Fade (audio engineering)

Possibly the earliest example of a fade-out ending can be heard in Joseph Haydn's Symphony No. 45, nicknamed the "Farewell" Symphony on account of the fade-out ending. The symphony which was written in 1772 used this device as a way of courteously asking Haydn's patron Prince Nikolaus Esterházy, to whom the symphony was dedicated, to allow the musicians to return home after a longer than expected stay. This was expressed by the players extinguishing their stand candles and leaving the stage one by one during the final adagio movement of the symphony, leaving only two muted violins playing. Esterházy appears to have understood the message, allowing the musicians to leave.[8]

Gustav Holst's "Neptune, the mystic", part of the orchestral suite The Planets written between 1914 and 1916, is another early example of music having a fade-out ending during performance.[9] Holst stipulates that the women's choruses are "to be placed in an adjoining room, the door of which is to be left open until the last bar of the piece, when it is to be slowly and silently closed", and that the final bar (scored for choruses alone) is "to be repeated until the sound is lost in the distance".[10] Although commonplace today, the effect bewitched audiences in the era before widespread recorded sound—after the initial 1918 run-through, Holst's daughter Imogen (in addition to watching the charwomen dancing in the aisles during "Jupiter") remarked that the ending was "unforgettable, with its hidden chorus of women's voices growing fainter and fainter ... until the imagination knew no difference between sound and silence".[11]

The technique of ending a spoken or musical recording by fading out the sound goes back to the earliest days of recording. In the era of mechanical (pre-electrical) recording, this could only be achieved by either moving the sound source away from the recording horn, or by gradually reducing the volume at which the performer/s were singing, playing or speaking. With the advent of electrical recording, smooth and controllable fadeout effects could be easily achieved by simply reducing the input volume from the microphones using the fader on the mixing desk. The first experimental study on the effect of a fade-out showed that a version of a musical piece with fade-out in comparison to the same piece with a cold end prolonged the perceived duration by 2.4 seconds. This is called the “Pulse Continuity Phenomenon" and was measured by a tapping-along task to measure participants’ perception of pulsation.[12]

An 1894 78 rpm record called "The Spirit of '76" features a narrated musical vignette with martial fife-and-drum that gets louder as it 'nears' the listener, and quieter as it 'moves away'. There are early examples that appear to bear no obvious relationship to movement. One is "Barkin' Dog" (1919) by the Ted Lewis Jazz Band. Another contender is "America" (1918), a patriotic piece by the chorus of evangelist Billy Sunday. By the early 1930s longer songs were being put on both sides of records, with the piece fading out at the end of Side One and fading back in at the beginning of Side Two. Records at the time held only about two to five minutes of music per side. The segue allowed for longer songs (such as Count Basie's "Miss Thing"), symphonies and live concert recordings.

However, shorter songs continued to use the fade-out for unclear reasons—for example, Fred Astaire's movie theme "Flying Down to Rio" (1933). Even using fade-out as a segue device does not seem obvious, though we certainly take it for granted today. It is possible that movies were an influence here. Fade-ins and fade-outs are often used as cinematic devices that begin and end scenes; film language that developed at the same time as these early recordings. The term 'fade-out' itself is of cinematic origin, appearing in print around 1918. And jazz, a favorite of early records, was a popular subject of early movies too.[13] The same could be said for radio productions. Within a single programme of a radio production, many different types of fade can be applied. When mixing from speech to music, there are a few ways that fade can be used. Here are three examples.

• Straight: the introduction has become a musical link between the music/speech that follows, additionally the first notes of the intro can be emphasized to make it ‘pop’ out more.[4]
• Cutting the introduction: Since the first word of the vocals has to follow promptly after the cue light, it could be used to move the recording onward.[4]
• Introduction under speech: The music is placed at the specified time on the cue, the level must be low in order for the vocals to be audible.[4] Here the fade-up generally occurs just before the final words in order for the cue to be given.[4] In stage productions the closing music is played from a predetermined time and fades up at the closing words in order to fit in exactly with the remaining program time.[4]

No modern recording can be reliably identified as "the first" to use the technique. In 2003, on the (now-defunct) website Stupid Question, John Ruch listed the following recordings as possible contenders:

Bill Haley's cover version of "Rocket 88" (1951) fades out to indicate the titular car driving away. There are claims that The Beatles' "Eight Days a Week" (recorded 1964) was the first song to use the reverse effect—a fade-in. In fact, The Supremes had used this effect on their single "Come See About Me", issued a little over a month before "Eight Days a Week".

More recently: "At the meta-song level, the prevalence of pre-taped sequences (for shops, pubs, parties, concert intervals, aircraft headsets) emphasizes the importance of flow. The effect on radio pop programme form [is] a stress on continuity achieved through the use of fades, voice-over links, twin-turntable mixing and connecting jingles."[14]

The shape of a regular fade and a crossfade can be shaped by an audio engineer.[2] Shape implies that you can change the rate at which the level change occurs over the length of the fade.[2] Different types of preset fades shapes include linear, logarithmic, exponential and S-curve.[2]

The simplest of fade curves is the linear curve and it is normally the default fade. It takes a straight line and introduces a curve.[2] This curve represents an equal degree by which the gain increases or decreases during the length of the fade.[2] A linear fade-in curve makes it sound as though the volume increases sharply at the beginning, and more gradually towards the end. The same principle applies on a fade-out where a gradual drop in volume can be perceived in the beginning, and the fade gets more abrupt towards the end. If in your audio there is a natural ambience or reverb present that one would like to reduce, the linear shape would be ideal because of the initial drop in perceived volume.[2] When applied it shortens the ambience. Also if the music requires an accelerating effect, this linear curve can also be applied.[2] This type of fade is not very natural sounding.[2] The principle of a linear crossfade is: at the beginning of the fade the perceived volume drops more quickly, one can see at the halfway point (in the middle of the crossfade) that the perceived volume drops below 50%.[2] This is a very noticeable drop in volume. Also if the control can move from position 0 to 100, and the percentage of the signal that is allowed to pass equals the position of the control (i.e. 25% of the signal is allowed to pass when the control is 25% of the physical distance from the 0 point to the 100 point). At the midpoint of the fade the effect of a linear crossfade is that both the sounds are below half of their maximum perceived volume; and as a result the sum of the two fades will be below the maximum level of either.[2] This is not applicable when the two sounds are on different levels and the crossfade time is long enough.[2] In turn if the crossfade is short (for example on a single note) the dip of the volume in the middle of the crossfade can be quite noticeable.[2]

Another type of curve is called the logarithmic ratio (also known as "audio taper"),[17] or an inverse-logarithmic ratio.[18] The log/audio-taper more closely matches human hearing, with finer control at lower levels, increasing dramatically past the 50% point. Importantly, the perceived volume of a sound has a logarithmic relationship with its level in decibels.[2] A fade that works on a logarithmic scale would thus counteract the curve of a linear fade.[2] In fade-ins the shape of the perceived volume curve looks like the level in decibels pulled towards the middle part of the line to the bottom right corner.[2] The fade-out in turn looks like it has been pulled toward the bottom-left corner.[2] A logarithmic fade takes a line that has already been curved and straightens it out [15]. The logarithmic fade sounds consistent and smooth since the perceived volume is increased over the whole duration of the fade.[2] This makes this curve very handy for fading standard pieces of music. It is best used on a long fade-out since the fade has a perceived linear nature.[2] Also a fade-out sounds very neutral when incorporated to parts of music with natural ambience.[2] In crossfades this type of curve sounds very natural. When this curve is applied the perceived volume of the fade's midpoint is at about 50% of the maximum – when the two sections are summed the output volume is fairly constant.[2]

The exponential curve shape is in many ways the precise opposite of the logarithmic curves.[2] The fade-in works as follows: it increases in volume slowly and then it shoots up very quickly at the end of the fade.[2] The fade-out drops very quickly (from the maximum volume) and then declines slowly again over the duration of the fade.[2] Simply stated a linear fade could thus be seen as an exaggerated version of an exponential fade in terms of the apparent volume.[2] Thus the impression that would be gathered from an exponential curve's fade would sound as though the sound was rapidly accelerating towards the listener.[2] Natural ambience can also be repressed by using an exponential fade-out.[2] A crossfade, in the exponential shape, will have a perceivable dip in the middle, which is very undesirable in music and vocals.[2] This depends largely on the length of the crossfade, a long crossfade on ambient sounds can sound perfectly satisfactory (the dip can add a little breath to the music).[2] Exponential crossfades (or a curve with a similar shape) have a smaller drop in the middle of the fade.

The S-curve shape is interesting since it has qualities that correlate with the previously mentioned curves.[2] The level of the sound is 50% at the midpoint, but before and after the midpoint the shape is not linear. There are also two types of S-curves. Traditional S-curve fade-in has attributes of the exponential curve can be seen at the beginning; at the midpoint to the end it is more logarithmic in nature.[2] A traditional S-curve fade-out: is logarithmic from the beginning up to the midpoint, then its attributes are based on the exponential curve from the midpoint to the end.[2] This is true for the situation in reverse as well (for both fade-in and fade-out).[2] Crossfading S-curves works as follows; it diminishes the amount of time that both sounds are playing simultaneously.[2] This ensures that the edits sound like a direct cut when the two edits meet - this adds an extra smoothness to the edited regions.[2]

The second type of S-curve is more apt for longer crossfades, since they are smooth and have the ability to have both of the crossfades in the overall level; so that they are audible for as long as possible.[2] There is a short period at the start of each of the crossfades where the outgoing sound drops toward 50% quickly (with the incoming sound rising just as fast to 50%).[2] This acceleration of sound slows and both sounds will appear as if they are at the same level for most of the crossfade (in the middle) before the changeover happens.[2] DAW's gives one the ability to change the shape of logarithmic, exponential, and S-curve fades and crossfades. Changing the shape of a logarithmic fade will change how soon the sound will rise above 50%, and then how long it takes for the end of the fade-out to drop below 50% once again.[2] With exponential fades the shape change will affect the shape in reverse, to the shape of the logarithmic fade.[2] In the S-curve's traditional form the shape determines how quickly the change can occur and in the type 2 curve the change can determine the time it takes for both the sounds to get to a nearly equal level.[2]

It is also possible to apply different fade times to the out and in portions; which a standard crossfade would not allow you to apply.[2] Appropriate fade-in time for a linear fade can be around 500ms; for the fade-out 500ms would also be affective.[2] By having this longer fade it makes sure that everything is gentle[2] as it gives the fade time to blend in and be less abrupt. To clear up plosive sounds created through vocals a fade-in can be used, but now it has to have a very short time of around 10ms.[2] The fade time can always be adjusted by the engineer in order to locate the best time. It is important that the fade does not change the intelligibility or character of the sound too much.[2] When the crossfade is longer than 10ms the standard linear fades are not always the best choice for music editing.[2]

When crossfading two signals that are being combined (mixed), the two fade curves can employ any of the shapes listed above (see #Shapes), such as linear, exponential, S-curve, etc. When the goal is to have the perceived loudness of the combined mix signal stay fairly constant across the full range of the mix, special shapes must be used, called "equal power" (or "constant power") shapes. Equal power shapes are based on audio power principles, particularly the fact that the power of an audio signal is proportional to the square of the amplitude. Many equal power shapes have the property that the mid point of the mix provides an amplitude multiplier of 0.707 (square root of one half) for both signals. A variety of equal power shapes are available, and the optimal shape will generally depend on the amount of correlation between the two signals. An example pair of curves that keep power equal across the mix are ${\displaystyle {\sqrt {mix}}}$ and ${\displaystyle {\sqrt {1-mix}}}$ (where mix ranges from 0 to 1).[24][25][26]

Equal power shapes typically have the sum of their curves (in the middle of the mix range) exceeding the nominal maximum amplitude (1.0), which may produce clipping in some contexts. If that is a concern, then "equal gain" (or "constant gain") shapes should be used (which may be linear or curved) that are designed so the two curves always sum to 1.

In the digital signal processing realm, the term "power curve" is often used to designate crossfade shapes, particularly for equal power shapes.

Digital faders are also referred to as virtual faders, since they can be viewed on the screen of a digital audio workstation. Modern high end digital mixers often feature "flying faders", faders with piezo-electric actuators attached; such faders can be multi-use and will jump to the correct position for a selected function or saved setting. Flying faders can be automated, so that when a timecode is presented to the equipment, the fader will move according to a previously performed path. Also called an automated fader, as it recalls the movement of the channel faders in time.[27] A full-function automation system will continuously scan the console, many times per second, in order to incorporate new settings.[27] While this scan is in progress, the stored representation of the previous scan will be compared to that of the fader's current position.[27] If the fader's position has changed, the new position will be identified, thus resulting in a spurt of data.[27]

The console's computer will update the console's controls on playback.[27] This will be done from memory at the same speed.[27] The advantage of working with mix automation is that only one engineer can perform the job with minimal effort;[27] it can be set up or recorded beforehand to make it even simpler.[16] An example of this is when Ken Hamman installed linear faders that made it possible for him to alter several channels with one hand while mixing, thus he assumed an interactive role in the process of recording.[28] This type of fader level adjustment is also called ‘riding’ the fader.[16]

Many DJ equipment manufacturers offer different mixers for different purposes, with different fader styles, e.g., "scratching", beatmixing, and cut mixing. High-priced mixers often have crossfade curve switches allowing the DJ to select the type of crossfade necessary. Experienced DJs are also able to crossfade between tracks using the channel faders.