Dynamic Range Processing and Digital Effects

Dynamic Range Compression

Compression is a reduction of the dynamic range of a signal, meaning that the ratio of the loudest to the softest
levels of a signal is reduced. This is accomplished using an amplifier with variable gain that can be controlled
by the amplitude of the input signal: when the input exceeds a preset threshold level, the gain is reduced. The
figure below is a device input/output graph that shows the basic adjustable parameters found on compressors:

1:1

Gain is the ratio of the output to the input, the slope of the line. For inputs above the threshold level, the
output level change for a given input level change is known as the compression ratio (the inverse of the above
threshold slope.) The knee is the point at which the gain changes. It can be changed all at once (“hard knee” as
in the figure above) or gradually with increasing input level (“soft knee”). Since this change is non-linear, it can
result in distortion that sounds slightly different when changed smoothly or instantaneously. The time required
for the gain to change after a signal crosses the threshold is referred to as the attack time and the time it takes
to returns to the original gain when the signal drops back below the threshold is the release time. Generally,
the attack is fast and the release is slow to prevent audible artifacts of the gain change, however too short an
attack time reduces the transient onset of a sound and noticeably alters the character of the sound: sometimes
this may be desirable, but often it is not. Overshoot is the result of the attack time setting, where the increased
input amplitude is not yet reduced by the circuit (see figure below.) This often retains the character of the sound
onset that identifies the type of instrument or source: a surprising amount of the distinctive character of an
instrument’s sound is contained in the initial transient, so allowing some time for the attack is often necessary.
This can be tens of milliseconds. Interestingly, the human auditory reflex displays attack and release times of
about 40 msec and 135 msec, respectively, and these values tend to sound natural when applied to electronic
compression.
You might think of the use of compression in terms of viewing the skyline of a big city: the taller buildings hide
the shorter ones behind them. This is analogous to how louder sounds mask softer ones in the same frequency
range. The effect of compression would be to shrink the taller buildings relative to the shorter ones so that
all the buildings attain more similar heights. You can imagine what this would look like: more of the shorter
buildings would be visible although some of them would still be blocked by others. Similarly, compressing
sounds makes them more equally prominent in the mix but there is still a limit to how much of the quiet sounds
we will be able to perceive.

There are two basic circuit topologies used in compressors - feed-forward and feedback. As the names imply,
the feed-forward devices use the input signal to generate the amplitude envelope signal while feedback circuits
measure the already-compressed output to derive the control signal. Operation of the feedback device is
simple: if the output is larger than the threshold it turns down the gain. That is all the circuit needs to do since
compression is already being applied to the signal being measured. In the case of feed-forward devices, you
need to know by how much the signal exceeds the threshold and turn down the gain by exactly the right amount.
This requires calibrated measurement and gain circuits to work together to produce the proper gain reduction.
The necessity of calibrating the measurement and gain circuits makes feed-forward devices inherently more
complicated than the feedback type. The topology to some extent influences the sound of the device, but so do
the type of gain control element used and the type of envelope tracking circuitry. In general, feedback designs
reduce the dynamic range demands on the control circuitry as the dynamic range of their input is already
compressed. The attack and release times of feedback compressors are affected by the compression ratio while
feed-forward compressors time behavior is determined mainly by the speed of the level sensing circuitry. Some
designs are program dependent, meaning their characteristics depend on the signal. Peak measuring circuits
provide a fast measure of the instantaneous signal amplitude, making sure the signal maximum is always
known. RMS and average circuits may more closely approximate how loud we would perceive a signal to
be, but may not track short peaks that could overload the circuit. Examples of feedback compressors include
LA-2A and 1176 models while dbx compressors are feed-forward types. Gain control elements include FETs
(1176), optical (LA-2A, LA-3A), VCA (dbx 160, API 2500), variable-mu vacuum tube (Manley Variable Mu,
Fairchild 670) and pulse-width-modulation [PWM] (Pye).

Compression may be employed at any stage in the recording process, although it can sound quite different
depending on where in the chain it is used. It may be desirable to compress signals before they are recorded,
especially if the recorder is noisy, so the reproduced signal will not also contain the compressed added noise
that would appear if the signal were compressed after recording. Unfortunately, a poor adjustment of the
compressor can cause undesirable changes to the signal that are then recorded, so compression must be used
carefully, if applied before a signal is recorded, to avoid an overly “squashed” sound. This usually means using
a threshold setting that results in only a few dB gain reduction on the signal peaks. It may also mean using a
low ratio; something like 2:1 to 4:1, for example. If the ratio is set to 10:1 or greater, the effect is considered
limiting because the output is effectively limited even if the input continues to increase. Limiting causes less
alteration of the loudness relationship balance between elements of a sound than does compression because it
leaves the signal unaltered until it hits a threshold higher than where a compressor’s threshold would be set to
achieve the same reduction of dynamic range. Limiters are often used on singers as they are being recorded to
prevent recorder overloads from vocal plosives.

The difference between a limiter and a compressor, while one of degree, is clearly audible. Compression alters
the loudness relationship between soft and loud elements of a signal more dramatically than does a limiter. By
setting a low threshold, a compressor is active on more of the dynamic range of the signal: it reduces the gain
over a relatively wide range of amplitudes. The limiter reduces the high amplitude elements of the signal, but
leaves the soft and mid-amplitude elements alone. This makes the limited signal sound more like the original
signal because only the peaks are reduced while the unaltered remainder of the signal is made linearly louder.

When a large amount of compression is desired, the limitations of the circuitry may be revealed. Because the
compressor relies on the envelope of the signal to create a control signal, the time response of the compressor
depends on the rate of change of the signal as well as on the time constant of the analysis circuitry. Many
compressors allow adjustment of attack and release times. Other circuits (like the dbx 166) use the signal itself
to determine the time response of the compression. In either case, a signal which has very large transients
may cause overloading of the envelope measurement circuitry and create an audible pumping effect as the
circuit recovers from the overload more slowly than the signal itself recovers. A way to avoid this is to use
small amounts of compression at two or more stages in the recording process. This will make the signal very
prominent without the overload problems because neither compressor is overdriven. Chaining dynamics
processors is also a common technique. Sometimes, a limiter with a high threshold is used to reduce the big
peaks in a signal followed by a compressor that reduces gain further without the overloading that the now-
removed peaks might cause. Multiple compressors can also be used - each contributes a little compression.
These techniques can produce a more natural compressed sound than a single processor alone.

Some compressors have a side-chain input. This allows a separate signal to control the gain of the compressed
signal. For example, a vocal might be used to reduce the gain of a rhythm guitar so that the vocal always
cuts through the mix. This technique is known as “ducking” and can be heard in commercials where the
announcer’s voice lowers the volume of the backing music. The effect allows for a sort of “automated mixing”
in certain circumstances. It depends on finding optimal attack and release times as well as appropriate threshold
and ratio to make the gain change sound natural, but sometimes this approach avoids the need for automation.
Side chain inputs, commonly called “keying” inputs, can be used to allow selected signals to control the gain of
several compressed signals. This technique is popular in electronic dance music (EDM), where the kick drum
modulates much of the mix.

Many compressors allow filtering of the control signal to produce special frequency-related compression effects.
The most popular of these effects is de-essing, which tends to reduce the sibilance from compressed speech or
singing. By increasing the side-chain (control) signal gain in the frequency range of sibilance, between 2 and 8
kHz, for example, the s’es will be more compressed than the remainder of the vocal sound, resulting in a more
natural sound. By decreasing the low frequency gain of the side-chain signal, the “pumping” effect of a bass
drum on snare compression may be reduced.

It should be mentioned that compression may cause a signal to sound quite “weird” if heard in isolation, while
in a mix it may still sound “correct”. One must make final adjustments within the entire mix in order to get
a complete picture of the overall sound. The masking phenomenon accounts for why this is so: parts of the
compressed sounds interact with the other sounds in the mix, resulting in some masking of the compressed track
itself. We do not hear this if we listen only to the compressed track in isolation. There is also the question of
compressing individual elements of a mix or compressing the entire mix together. Both approached have their
merits, but the end result will sound different and often both are applied. Compressing individual tracks makes
them more “focused” in the mix, while compressing the whole mix tends to bring the elements together.

Compression is one of the more powerful techniques for manipulating sounds in a mixing environment and it
is important to become familiar with its uses and limitations. Experimentation is the best way to appreciate
what we can expect from dynamic range manipulation and what we cannot realistically achieve. In popular
music, a lot of what separates professional-sounding mixes from more amateur-sounding ones is the effective
use of compression and the use of recording techniques that are compatible with the compression that is to be
used in mixing. There are a lot of compressors available, both in hardware and software, and each type has a
distinctive sound. Solid-state devices like the Universal Audio 1176 compressor/limiter are good for aggressive
drum compression while vacuum tube, optical gain cell devices like the LA-2A are favored for vocals.
Although certain compressors are favored for a particular application, there are no absolutes as far a selecting an
appropriate device.

Compressor plug-ins often try to emulate the behavior of the well-known analog compressors. As these
approximations get better, they are growing in popularity, not unreasonably since you buy one plug-in and may
then use several instances of it - something not possible with expensive hardware. Another advantage of digital
compression programs is the ability to “look-ahead” at the signal to anticipate rapid changes in amplitude.
Since there is inherent delay in all software effects, the mixing system must compensate for the delays so that
all tracks in a mix occur at the same time at the output. While the entire output is slightly delayed, the individual
elements remain in synchrony. This delay means that there is time to process a signal based on its future
behavior, something not possible in the analog domain.

So how do we choose the right compressor when we have so many choices? When studios used hardware units,
the decision was often made for us by the limited availability of devices. With plug-ins, there are so many
available the choice can be baffling. The best long-term solution is to try the various options and decide which
you like best for the range of signals you wish to mix. There is a lot of preference information on the internet,
but there is no short cut from experimenting with the various possibilities to decide which implementations
work best for the specific type of music you are working with at the moment. There are plenty of recording-
related web sites that offer personal recommendations and these can be a clue about which software
compressors might be worth consideration but they is no guarantee you will share the same opinions. As you
try various plug-ins, you will soon find those that work best for you. You will also find that it is a moving target
as new versions become available and offer different ways of treating the same tracks. The ability to adapt is an
asset as technology moves rapidly and we do not wish to be trapped by options that disappear overtime as newer
programs change what is available to us. It is the basic process we want to learn, not one particular method of
accomplishing the task at hand.

Dynamic range processing gives the mix engineer great power to sculpt the individual sounds and the final
combination as well. Learning how to maximize the effectiveness of these techniques requires experience and
patience. The reward will be worth the effort.

Dynamic Range Expansion

Expansion is the complementary process to compression: it actually increases the dynamic range of a signal.
When a signal falls below a threshold level, the gain is decreased. This causes quiet sounds to become even
quieter, or inaudible, in the case of the noise gate. The most common form of expansion is the noise gate, which
simply switches off signals that fall below the threshold. It is used to eliminate low-level noises from signals.
The noise can be generated by the performance (as in breathing, rustling, etc. from human performers or noise
from electronic sources like synthesizers or guitar amplifiers and pedals) or by imperfect recording techniques
(tape noise or 60 Hz noise). Expanders also have attack and release controls, but here a fast attack opens up
the gate immediately after a sound crosses threshold and release time reduces the gain of the sound some time
after it drops below threshold. While a gate is simply on or off, expanders allow you to adjust the time course
of the gain reduction and adjust how low the gain becomes while the signal is below threshold (it needn’t be
completely off).

While noise gates may be used at any stage of the recording chain, it can be risky to gate input signals. This is
because misadjusted thresholds can cause the signal to be gated off when it should be allowed to pass through,
clipping the beginning of phrases or sounds. Unlike compression, if expansion is applied after recording, it
will eliminate some of the noise added in the recording process rather than increase it. Expansion can also
be used to “tighten up” a signal by automatically shortening the decay time of a sound. Also, sends to reverb
processors may be gated to create a special gated reverb effect or to eliminate low level parts of the signal from
being processed. This can be helpful when applying heavy effects to a snare drum while there is significant
bleed from the high-hat, for example. Expanders with selectable expansion ratios may be used to alter the
dynamic range somewhat like a compressor, only in the opposite direction. You can set the gain for signals
that drop below the threshold from none (as in a gate), to a ratio that only slightly reduces the gain of low-
level sounds. When combined with a compressor or limiter, an expander gives an engineer the ability to finely
sculpt the dynamics of sounds to better fit into a mix. Only through experimentation will you find the optimal
combination of dynamics processing for a specific job, and this type of dynamics processing is not always
necessary or appropriate.

Creative Uses of Dynamic Range Processing

In addition to allowing the engineer to make relatively natural-sounding changes to the dynamics of a program,
these effects may also be used for special effects that are not necessarily natural. For instance, by feeding the
control signal of a compressor from a different signal, one sound may be used to modulate the amplitude of
another, a process known as “ducking”. Since the compressor gain is controlled by the level of the second
sound, the compressed signal’s gain is now controlled by that other signal instead of its own level as in normal
compression. While this effect is commonly used to allow an obnoxious announcer to hawk a product while
the cheesy music’s amplitude is automatically turned down as the words are spoken, it can be used to allow a
vocalist to ride just above the backing instrumental tracks without having to carefully ride the gain of the music
tracks. Using a similar technique with the gate allows one signal to gate another, an effect known as “keying”:
in fact, the control signal external input is often called the key input. This technique can tighten up a group of
background singers, for instance.

Another interesting use of the control signal in a compressor is to insert an equalizer in series with the control
signal input. This allows frequency-selective compression, since the equalizer can alter the frequency response
of the control signal so as to allow different frequencies to cause more or less compression that they would
otherwise. As previously mentioned, this can be used for de-essing and to prevent some frequencies from
causing compression while other frequencies are compressed normally.

And inserting delay devices and distortion pedals into the control circuit can result in some “interesting” effects,
although not generally as useful as the above techniques.

Digital Delay Effects

The advent of digital audio has, in addition to revolutionizing signal flow and recording processes, allowed for
the development of sophisticated delay-based audio effects. These include reverberation, echo, chorus, and
flange effects, all of which depend on signal delay. Before the availability of digital circuits to create these
effects, they could only be synthesized using tape recorder delay and cumbersome mechanical devices like
large tensioned steel plates with transducers to excite and capture the mechanical vibrations. Of course, the
idea of using a real room as a reverberator, via speaker and microphone, is still of interest, but few recordists
have that luxury. Even without a detailed understanding of digital signal processing theory, it is possible to use
these devices, since their operation can be understood simply by examining the effect of combining direct and
delayed signals.

Digital delay

The simplest of these devices is the digital delay or delay line. The time delay can be accurately determined and
synchronized to the tempo of the music. The device samples the input signal and replays a delayed copy of the
signal in combination with the direct (undelayed) signal. This provides a single discrete echo. If the delayed
output is fed back to the input, the familiar series of decaying repeats is produced. The amplitude of the signal
returned to the input determines how many repeats are heard. Digital delay is often used on vocals in a mix to
give a larger vocal image. It is also commonly used on solo instruments like saxophone or guitar. The delay
time can be adjusted to the tempo of the music, so that a discrete echo falls on an eighth note, for example. This
results in an echo that is not obtrusive and produces a sense of space without being obvious. To determine the
delay time (in milliseconds) for an eighth note, divide 30,000 by the tempo (in beats[quarter notes]/minute).
Delay times of quarter notes, sixteenth notes, and eight-note triplets may sometimes be used, depending on
the program material. Stereo delay can create a simulated stereo image by delaying the signal to the left and
right channels by different times. Delay times in the range of 20-50 milliseconds work well for this on rhythm
instruments, but longer times can be used as special effects on solos. It is recommended to check the effect in
mono as it is possible to create some odd-sounding results when the multiple delays are mixed in the system
rather than in the air as they are generally combined in the air between two loudspeakers.
 Flanging

As we remember from multiple microphone technique, when time-delayed and direct signals are combined,
a comb-filter effect is created. Generally, this is undesirable; however, the effect can be used to “spice-up”
certain sounds. If the delay time is constantly altered slightly, a rich sweeping filter is created. This is known
as flanging. The name derives from the original way of creating the effect: using a second tape recorder to
delay the sound and slightly slowing the machines by placing hands on the reel flanges. Now, digital delay
devices allow a slow oscillator to control the delay time. In addition, the depth of the effect can be controlled
by changing the balance of the delayed and direct signals. For flanging, the delay time is in the range of .5 to
35 milliseconds and the modulation rate (which changes the delay time) is in the range of 1 to 10 Hz. A stereo
effect can be created by placing the direct + delayed signal in one channel and direct - delayed signal in the
other.

Chorus

The chorus effect is qualitatively similar to the flange effect; it attempts to simulate the effect of several distinct
sound sources producing nearly the same sound, like a choir does with multiple singers in unison. Unlike
flanging, chorusing often employs amplitude modulation as well to simulate the way singers’ volumes vary in
time. Electronically, it is achieved using small random variations of the time delays and amplitudes and uses
several separate such channels which are recombined in stereo to produce a very rich sound. Modulation rates
are longer than for flanging, typically 0.1-0.5 Hz, with similar delay times of 1 to 50 milliseconds.

Reverberation

By far the most complicated of the digital effects is simulated reverberation, which attempts to create artificially
the complicated sound field created by sounds reflecting off the walls, ceiling, and floor of a room. Sound,
traveling at about 1100 feet/second, bounces off surfaces and slowly decays, producing what we recognize
as the sound of a room. Each bounce alters the spectral content of the sound, as frequencies are absorbed or
reflected depending on the physical nature of the reflective surfaces. Just after the onset of a sound, discrete
echoes, known as early reflections, are audible. Soon, however, these reflections build into a less discrete, but
denser sound. The complexity of this process has made reverberation programs extremely complicated, and
early attempts at synthetic reverberation were obviously poor imitations of the real sound. Newer devices
have improved dramatically, to the point where it is difficult to tell whether real or synthetic reverb has been
employed in a recording. Digital reverbs allow the user to choose between different programs, each of which
seeks to duplicate the behavior of one type of sonic environment, such as room, hall, auditorium, etc. Within a
program, adjustment of decay time, early reflections, pre-echo time, reverberation equalization and many other
parameters may be used to tailor the sound, however the basic program still simulates the same physical space.
While the spatial size may be adjustable, the basic character of the sound remains similar within a program.
The time required for the reverberations to fall to -60 dB is known as the RT60 and is often referred to as the
reverberation time.

An alternative to simulating the physics of a reverberant space is sampling the impulse response of real spaces
and using a digital process called convolution to simulate the sound. Convolution mathematically combines
the input signal with the sound of the space captured by the impulse response to produce the reverberation
that would have occurred in the actual space. Some of the character can often be adjusted in the process.
Convolution reverbs are increasingly popular.

In addition to digital reverberators, there exist mechanical reverberation systems, using springs, plates, and
chambers to create reverb effects. These systems were popular before digital systems became available and
are frequently simulated using digital techniques. The simplest systems use springs connecting transducers to
create the boingy reverb still used in many guitar amplifiers. More sophisticated plate reverbs use tensioned
steel plates (several feet in length) with transducers carefully placed to drive the plate and convert the vibrations
back into electronic signals. They are very heavy, bulky, and cannot be located where there are physical
vibrations or extraneous sounds. They often sound good on vocals and drums, so they are also simulated
on many digital reverberation systems. Chamber reverbs used actual chambers to create reverberations, via
speaker and microphone. These have a characteristic sound that is sometimes desirable and can be selected on
many digital reverbs.

Precisely which controls are present on reverb processors vary according to the complexity of the device,
often in proportion to the price. Many low priced machines allow for simple programming of reverb time,
equalization, and pre-delay only, while more expensive machines allow many, more complex parameters to be
adjusted. It is important to understand thoroughly each specific device in order to maximize the performance
of the reverberation algorithms employed. Reverberation can be used to simulate realistic sounds, as in the
case of enhancing live stereo recordings, or it can be used as a special effect in multi-track recordings and on
synthesized sounds to impart a sense of realism. Each of these applications requires a bit of experimentation to
obtain the best possible sound.

Making choices with effects

There is a tendency to think that if an effect sounds good it should be used everywhere. The truth is that sparing
use of effects is often more noticeable and preferable to their overuse. Dynamics processing is widely used
but the intent is often to make things sound natural rather than obviously enhanced. Particularly with the wide
availability of plug-in effects, it’s tempting to over apply them. How we listen to our mixes also affects how
much of an effect we choose to use - headphones enhance our perception of reverberation, for example, leading
to less reverb in the mix than we would choose if listening on loudspeakers. Deciding on which effects to use
and how much is an artistic decision that often changes as we become more adept at mixing and learning from
our past choices (gated snare reverb, anyone?) It is an art we can continue to refine though out our lives.

© 2006-2018 Jay Kadis (For educational use)