Silk Oud Strings - Making Sense of the Historical Data

The attached images are of a modern replica of an early wound silk Chinese qin (zither) string (I don't know who made it). The string is worn out so the windings loose and separating but, nevertheless, it gives a good idea about the string construction.
It can be seen that the big difference between a modern wire wound string and the early Chinese silk wound strings is that the silk winding of the latter is flattened (like a tape) producing a smooth surface as well as adding mass to the string.

The string wrapping tool used by the early Chinese string makers is represented by the attached sketch - a copy of a low resolution image of the original wood cut for clarity.
The tool is carved from a heavy hardwood (with a metal weight added if necessary (for added inertia). The wrapping material (made of six threads) is wound onto a bobbin and is fed between two round metal rods (mounted closely together) in order to flatten the wrapping as it is wound onto the string core - very important according to the early Chinese texts "as everything depends on this". (The wrapping must not be round in section).

In practice, the string core - about 12 feet (3.7 metres ) in length - is stretched tight between two posts with the wrapping tool riding on the string. The tool is then made to rotate around the string by shaking the string (like a skipping rope) causing the wrapping to be wound tightly over the string core.
Clearly, making a wound string with this simple tool must have been a highly skilled operation in order to achieve the required dimensions and tension of the wrapping to produce a string that was "uniform without scars".

The ancient Chinese were familiar with geared drives so one has to wonder why they did not wind the strings on a machine similar to those used by modern string makers? Possibly because the winding had to proceed relatively slowly and exactly to ensure a uniform, smooth string surface and avoid over stressing and breaking the wrapping?
(The early texts provide instructions on how to repair the wrapping in the event of breakage).

The early Chinese texts and translations can be found on John Thompson's website at

http://www.silkqin.com

Making the basic liquid glues using acetic acid (vinegar) is fairly straightforward.

Hide glue and gelatin are sold in various strengths (known as Bloom strength) - the higher the Bloom number the stronger the glue and the quicker it will set or 'gel'. A strong granular hide glue suitable for woodworking would be about 260 Bloom strength. Pearl hide glue has a lower Bloom strength and food grade gelatin has a higher Bloom strength.
The greater the Bloom strength the more acetic acid is required to make a glue mixture liquid at room temperature.
The vinegar strength used for these trials is household food grade - strength 5% by volume. Presumably if vinegar of a higher strength was used, less would be required.

The food grade gelatin used for the trials is made by "Knox" - readily available from local grocery stores. This gelatin is derived from bone marrow.

The first step is to add sufficient vinegar to cover the dry glue that is then left overnight for the liquid to be fully absorbed. Gelatin will absorb a much greater volume of liquid proportionally than hide glue.
The glue mixture (in a suitable glass container) is then heated on a water bath maintained at a temperature of 140 F (60 C) until it becomes fluid (do not exceed this temperature or the glue will spoil).
Next add more vinegar little by little - stirring the glue - until on cooling the glue does not gel but remains a liquid. This is a trial and error procedure so if the glue starts to gel on cooling bring it back up to temperature, add more vinegar, and repeat the process as necessary.

The liquid glue made for the trials gelled at room temperatures below about 70 F (21 C) but was easily liquified again by warming the glue container in a pan of warm water. Probably by adding a little more vinegar the tendency to gel would have been eliminated.

The attached image compares the liquid glue samples. The gelatin makes a clear, transparent glue.

The strength of the glues was tested subjectively by gluing together pieces of cotton cloth and then trying to 'peel' them apart after the glue had dried. Both glues were judged to have a satisfactory peel strength.

The purpose of the glue binder on silk strings - made from continuous threads of degummed reeled silk - is to replace the sericin gum, to cement the individual filaments together and to add mass.

Likewise for strings made from degummed spun silk yarn. In addition the glue increases the tensile strength of the twisted string assembly by cementing together the intertwined short fibres of the spun yarn preventing longitudinal slipping of the fibres when under tension.
In this case it is the shear strength of the glue binder that provides additional strength after the yarns are brought into intimate contact by being twisted together.

For information, the attached image shows a section of a single thread of raw reeled silk as seen under a Scanning Electron Microscope at X1300 magnification (the image is from the 'Silk Reeling and Testing Manual' by Yong-woo Lee and is reproduced here by kind permission of the Food and Agriculture Organisation of the United Nations). I have added the text for clarification.
There is no scale on the image but the 'diameter' of each Brin filament would be about 5 -10 microns so this thread, made from, say, 16 individual fibroin filaments might measure about 30 microns in diameter (or 0.03 mm). Each Brin is roughly triangular in section enabling the filaments to 'nest' closely together, cemented as a whole by the Sericin gum as a matrix.
The many threads required to make a string are forced into even closer contact when twisted together.

The properties and shear strength of the liquid hide glue used in these trials is not at present known but each string will eventually be tested to destruction to determine relative tensile breaking stress.
Commercial liquid hide glues (such as the 'Franklin' brand) contain additives that make them weaker than hot hide glues at increasing levels of relative humidity. For example trials run by Susan L. Buck, Art Conservation Fellow, University of Delaware have shown that at 50% RH both liquid hide and hot hide glues were similar in shear strength with a mean value of 4174 psi and 3640 psi respectively (but Bloom strength of the hot hide glue was not specified) but at 85% RH the values dropped to 1143 psi and 2637 psi respectively. For this reason the Franklin company does not recommend using liquid hide glue for applications above 75% RH - not a problem for restoration of Museum artifacts stored under controlled humidity conditions of around 50% RH but luthiers beware!

The performance of the various slower drying, flexible liquid glue binder formulations to be used for these string making trials remains to be seen.

Preparation of the yarn bundles for twisting is a straightforward operation taking only a few minutes using the string winding apparatus details of which have been posted earlier in this thread.
To facilitate handling of each yarn bundle the yarns are secured to metal hooks (for mounting on the twisting rig) using waxed thread (waxed thread does not slip). Once secured in this manner each bundle may be handled without the yarns tangling.

For string #1, after mounting on the twisting rig and loading with a weight of 1.84 Kg, the liquid glue was applied to the yarn bundle with a cotton cloth and then the bundle twisted 175 turns. Surplus glue was wiped off with a cotton cloth and the string allowed to dry on the rig for 24 hours.

To ensure maximum liquidity of the glue, the glue container was set in a pan of warm water for a few minutes before being applied to the yarn bundle. For strings #2 to #8 (now completed) it was found easier to place the yarn bundle on a plastic sheet, pour a small quantity of glue onto the bundle and then work the glue into the silk with a small artists paint brush. This ensured full saturation of the bundle with glue.
Each string was then twisted 175 turns with an initial load of 0.4 Kg, all surplus glue wiped off with a paper towel and then the full load of 1.84 Kg applied. After a final wipe with a cloth, each string was allowed to dry for 24 hours.

The number of turns and loadings are a best guess at the present time so may be subject to modification as experience is gained.

All strings are uniform in diameter and fairly smooth but with some slight roughness to the touch due to some small hairs of silk that were not smoothed down by the glue. Some additional finishing operation - yet to be determined - may be required to correct this situation.

String #1 was partially untwisted at one end to confirm full saturation of the string by the glue binder (note the little hairs of silk that have not been stuck down by the glue). It can be seen from the attached image that the string is very flexible.

Now for testing the strings.

The first test is for a Mathna string, 0.6 mm diameter made using a flexible hide glue/acetic acid binder.

All strings will be tested on one of my lutes - a reconstruction of an early 16th C bass lute by Laux Maler that I made in the 1970's. String length is 73 cm with six courses - the bottom three courses being octave tuned. The lute is currently fitted with 'Pyramid' nylon lute strings in 'D' tuning. The nylon strings allow the pitch to be raised to 'E' tuning, however, 'D' (at A440 standard) is the proper pitch for all gut and silk stringing.

The test string has been fitted at the sixth course position and tuned to g (196 Hertz frequency at a'=440 standard). The 5th course treble in plain nylon (0.6 mm diameter) is also at the same pitch with the overspun bass an octave lower (G 98 Hertz) so the performance of the silk test string with can be directly compared with modern nylon.

String #1 came up to stable pitch quite quickly and has been held at this pitch for 7 days at about 60% Relative Humidity. Assuming silk strings are similar to gut, the tension of the string at this pitch (using a 'Pyramid' string calculator for convenience) is about 3.2 Kg.
By comparison the nylon strings are at a tension of 2.4 Kg.

The attached sound file (compressed to MP3 so there are some significant audio losses) gives an idea of the relative performance the sequence - plucking the strings with a fingertip - is nylon overspun bass G then silk g and finally plain nylon g.
Audible sustain for all the strings was about 7 to 8 seconds.

The next step will be to raise the pitch of all strings by a semitone to g#208 Hz to see what happens. String tension of the silk string is about 3.6 Kg at this pitch. The tension of the nylon strings will increase to about 2.7 Kg which is about their optimum tension.

The sound clip was made with a Zoom H2 digital recorder positioned about 60 cm from the lute sound board.



Attachment: String #1 finger tip.mp3 (375kB)
This file has been downloaded 478 times

The new string test rig is now essentially complete apart from a few minor additions.

Made from scrap/recycled materials it is comprised of a solid pine wood frame on which is mounted a simple sound box made from Yellow Poplar with a Sitka Spruce sound board. The bridge is guitar style with ebony tie block and bone saddle. The nut can be moved to provide a vibrating string length ranging from 50 cm to 75 cm. Each test string passes from the nut over a plastic pulley where a weight is applied to provide a measured string tension (ranging from 1.5 Kg to 4 Kg).
Between the nut and pulley each string passes over a small block of UHMW (ultra-high molecular weight) polyethylene - a dense rigid plastic with a very low coefficient of friction. Not shown in the attached images of the rig is a simple clamp device that will lock each string to the plastic block during testing.
The sliding nut will also be locked in place with a screw during testing (yet to be added). Although the nut position is infinitely variable, strings will be tested with vibrating string lengths ranging from 50 cm to 75 cm in increments of 5 cm for convenience.

After nearly 3 months, have finally found time to run some tests on the first batch of strings made with 120/2 de-gummed spun silk, using acid based, hide glue binders - as previously reported.

The acid based binders have proven to be resistant to bacterial attack without requiring refrigeration - unlike untreated hide glue or the experimental alkaline based glue binders that both began to decay after a month.
The acid based glues are also flexible when dry so will be used exclusively for further experimentation.

Preliminary comparative testing is being undertaken on the string test rig in order to measure string sustain and linear density of each string. The 'loading' of a string (due to weight added by the binders) is calculated indirectly by measuring the frequency of vibration and, knowing string length, diameter and string tension, using the Mersenne-Taylor relationship for vibrating strings to determine the linear density. Alternatively, weighing each string can only be achieved using an accurate (and costly) analytical balance.

Each string when mounted and loaded with measured weights on the test rig was sounded with a leather 'pick' and the frequency of vibration measured with a low cost ($10) but effective digital tuner. The sustain of each string was also measured from the instant a string was sounded to the point at which the tuner pointer returned to zero.
The attached image shows the setup. Here the tuner is indicating a pitch of about C sharp plus 20 cents at A440 standard - equivalent in this case to a frequency of nearly
140 Hz.

Preliminary results indicate that the glue binder results in a modest loading of the silk from an untreated silk linear density of about 1.2 gm/cc increasing to an average of about 1.35 gm/cc (measured range 1.25 to 1.45 gm/cc) due to the glue binder composition and loading applied during twisting of the string.
For comparison, the linear density of a plain mono filament nylon string (Pyramid) was measured on the test rig and calculated to be about 1.14 gm/cc.

String vibration sustain for string lengths between 55cm and 70cm - plain nylon and silk - was relatively consistent, measured at 5 to 6 seconds duration. String tension on test ranged from 1.5 Kg to 4 Kg in 0. 5 Kg steps and string length from 70 cm to 55 cm in 5 cm steps.
String diameters initially tested were in the range 0.6 mm to 0.66mm (Mathna) and 0.9 to 0.92 (Bamm)

The increasing of string loading due to the glue binder more than compensates for the weight loss in the silk due to removal of the Sericin in the de-gumming process. Nevertheless, a more significant (and desirable) increase in loading - particularly for bass strings - will necessitate a different approach.
More to follow.

It has been necessary to modify the string test rig because of inconsistent measurements. The main problem was that, due to considerable friction in the pulley assembly and at the nut, the tension in the string was less than the measured dead weight. This has been corrected by :
1) dismantling the pulley assembly, cleaning and lubricating the bearings so that the pulley wheel freely rotates.
2) raising the pulley wheel level with the nut to eliminate any downward force of the string on the nut.
3) mounting the test rig at an angle of 45 degrees to further reduce loading on the pulley wheel bearing (and hence bearing friction by about 50%).

The modified rig has been 'calibrated' using 'Pyramid' monofilament plain nylon and 'Pyramid' gut with dead loads of 2.5, 3.0, 3.5 and 4 kg. This has returned consistent results for calculated string density of 1.1 gm/cc for nylon and 1.3 gm/cc. These are reasonably accurate results - to the first place of decimal - which is about as close as can be expected given this relatively unsophisticated apparatus.
As the objective is to compare the relative densities of the experimental silk strings under test, this level of accuracy will be good enough.

The attached image shows the procedure for calculating string density from the test rig measurements - derived from the basic Mersenne-Taylor Law for vibrating strings.

Now that the modified rig has been satisfactorily calibrated, the first batch of silk strings will now be retested to determine the calculated string densities.

Silk strings can be made heavier ('weighted) by the addition of the glue binder as well other materials that are more dense than the silk filament.

Silk will readily adsorb metal salts so the traditional weighting method used by the silk fabric industry is to immerse the silk in a bath of a metallic salt solution followed by immersion in another salt bath that will react to form an insoluble precipitate of the metal salt either in or on the silk filament.
Heavy metal salts of tin and lead were commonly used for weighting and reagents creating the desired precipitation can include organic materials such as gelatin and tannin.

The other method of weighting strings is to simply mix heavy powders with the glue binder - such as copper powder or oxides of iron.

Due to the toxicity of heavy metals (not a problem in earlier times - but lead and mercury are not an option here!), these string weighting trials will be restricted to the metals tin, copper and iron as well as the organic materials gelatin and tannin. A further restriction is that the weighting materials or chemicals must have been available to the ancients.

The procedure for making the heavy metal salt Stannous Chloride has been described in a previous post - easily made - and to be used in these trials.

Next to make fine copper powder - as an additive to the glue binder. Again this is easily made by reacting a Copper Sulphate solution with iron to form metallic Copper and Ferrous Sulphate.

Here a convenient source of iron is fine grade 'steel wool' readily available from any hardware store (the kind that is oil free - used for finishing painted surfaces). This is immersed in a saturated solution of Copper Sulphate (blue coloured) and allowed to stand until the reaction is complete. The heavy metallic Copper falls to the bottom of the container and the green coloured liquid remaining is Ferrous Sulphate (that will be used as another weighting chemical).
The copper powder is collected by straining the liquid though a fine filter (paper coffee machine filter) and after washing with water allowed to dry.
The dried powder is then ground up in a mortar and will be sifted through a fine sieve to remove any 'large' residual iron wire particles prior to mixing with the glue binder.

Another weighting material that will be tested is iron oxide - readily available and safe. Iron oxide comes in many complex compositions (such as 'rust' that forms on iron in damp conditions) and is widely used in commerce from making iron to paint pigments to cosmetics. It occurs naturally in minerals such as Hematite that contains over 90% of iron oxide. It is red in colour but other forms of the oxide are coloured black, brown yellow, green and blue.

The attached image shows a sample of 'red oxide' finely ground and sold as an artist's pigment to colour paint and varnish. The same stuff is also sold as 'Jewellers Rouge' - a very fine abrasive polish for gold and other metals.

Iron Oxide has a density ranging between about 4 and 5 gm/cc so is about 4X more dense than degummed silk filament. However as Iron oxide and copper powders will be added to the glue binders for added mass (and not adsorbed by the silk) the possibilty for significant increase in string mass is likely limited.

The most promising weighting procedure is likely to be by chemical adsorption - as employed by the silk fabric industries. Unfortunately there is little published information available on details of the procedure, chemicals used etc - probably because this is proprietary information (trade secret).

For this investigation the weighting chemicals to be tested will be tannin, copper sulphate, ferrous (iron) sulphate and stannous (tin) chloride - all chemicals (among many others) familiar to the ancients.
For weighting powders - glue binder additives - Iron oxide and copper powders will be tested.

From the batch #1 string tests the glue binders having the greatest elasticity are the acid based liquid glues modified with the addition of tannin (equivalent to the softening of tanned leather) combined with pretreatment of the silk fibre by soaking in a tannin solution.

The 8 batch #1 test strings (0.62 to 0.68 mm diameter) have now been subject to destructive testing to determine approximate Ultimate Tensile Strength.
Each string was simply loaded using a spring scale until it broke - repeated twice for each string. The results were fairly consistent regardless of the composition of glue binder to give an approximate U.T.S. value of 31,000 psi (2180 Kg/cm2).
This is about 40% of the value reported by the textile industries for raw filament silk. The reduced value may be due to the use de-gummed spun silk rather than continuous filament raw silk, the degree of twist in the fibres of the string and the acidity of the glue binders - all of which may contribute to a reduction in string breaking strength - particularly the twist factor (more twist = less tensile strength).
These results indicate that strings made from spun silk may not be suitable for making the top two treble courses - for which the stronger filament silk must be used to avoid frequent breakage under tension

It is interesting to note that the batch #1 experimental strings broke with quite a uniform 'clean break' - a good sign indicating homogeneous construction of the string - see attached image. On the other hand this may be due to brittle failure of the silk fibres due to acidity of the glue binder - not so good!

'Plictho Dell'Arte de Tenori' is a book in four parts about the dyeing of fabrics published in Venice in 1548. The author - Giovan Ventura Rosetti - provides many formulae for the dyeing of silk in the third part of his book including procedures for stiffening or weighting of silk - of particular interest here. These include application of Gum Arabic, use of tannins in combination with animal glues, and a pre-tannin soaking followed by a treatment with an iron salt - all part and parcel of these planned trials for the weighting of silk strings.

The silk strings with the most elastic properties (potentially at least) are those made like a small rope - from 3 or 4 twisted strands . The Chinese have been making this type of instrument string for centuries and still are today. I have an assortment of these strings purchased over 30 years ago in London, England as well as some recently purchased by my son on a business trip to the Far East.
The strings are quite stiff (not very pliable), supplied in coils, so are probably made from raw silk with the sericin gum left as a binder (but not sure about that).
The plan is to use the strings as a basis for weighted 6th and perhaps 5th course bourdons by first de-gumming the strings followed by treatment with tannin/ ferrous sulphate and using a flexible hide glue binder loaded with iron oxide powder. Alternatively a flexible hide glue loaded with copper powder will be tested.

As an initial test, a metre length of Chinese 'roped' silk string - measuring 1.1 mm diameter - has been de-gummed prior to being subject to loaded with Tannin and Ferrous Sulphate solutions.
The length of string was de-gummed - simply tied in a coil and immersed in a solution of Sodium Carbonate (Washing Soda) and dish washing detergent held at a temperature just below boiling point for 45 minutes. (Alternatively, a solution of wood ash in water could have been used for de-gumming - more historically correct but less convenient). The proportions of the de-gumming bath were 100 c.c. Detergent, 50 c.c. Sodium Carbonate powder to 3800 c.c. of water (1 U.S. gallon). The coil twisted around itself as the silk absorbed moisture and expanded - as seen in the attached image - but was easily untangled after de-gumming.
After de-gumming the string was washed in fresh water and left to dry under a nominal load of 0.4 Kg - to ensure that the string remained straight and uniform without kinks.
The extent of the de-gumming will be assessed tomorrow once the string has dried.

The dried string after degumming and under a 0.4 Kg load stretched from 100cm to 102 cm and the nominal diameter reduced from 1.10 mm to 1.02 mm - most likely due to loss of the Sericin gum - as well as some partial unwinding of the plies of the string. The dried string is very pliable (like a rope) and white in colour so is probably almost fully degummed.

The string has been immersed in a hot tannin solution for 90 minutes and allowed to remain in the solution for another
4 1/2 hours - then washed in fresh water. The tannin had been adsorbed by the silk to become brown in colour.

The string was then immersed in a hot solution of Ferrous Sulphate for 90 minutes and then washed in fresh water. The string had turned black in colour due to the reaction between the Tannin and Iron salt forming the complex compound Iron Tannate within the silk (hopefully).
The string has been left to dry under a nominal 0.4 Kg load.

It is interesting to note that the Bamm bass string of a four course oud in Ziryab's time (9th C) was coloured black (the other strings being coloured white Mathlath , red Mathna and yellow Zir) - suggesting that metal salts may have been used to weight the larger diameter strings at least.

The next step is to treat the dried string with a flexible acidic glue binder to complete the weighting process.

After treatment with the acidic glue binder the string diameter was restored to 1.1 mm diameter. The dried string was flexible - not with a brittle finish like a string made with a Sericin or Gum Arabic binder - but still quite stiff. The string was wrapped several times around a small diameter rod to improve flexibility and then subject to testing on the test rig. Sustain was modest at about 5 seconds and calculated relative density about 1.14 gm/cc.
Acoustic performance - tested on a lute tuned to E at A440 - as a fifth course bourdon with nylon octave (A 110 Hz at string length 67.5 cm - tension about 2.5 Kg) - was judged to be poor, lacking in sustain and dull in tone colour (at least compared to modern nylon strings). Disappointing!

The attached image shows the finished string.

Silk filament is not perfectly elastic - i.e. once stretched it does not fully return to its original length. The attached plot shows the recovery of de-gummed silk filament after being stretched under load from its original length of 500 mm to 525 mm. (i.e. 5% extension). When the load is released the length returns to 508 mm after 15 seconds but the filament then remains permanently stretched to 506 mm.
Gut strings have similar inelastic properties.

Measurements on the test rig to determine string elongation under load were quickly done in order to get some idea of the calculated relative Modulus of Elasticity (or stiffness - Young's Modulus) of each string under test - order of magnitude for comparison purposes rather than high precision.

For information, Young's Modulus (E) calculated from these test rig measurements - expressed in International Standard units (Gigapascals or GPa - a unit of pressure) - ranged from an E value of about 6.4 to 8.9 for the 10 batch #1 experimental strings. This compares to a commercial E value for silk filament of between 8.4 to 12.9 GPa - so the twisted batch #1 experimental strings (made from spun silk and glue) are a bit more elastic than a string made from untwisted silk filament.

The original Chinese silk 'roped' string had a calculated E value of about 3.5 GPa. For the experimental modified 'weighted' Chinese string the E value is 4.2 GPa - a bit less elastic than the original due to reduction in degree of twist.

For comparison, historical string maker Eph Segerman (Northern Renaissance Instruments) reports that measured E values for his gut strings range from about 5 to 6 GPa for low twist plain gut to about 1 to 2 GPa for roped gut strings.

Kanz at-tuHaf on silk strings

Quote: Originally posted by danyel

He does not seem to understand that Urmawî devised a practical fretting for the oud (seven frets LLCLLC) ....... Even in his articles on Marâghîs notations he spoils the only extremely clear aspect, viz. the fretting referred to by Marâghîs notation (LLCLLC, being an Urmawî disciple) …

Quote: Originally posted by danyel

L=limma (ca 92 cents), C= comma (ca 23 cents)

Quote: Originally posted by jdowning

So did the Ihwan al-Safa get their geometry wrong in trying to demonstrate the universal application of the 4/3 ratio - or is there another explanation for the apparent discrepancy? More to follow