68

What secrets lie beneaththe surface of violins bythe great del Ges?Here, Jeffrey S. Loen,Terry M. Borman andAlvin Thomas Kinguse the latest techniquesto reveal the hiddencontours of wood density

The Strad September 2005

ABOVE mapping the graduations of an instrument such as the c.1730 Kreisler del Ges can begin to yield its acoustical secrets

Path through

068-75 Del Gesu 7/27/05 3:51 PM Page 68

Familiar iconic images of old Italianviolins show only the outsidesurfaces of remarkable three-dimensional creations. We naturallywonder what lies beneath thesurface of a great Guarneri delGes? Is the wood light or heavy?How do the graduations pinch and swell beneath the smoothlymodelled arching and the 270-year-old patina?

In our research, we have usedtwo methods of revealing hiddenfeatures: CT scans and thicknesscalipers. Our CT scans showchanges in shape and wood density.The separate colour contour thick-ness graduation maps make itpossible to compare plate structuresof different instruments visuallyand intuitively.

Thinking about wood thicknessas shown in maps is becomingfamiliar to violin makers. However,thickness without density meansnothing to the working luthier. A light piece of spruce will need to be thicker than a heavy one. In addition, a material withuniform density, such as willow or carbon fibre, does not begin to approximate the acoustical prop-erties of spruce. In our opinion, the

characteristics of spruce (ie softspring growth and hard fall growth)contribute significantly to theacoustical properties and strengthsof individual logs. For that matter,individual sections of logs on asteep hillside will develop differ-ences in the uphill part comparedwith the downhill section. SeeFigure 1 for two examples of sprucegrain in del Ges top plates.

CT SCANSWe have been investigating dimen-sions, design and wood densityusing computed tomography imag-ing technology, commonly knownas CT scans. In CT imaging, X-raysprojected through an object arereceived on a processing plate orreceptor array. The differencebetween the amount of energy sentand received is referred to as atten-uation. Denser objects absorbmore of the energy transmitted andappear lighter, whereas less denseobjects appear darker. The scaleused is the Hounsfield Unit (HU)whereby -1000 represents the atten-uation of air and zero represents theattenuation of water. This is easilyconverted to the standard formatfor density, which is the ratio of an

69

Photo

op

po

site: Stew

art Po

llens, c

ourte

sy of Pe

ter Bid

dulp

h/Giuse

pp

e G

uarne

ri de

l Ge

s; all o

ther im

ag

es c

ourte

sy of the

autho

rs

The Strad September 2005

objects mass to volume (air = zero;water = 1.0g/cm3).

CT has advantages over standardstudy techniques, which ofteninvolve handling and disassem-bling of instruments. It isnon-invasive, rapid and accurate. A typical CT scan of a violininvolves removing the instrumentfrom its case, placing it on a paddedsupport in the machine for twominutes and then returning it to itscase. In that time, the machineacquires as many as 1,500 profilesthrough the instrument. Theseimages can be displayed as trans-verse or longitudinal profiles, or as three-dimensional models (see figure 2, overleaf).

The transverse CT profile showsthe alternating lightdarkspringfall wood-grain structure inthe spruce top, and the morehomogenous nature of the mapleribs and back. We have measuredthe density of each grain line (seedensitometer graph, figure 3). Thesedensities have then been averagedto arrive at a mean spring and fallgrowth, and a springfall ratio hasbeen calculated. For the 1735Guarneri top shown in figure 1,the springfall ratio is 1:1.275

the woods

Figure 1 photographs of spruce grain in two 1735 del Ges tops

068-75 Del Gesu 7/27/05 3:51 PM Page 69

70 The Strad September 2005

and the overall density is -633 HUor 0.368g/cm3. Hence, the fall wood is almost 30 per cent denser thanthe spring wood, although the overall density is relatively light.Another scanned Guarneri top, alsofrom 1735, has a lower springfallratio of 1:1.241 and a heavier overalldensity of -619 HU or 0.382g/cm3.In comparison, a 1736 Stradivariviolin top shows a springfall ratioof 1:1.153 and an overall density of -574 HU or 0.427g/cm3, which isrelatively heavy.

We believe that these preliminarydata are simply an affirmation ofthe ancient technique of testingwood by dragging ones thumbnailperpendicular to the cross-grain ofspruce: winter grains create a wash-board effect depending on theirhardness, spacing, width and rela-tionship to summer growth. Thistechnique also gives an idea of

general density. Developing this sortof feel with fingers contributesgreatly to a luthiers choice of wood,and approximates the results of CTdensitometry. Further CT work willcompare the wood choices of classi-cal makers in other areas to thesebaseline examples as well as woodchosen by modern makers.

THICKNESS MAPSWe have compiled over 4,500 thick-ness measurements from 46 violinsmade by del Ges. Eight instru-ments, all outstanding examples,are shown here (see figures 5 and 6overleaf). Thickness values weremeasured using magnetic thicknessgauges. Data compiled from various sources (violin shops, Strad posters and museums) were systematically contoured and coloured using a computerisedgeographic information system

(see our article on Stradivaris platethicknesses in The Strad, December2002, for details of how these mapsare produced). Note that the mapsare based on the instrumentscurrent condition and we do notknow if thicknesses are original.Were also unable to say whetherthe maps depict original wood or multiple layers, as in patches. The resulting maps, coloured in anintuitive way (hot colours for thick,cold colours for thin; see figure 4 fora colour bar) allow us to comparethe overall structures of the plates.Positions of data points are shownby black dots, although values areomitted in order to emphasisestructural patterns. All plates areviewed from the outside.

The earliest violin depicted,made in 1728, shows zones of minimum thickness between the f-holes. The centre-of-back

Figure 4 colour bar used in wood thickness graduation maps

Figure 2 transverse CT profile of a 1735 Guarneri del Ges at the widest section of the lower bout. Note end of bass-bar (within body) and part of tailpiece (above body)

Figure 3 CT densitometer graph for the 1735 del Ges top shown in the CT scan image in figure 2. Peaks represent harder fall growth and troughs represent softerspring growth. Higher densities at the two extremities are due to purfling, glue infiltration and an edge underlay on the right-hand side

068-75 Del Gesu 7/27/05 3:52 PM Page 70

thickness is slightly lower than thecentre of the plate, and the thicknesspattern is somewhat square.

The 1730 Szymon Goldbergviolin again shows a zone of minimum thickness between the f-holes, but both the top and backplates are more variable than the1728 violin. The centre-of-backthickness is irregular and suggests a linear pattern.

The top plate of the c.1730Kreisler violin is thinner in thecentral areas than near the edges,block platforms and f-hole exteri-ors. A zone of minimum thicknesssweeps diagonally through the f-hole region, from the upper left to the lower right. The back, like the top, has a highly asymmetricalthickness pattern. The back plateshows a rather small central thickness zone running parallel to the centreline.

The 1734 violin again showsa diagonal zone of minimum thick-ness that passes through the f-holeregion (a soundpost patch hasincreased the thickness near theright f-hole) although this timefrom the upper right to the lowerleft. The back plate is stronglyasymmetrical, including the posi-tion of the centre-of-back thickness,which is shifted significantly to the bass side.

The c.173940 violin has a ratheruniformly graduated top. The backplate has a prominent oval zone ofgreater thickness centred near thehalfway point.

The 1742 Alard violin has auniformly graduated top plate. The back plate is rather symmetri-cal and the centre of thickness isnear the halfway point.

The 1742 Lord Wilton violin isgraduated quite uniformly, exceptfor one slightly thicker valuebetween the upper f-holes. This topplate is rather thin in the centre andattains maximum values just insidethe linings of the upper bouts.The back plate shows a large trian-gular zone of increased thickness.The centre of thickness seems to behigher than the halfway point.

Paganinis 1743 Cannon isthought to be in a nearly unmodifiedcondition. The top plate is especiallythick between the f-holes and theback plate is extremely robustthroughout its length. Regrettably,data do not agree from the threetimes that the Cannon has beenmeasured. Candi collected meas-urements along the centre-lines ofthe top and back plates in 1937,using a caliper on free plates(see The Cannon and TypicalFeatures of Guarneris Instrumentsfrom Paganinis Violin by AlbertoGiordano, 1995). Candi documenteda thicker top plate (max. 4.3mm)and a thinner back plate (max.5.4mm), compared with morerecent attempts (top plate 3.4mm;back plate 6.2mm). We have no

73The Strad September 2005

explanation for these differences.A definitive measurement isneeded, in which sufficient datapoints are collected using a well-calibrated gauge, before we canmake firm conclusions about thethickness of the Cannon.

Looking at these, and other mapsin the database, we can makegeneralisations about how theseviolins compare with otherCremonese masters and master-levelwork of modern violin makers.Firstly, most plates are slightlythinner than modern makers aretaught to use. Secondly, the rangebetween thick and thin is greaterthan we are often led to believeis acceptable. Finally, plates arequite irregular by todays standards,showing considerable pinching

Figure 5 graduation contour maps for top plates of eight violins. Positions of data points are shown by black dots, although values areomitted in order to emphasise structural patterns. All plates are viewed from the outside. See colour bar below for reference

1728 1730 Szymon Goldberg

c.1730Kreisler

1734

c.173940 1742Alard

1742Lord Wilton

1743Cannon

068-75 Del Gesu 7/27/05 3:52 PM Page 73

and swelling. Such variable patternsare typical of other Cremonesemakers (including Stradivari).

Top plates mostly have amembrane-like structure, a fewof which are slightly thicker inthe centre between the f-holes, chiefamong these being the Cannon;although most are thinner in thecentre than at the edges. Thicknessesare commonly in the 2.22.9mmrange, although outside marginsof f-holes and block platforms typi-cally are 3.03.5mm. The variabilityof adjacent values is moderate,often showing changes of0.20.6mm within distances of2030mm. Spots as thin as 1.5mmoccur in several cases, especially inearly work before 1730, althoughthe usual minimum values on topplates are 2.12.3mm. Severalviolins have known or suspectedbreast patches (a restoration tech-nique used to correct andstrengthen extensively repairedareas), which could also have beenexecuted because the plate wasjudged as too thin in the centre.

Back plates all have a concentricstructure, with a strong central zoneof greater thickness. Maximumthickness values in the central zonerange from 3.56.5mm (average4.8mm), and the shape of the centralzone is variable, including round,oval, rectangular, triangular andlinear patterns. The upper and lowerbouts tend to be of uniform thicknessin the range 2.03.0mm, althoughthin spots in the range 1.42.0mmare common, particularly near edges.

We will never know the originalcharacter of most of the violins bydel Ges (with the likely exceptionof the Cannon). One could speculate that they were originallyall like the Cannon. However, it is just as possible that the Cannonwas one of few violins that weremade exceptionally thick, whichmight explain Paganinis troublefinding similar Guarneris duringthe early 1800s. Perhaps the mostsurprising aspect of these maps isthat there is so much variation, andyet sound quality is consistently high.

This lack of correlation betweenthickness patterns and soundquality suggests to us that thicknessgraduations do not play as greata role as we thought relativeto acoustical performance.

In conclusion, plate graduationsappear to be highly variable in theviolins of del Ges. Substantialthickness variations and numerousgraduation patterns occur, and yet they retain their high level ofperformance; this despite possiblyextensive modifications over theyears. We think that perhaps themost significant observations comefrom the CT densitometry scans,although we need to scan moreinstruments to arrive at definitiveconclusions. We believe that we areone step closer to getting inside themind of perhaps the greatest violin

75The Strad September 2005

maker of all time, by taking intoaccount the densities of the woodthat he chose and getting a feel forhis preferential springfall ratios.

The c.1730 Kreisler and the 1742 Alard

violins appear in The Strad calendar 2006,

available from www.thestrad.com or by

calling +44 (0)1371 810433.

Thanks to Dr Ronald Glass and the faculty

and staff of the Mount Sinai School of

Medicine; and Angie Ackerman and Scott

Carlson at the McKay-Dee Hospital Center

for help with CT scans. CT scanners used

were Siemens Somatom Sensation 16.

Thanks also to Gregg Alf, Gary Frisch, Gary

Sturm and Peter Westerlund for help with

graduation data. Data for The Strad posters

of the 1742 Alard, the 1742 Lord Wilton

and the 1743 Cannon were compiled by

John Dilworth and Roger Hargrave.

S

Figure 6 graduation contour maps for back plates of eight violins. See colour bar below for reference

c.173940 1742Alard

1742Lord Wilton

1743Cannon

1728 1730 Szymon Goldberg

c.1730Kreisler

1734

068-75 Del Gesu 7/27/05 3:52 PM Page 75