REVERB DESIGN
Author: Urban Schlemmer
Reverb Design is about creating an aesthetic appealing spacial impression aimed at sweetening the listening experience for a given context. It is thus strongly dependent on the type of music and may be an integral part of a composition.
Natural sounding artificial reverberation (Schroeder 1960) tries to link the listening experience to real venues. Further developed by Gardner 1992 and implemented by Griesinger in the Lexicon 480L unit, these recursive designs are still in use today.
Brief Introduction to Psychoacoustics
20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)~~
To design a reverb for a given sound, some knowledge of psycoacoustics will be of interest:
What we hear is subject to cognition. Research in the field of the precedence effect by Clifton, Freyman, Shin-Cunningham et al. (1989-1992) shows that the neural system permanently builds expectations of what will be heard. Longterm memory and (short term) learning to adapt to a reflective environment equally play important roles for these estimations. Top-Down processes compare the expected to the real sound event and adjust the auditive sensation accordingly. This is particularly true for echo perception and spacial impression.
When reproducing a concert hall with loudspeakers, different cognitive criteria may be used to form expectations compared to the real concert event (because eg. important visual cues are missing). Consequently, exact reproduction of the soundfield of the concert hall may not be the best option for the reverb designer.
Algorithms specialized in the creation of an auditory illusion outperform the physical model in an aesthetic sense due to psychoacoustic facts. Memory of previous similar experiences tells the listener to expect a super-real, exiting room simulation, where contradicting spacial cues may coexist - such as beeing extrem close to the sound source and far away at the same time. This leads to hightened attention to these spacial attributes.
How may "natural sounding" artificial reverberation differ from a perfect reproduction of the original sound field (eg. provided by binaural technique or wave field synthesis)? This is also an aesthetic question, however, the evolution of measurement standarts in concert halls addressing clarity, intimacy, listener envelopment (LEV) and auditory source width (ASW) gives important hints. The following parameters are independently controlable in artificial reverberation while preserving spacial impression cues of real halls:
- Reverb time (RT) relates to perception of room size, damping, LEV and distance perception. Long reverb tails are typically used in classical music production even if the recorded ambience was tiny. This is possible because the reflections of small rooms may also occur in larger halls.
- "Intimacy" has found to be related to the initial time gap (ITG) in real halls, a parameter controlable as "predelay" in reverb design.
- In the aproximate time range between 60...250 ms (baroque) concert halls show characteristic coloration as well as late reflection patterns. For wet recorded sound it is possible to modify shape, density and coloration of the impulse response (IR) in this time range to form an optimal extension to the pre-existing material.
- ASW and LEV ("spaciousness") are independent of early reflection densitiy. They depend, however, on the total (lateral) reflected energy (Barron 1972). This enables the reverb designer to improve clairity of the reproduced sound by reducing the number of early reflections to those most salient in a modeled room.
Salient Properties of the Vienna Musikvereinssaal 20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)Bjoern 20:12, 20 May 2011 (CEST)
What are salient reflections of good sounding concert halls ready to be modeled? Barron (2002) investigated several halls including Vienna Musikvereinssaal (known by the author from own recording experience) by sending questionaires to musicians and directors performing in these venues. Typical architectural properties were found, although depending on the type - and even the piece - of music:
- size of 15.000 - 17.500 m3 - shoebox shape with stage housing and slightly concave front, forming an acoustic lense - slightly concave ceiling with rounded corners - balcony - colums and sound scattering, reflective surfaces in the front and to the sides - plaster and wood are preferred construction materials - absorbing public on the main floor is part of the design
Implementation 20:12, 20 May 2011 (CEST)20:12, 20 May 2011 (CEST)Bjoern 20:12, 20 May 2011 (CEST)
These properties were implemented in a room model to calculate delaytimes, spherical coordinates and dumping of dedicated source image models. The choice of 12 reflections is left to the user, although standart presets are available. Reflections from the side walls and ceiling-to-balconies are a standart for most presets. Geometry and room size are adjustable by the user.
Micromovements of the sound source are calculated in real time leading to realistic doppler effects and time variing changes in spectra. (Called "fluctuations" by Griessinger).
Previous studies have shown that the directivity of orchestral instruments is an important property of the frequency response of these calculated reflection patterns (Schlemmer 2006). Therefore the directivity data of 13 instruments was made available as a PD external. The audio part consists of three delay lines plus allpass section per channel. A feedback matrix forms a feedback delay network (FDN) for precise control of diffusivity and reflectivity of sound scattering surfaces. A filterbank was built in FAUST to take advantage of the directivity data.
Although this FDN may generate late reverberation as well, it is mainly used to design convincing ER patterns. It was tested with a stereo setup. In previous studies, five and eight speakers were used sucessfully.
The late reverberation part was inspired by Gardner's and Griesinger's delay lines with nested allpass sections (Gardner 1992). Input diffusion stages constist of short, time-variable feedback delays followed by a Schroeder section implemented in FAUST as PD externals.
To combine early and late parts the input of the Schroeder sections can be controlled in various ways: ER and direct sound may be blended - which may be seen as an unusual approach for diffuse input - so as to design onset, shape and density differences in the resulting IR. Parameter changes can be viewed in realtime (both grafically and auditory) to learn the handling of the complex interaction of parameters. Examples of this interaction will be shown in the talk and typical cave eats will be discussed.