Since the dawn of recording, engineers and musicians have been manipulating sound using various mechanical and electrical means. Analog was the only show in town for the first 50 years of electrical-mechanical recording. Back then, all the gear - filters, EQs, compressors, limiters, reverbs, tape machines, etc. - was analog.
Recordings made in analog’s hayday (1950s through the 1960s) attest to the sonic excellence of the industry’s equipment. Analog circuits made high fidelity and stereo possible and a wide variety of EQs, compressors and limiters were widely available in studio rack mount format. Eventide’s Omnipressor® (1974), which introduced the concept of a ‘side chain’, is an early example of analog gear specifically designed to alter the dynamics of an audio signal in ways beyond what had been possible with traditional compressors.
But analog has a number of fundamental limitations. For example while an analog compressor can be adjusted to reinforce the attack of an audio signal, any modification of the transient will, to a lesser or greater degree, change the tonality of the overall signal.
Audio is now digital. The transition from vinyl to cds, from tape to memory, was a gradual one which started with the primitive DDL - the Digital Delay Line - which did dangerously close to nothing. In fact, in 1972, Eventide’s Model 1745 DDL did nothing more than delay audio just enough to allow studios to retire the tape machines used for “tape delay.” The rest is history.
Continued advances in signal processing and semiconductor technology over the last 45 years has made it possible to manipulate sound digitally with ever greater precision and in ways not possible in the analog domain. With each advance, newer and more powerful methods become practical.
Thousands of plug-ins are available to today’s audio engineer. Some plug-ins are digital ‘emulations’ of rackmount analog gear which attempt, with varying degrees of success, to model classic analog EQs, compressors, limiters and transient shapers. Others, like reverb and pitch change, have broken new ground by utilizing methods not possible in the analog domain. While the large majority of plug-ins could be considered variations on a theme, every now and then something truly ‘different’ comes along making it possible to take the next step in audio processing.
The Structure of Sound
Eventide’s patent-pending technology, Structural Effects, does something new. It separates a sound into its transient (impact/unpredictable/jagged) and tonal (sustaining/stable/smooth) parts more effectively and accurately than previous methods. Take, for example, a snare drum. We perceive the smack of the stick hitting the drum head and the resulting snare sizzle as the “transient” part of the sound. The ringing of of the body of the snare is perceived as “tonal.” Our new method employs a technique which can be broadly classified as morphological component analysis to cleanly separate these "sonic shapes."
Structural Effects technology is different because it works in a fundamentally new and different way. It’s not an EQ or compressor or limiter or transient shaper, per se. It's a a method for precisely deconstructing sound by using what can be described as a sonic ‘sieve’ in the sense that the audio is ‘filtered’ not by frequency but rather the audio is ‘filtered’ based on ‘shapes’.
The method splits a sound into two separate streams: transient and tonal. These two streams can then be independently processed, manipulated and generally messed with. These streams can then be recombined to produce a new sound. Eventide anticipates developing a family of plug-ins based on the method of splitting sound into its structural components.
Fig. 1 Signal Flow of Structural Effects Method
Structural Effects, with its emphasis on the transient nature of sound, will naturally be compared to traditional transient shapers which focus on the dynamics of a sound’s transients. While transient shapers can be used to great effect, their separation of fundamental sound components is imprecise, limited and comes with a steep learning curve. Furthermore, transient shapers typically only allow the manipulation of gain and decay, and usually only in the vicinity of prominent transients.
If transient shapers are a bludgeon, Structural Effects is a scalpel. Before now, attempting to modify the transient part of a sound without affecting the tonality often resulted in the components, the tonal and transient parts, bleeding into one another, introducing fading or phasing or slicing artifacts. Structural Effects completely separates transient and tonal, allowing each component to be processed independently.
This opens up new possibilities. For instance, tuning the tonal “ring out” of a snare without changing the pitch of the strike. Or removing the pick attack of a guitar note to make it sound bowed. Simply adding short delays to either advance or retard the transients relative to the tonal component before recombining can make subtle changes to the timbre. Extreme effects (delay, reverb, tremolo, etc.) can also be applied separately to the tonal and transient audio components, enabling the creation of completely new sounds.
The Structural Effects method creates entirely complementary pieces that can be manipulated independently and then recombined with minimal glitches, noise or other artifacts. Here’s an overview of the method:
First the waveform is divided into frames of ~40ms and is analyzed to find stable and/or predictable patterns. Stable sound components are then classified as tonal elements of the sound. There can be subtle or radical changes from frame to frame, but as long as those changes are predictable in the short term we choose to classify those elements as tonal. For instance, the decay of a note on a piano is a fairly predictable event even though the harmonics may shift slowly and the exact decay envelope is unknown. By analyzing individual frames in the context of the overall sound we are able to classify elements as tonal even as they vary over time.
Once the analysis has identified the portion of the sound that it ‘determines’ to be tonal, it then gracefully eliminates those portions leaving only transients. In a sense, the method ‘sifts’ the audio, separating the smooth gold (tonal) from the jagged sand (transient). The result is two completely separate, yet recombinable, streams of audio: a tonal stream and a transient stream.
Perception is, of course, Subjective
Of course, there is no precise dividing line between these ‘parts’ of a sound. Our method includes a decision making process which takes into account multiple factors and relentlessly splits the audio stream in two. While most of the determining factors are hidden ‘under-the-hood’, a select group of controls is provided which allow the user to help guide (and even force) the division decision.
Selecting SOURCE TYPE sets the under-the-hood expert tunings that prime the analysis and decider for a natural sounding split of a given source.
The user can then adjust the FOCUS contol to ‘fine tune’ or ‘guide’ the parsing of the sound based on contextual and musical needs/desires/tastes. FOCUS forces the method to favor one stream over the other. For example, changing the focus toward tonal will result in a more staccato transient stream as more transient tail energy flows into the tonal stream. However, changing the focus toward transient will result in more tonal bleed into the transient stream, causing heavier transient tails but also auto-swelling in the tonal stream. Extreme settings of focus will push the entire signal to either the transient or tonal stream. The focus control can be used to make subtle modifications (e.g. de-essing, de-reverberating or reducing pick noise) when combined with small changes to individual tonal or transient gain.
Additionally, the method provides SMOOTHING and TRANSIENT DECAY controls allowing the user to manipulate other aspects of the splitting process.
SMOOTHING essentially controls the speed of the decision making process, preventing it from making too fast a decision either way. This is useful for trimming out unwanted transient chirps or smoothing rough transitions.
Turning up TRANSIENT DECAY preserves energy in the transient stream, further thickening up the transient tail as well as increasing the auto-swell on the tonal stream.
Below is a screenshot of the STRUCTURAL SPLIT showing the layout of controls and a display of the sifted audio streams (transient in blue and tonal in green).
It’s important to keep in mind that what is added to one stream is subtracted from the other. This is the complementary nature of the method and the reason that a natural, musical division results. No portion of the source audio is discarded. This allows for near perfect reconstruction of the original audio.
We believe that Structural Effects is a breakthrough technology and hope this brief dive into the details was helpful. We’re eager to see how its potential can be tapped and exploited by artists, engineers and musicians. Our first Structural Effects based product is a plugin named Fission and will be released 1Q 2017.
Analog only for 50 years: 1925-1945 “Electrical” era +1945-1975 “Magnetic” era.