Programme Guide

Download the 'Programme at a Glance' giving the complete schedule of conference events and venues.

Introductory Remarks - Mark Sandler, Head of the Centre for Digital Music at Queen Mary, University of London, Josh Reiss, General Chair of AES31

1-1 Creating and Delivering High-Resolution Multiple 5.1 Surround Music Mixes—Mark Waldrep

The future of music reproduction is multichannel and high-resolution. However, the limited commercial success of recent high-resolution, surround audio optical formats demands that producers, engineers, and recording companies define exactly what is meant by high-resolution, surround music and that consumers experience content that demonstrates its full potential. Engineer and educator Mark Waldrep has built a unique library of award-winning, high-resolution, surround music titles that include multiple surround mixes created from dozens of stereo microphone pairs. His hybrid recording methodology brings studio techniques and “live” concert approaches together in a unique way.

1-3 Precision Measurement of ADC Effective Number of Bits Using Multitones—Raymond Belcher, Jonathon Chambers, Cardiff University, Cardiff, UK

A pure sine wave is the conventional test signal used for measuring the effective number of bits (ENOB) of an analog to digital converter (ADC) integrated circuit. Sine wave testing of ADCs weights the result so that peak level distortion is highlighted, and this is known to be less appropriate for audio. This paper demonstrates that a multitone test signal can be used to give most weighting to the central region and therefore produce an ENOB more relevant to audio, using less samples and without the difficulty of generating pure test signals.

2-1 Which of the Two Digital Audio Systems Meets Best with the Analog System?— Wieslaw Woszczyk,1 Jan Engel,2 John Usher,1 Ronald Aarts,3 Derk Reefman3
1McGill University, Montreal, Quebec, Canada
2Centre for Quantitative Methods CQM BV
3Philips Research, Eindhoven, The Netherlands

In this listening test, two digital audio systems (B and C), and one analog system (A) were tested by 10 test persons who listened to a surround sound scene “live” (without recording). The main question to be answered was: “Which of the two digital systems meets best with the analog system?” Both digital versions had 24-bit dynamic resolution but differed in sampling rate with which the analog signal was sampled. One version (C) was sampled with a CD rate of 44.1 kHz, the other (B) 8 times faster. There were also two test conditions, where in one condition there was a bandwidth cut off at 20 kHz instead of the 100 kHz that was possible with special 100 kHz microphones and added super-tweeters. For each subject, the experiment was replicated six times, in each of the two conditions. The outcome of each experiment was a 0 or 1, where the 1 means that the, technically best, digital system B has been chosen as meeting the analog quality. The paper describes the test and the outcome.

2-2 A Comparative Study of the Performance of Spatialization Techniques for a Distributed Audience in a Concert Hall Environment—Gavin Kearney, Enda Bates, Dermot Furlong, Frank Boland, Trinity College, Dublin, Ireland

The performance of various spatialization techniques is evaluated for a distributed audience using the non-ideal speaker arrangements found in medium-sized concert halls. The spatialization methods are
assessed in terms of their localization accuracy and listener envelopment. The data is presented by comparison of empirical binaural measurements and perceptual listening tests to simulations of the speaker arrangements in an equivalent acoustically modeled environment.

2-3 Audio Quality on the Air in DAB Digital Radio in Norway—Sverre Holm, University of Oslo, Oslo, Norway

We have analyzed the audio quality as recorded on the air for stations in the Norwegian DAB network. When the capacity is fully utilized, most stations with music transmit at a capacity of 128 kbps using MPEG1 layer II audio coding. By analyzing the mono component and the stereo component we have found that the audio is characterized. by a smeared stereo image due to the use of the heaviest from of joint stereo coding. There is also a lack of treble as the upper limit is usually 14 kHz. The result is a loss of brightness and a veiled sound stage which is particularly noticeable to young people.

P1-1 Practical Design of Circular Microphone Arrays: The Analysis of Spatial Aliasing Error and Microphone Placement Error in Circular Microphone Arrays—Abhaya Parthy, Craig Jin, Andre van Schaik, The University of Sydney, Sydney, NSW, Australia

We present a methodology for analyzing the spatial aliasing error, and the microphone placement error of both an open, or unbaffled, circular microphone array, and a circular microphone array that is baffled, ideally, by an infinite-length rigid cylinder. The methodology describes the practical design of a circular microphone array, with a specified number of microphones, which satisfies a specified threshold noise-to-signal ratio, beamforming order, and frequency range. Two practical designs have been constructed, and results of their analysis are presented.

P1-2 The Localization of a Sound Source in a Reverberant Room Using Arrays of Microphones —Simon Roper, Timothy Collins, The University of Birmingham, UK

This work is aimed at the problem of continuously measuring and predicting the impulse response within a reverberant room using sources-of-opportunity. The target application is that of compensating a sound reproduction system for variations in room acoustics and the non-ideal placement of the loudspeakers. The prediction of the impulse responses corresponding to the locations of an arbitrary number of listeners requires knowledge of the room geometry and the acoustic impedance of the boundaries. The room geometry may be obtained by estimating the image locations of a source. This has been addressed by the use of small arrays, with novel geometry, and processing both angular and temporal measurements. Preliminary practical results are presented and compared to theoretical predictions.

P1-3 Modeling and Control of Class-D Power Amplifiers for Vented-Box Loudspeaker Systems—Fran Gonzalez-Espin,1 Emilio Figueres,2 Gabriel Garcera, 2 Jesus Sandia2
1VMB, Valencia, Spain
2University Politécnica de Valencia, Valencia, Spain

The purpose of this paper is to analyze the stability of closed loop switching power amplifiers using an accurate model of a vented-box loudspeaker system instead of the commonly used resistive model. The model takes into account the electrical-mechanical-acoustical parameters of the transducer as well as the voice-coil loudspeaker’s nonlinear behavior, avoiding stability problems when trying to compensate for the output filter stage, thus obtaining better THD figures. This model has been experimentally tested and then used to design a regulator for Voltage-Mode Control method by means of MATLAB software. The regulator has been tested using a full-bridge power converter along with the proposed model of the transducer.

P1-4 Improved Psychoacoustic Noise Shaping for Requantization of High-Resolution Digital Audio—Christian R. Helmrich, Martin Holters, Udo Zölzer, Helmut-Schmidt University Hamburg, Germany

The popularity of high-resolution digital audio systems has renewed the interest in psychoacoustically optimized wordlength reduction. In this paper we examine recent approaches in fixed (time invariant) and signal-adaptive (time variant) psychoacoustic noise shaping. For the fixed case, we identify problems occurring when equal-loudness contours such as the threshold of hearing defined in ISO standard 226 are used as the basis for psychoacoustic noise shaping. For the signal adaptive case, we propose a noise shaping solution based on work by Verhelst and De Koning with improvements in the computation of the time-variant noise shaping filter. The paper concludes with a comparative evaluation of fixed and adaptive noise shapers based on listening tests in different environments.

P1-5 The Phase Amplitude Control Bit Stream Adder: A One-Bit Processing Structure for Phasor Manipulation of Oversampled Sinusoids—Enrique Perez Gonzalez, Joshua Reiss, Queen Mary, University of London, London, UK

This paper introduces a one-bit signal processing structure called the phase amplitude control bit stream adder. The proposed method is a digital filter structure, which is capable of controlling amplitude and phase over a one-bit sine wave without the need of directly multiplying the one-bit stream with a floating-point constant. It has applications in precision variable oscillator control and in oscillator bank additive synthesis reconstruction models. The research also explores a method for one-bit oscillator bank synthesis, without using intermediate multi-bit stages.

3-1 Segmented Dynamic Element Matching using Delta-Sigma Modulation—Ivar Løkken, Anders Vinje, Trond Sæther, Norwegian University of Science and Technology, Trondheim, Norway

In multibit delta-sigma digital-to-analog converters (DACs), the distortion from physical element mismatch can be spectrally shaped using dynamic element matching (DEM). A problem with all DEM schemes is that the complexity increases very rapidly with the number of levels in the DAC. To reduce DEM complexity for DACs with many bits, DEM segmentation using a dedicated segmentation delta-sigma modulator (DSM) has previously been suggested. Published segmentation DSMs have usually been first-order error feedback designs, to maximize the DEM complexity reduction and to minimize the analog overhead. In this paper high-order segmentation DSMs will be investigated and improved solutions proposed.

3-2 Energy Balance Decision Threshold in SDM Systems
Malcolm Hawksford, University of Essex, Essex, UK

Conventional SDM employs an amplitude comparator to convert sampled multilevel signals to a binary level bit stream where negative feedback is used to shape the output distortion spectrum. A modified threshold decision process is proposed that takes a holistic view of the state of the loop. Here a measure related to stored energy within the loop is calculated over a short look-ahead period and then used to select the polarity of the output code at each sampling instant. The motivation is to improve loop stability for high order coders especially for high amplitude input signals as encountered in switching power amplifier applications.

In the Listening Room, throughout the day, will be demonstrations of high resolution audio by Meridian Audio

4-1 Perceptual Investigation into Envelopment, Spatial Clarity, and Engulfment in Reproduced Multichannel Audio— Robert Sazdov, University Western Sydney - MARCS Auditory Laboratories, Penrith South DC, NSW, Australia

Composers of electroacoustic music have engaged with 3-D sound since the first performances of these works in the 1950s. Currently, the majority of electroacoustic compositions continue to be presented in 2-D. Human auditory perception is 3-D, however music composition has not adequately exploited the creative possibilities of this dimension. It is argued that ecologically valid perceptual experiments are required when attempting to formulate compositional techniques for electroacoustic music composition. Further, the paper presents a novel research method for the perceptual evaluation of 3-D multichannel electroacoustic music. The spatial attributes of envelopment, spatial clarity, and engulfment, are employed to evaluate composed multichannel 3-D sound executed within a concert hall environment. Results support various findings within the related disciplines of concert hall acoustics and multichannel reproduced audio.

4-2 Musical Attractors: A New Method for Audio Synthesis—Eric Nichols, Ian Knopke, Indiana University, Bloomington, IN, USA

In this paper we use mathematical tools developed for chaos theory and time series analysis and apply them to the analysis and resynthesis of musical instruments. In particular, we can embed a basic one-dimensional audio signal time series within a higher-dimensional space to uncover the underlying generative attractor. Röbel (1999, 2001) described a neural-net model for audio sound synthesis based on attractor reconstruction. We present a different methodology inspired by Kaplan and Glass (1995) to resynthesize the signal based on time-lag embedding in different numbers of dimensions, and suggest techniques for choosing the approximate embedding dimension to optimize the quality of the synthesized audio.

4-3 Wavelet Based High Resolution Audio Texture Synthesis—Deirdre O’Regan, Anil Kokaram, Trinity College Dublin, Ireland

Audio (or Sound) Texture Synthesis is performed by application of a well-known 2-D image texture synthesis algorithm to one dimension. This nonparametric, statistical approach is used to synthesize long, acoustically similar, high resolution audio textures from much shorter examples of both stochastic and quasi-periodic audio samples, which include a variety of ambient sounds (e.g., crowd noise, a baby crying). The process employs the Dual-Tree Complex Wavelet Transform (DT-CWT) to minimize computational load and maintain both temporal and spectral coherency in the synthesized audio. The results of this method are compared and contrasted with other state-of-the-art algorithms of a similar nature.

5-1 Horizontal Plane HRTF Reproduction Using Continuous Fourier-Bessel Functions—Wen Zhang, Thushara Abhayapala, Rodney Kennedy, Australian National University, Canberra, ACT, Australia

This paper proposes a method to reproduce the Head-Related-Transfer-Function (HRTF) in the horizontal plane. The method is based on a functional representation for HRTFs being a conventional Fourier series expansion for spatial dependence and a Fourier Bessel series expansion for the frequency components. The proposed representation can be used to predict HRTFs at any azimuth source position and at any frequency point from a finite number of parameters. Measured HRTFs from a KEMAR are used to validate the fidelity and predictive capabilities of the method. Errors between measured and modeled HRTFs are generally less than 2 percent.

5-2 HRIR Customization in the Median Plane via Principal Components Analysis of Head-Related Impulse Responses—Sungmok Hwang, Korea Advanced Institute of Science and Technology; Youngjin Park, Korea Advanced Institute of Science and Technology, Daejeon, Korea

A principal components analysis of the entire median HRIRs in the CIPIC HRTF database reveals that the individual HRIRs can be approximated as a linear combination of several orthonormal basis functions. The basis functions cover the inter-individual and inter-elevation variations in HRIRs. There are elevation-dependent tendencies in the weights of basis functions, and the basis functions can be ordered according to the magnitude of standard deviation of the weights at each elevation. We propose a HRIR customization method via tuning of the weights of three dominant basis functions at each elevation. Subjective evaluation results show that all subjects perceive the elevation angles more accurately with the customized HRIRs than the non-individualized and individual HRIRs.

5-3 The Generation of Panning Laws for Irregular Speaker Arrays Using Heuristic Methods—Bruce Wiggins, University of Derby, Derby, UK

In this paper an automated decoder optimization system using heuristic methods will be presented that will be shown to be robust enough to generate higher order Ambisonic decoders based on the energy and velocity vector parameters as proposed by Gerzon (Gerzon, 1985). This method is then analytically compared to Craven’s decoder (Craven, 2003) using both energy/velocity vector and head related transfer function based methods including analysis of head turning.

P2-1 A New Approach for CD 16-Bit Audio to High Resolution 24-Bit Audio—Prathibha Dhanushkodi

This paper analyzes the improvement of CD audio quality by using the differential evolution (DE) algorithm, which is a simple efficient adaptive scheme based on vector differences. It is used for high quality resampling of audio with the improvement of bit depth. A small amount of colored dither has been added to improve the value of SNR. Matlab results have been shown for resolution using this interpolation and the increase in the dynamic range of the signal.

P2-2 A Discussion about Subjective Methods for Evaluating Blind Upmix Algorithms—Nicolas Chétry, Grégory Pallone, Marc Emerit, David Virette, France Télécom R&D, Lannion, France

In this paper we discuss the problems that arise when one wishes to evaluate the performance of blind upmix algorithms. Based on the characteristics and spatial attributes that an ideal upmix should exhibit, we first discuss several evaluation methodologies published in literature. Second, we report internal listening test results during which the performance of four upmix algorithms has been evaluated. The dependence upon test material and listener expertise is highlighted. In order to complement already published research works, we present our experimental results and thoughts on this evaluation process.

P2-3 Visual enhancement using multiple audio streams in live music performance—Rozenn Dahyot, Conor Kelly, Gavin Kearney, Trinity College Dublin, Ireland

The use of multiple audio streams from digital mixing consoles is presented for application to real-time enhancement of synchronized visual effects in live music performances. The audio streams are processed simultaneously and their temporal and spectral characteristics are used to control the intensity, duration, and color of the lights. The efficiency of the approach at various audio resolutions is tested on rock and jazz pieces. The result of the analysis is illustrated by a visual OpenGL 3-D animation illustrating the synchronous audio-visual events occurring in the musical piece.

P2-4 Object-Coding for Resolution-Free Musical Audio—Steve Welburn, 1 Mark Plumbley, 1 Emmanuel Vincent2
1Queen Mary, University of London, London, UK
2IRISA-INRIA, Rennes, France

Object-based coding of audio represents the signal as a parameter stream for a set of sound-producing objects. Encoding in this manner can provide a resolution free representation of an audio signal. Given a robust estimation of the object-parameters and a multi-resolution synthesis engine, we can “intelligently” upsample a signal, extending the bandwidth and getting best use out of a high-resolution signal-chain. We will present some initial findings on extending bandwidth using harmonic models.

P2-5 St. Kliment (2006/7) 05.40—High Resolution Music Composition—Robert Sazdov, University Western Sydney, MARCS Auditory Laboratories, Sydney, NSW, Australia

The proposed high resolution music composition is a 3-D multichannel work that demonstrates novel compositional techniques based on audio perceptual research into spatial attributes.

6-1 System Configuration for High Quality Audio Capturing in a Large Microphone Array—Ines Hafizovic,1,2 Morgan Kjølerbakken,1 Vibeke Jahr1
1SquareHead Systems AS, Oslo, Norway
2University of Oslo, Oslo, Norway

In this paper we describe a speech acquisition system developed and manufactured for directive audio recording in outdoor arenas. Core technology, initially developed for audio production of sport meets the high requirements of the broadcasting industry. The system differs from the present audio recording solutions in its ability for user controlled focusing and steering of the sound recorded with a large, highly directive microphone array. Multi-channel real-time (RT) audio output presented together with video allows for a new multimedia experience in broadcast applications and other areas where remote speech acquisition is desired.

6-2 Digital Microphones for High Resolution Audio—Martin Schneider, Georg Neumann GmbH, Berlin, Germany

Microphones with a digital output format have appeared on the market in the last few years. They integrate the functions of a microphone, preamplifier, and analog-to-digital converter in one device. Properly designed, the microphone dynamic range can thus be optimally adapted to the intended application. The need to adjust gain settings and trim levels is reduced to a minimum. Dynamic range issues inside and outside the microphone are discussed. Advantages of microphones with a wide dynamic range and 24-bit resolution, according to AES 42 are shown and compared to simpler realizations with 16-bit resolution only.

Panel Discussion — Design Issues in High Quality Integrated Audio Systems
Panelists will include John Dawson, Arcam, Philip Hobbs, Linn, and John Atkinson, Stereophile. Additional panelists to be announced.

7-1 The Design of Improved First Order Ambisonic Decoders by the Application of Range Removal and Importance in a Heuristic Search Algorithm—David Moore, Jonathan Wakefield, University of Huddersfield, Queensgate, Huddersfield, UK

This paper presents improvements to previous work on deriving first order Ambisonic decoders for ITU 5.1. The decoders are derived using a heuristic search method with an objective function based upon Gerzon’s metatheory of auditory localization. An analysis of previously derived decoders shows that they are biased toward particular design objectives due to the nature of the multiobjective function guiding the search. This paper applies a technique called range removal to systematically and logically remove this bias that leads to improved decoder coefficients that better meet all of the objectives. A further technique known as importance is introduced that enables the logical biasing of range-removed objectives. A case study to develop a “max RE” decoder demonstrates this technique in action.

8-1 Achieving Real Bandwidth Beyond 20 kHz with a Loudspeaker System—Neil Harris, NXT, Huntingdon, UK

There are now quite a number of so-called “super tweeter” products on the market, all claiming bandwidth beyond 20 kHz. Frequently, however, this bandwidth is achieved only within a few degrees of the loudspeaker axis, and the read power bandwidth is much lower. The limiting factor in all piston-based designs is the “ka = 2” relationship between wave-number and radius. At 20 kHz, this results in the requirement that a is about 5 mm. By removing the requirement for the diaphragm to move as a piston, the “Balanced Mode Radiator” allows the use of larger diaphragms, making the possibility of real power bandwidth beyond 20 kHz a practical reality.

8-2 All Digital High Resolution Class D Amplifier Designs Using Power Supply Feed-Forward and Signal Feedback—Steven Harris, Jack Andersen, Daniel Chieng, Jeff Klaas, Michael Kost, Skip Taylor, D2Audio Corporation, Austin, Texas USA

This paper describes a digital input Class D amplifier that uses an integrated circuit controller. Sophisticated digital pulse width modulation, combined with digital feed-forward and feedback paths, yields high resolution amplifier designs. A powerful DSP is included in the controller to support amplifier control and allows comprehensive audio signal processing, including loudspeaker load compensation, EQ, time alignment, room acoustics compensation, bass enhancement, loudspeaker driver protection, virtual surround, and other audio signal processing tasks. Power supply feed-forward and closed-loop feedback technology correct for nonlinearity and other distortion-inducing mechanisms.

8-3 A Digital Amplification Technology to Optimize Performance with High-Resolution Audio—Craig Bell

This paper describes a performance-orientated closed-loop digital amplifier architecture that solves the traditional problems facing digital-input open-loop amplifiers and is shown to have performance advantages over equivalent systems implemented using conventional analog amplifier technology. Through the use of a high-resolution data-path and the global feedback structure, significant improvements in retained resolution at the amplifier output are evident.

9-1 Processing Techniques for the Recovery of Audio from Edison Cylinder Recordings, via Noncontact Surface Measurement—Antony Nascè, John McBride, Martyn Hill, Peter Boltryk, University of Southampton, Southampton, UK

A noncontact method for the recovery of sound from an Edison cylinder record is presented. The cylinder surface is scanned via white light displacement
sensor, capable of submicron axial resolution. Sound recovery is achieved by estimating the trajectory of a playback stylus over the measured surface, by tracking the central axis of the grooves. The processing methods required to extract audio from a discrete height map are described. We examine the signal to noise ratio as a function of position across the groove cross-section for different data sets, and compare spectra from the non-contact and stylus transfer methods.

9-2 MPEG-A Professional Archival Multimedia Application Format (MAF) Under Development—Noboru Harada, Yutaka Kamamoto, Takehiro Moriya, NTT Communication Science Labs., Atsugi, Kanagawa, Japan

This paper describes MPEG’s latest specifications of the Professional Archival Multimedia Application Format (MAF) that is currently under development, and describes proposed extensions to MPEG-4 Audio Lossless Coding (ALS) in order to support the Sony Wave64 Format and the Broadcast Wave File Format (BWF) with RF64, which handle large data size
exceeding 4 GB. The Professional Archival MAF compresses and archives files and folder structures into a single archive file so that recorded digital audio projects including meta-information and digitized traditional documentation (e.g., tracking sheets, lyrics, and engineer notes) are archived losslessly in the standardized manner. The format is sufficient for the future-proof archiving tool.

The following demonstrations will be held throughout the Conference: sound installation by The Illustrious Company; loudspeaker demonstration by Dyer Audio; original high resolution audio recordings; various demonstrations and exhibitions.