weaver

Analysing the impact of reverberation on expressivity in ensemble performance

James Weaver

Music & Speech Modelling 2018

Abstract

It is widely observed that reverberation time has an impact on musical performance, and while it has been analysed objectively for soloists,  ensemble performances are mainly analysed via subjective self-report. This project presents a study of two keyboard players performing in ten simulated reverberation conditions, and analyses the changes in tempo with a methodology that can be used to analyse other changes in performance. The study presents an interesting finding on error and adjustment in performance where errors increase in dry and highly reverberant conditions. 

Introduction

Performing musicians experience variable acoustics from venue to venue, and acoustic design has historically prioritised the audience over the musician to the detriment of the working conditions of performers. While stage and in ear monitoring mitigates against many of the issues for amplified performers, an acoustic ensemble remains at the mercy of the natural room acoustic influencing many factors including Agogic, Dynamic Strength, Dynamic Bandwidth, Tempo and Timbre. 

For this study we will be particularly looking at Tempo as a measure – This will be the tempo (speed) chosen by the duo for the piece.  From a previous study of a duo conducted by Kato et al[1] it is shown tempo was reduced for short and long reverberation times. While this is logical when thinking about the idea of ‘comfort reverb’ used in say recording studios which utilise a certain degree of reverberation to support performers does it hold true in more extreme reverberation environments and when performers are located in the same space?

Literature Review

In acoustic design of performance spaces the majority of work focusses upon the audience, and work that does look at the impact of acoustic environments upon performers is either from a subjective viewpoint, or concentrates on solo instrumentalists [2]–[5]. In the literature reviewed for this project studies generally only analyse the performance of solo musicians in a non-subjective manner, with larger ensembles only analysed through reflective analysis. A key study in this area by Ueno  [6], [1], [7] looks at the playing of soloists in a range of simulated acoustic environments and attempts to understand in a quantifiable way the adjustment of tone, tempo, the length of notes, the silence between notes, articulation, dynamic level, dynamic range, harmonics strength, pitch tuning, and/or vibrato extent in relation to acoustic environments. This particular study goes into some interesting detail, and correlates it to the subjective opinion with the changes in the audio showing considerable agreement with the statements from the performers. The paper found that tempo, note length, silence, articulation, dynamic level, dynamic range, harmonics strength, pitch tuning, and vibrato all varied significantly with room acoustic conditions.

However, there are limitations with this study, along with others by Kalkiandjiev [8], [9] and Amengual Garí [10]. The first is the focus on soloists rather than ensembles. The area of the soloist is widely covered, but music often relies on the ensemble for its performance, and has not been subjected to the same level of study.  The second limitation is that where multiple performers have been studied such as in Kato et al[1] they have been located in separate spaces. While this makes logistical sense when using small anechoic chambers there is an issue here in ensuring ecological validity, as most musical performance is conducted in the same physical space – which is something I’ve looked to address in this study.

Procedure for Data Collection

For this study two experienced pianists were recruited from the Centre for Digital Music at Queen Mary University of London. They were given 48 hours to prepare a duet of GershwinPiano Concerto in F II. Adagio - Andante con moto (in an arrangement for 2 pianos)[12]. The musicians were asked to learn bars 0-63.

In order to make a replicable study easily available commercial equipment was used.  The keyboards used were  M-Audio Axiom 61 keyboards using the Simpler Grand Piano plugin in Ableton Live 9 Suite. This plugin was used as it is a polyphonic sample based plugin, and provides a realistic piano effect.   The ten different reverberation conditions were presented through the Convolution Reverb Pro Device for MAX/MSP[13] and a dry condition presented through two Beyer Dynamic DT 150 Headphones. These were patched through a Focusrite Scarlett Solo audio interface and a Samson S-Amp to ensure participants and investigator could hear proceedings.  The Impulse Responses were selected from the OpenAIR Library[14] and were chosen to represent a wide range of conditions for the musicians. These are presented in the appendix.

The conditions were delivered in a randomised order, as due to the unfamiliarity of the piece of music being performed there was a concern their improving performance could confound the results if the conditions were presented from (for example) no reverberation time to a full reverberation time.

The performances were captured in MIDI in separate channels create a discrete channel of data for further analysis. The MIDI data was then synchronised and converted to audio to allow annotation.

Due to time constraints the annotation was conducted by the investigator using Sonic Visualiser[15] an plugins ‘BeatRoot’[16] (to analyse the timings of the performances ). Data was then exported and analysed in MATLAB using bespoke code.  

Results

Initial analysis was conducted on the overall tempo of the performances to compare all the different conditions. This was outputted in performance time, and also score time in order to check comparison. It is clear that, despite the observations of the performers and the investigator that there was very little significant change to the overall length of the performances. However, there were clearly a range of different tempo changes in the piece, and analysis of the tempo was conducted to gain maximum, minimum and mean tempos for each piece to give an overall picture of the performances.