeaker systems as they interact with the room and other loudspeaker systems, equalization may be needed to alter the frequency response to accommodate requirements for a particular recording medium. Use it wisely. It's a good thing. [Editor note: STOP QUOTING MARTHA STEWART!]
Reverberation is the natural phenomenon that occurs when sound pressure waves are reflected and reinforced and blended with the original, direct sound. All concert halls and theatres exhibit different reverberation characteristics. The size and shape of the venue coupled with any obstacles or sound-absorbing material all affect the way in which the sound waves interact. Reverberation Time is a measurement of reverberation characteristics-- abbreviated RT60, it is the time it takes for the reflected sound pressure levels to decay to one-millionth of its original value (after the original sound source has stopped). The ratio of one-millionth measured in decibels is interpreted as a 60 dB reduction; hence, RT60. Concert halls designed for classical music often have somewhere in the neighborhood of 2.50 seconds of natural reverberation time; churches may have as much as 5.00 seconds. "Dead" theatres may only have about 1.00 seconds of reverb time.
Reverb processors are electronic devices which emulate this effect. Originally designed for the recording industry to simulate different types of sound locations, they have also found a home on the reinforcement side to give more "fullness" to program material or to be utilized as special effects, giving the listener the illusion of distance or space. Reverberation is often confused with echo and delay, and while both of these effects comprise reverberation, they are not the same thing. Delay refers to one or more distinct sound images-- echoes. True reverberation usually begins with a few closely-spaced echoes called early reflections, but the final product is influenced by more and more reflections blended together, providing a more homogeneous sound field.
In the olden days, reverberation units were large and cumbersome machines in which a transducer was attached to a metal coil; the loudspeaker twisted the spring, and the sound traveled up and down the spring. A microphone or pickup attached to the other end of the spring converted these mechanical reflections back into an electrical signal. This design was called the spring-type reverb. A similar invention, called the plate reverb, used a large metal plate to reflect and modulate the sound energy. Digital technology has eliminated the need for such analog devices by processing the original sound using a variety of mathematical algorithms that examine frequency, phase, amplitude, and allow the engineer or designer to adjust parameters, such as reverberation time, equalization, and pre-delay in real-time. Reverb processors can be found on most if not all popular music recordings, often applied directly to the vocals in varying amounts. In theatrical sound reinforcement, some designers prefer using digital reverb on sung vocals, or globally over the entire orchestra in order to provide a fuller, more natural sound. The orchestra in many a theatre production is usually close-miced, which provides a very clean input signal but not a great deal of ambience; adding reverb in judicious amounts can help restore a "natural" quality to the amplified sound.
Most modern reverb processors can do more than just supply digital reverb-- many incorporate other digital effects such as chorus (which replicates the original sound and frequency-shifts the copies), or pitch-shifting, or distortion, or radical equalization, or distinct echoes-- the list goes on. Special effects can be operated and processed in real time using such devices, and, used judiciously, can provide a variety of acoustic environments by which the audience's perception can be altered.
Delay units do precisely that-- take an input signal and delay the output by a user-defined amount of time. In the old days, delay units were tape-based; an infinite loop of magnetic tape was fed through a recorder; the record head constantly updated the information on the tape, and a series of playback heads spaced away reproduced the recorded signal a given amount of time later. This process was inefficient and prone to extensive noise problems. Modern delay units use digital technology to convert the analog signal into digital data, which is then stored in RAM registers. An internal clock generates synchronization pulses that tell the RAM register to expel the data, which is reconverted back into analog audio. The recording industry uses delay units as special effects; in sound reinforcement, we use digital delay devices to delay the audio signal feeding a given loudspeaker system such that sound waves from two spaced loudspeaker systems will arrive at the same location at the same time-- but we'll cover that in another section.
Compressors and their older brothers, limiters, are signal processors that reduce the dynamic range, or loudness, of the audio signal. Limiters are in fact a kind of compressor; the difference in nomenclature refers to the difference in application of the unit. The compressor was first developed early in phonograph recording and the telecommunications industry, applications in which the audio signal had a greater dynamic content (difference between the loudest peaks and the softest lows) that the medium which was to distribute it. Most commonly using voltage-controlled-amplifier, or VCA, technology, it was possible to constantly monitor the input signal and set a threshold at which the compressor would reduce the level of the output signal by a specified ratio. The threshold of the compressor, a user-selectable value, dictates at what point, in decibels, the audio signal should be reduced. The ratio of the compressor dictates the amount of reduction of the audio signal. It is expressed as a ratio of decibels: a compressor with a ratio set to 1:1 will not compress at all; with a 4:1 ratio, for every four dB of gain over the threshold value, the output signal is allowed only one dB of gain increase. A variety of other controls are usually available to affect the way in which the compressor acts upon the audio signal: attack time and release time, for instance, set the speed at which the compressor reduces the output signal. Compressors are widely used in popular music recording, where they are used both on individual instruments ("tightening" drum tracks, for instance) or on the entire mix (most prerecorded popular music is compressed within an inch of its life). In theatrical sound design, compressors may be found inserted on individual orchestra inputs, but very rarely are they used on vocal material; designers shy away from the use of compressors as it defeats the purpose of a human operator and generates a non-linear audio signal. Live sound for popular music often use compressors in much the same way that a recording studio might, in order to "fatten" up vocal or drum tracks, and used in the extreme, to prevent the loudspeaker system from mechanical destruction due to a dropped microphone, et cetera.
Compressors with a ratio of infinity:1 do not allow program material to increase above the set threshold level, and these compressors are called limiters, or peak limiters. FM radio broadcasts, for instance, are usually put through a limiter before they are sent to the actual RF broadcast system so that any high peaks in program material are flattened out; these high peaks can cause distortion in RF broadcasts. Some active loudspeaker systems include in-built limiter circuits to prevent transient peaks from harming the loudspeaker drivers. All RF wireless microphones use some form of compression in order to squeeze the required dynamic range into a given radio-frequency bandwidth.
Most compressors also provide separate inputs, called a sidechain input. Whereas in normal mode, the compressor examines the input signal for levels above a given threshold and changes the output gain accordingly, when a compressor uses the sidechain inputs, the compressor examines this external input for levels above a given threshold and process a different audio signal entirely. Applications for the sidechain function are manifold: a ducker is a compressor commonly used in radio broadcasting and background music systems in which an audio signal, such as music, is fed through a compressor. A paging microphone or broadcast announcer microphone is split into two outputs; one output is mixed with the music program after the compressor circuitry, and the other output is fed to the compressor's sidechain. When the compressor circuitry senses program from the announce microphone, it reduces the gain of the music program, and thus provides for increased intelligibility of the announcement. A de-esser works in a similar fashion; the unit is designed to reduce the more sibilant vocal sounds, such as "s," "sh," or "ch" sounds. In essence, the de-esser is a compressor with an equalizer at its sidechain input. The vocal signal is split into two outputs-- one output feeds the input of the compressor, and one output feeds an equalizer which is set to boost any of the offending frequencies (usually in the 2kHz - 5kHz range). The output of the equalizer is connected to the sidechain input of the compressor. If the vocal program remains in the unboosted range (i.e. below 2kHz), the compressor is not activated, and the output signal remains unchanged. However, if the vocal program enters the range set by the equalizer, the equalizer outputs provide a higher output signal to the sidechain, which tells the compressor to activate, thus lowering the output gain.
Noise Gates and their older brothers, expanders, work on the opposite end of dynamic range. A noise gate is a dynamics processor that mutes or significantly lowers the audio signal output when the input signal level falls below a user-defined threshold. The original intent for the noise gate was to eliminate background hiss, noise, or leakage from contaminating the audio program when the primary signal was not present. However, since noise gates mute the signal when the input signal is below the threshold level, the audible result can be disturbing or annoying as the unit mutes and un-mutes the output signal. To alleviate this sudden, binary on-off switching noise, many noise gates simply reduce the output signal level by a finite amount (they lower the gain) when the input signal falls below the threshold. This process effectively reduces the noise floor, but does not produce a sudden and often audible change. Many units also provide adjustable time parameters which allow the user to set the amount of time it takes for the noise gate to activate and deactivate.
We call the electronic circuit that reduces the gain an expander because the unit lowers the noise floor of the program material, effectively providing us with expanded dynamic range, but nomenclature gets confusing when describing the actual piece of equipment. When the unit is set to work only below a set threshold and reduce or mute the output signal below this threshold, we call this device a noise gate. If the unit acts as a noise gate below a set threshold, and acts in the opposite fashion above the threshold (i.e. any input signal above the threshold is effectively made louder), we call this device an expander.
Noise gates are often used for automatic microphone mixing situations. Dan Dugan, a pioneer in the field of automatic mixers, developed a system in which microphones were automatically muted if no primary signal was detected, which reduced the amount of background noise in the system and also increased the available gain-before-feedback in a reinforcement system. In recording, noise gates are often found on individual drum inputs or electric guitars to provide a clean, undisturbed signal.
Expanders find their homes in most noise-reduction and broadcast systems. Upon recording, the noise reduction system compresses the audio signal in order to fit the original dynamic range into a smaller dynamic range for recording. On playback, the noise reduction system decompresses (expands) the compressed signal, which restores the original dynamic range of the recording and "pushes down" any inherent tape hiss or noise below the program noise floor. Radio-frequency wireless microphones also use compressors and expanders to provide a larger allowable dynamic range: upon conversion from audio to radio-frequency, the audio signal is compressed so that it will "fit" into the small bandwidth provided by the RF system; upon reconversion from RF to audio, the signal is expanded to restore original dynamic range and reduce RF-induced noise. Systems that integrate compressors and expanders for this purpose are often referred to as companders.
OTHER NEAT-O EFFECTS
Other effects units include phasers, flangers, and exciters. Phasers and flangers produce the same-sounding effect via different means. The final effect of a phaser or flanger is a hollow-sounding sound caused by phase differences between copies of the same input signal. Flangers originated from the use of two tape playback machines playing the same signal. By mixing the two outputs and alternately slowing down one machine, then the other, different phase cancellations occurred. The slowing down of the machines was acheived by applying slight hand pressure to the flanges of the tape supply reels hence the name. Modern flangers use electronics to simulate this effect. Phasers, or phase-shifters, use one or more deep, high-Q filters. The input signal is split two ways: one bypasses the filter, and one flows through the filter circuit. Because of the capacitative nature of all analog filters, phase shift occurs when a signal is passed through a filter on either side of the notch frequency. By varying the notch frequency through the audio frequency spectrum and mixing the resultant signal back with the original, a series of phase-cancellations results.
Exciters are units designed to add "punch" and/or "air" to the overall program material without changing system gain. The exciter circuits often use a set of filters and equalization coupled with harmonic manipulation to produce a "phatter," more intelligible sound. The Aphex corporation introduced the first exciter in 1975: the Aural Exciter, and many companies have followed suit. Exciters are sometimes used in the penultimate mixdown process of popular music and are often used in dance-club reinforcement systems. Only rarely are they used in theatrical sound design.
Be aware that any frequency-manipulating device will introduce phase-shift into the audio circuit, which may have detrimental results on the frequency response of loudspeaker systems in the venue! On the other hand, frequency-manipulating devices and other effects processors can aid the designer in creating an aural soundscape that transports listeners into another environment. Let your ears be the judge!
Briefly, one last use of digital signal processing devices is acoustic enhancement, or the artificial electronic treatment of the acoustical characteristics of the performance venue itself. It is akin to providing the natural reverberation qualities of a concert hall with copious amounts of Viagra or Qualuudes, depending on the eventual goal. The particulars vary, but the basic concept is to place microphones around the venue and add electronic reverberation to a system of loudspeakers in the venue to create the illusion that the hall exhibits different reverberation characteristics. I wrote an article for Entertainment Design magazine that goes into more detail. Since the copyright to the piece technically belongs to them, I will merely provide a link to the article on-line. Here it is: <http://industryclick.com/magazinearticle.asp?releaseid=5643&magazinearticleid=66853&siteid=15&magazineid=138>.
Continue to Loudspeakers. Return to the Sound Index.
Comments, Questions, and Additions should be addressed via e-mail to Kai Harada. Not responsible for typographical errors.
http://www.harada-sound.com/sound/handbook/process.html - © 1999 - 2002 Kai Harada. 19.05.2002.