Model-Based Synthesis of the Clavichord Vesa Vlimki 1 , Mikael - PowerPoint PPT Presentation

ICMC 2000, Berlin, August 2000 HELSINKI UNIVERSITY OF TECHNOLOGY Model-Based Synthesis of the Clavichord Vesa Välimäki 1 , Mikael Laurson 2 , Cumhur Erkut 1 , and Tero Tolonen 1 1 Helsinki University of Technology, Laboratory of Acoustics and Audio Signal Processing (Espoo, Finland) 2 Sibelius Academy, Centre for Music and Technology (Helsinki, Finland) Välimäki et al. 2000 1

HELSINKI UNIVERSITY OF TECHNOLOGY Model-Based Synthesis of the Clavichord 1. Introduction 2. Acoustics of the Clavichord 3. Synthesis Model 4. Musical Example 5. Conclusions Välimäki et al. 2000 2

HELSINKI UNIVERSITY OF TECHNOLOGY 1. Introduction • The clavichord is one of the oldest musical instruments • Sound is pleasant but quiet – Maximum SPL only about 50...60 dB • Only useful in intimate performances • We propose a digital synthesis method to generate clavichord tones • SPL can be amplified using the volume knob Välimäki et al. 2000 3

HELSINKI UNIVERSITY OF TECHNOLOGY 1. Introduction (2) • Mikael has a clavichord Välimäki et al. 2000 4

HELSINKI UNIVERSITY OF TECHNOLOGY 1. Introduction (3) • The instrument used in our measurements (Heugel, Paris) Sound example: A short excerpt of a piece by J.S. Bach Välimäki et al. 2000 5

HELSINKI UNIVERSITY OF TECHNOLOGY 2. Acoustics of the Clavichord • String is excited with a simple lever mechanism • Enables mechanical aftertouch (from T.D. Rossing, “The Science of Sound”, 2nd Ed. Addison-Wesley, 1990) Välimäki et al. 2000 6

HELSINKI UNIVERSITY OF TECHNOLOGY 2. Acoustics of the Clavichord (2) • Waveform shows the attack transient and a noise burst at the end (A3, 197.5 Hz) • Sound example: this signal 3 times 0 0.5 1 Time (s) Välimäki et al. 2000 7

HELSINKI UNIVERSITY OF TECHNOLOGY 2. Acoustics of the Clavichord (3) • Pitch decays during the attack – Player’s hand probably affects the string tension – Contribution of tension modulation effect is weak • Should be accounted for in sound synthesis Frequency (Hz) 200 199 198 197 0 0.5 1 Time (s) Välimäki et al. 2000 8

HELSINKI UNIVERSITY OF TECHNOLOGY 2. Soundbox • The bridge was excited with an impulse hammer – Most prominent modes can be found – Excitation point was varied to see how mode amplitudes change Välimäki et al. 2000 9

HELSINKI UNIVERSITY OF TECHNOLOGY 2. Soundbox (2) • Modes are Magnitude (dB) excited in a 60 High end different way 40 • The soundbox 20 response is an 0 100 200 300 400 500 essential part of Magnitude (dB) 60 clavichord tones Low end • Sound example: 40 1) Original 20 2) Fade-in to avoid 0 100 200 300 400 500 metallic attack Frequency (Hz) (Played 2 times) Välimäki et al. 2000 10

HELSINKI UNIVERSITY OF TECHNOLOGY 3. Synthesis Model • Commuted Waveguide Synthesis (Smith, ICMC’93; Karjalainen et al., ICMC’93) – Partials are extracted from recorded tones (inverse filtering or subtraction of a sinusoidal model) – These are excitation signals for a string model • Currently, we have a fully automatic analysis (Erkut et al., 2000, AES 108th Conv.) • Sound example – A chromatic scale “without partials” but including the soundbox sample Välimäki et al. 2000 11

HELSINKI UNIVERSITY OF TECHNOLOGY 3. String Synthesis Model • Invented by Jaffe and Smith (1983) based on the Karplus-Strong algorithm (1983) • Can be derived from the solution of the wave equation (Smith, CMJ 1992) Input O ut p u t F( z) D e la y l i n e H l ( z) F ra c t i on a l D e lay Loo p F i l t e r Decay r a t e Fundam en tal f r e q uency Välimäki et al. 2000 12

HELSINKI UNIVERSITY OF TECHNOLOGY 3. Modified String Model • Modification to control the sharpness of the attack – Controls the contribution of the hammer noise – Needed particularly at high frequencies • A sharper attack is obtained when g dir > 1 A t tack sharpness g d ir Input F( z) D e la y l i n e H l ( z) O ut p u t Välimäki et al. 2000 13

HELSINKI UNIVERSITY OF TECHNOLOGY 3. Modified String Model (2) • Sound example (A4, 392 Hz) – g dir = 1.0 – g dir = 2.0 – g dir = 3.0 – g dir = 4.0 (played 2 times) g d ir Input F( z) D e la y l i n e H l ( z) O ut p u t Välimäki et al. 2000 14

HELSINKI UNIVERSITY OF TECHNOLOGY 3. Model Structure • Two string models coupled together for each string • Resonators simulate the ringy modes of the soundbox g g i 1 o 1 In Out ( ) S 1 z c 1 2 g g i 2 o 2 ( ) S 2 z R 1 z ( ) ... ( z ) R K Välimäki et al. 2000 15

HELSINKI UNIVERSITY OF TECHNOLOGY 3. It’s Not Really Physical Modeling • Several samples are used 1. Excitation – Processed recordings – Fed into the string models 2. Hammer noise for ends of notes – Can be extracted from recordings of very long tones 3. Soundbox response – Processed impulse response of the soundbox – Triggered at a low level for each note Välimäki et al. 2000 16

HELSINKI UNIVERSITY OF TECHNOLOGY 3. Implementation • Implemented using ENP – Extended standard musical notation and synthesis control (see Laurson et al ., ICMC’99) • Real-time synthesis with 6-voice polyphony on a Macintosh G3 laptop • Recently, tension modulation nonlinearity was included (Tolonen et al., 2000) – Produces “less synthetic” sound Välimäki et al. 2000 17

HELSINKI UNIVERSITY OF TECHNOLOGY 4. Musical Example • Renaissance music produced by Mikael Laurson: Guillaume Dufay (1400-1474), “Par le regard de vos beaux yeux” Välimäki et al. 2000 18

HELSINKI UNIVERSITY OF TECHNOLOGY 5. Conclusions • A simple clavichord synthesis model was proposed • Combines Commuted Waveguide Synthesis and Sampling – String synthesis using coupled digital waveguides – Repreduction of hammer noise and soundbox responses using special and processed recordings • Demos and musical examples will be available at: http://www.acoustics.hut.fi/~vpv/publications/icmc00.htm Välimäki et al. 2000 19