Anda di halaman 1dari 6

Analysis Study of Mode Frequency and Damping Changes

due to Chemical Treatments of the Violin Bridge

Dr. George Bissinger


Physics Department
East Carolina University
Greenville, NC 27858

Mach&e Bailey
Physics Department
East Carolina University
Greenville, NC 27858

ABSTRACT
Twenty maple violin bridge blanks - six with a
f o r m a l i z a t i o n c h e m i c a l t r e a t m e n t , six
with
a
rcsorcin/formaldehyde treatment and eight untreated were
studied to compare UK. effects of chemical treatments on
free-free vibrational mode frequency and damping.
Hammer-impact modal analysis was performed on one
bridge of each group, both in-plane and out-of-plane over a
0 20 kHz range, using a microphone as a no-load vibration
transducer. All the bridges exhibited three reasonably
strong modes in the IP modal shapes while six modes were
seen in the OP analysis. Damping measurements made on
all twenty bridges with one excitation IP point and one OP
point gave frequency averages with -0.4% errors and
damping averages witb -3% errors. Significant frequency
changes baween UN and FA or RE bridges were observed
in the IP mode at -10.6 kHz of -1.4% (FA) and +1.4%
(RE). For the OP mode at -3.0 kHz FA and RE! bridges had
frequency increases of 1.4 and 1.9%. respectively.
Signilicant changes in damping were seen for tbe RE IP
nwdc at -10.6 kHz, for tbe FA OP modes at 2.64, 3.01 and
4.72 kHz, and the RE OP modes al 2.64,4.71 and 5.66 kHz.
NOMENCLATURE/ABBREVIATIONS
FA Formalization treated bridge
RE Rcs~)r~itl/foormaldehyde
treated bridge
U N Untrcatcd bridge
IP In-plant (vibrations)
OP Out-of-plane (vibrations)
,,I Mas
FE Finite clement (analysis)
MA (cxpcrimental) Modal analysis
DOF Dcgrce of frwdom
< %&critical damping
RH Relative humidity (in %)
FRF Frcqucucy
response function
FFT Fast Fourier uansform
STDEV Standard deviation of mean
MAC - Modal Assurance Criterion

WIRODUCIION
Bridges are a crucial component io the performance of a
violin. Their main function is to couple the energy in the
strings to the top plate and thence to the rest of the
instrument, the corpus, where it is radiated. Felix Savart
noted in 1840 that, Many attempts have been made Lo
perfect [the violin bridge]; the result has been that we cannot
modify it [physically] without taking away many qualities
from the. sounds and from the instruments [I I. But now an
old idea has taken on a new twist because it appears that the
bridge can be further modified chemicall.v to aid modern
luthiers 121. The formalization treatment used on one set of
the bridges appeared to improve the sound quality of various
violins including mediocre, modern violins as well as old,
high-quality, Italian violins. Blind sensory tats conducted
with professional string players to evaluate the sound quality
of each instrument returned judgments tbal the violins with
the treated bridges tended to be brighter, more sonorous,
less poor and less dull when compared to the same violin
with an untreated bridge 121.
To understand these perceived effects a sludy of twenty
maple violin bridge blanks has been conducted. Hammcrimpact MA measuremenls on one bridge of the untreated
and treated formalizaliou and resorcia-formaldehyde types
were used to identify IP and OP modes in the vibrational
spectra and to extract mode shapes, frequencies and
dampings to characterize tbc Ilexures. In conjunction with
the MA, vibrational response measurements at one IP and
OP point were made for every hridgc to extract frcqucacics
and dampings for modes below 11 kHz lo look for trends
associated with each chemical treatment.
EXPERIMENT
1. Phvsical Prooerties of Bridges
Six FA bridges, six R!? bridges and tight UN hridgcs ol
similar maple wood and prc-lrcatmcnt mass were studied.
All were blanks; i.e. they were unfitted to any particular
violin and hencc were all the same shape. A tilted hridgc is

828

II. Chemical Treatments of Bridges


A. Formalization treatment
The formalization treatment was conducted by Yano as
follows: 6 bridges and 0.003 kg of textraoxane, a reagent
source of formaldehyde (giving a formaldehyde
concentration of about 0.03 mol/dm3 when depolymerized
completely), were put into a 3.5 liter glass vessel and the
vessel was evacuated. After a vaporous sulfur dioxide
catalyst of 4.0*10w3 mol/dm3 concentration was introduced
volumetrically, the whole vessel was heated at 120-C in an
oven [3].

one in which the feet, arch and eye holes have been
trimmed. All bridge masses were measured at 28% RH.
Treatment caused a definite mass change; the treated bridges
were heavier, as expected, with the RE bridges being
heaviest, as seen in Figure 1. The UN bridges had m =
(2.358 f 0.024) x 1O-3 kg (average + STDEV). The FA
bridges had an average mass of (2.423 * 0.022) x 10m3 kg, a
mass increase of 2.8%, with almost overlapping errors. The
RE bridges were significantly heavier than the UN, with an
avcrage mass of (2.573 * 0.024) x 10e3 kg, 9.1% larger than
the UN bridges. The chemical treatments both appear to
change the mass of the bridges, and particularly so for the
RE bridges.

B. Resorcinlformaldehyde treatment
The resorcin/formaldehyde
bridges were treated by soaking
6 bridges in a 3% aqueous solutions of resorcin for 10 days,
air drying for one week and then vacuum drying at WC for
24hrs. After conditioning at 2rC and 65% RH for 7 days,
the bridges were subjected to an uncatalyzed
paraformaldehyde
treatment in a glass vessel for 24hrs at
12or [3].

2.9
2.8
2.7
s
g 2.6
b
5 2.5
g 2.4
ii
2.3
2.2
2.1
0

Bridge Number
Figure 1 - Bridge blank mass vs. treatment for UN, FA and
RE bridges.

Each bridge was suspended from both feet and the heart
shaped center cut with light-weight thread. Rubber bands
were used to decouple vibrations of the suspended bridge
from the frame. Small foam fingers were used to dampen
the large magnitude, low frequency OP motions. The
excitation was provided by an electro-mechanical device
attached to a PCB 086C80 force hammer [4]. Among the
hammers features, there was a choice of using a manual trip
or an external pulse for an automatic trip. The latter was
generally chosen and the hammer was triggered by a
rectangular pulse from a signal generator. The metal
hammer tip was slightly blunted. This modification gave
the advantage of the cleaner spectrums at higher frequencies
observed with metal tips without damaging the bridge. The
hammer signal was channel X in the FFT analyzer. Channel
Y input was from a 6 mm Radio Shack elecuet lapel
microphone used as the response transducer. The
experimental setup can be seen in Figure 2.

X
Input

Y
Input
Figure 2 - Apparatus setup for free-free violin bridge blank hammer-impact modal analysis with microphone vibration
transducer.

829

For data acquisition, a CSI WAVEPAK dual m signal


analysis system housed in a Z-slot expansion crate on a
COMPAQ III computer was used. The system setup for the
modal analysis used 800 lines over the range 0 - 20 kHz.
Each measurement averaged five hammer strikes in a
uniform window with delay and a trigger set at -2% of the
maximum force signal.
The IP damping measurements used 800 lines over 5-12
kHz; the OP measurements used 800 lines over 5-15 kHz.
The 6 mm microphone was placed as close as possible
without being struck by the bridge during the rebound from
the hammer strike (-3 mm). Microphone position was 82
for IP and 46Y for OP measurements. The damping
analysis used one excitation point in each plane; 1X for the
IP, 36Y for the OP measurements.

Figure 3 Bridge test points: X, Z labeled points for IP


measurements; Y points (shown as dots but numbered only)
for OP measurements are perpendicular to page.
Due to the extreme lightness of the bridge blanks it was not
possible to make reliable hammer strikes for OP
measurements along the very edge of the blank.
Consequently only interior points were used for these, hence
the bridge geometry used two sets of test points. For the IP
set, there were fourteen points defined, versus seventeen
points for the OP set, as shown in Figure 3.
V. Microohone Response
The mass loading of any accelerometer presently available
would be unacceptable on these low mass blanks, hence a
microphone was chosen as the motion transducer. At close
range, microphones have proved to be an acceptable
substitute on most objects for uncalibrated FRF
measurements and were adequate for the purposes of this
study [5,61. With such a small object near-field radiation
effects would be expected, particularly for the IP case where
the bridge foot is not as wide as the microphone. To
investigate this the microphone was placed in the IP mode
measurement configuration again; the bridge was impacted

at DOF 5X and its response measured at DOF 82 with micbridge separations varying from 2 mm to 16 mm. At 2 mm,
the real and imaginary parts of the response spectrum had
mixed characteristics, i.e. at some times, the real part looked
imaginary and vice versa. At 16 mm, the real and imaginary
parts both appeared relatively normal. Unfortunately the
spectrum at 16 mm was considerably noisier, as expected.
Going from 2 to 16 mm the pressure magnitude dropped to
0.223 of the original value for the mode at 6.25 kHz and to
0.101 for tie mode at 10.7 kHz.
VI. Relative Humiditv Effects
When working with wood, humidity is always of
importance. Because the measurements were taken in the
winter and spring, the relative humidity was low. During IP
modal analysis the RH was always 34%; for FA and RE OP
modal analysis the RH was 34% while the UN bridge was
measured at 48% RH. All the damping measurements were
made at 31% RH. A humidity effect comparison on IP
modes at 34% and 48% RH (with a 3 week interval to
stabilize changes) was made on one bridge with each
treatment. The UN bridge showed the largest frequency
drop of 0.9% as humidity increased, while the RE bridge
showed only a 0.4% drop. Some small differences between
spectra were observed but these could fall within normal
reproducibility tolerances for a non- rigid positioning fixture
such as shown in Figure 2.
VII. Data Analvsis
SMS STAR was used to analyze all the FRF data. The data
came in two separate Sets - the first for the complete modal
analysis on one each of the three differently treated bridges
and the second for the damping measurements made at one
IP or OP point involving all twenty bridges. For intercomparison between bridges the Magnitude2 IP and OP
Modal Peaks spectra from STAR are shown in Figure 4. It
was impressive how little crosstalk there was between the
IP and OP modes as seen in their respective measurements.
The microphone was easily able to give reliable frequency,
damping, and relative behaviors among modes, but not good
calibrated results [5,61.
Generally, for good quality FRFs Global curve fitting was
used to extract modal shapes. Using the Modal Peaks
spectrum for the entire bridge as a guide, the Global
Frequency and Damping values were determined for the
entire data set. The Global Residues of each peak were
extracted using the frequencies and dampings already
calculated in the previous step, and each tit was observed
visually for quality. At the high frequency end of the scale
typically the FRFs were noisy, so only FRFs with acceptable
peaks were selected for the Global Frequency and Damping
analysis. A Polynomial tit was used as a last resort when
the Global method gave unsatisfactory residue fit results.
On the other hand statistical information for the damping
parameters was always extracted from each bridges FRF

RESULTS
I. Mode Shw
In terms of importance to the violin during playing it would
be expected that the IP bridge motions would be most
important, however OP motions were studied to look for
effects here also. The UN bridge was used as the standard
to which the FA and RE bridges were compared. The
modes found in the UN bridge were labeled A, B, C, etc.
with increasing frequency. Corresponding modes in treated
bridges were then labeled accordingly. The UN bridge IP
mode MAC values for adjacent modes were all < 0.17, and,
except for MAC values between the weak mode B at -8.6
kHz and A (FA) of 0.28, or C (RE) of 0.52, all were below
0.13 The mode shapes are shown in Figure 5 (contour plots)
for IP vibrations and Figure 6 (delineated by nodal lines) for
OP vibrations for UN, FA and RE bridges. Generally, good
agreement was observed for mode shapes among rhe various
treatme&..
IP modes were characterized as having combinations of two
top motions, rocking (side to side) and breathing (entire
top expands/contracts), and two foot motions, jumping
(both feet in phase) and stamping (out-of-phase). Mode A
had a rocking top and stamping feet, mode C had H
breathing top and jumping feet, and mode E had a rocking
top and jumping feet, as seen in Figure 5.
2000

7000

UN

12000

F A

RE

W-W
Figure 4 - Modal peaks spectra for UN (solid heavy curve),
FA(solid light curve), and RE (short dash) for IP modes
(upper group) and OP (lower). (OP magnitudes divided by
10 to separate curves)

using the Polynomial tit function. Each peak was tit three
times, nominally at 112, Kith and 1110th maximum (for IP
modes only also over the entire band including modes A and
C). To understand the influence of fit band on the frequency
and damping these analysis results were used to calculate tit
means and standard deviations (STDEV). Average
STDEVlmean values were 0.4% for frequency and 3.2% for
damping.

Figure 5 In-plane mode shapes for UN, FA and RE


bridges.

Once the intra-treatment fit consistency was understood,


means and STDEV were then calculated for every
reasonably strong mode for each treatment to understand
whether the observed variations arose from the natural
variability of wood or from the chemical treatment. Some
modes did not exhibit good fit results due to being weak,
noisy, or highly coupled, leading to large standard
deviations in the damping parameters. If they were too
difficult to fit consistently they were not included in the
table. A complete listing of fit results for qualifying modes
was collected in Table 1.

Some modes observed in this study appeared complex in


animation. Those modes appearing complex that are listed
in Table 1 are shown with bold frequency values. The UN
and RE bridges were identical with respect to which modes
were complex and which were normal. IP, mode A, in these
two bridges, was normal while modes C and E were
complex. The FA bridge differed somewhat; mode A
became complex with this treatment, mode C became
normal and mode E remained complex.

831

Table 1 - Average in- and out-of-plane frequency &Hz) and damping (% critical) results for modes below 11 kHz for 8 UN, 6
FA and 6 BE bridges. (Bold frequency value indicates complex mode - see text.)
In-F-hoe
MC&
A
C

Mock%

A
B

Mean

SIDEV
0.020
0.015

6.261
0.860

0.017
0.017

6.416
0.858

to.659
0.798

0.025
0.009

10.519
0.838

0.032
0.017

10.808

0.045

0.869

0.024

f
r

2.648

0.007
0.010

2.636
0.537

0.009

2.640
0.556

0.013

0.021

2.967
0.472

0.017

0.008

3.023

O.MO

3.008
0.349

0.028

0.480

0.013
0.038

f
c

4.730

4.722
0.456

0.045

0.022

4.714
0.514

0.028

0.583

0.021
0.009

5.674
0.688

0.028
0.013

5.670
0.676

0.021
0.012

5.655
0.755

0.072
0.014

7.724
0.828

0.037
0.019

7.790
0.798

0.043
0.020

7.736
0.849

0.035

9.116
0.848

0.032
0.083

9.115
0.854

0.025
0.023

9.106
0.977

0.042
0.074

Modes:

f
c

J/K

SIDEV

0.010
0.009

Mean

0.870

BE

SIDEV

0&.+Phae

Ei

Parameter Mean
f
6.300

f
5

0.599

OP mode JiK at - 9.1 kHz occurred as one mode in the UN


and FA bridges but split into two distinct modes in the RE
bridge. The OP modes J/K appeared complex in all the
bridges. Only the RE bridge FRFs displayed modes J and K
as two distinct peaks. Although nodal lines are not well
detined for complex modes - hence Figure 6 has different
nodal hatching patterns for the J/K modes - none-the-less
nodal lines were drawn in Figure 6 to guide the eye to
regions showing little motion and a fairly clear phase
reversal between adjacent eigenvectors.
II. Mode Freauencv and Damoing
A. In-plane
Three strong distinct modes, labeled A, C, and E , were seen
in all the bridges. Analysis results for IP mode E at -14.5
kHr, due to spectral noise, coupling with other modes, and
possibly two almost degenerate peaks, were not considered
reliable enough for intercomparison. Significant frequency
changes between UN and FA or RE bridges were observed
in the IP mode C at -10.6 kHz of -1.4% (FA) and

0.010

0.011

0.008

+1.4%.(RE). Significant changes in damping were seen for


the RE IP mode C at -10.6 kHz.
B. Out-of-plane
Six (possibly seven) reasonably strong OP modes A, B, D,
F, I and In< - found below 10 kHz were also studied for
damping purposes. The excitation point chosen for this
damping analysis appeared to very weakly excite the J mode
in the JiK doublet in the RE bridge. Fits to the J/K region at
-9.10 kHz with two peaks were definitely not as reliable as
most others and the damping value errors reflect this. This
region probably contained a doublet in all bridges although
only the RE bridge showed two distinct peaks. A complete
listing of the OP fit frequency and damping results are
presented in Table I. A comparison of UN with FA or RE
bridges indicates the OP mode at -3.0 kHz had frequency
increases of 1.4% for FA and 1.9% for RE bridges.
Compared to the UN results, significant damping changes
were observed for the FA OP modes at 2.64, 3.01 and 4.72
kHz, and the RE OP modes at 2.64,4.71 and 5.66 k?lz.

832

UN

FA

observed feet motions did not match these. However,


predicted motions of the top were similar to our
experimental results: FE Modes I and III, corresponding to
our A and E respectively, had rocking tops, while mode II
(C)had a breathing top.

RE

A:

D:

Neither the IP nor OP frequency or damping analysis results


for the chemically treated bridges showed major changes
from the untreated bridges for any of the reasonably strong
modes listed in Table 1; the maximum frequency variation
was 1.9%, the maximum damping variation was 26% (but
this was for OP mode B, coupled with Mode A see Figure
4). Perhaps the changes we observed were sufficient to
cause the acoustic changes noted [2] when these bridges are
fitted to a violin; perhaps the changes are amplified when
the bridge is fitted and trimmed to the optimum for the
individual violin.

F:

J/K*:

M:
Figure 6 Out-of-plane mode shapes for UN, FA and RE
bridges. (* on J/K denotes complex mode see text.)
DISCUSSION
As expected the microphone was able to provide consistent
frequency and damping values in these bridge
measurements, and the mode shapes appeared reasonable
without being definitive. When FE calculations are done for
the violin bridge with free-free boundary conditions it will
then become possible to judge the quality of mode shapes.
The presence of traveling waves on animation of mode
shapes might be an artifact of the response transducer.
Clearly, use of another response transducer such as a laser
vibrometer would have been desirable in this experiment to
achieve calibrated results.
On the other hand the hammer-impact modal analysis
technique has shown here its adaptability to- and capability
for non-loading excitation of very small structures; this is
an extremely useful property for working with musical
instruments such as the violin where the effects of small
mass loading at critical places such as the bridge can he
immediately apparent acoustically.
It should he noted that FE analysis for a fitted bridge, not a
bridge blank as here, and with boundary conditions of
anchored feet and the string contact areas, has found three
major IP modes below 10 kHz [71. Not surprisingly our
experimental values for the frequencies of the modes and

However, there is another parameter here that we should


consider - the bridge stiffness. We noted that the RE
chemical treatment increased the mass of the bridges (Fig.
1) by an average of 9.1%. hut the IP mode frequencies
actually increased a little under 2% with this treatment.
Qualitatively the relationship 0 = Jk/m predicts an
increase in the stiffness of the material on the order of the
mass increase. The increased mass and stiffness of RE
bridges together might he sufficient to cause the changes
noted in sound produced by violins fitted with these bridges.
We wish to thank Dr. Yano for graciously providing a large
collection of treated and untreated bridge blanks for us to
measure.
REFERENCES
1. F. &wart On Musical Instruments. MusicaJ Acoustics, Pan 2:
Violin Family Functions, pg. 17. Dowden, Hutchinson and
Ross, Inc., Stmudsburg, Pennsylvania, 1976.
2. H. Yano. Improvement of the acoustical and hygroscopic
properties of wood by a chemical treatment and application to
the violin parts, J. Acoust. Sot. Am., 9.2 (3) September 1992.
pp. 1222.1227.
3. H. Yano. private cmmmmication.
4. K. Jacobs, Comprehensive Invesrigafion of the violins
Acourricdly Impon~r Corpus, Cavity, and Corpus-Cavity
Modes of Vibrarion Below 2 kHz. Masters Ibesis, East

Carolina University. Greenville. NC, 1993.


5. G. Bissinger and M. Chowdhury, Comparison of Hammerimpact modal analysis measurements with cardioid and
omnidirectional micmphone in the near- and far-field
regions, Pmt. 10th Intern. Modal Analysis Conf., 1992, pp.
394-399.
6 G. Bissinger and M. Chowdhury Comparison of modal
analysis measurements with microphone and accelerometer on
hammer-impacted structures, Proc. 8th Intern. Modal
Analysis Conf., 1992, pp. 667.672.
7. 0. Rodgers and T. M&no. The effect of wood removal on
bridge frequencies, Catgut Acoust. Sot. I., Vol.1, No.6
(series It) November 1990, pp. 6-10.

833

Anda mungkin juga menyukai