Anda di halaman 1dari 4
Audio Engineering Society Convention Paper Presented at the 128th Convention 2010 May 22–25 London, UK The

Audio Engineering Society

Convention Paper

Presented at the 128th Convention

2010 May 22–25

London, UK

The papers at this Convention have been selected on the basis of a submitted abstract and extended precis that have been peer reviewed by at least two qualified anonymous reviewers. This convention paper has been reproduced from the author's advance manuscript, without editing, corrections, or consideration by the Review Board. The AES takes no responsibility for the contents. Additional papers may be obtained by sending request and remittance to Audio Engineering Society, 60 East 42 nd Street, New York, New York 10165-2520, USA; also see All rights reserved. Reproduction of this paper, or any portion thereof, is not permitted without direct permission from the Journal of the Audio Engineering Society.

String Ensemble Vibrato: a Spectroscopic Study

Stijn Mattheij 1 ,

1 AVANS University, Breda, 4800 RA, The Netherlands


A systematic observation of the presence of ensemble vibrato on early twentieth century recordings of orchestral works has been carried out, by studying spectral line shapes of individual musical notes. Broadening of line shapes was detected in recordings of Beethoven’s Fifth Symphony and Brahms’ Hungarian Dance no. 5; this effect was

attributed to ensemble vibrato. From these observations it may be concluded that string ensemble vibrato was common practice in orchestras from the continent throughout the twentieth century. British orchestras do not use much vibrato before 1940.


Commercial audio recordings, especially the oldest, are interesting documents, as they inform us about the development of performance styles in the twentieth century. The study of performance characteristics of violin playing such as timing, rhythm, vibrato and portamento has gained considerable attention in recent years [1, 2]. Often listening is used as a method of extracting information from these sources. However, it is known that vibrato is often inaudible to the human ear below a certain threshold or in the presence of noise due to partial masking, making the effect difficult to observe on old noisy 78 rpm recordings from the early twentieth century [3, 4]. This is one of the reasons that both solo and ensemble vibrato are controversial subjects in the field of musicology [5]. Solo vibrato was the subject of a previous study [6]. The present study is of particular importance for conductors specializing in the late

romantic symphonic repertoire: did the strings in the orchestra employed vibrato around 1900? [7]

Vibrato is produced by a periodic motion of the left hand of the performer’s finger on the string, resulting in a small modulation of its effective length. This causes periodic variations the vibration frequency of the string, which in turn results in variations of the sound amplitude and the directional pattern [8].


In the previous study solo vibrato was investigated [6]. Time dependent spectral analysis using Morlet wavelets was used, resulting in a time resolution of 0.01 second and a pitch resolution of about 10 cents.

Ensemble vibrato produces a sound which is quite different from solo vibrato. If ten to twenty violin players play a tone with vibrato, they will generally not



synchronize the rocking motion of their left hand. Their individual sounds will blend into a thicker sound (this produces a chorus like effect). As a result, the pitch fluctuations which are a characteristic of solo vibrato cannot be heard. A further complication results from the fact that in an ensemble with multiple voices playing various pitches to build a harmony, the higher partials will mix, so for these harmonics it is impossible to determine whether they stem from instrument A or B. A wavelet analysis of the musical signal then still shows what actually happens, but the timing information has no meaning because of the different timings of the individual players. Because of the absence of synchronization and acoustic reflections the time frequency plots are very hard to interpret and cannot be used to measure the width of the vibrato [9]. Therefore time averaged acoustic spectra were employed to obtain information about the presence of vibrato. This restricts the analysis to notes of some length. The method is illustrated in figure 1, where the difference between solo and ensemble tones, with and without vibrato is shown. These graphs were obtained by performing a Fast Fourier Transform on a musical sample of the signal with a length 1.5 seconds (giving a frequency resolution of 0.7 Hz). In the case of a solo performance the width without vibrato is about 10 cents at the vibrato fundamental frequency. If vibrato is applied, the width increases to 60 cents. To illustrate the effect of ensemble playing, spectra from the G5 in bar 21 of the first movement of the Fifth Symphony by Beethoven are shown (G5, fundamental at 784 cycles/second, this tone is played by the first violins only). Here we compare a performance which is believed to employ minimal vibrato (Norrington) with a the performance which employs maximum vibrato (Horenstein). The performance by the Vienna Pro Musica Orchestra, conducted by Horenstein (1956) has a spectral width of 149 cents (fundamental) and 110 cents (first harmonic), for the performance by the Stuttgard Radio Symphony Orchestra conducted by Norrington the corresponding figures are 54 and 72 cents respectively. If we take the average over the first six harmonics, the results are 116 cents (Horenstein) and 44 cents (Norrington). We conclude that the spectral width of these tones reflects the extent of the vibrato.

We have one complication here: though we have shown that vibrato is a factor which determines the shape and the width of the time averaged spectrum, it is not the only factor that plays a role. Variations in the rotational speed of the recording and/or playback apparatus will produce spectral line broadening: a variation of 1

Mattheij STRING ENSEMBLE VIBRATO: A SPECTROSCOPIC STUDY synchronize the rocking motion of their left hand. Their

Figure 1 Normalized spectral lineshapes of violin vibrato tones; upper trace: G 5 (fundamental), with and without vibrato; lower trace: G 5 from Beethoven, Symphony no. 5 first movement bar 21 conducted by Horenstein and Norrington.

percent will produce a line broadening of 10 – 20 cents. Another factor is the spread in intonation of the players. If we have 10 to 20 first violins, their intonation will not be exactly the same. A spread in intonation is acceptable in a limited bandwidth and is not observed by the auditory system as playing out of tune. A systematic study of the acceptable spread in intonation showed that for tones, not modulated by vibrato, a spread in pitch of

  • 24 cents is acceptable; in case that vibrato is used, a

spread of 34 cents is acceptable [10]. If we associate the pitch of a tone with the maximum in the spectrum,

then we might expect an overall spectral width of 24 cents from the spread in intonation plus 10 cents

because of the spectral width of a tone produced by an individual player; thus the spectral width of an ensemble of unisono players not using vibrato is about 24 + 10 or

  • 34 cents. In the case of an ensemble vibrato tone, the

width will be at least 34 + 10 or 44 cents or more. In general, there is no simple algebraic relation between

the actual spread in intonation, the width of the vibrato and the overall spectral width. However, for the maximum spectral width, the situation is easier: in case of a wide ensemble vibrato, the maximum width may be estimated to be 34 cents plus the vibrato width of the player with the widest vibrato.



To check these figures, the audio signals from two video recordings (where the use of vibrato can be checked by visual means) were studied. Two fragments were selected according to the following criteria: 1) only the first or second violins must play alone; 2) the recording should show the violins in close up, to allow for a careful visual observation of the motion of the left hand of the performers. In the video recording of Tchaikovky’s 6 th symphony Norrington explains the arrangement of the first and second violin parts of the first phrase of the last movement, the Adagio Lamentoso: both do not play the melody individually, but the melody can be heard if they play together. Here we can hear and see the first or second violins play their part alone. The observed spectral width for the first violins equals 36 cents, there is no vibrato visible on the video recording. With the G5 in bar 21 from Beethoven’s Fifth performed by the Berliner Philharmoniker conducted by Herbert von Karajan the first violins employ a clearly visible vibrato; it has a spectral width of 82 cents. We conclude therefore, that a spectral width of less than about 40 cents corresponds to no vibrato; 50 cents or more indicates some vibrato (in the case some players employ it), and 80 cents or more indicates a strong vibrato. Sometimes it is possible to check a recorded performance for the presence of vibrato if the violins play an G2 somewhere in the piece: it has to be an open G string, so there is no line broadening because of vibrato.

Mattheij STRING ENSEMBLE VIBRATO: A SPECTROSCOPIC STUDY To check these figures, the audio signals from two

Figure 2 Vibrato width in cents in four recordings of Brahms’ Hungarian Dance no. 5; in every pair the left bar corresponds to the A 5 in bar 8 and the right bar to the G 3 (open string) in bar 49


Two musical fragments, spanning about a century of recording history, were selected with long (> 1,5 seconds) tones played preferably by the first violins only. The width of the vibrato was measured for several long notes about 1 second or more in length. Here some representative results are shown for two pieces. The

results for the first piece, the Hungarian Dance No. 5 by Brahms (in G minor for the orchestral version) are shown in figure 2. To demonstrate the effect of ensemble vibrato, the spectral width of the A5 in bar 49

is compared to the G3 (open G string) in bar 8 for four recordings. The spectral width of the vibratoless open G string is always significantly lower than the spectral width of the A5. The latter fluctuates over a whole century within a bandwidth of 80 +/- 10 cents. The spectral width of the open G string fluctuates between 40 and 50 cents. The earliest recording was made by


and his orchestra (with the same name) in

Paris in 1906. It follows that even in Colonne’s recording a rather wide ensemble vibrato was present. The other recordings behave as we would expect. In figure 3 the results for the opening phrase of Beethoven’s Fifth Symphony (the G5 in bar 21 of the first movement) are shown. Vibrato is already observable in the oldest recording (Nikisch, Berliner Philharmoniker); sometimes it is even quite strong (Strauss). In the recordings between 1930 and 1970 it has a width of about 100 cents. The recordings after 2000 conducted by Norrington and Weil display no vibrato, as these recordings were ‘historically informed’.

Mattheij STRING ENSEMBLE VIBRATO: A SPECTROSCOPIC STUDY To check these figures, the audio signals from two

Figure 3 Vibrato width in recordings of Beethoven’s Fifth Symphony (G 5 in bar 21 of the first movement)

The recordings by Weingartner deserve special attention. The earliest from 1926 was recorded by the British Symphony Orchestra, the one from 1930 by the London Philharmonic. The first shows a spectral width of 45 cents, indicating no vibrato at all, the latter a width of 60 cents or some vibrato. Are the British orchestras different from the continental orchestras? To explore this point, two video recordings available from the Youtube website of Elgar’s Land of Hope and Glory were studied: one conducted by Elgar himself in 1931 and the performance of 2008 by the London Philharmonic conducted by Norrington. On the Elgar



recording, some individuals can be seen playing with vibrato, on the Norrington recording one violinist can be seen, obviously not participating in the vibrato ban. The analysis of the vibrato of the A3 in bar 4 give 50,3 cents for the Elgar recording and 57,3 cents for the Norrington recording. So we may –quite surprisingly– conclude that Norrington’s performance is indeed ‘historically informed’, and that around 1930 the British orchestras employ a vibrato which is at least significantly shallower than the orchestras on the continent. When the Berlin Philharmonic visited London, letters to the editor of the Musical Times were sent complaining the deplorable state of British Orchestras [11, 12]. Some ask for an orchestral reform [13]. Though these sources do not mention string vibrato explicitly, they indicate that the level of training of orchestral musicians was inadequate. In 1977 Sir Adrian Boult, in an interview with Peter Wadland (who remembers older orchestral members playing straight and the younger ones using vibrato), commented on the rise of the continuous vibrato: ‘yes, I suppose that happened. It seemed to blend very well in the end, but I wasn’t really conscious that that the thing was changing very much’ [14].


Ensemble vibrato primarily has an effect on orchestral timbre: the bandwidth of the harmonic partials increases with vibrato extent, resulting in a sound that may be perceived as ‘warmer’ or ‘thicker’. It may or may not add to the expression of the sound (dependent on the perception of the listener). If vibrato is used, the sound is more tolerant with respect to inaccuracies in intonation. There is no hard evidence from historical writings that ensemble vibrato was out of fashion in the late romantic period. It is present in early recordings (with recordings made in England being an exception). The results suggest a dynamic picture: it may pass in and out of fashion, and local variations are possible (certainly in the time preceding the use of mass media).


The author wishes to thank Dr. George Mowat-Brown for his stimulating discussion of this work.


[1] Clive Brown: Classical and Romantic Performing Practice 1750-1900. Oxford University Press, Oxford, 1999.

[2] Daniel Leech-Wilkinson, The Changing Sound of

Music: Approaches











intro.html, last acces on 01.09.2009.

[3] Carl Seashore: Psychology of Music. McGraw-Hill, New York, 1938.

[4] Hugo Fastl, Eberhard Zwicker: Psycho-Acoustics. Springer, Berlin, 2006.

[5] Frederick






Performance Practice Review 4 (1991) 14-27.











Recording: a Wavelet Study. Proceedings of the NAG/DAGA, International Conference on Acoustics, Rotterdam (2009) 1490-1493.

[7] Roger Norrington: The Sound Orchestras Make. Early Music 32 (2004) 2-5.

[8] Judith C. Brown: Frequency ratios of spectral components of musical sounds. J. Acoust. Soc. Am. 99 (2) (1996) 1210-1218

[9] Jürgen






Streicher-Vibratos. Acustica 76 (1992) 283-291.

[10] R. Van Besouw, J. Brereton, D. Howard: Range of Tuning for tones with and without Vibrato. Music Perception 26 (2008) 145-156.

[11] Frank Howes: The Berlin Orchestra. The Musical Times 73 (1932) 448.

[12] J.B.




Times 73 (1932) 442

Orchestra. The Musical

[13] Henry Welsh and F.G. Rendall: Orchestral Reform. Music & Letters 12 (1931) 21-29.

[14] Bruce Haynes: The End of Early Music. Oxford University Press, Oxford, 2007.