Marinated Wood - an experimental investigation into fluted ribs

Thin and flat, (under 1.5 mm thickness) lute ribs, when bent to shape on a lute bowl mold, will take up a natural, shallow curve, or 'fluting' in cross section. This is due to the transverse and longitudinal stresses in the bent rib coming into an equilibrium, or balance, state. This feature may be less pronounced in ribs made from thicker material - as might be found in modern oud construction.
The attached image shows the fluted rib effect represented in a late 16th C representation of a lute (c. 1595) by Caravagio in the Hermitage museum.

Thankyou suz_i_dil. I have not seen the picture of the Yayyar oud but will look it up. This subject of fluted ribs has been discussed at some length on the forum (see the topics "Awsome Oud ribs" on the Projects Forum and "Oud with fluted ribs" on the Ouds,Ouds, Ouds forum). However, I want to start this investigation afresh with a 'clean sheet of paper' and as few as possible preformed ideas - just to see where it might lead.
My objective in this topic is to investigate how far and by what method wood can be bent and formed into deeply fluted ribs without resorting to carving. As this is a hands on experiment I have no idea, at this point in time, how these experiments will turn out but - success or failure - it should be a useful learning experience.

Deeply fluted ribs can be found on surviving guitars, vihuelas and small descant lutes (mandolinos) dating from around the end of the 16th C. On these instruments, the fluting may created by bending thin ribs or by carving thick ribs after bending. The attached image shows a section of carved fluted ribs on an elaborately decorated early 17th C vaulted back guitar by Magno Stegher dated 1621 (Charles van Raalte Collection, Dean Castle, Kilmarnock, Scotland). Typically, the body of the instrument is made from many, narrow ribs measuring about
7 mm maximum width.
Fluted ribs, formed by bending, may be found on two, well known, surviving instruments - a guitar by Belchior Dias, dated 1581, in the Royal College of Music, London and a vihuela cat. E7048 in the Cite de la Musique museum in Paris. (There is some dispute about the designation of these instruments as guitars or vihuelas but these arguments are irrelevant to this investigation). The attached image shows a section through the central rib of the Dias guitar drawn to scale having a radius of the fluting about 18 mm and a reverse curve longitudinally having a radius of about 33 inches (838 mm).
This is a shallow longitudinal curve compared to an oud or lute rib, however, the guitar/vihuela ribs have sides that are relatively parallel whereas lute and oud ribs taper sharply longitudinally. This sharp taper might make it possible to form fluted oud and lute ribs by bending alone although I am not aware of any surviving early lutes or ouds that have this feature
The current wisdom, however, is that deeply fluted oud or lute ribs can only be made by carving and there are examples of modern luthiers making instruments in this manner.

Thinness of the bowl seems to have been a feature of early ouds as it was for lutes according to Arabic/Persian accounts during the Mediaeval period in Europe. The accounts do not state 'how thin is thin' except to record that the ribs were made thinner than the soundboard. This may imply a rib thickness of under 2mm. This does not necessarily mean that early ouds had faceted bowl exteriors like the lute - they could still have been made like modern ouds starting with relatively thick ribs that are then shaped to give a smooth rounded bowl exterior. If this was the case then the bowl might then have been shaped on the interior as well - to produce a uniform thin wall thickness.
BTW, I was unable to locate a picture on the forum of a Yayyar oud with a fluted rib bowl - all of the images posted seem to be of ouds with a conventional smooth rounded profile.

The first stage of this investigation will be to establish a method for bending and forming the "inverted saddle", fluted form of rib found on the 16th C vaulted back guitar/vihuela examples posted . The attached image is a sketch representing the required general shape of this type of rib.

Constraints to be applied in the investigation are:
- the method must be one that could possibly have been used by luthiers of the 16th C.
- the method must be a process that will not result in deterioration of the structure or physical properties of the wood. However, removal of some residual materials from within the cell structure of the wood (extractives - waxes, resins, mineral salts etc.) may be acceptable if found to be beneficial in improving bending properties.
- after bending, the geometry of the wood shall be fixed and permanent - within a reasonable tolerance (recognising that all wood expands and contracts with humidity/temperature variation)

There are a number of possible bending procedures which may or may not work successfully. These are to be considered, tested and evaluated separately.
All methods will require the wood to be temporarily softened and/or made 'plastic and remain in that state long enough for the wood to be clamped to a mold of required shape without splitting or breaking.

Wood is complex structurally but simply put, is comprised of longitudinal fibres which are hollow, tubular cells with the cell walls made of cellulose compounds - a kind of sugar. The cells make it possible for transfer of nutrient containing sap within a tree. The cells are joined to each other with 'lignin'. Lignin, softens at temperatures above 170 C/ 340 F but in the presence of moisture, the softening temperature of lignin is lowered to below the boiling point of water (100 C/212 F).
In a softened state, the wood cells may be compressed considerably but stretched very little before failing.
Lignin may also be temporarily softened with chemicals.

The objective of this first part of the experimental process is to try to get a general feel, quickly, for what bending procedures may or may not work in practice. Samples of seasoned American walnut and Maple - quarter sawn - measuring about 35 mm wide by 135 mm long by 1.2 mm thick (1.375 X 5.30 X 1/16 inches) were prepared, planed smooth both sides.
A 'male' mold, (the type used by luthier Dan Larson), on which the test pieces will be formed - representing the geometry of the central rib of the 16th C Dias guitar - was cut from a piece of pine. The mold is about half the length of the Dias rib for preliminary test purposes - to save time and material. This mold configuration is the easiest to manufacture - quickly made by cutting the longitudinal curve on a bandsaw and then shaping the cross section with a rasp. The longitudinal curvature of the mold was made with a radius of 27 inches (656 mm) - tighter than that of the original Dias rib (33 inches/ 838 mm), to allow for some anticipated 'spring back' of a bent rib after release from the mold. The longitudinal curvature was accurately measured using a set of standard draughting curves.
The attached image shows the completed test mold. The nails in the side are convenient anchor points for the string that will be used to tie down the sample to the mold
So - ready to go for the first test.

Wood, when freshly cut and full of sap ('green' wood) - dependant upon the species, - may be soft and flexible enough to be formed to shape on a mold where it will permanently attain (within limits) the shape of the mold on drying and seasoning. The disadvantage of this method is that it is slow and the wood may be subject to cracks and splits during the drying process unless carefully controlled. Also, slab cut wood is less stable than quarter sawn under variable humidity conditions.

For information and interest, here is an image of a sample of wood (vinewood) that was slab cut from a 2 inch (50 mm) stem earlier this year and allowed to air dry. Note that the sample, on drying, has naturally attained (almost!) the ' inverted saddle shaped' geometric profile that is the objective of these test - without using a mold!!
Also, this vinewood has no visible seasoning splits or cracks.
The reason that this sample has taken a fluted profile is because of the differential grain length across the slab cut section - shorter at the bottom of the sample than at the top so that the shrinkage on drying is less at the bottom than on the top of the section (see the bottom two images of end grain). The longer the grain length the greater the amount of shrinkage. Also, as the transverse stresses induced by this natural curvature must be balanced by an equal and opposite stress, an opposing longitudinal curve can be seen in the sample.

This is test #2 which may have some useful future application.

With the pre-soaking tests in progress, preparations for the second phase of this investigation - to make full sized fluted oud or lute ribs - have gone ahead with the making of the molds. I plan to test both male and female type molds into which the softened rib blanks will be clamped and allowed to re-harden.
The profile of an early 17th C lute has been used for the molds. Both molds were straightforward to make. The male mold version was made from a piece of elm wood with the rib profile cut out and the curve of the flute cut using a rounding over bit in a router. The female mold was a little more complicated as it has to be made in two halves. Two pine boards were screwed together and cut to the required rib profile using a bandsaw. After smoothing the profile, the two halves were separated and the inside curve of the flute cut with a coving cutter using a router. The two halves were then screwed back together - the screws providing perfect registration and fit of the parts.
In my haste to make the molds, I forgot to provide extra height to the profile of the female mold to allow for the depth of the flute so this mold is smaller than it should have been. Not to worry, this is only a mold designed to test the feasibility of making fluted ribs. If it works using this mold - the method will also work for larger or smaller molds.
Additional nails or screws will be fitted to the male mold to provide anchor points for the string windings that will be used to provide the clamping force on the rib. The clamping force on the female mold will be provided with a rope or fabric tape under tension or using a series of shaped wooden blocks clamped in place. The method that works best under test will be the one chosen.

The rib blanks will be cut a little oversize to the required profile (to allow for final trimming) - ie tapering to a point at each end. This will be necessary to minimise the stresses under tension generated along the outer edges of the rib when clamped in place. This is possible because the depth of the flute diminishes from the centre of the rib to the extremities. Attempting to bend a parallel sided rib blank would be more likely to fail due to the greater tensile stresses generated along the edges. If the methodology proves to be successful, most of the fluted shape will be achieved under compressive stress loadings that the wood cells are able to tolerate much better than if subject to tensile loadings. At least, that is the theory - the 'proof will be in the pudding'!

The images show rib blanks positioned on the molds as they would be prior to clamping in place following softening of seasoned wood (or when 'green').

As I have 'green' vinewood to hand that I know is very flexible it was decided to try to try to 'cold' bend a sample on the male mold. At the same time, advantage was taken of the natural curvature in the vine stem selected. The stem was first cut longitudinally through the centre and then sliced into curved strips, about 3 mm thick using a bandsaw and temporary wooden guide clamped to the bandsaw table.
The outermost strip was chosen for the test - this being slab sawn so that the grain would cause the strip to 'cup' in the direction of the flute curavture on drying.
The strips were very flexible longitudinally and there was moisture (sap) visible on the cut surfaces.

The strip selected was so flexible that it could be flattened with the weight of a small plane. The top surface of the strip was, therefore, planed smooth - this surface being in contact with the mold. The thickness of the strip was about 2mm after planing one side. The opposite side was left with the sawn surface.
The strip was cut to the dimensions of a rib ready for clamping to the mold.

The rib was clamped to the mold using string wound tightly around the mold. A quick convenient method that applies a gradual increasing clamping force as the string winding progresses. The rib blank readily conformed to the longitudinal curve of the mold but failed by cracking along the centre line as loading was applied by the string. Clearly the combined transverse and longitudinal tensile stresses (with greatest stress along the centre line) exceeded the tensile stress limit of the wood.
What can be learned from this failure? The vine rib blank was not clear and straight grained but contained knots and crooked grain. The cracking initiated in the grain irregularities surrounding the knots. The rib blank measured 2.3 mm thick which is likely much too thick. The string winding does not provide uniform support of the surfaces under tension. A series of metal (or woven fabric) bands for clamps might be a better arrangement for providing even support.
For this procedure to have a chance of succeeding, a rib blank must be straight grained and free of flaws, thin (under 1.3 mm say), and be fully supported during bending against tensile stress failure. It must also, of course be soft and flexible prior to bending.
It is uncertain at present if a blank should be quarter sawn or slab cut for best effect.
The initial pre-soaking trials of phase 1 of this investigation will be completed and reported on tomorrow.

Pre-soaking/ Boiling Bending Tests.

Test samples were prepared from Black Walnut and Maple, each, quarter sawn, measuring about 35 mm wide by 135 mm long by less than 1.5 mm thick - planed smooth on both sides. The Walnut was wood recovered from a 19th C reed organ so has been drying for about 100 to 150 years. The Maple has been drying for at least 30 years - but probably a lot longer than that - so both species are well seasoned.
The saddle shaped wooden mold previously posted has been made to the same rib shape as the 16th C Dias vaulted back guitar i.e. with longitudinal curve measuring 27 inches radius (686 mm) and a flute radius of 18 mm (less than 3/4 inch). After soaking in the chosen marinade for several days, each sample was boiled in fresh water for 5 minutes before being quickly clamped to the mold by binding it with string. The string binding method is very simple and applies a gradually increasing clamping force. The main purpose of boiling was to cleanse the wood pores of any residual marinade and to heat the wood in order to accelerate drying of the sample in the mold - although some softening of the wood, due to boiling, probably occurred as well.
Each sample was left to dry, clamped to the the mold, in a warm room (at about 20 C/ 65 F, relative humidity about 65%) for 24 hours. After being released from the mold each sample was allowed to dry for a further 24 hours before being measured dimensionally.

The first test, as a 'benchmark', was to try to bend a sample - without any pre-soaking - simply by boiling it in fresh water for 5 minutes. This Walnut sample failed by splitting down the centre well before the string winding could be completed and full pressure applied. The test was not repeated using a Maple sample as boiling alone, to soften the wood, is clearly not going to be an effective procedure for these tests.

The marinating of the samples was accomplished by placing a Walnut and Maple sample, together, in a small 'Ziploc' type plastic bag with about 15 to 20 ml (cubic centimetres) of the chosen marinating fluid. The bag with samples was then sealed and left to stand in a cool room at 10 C/ 50 F.
The first samples to be tested were those marinated in Sake (Rice wine) for a period of 4/5 days. The attached images show that both samples were successfully clamped to the profile of the mold after boiling. The Maple sample cracked at both ends - the cracks running for 25 mm and 15mm in length respectively. I am not too concerned about the end cracking - which is due to intensified tensile stress at the free ends of the sample, stresses most likely further amplified by the rough, sawn, edges - the saw cuts being 'stress raisers' or local points of high stress concentration. This fault may be minimised in future full scale tests by ensuring that the ends of the rib blanks are cut smooth and by leaving extra length to the blank so that such faults may be cut off.

So far so good. The walnut sample after release from the mold retained the desired inverted saddle shape. The dried, bent sample measures about 1.1 mm thick. The longitudinal radius has 'sprung' on drying from a radius of 27 inches to 34 inches (686 mm to 854 mm) and the flute inside radius is maintained at 18 mm.
The thicker Maple sample at 1.3 mm thick also achieved a longitudinal radius of 34 inches but the inside radius of the flute opened up to 24 mm (about 15/16 inch). Not quite so good.
There are a few variables to take into account here, but thinner would seem to be better as far as three dimensional bending of the samples is concerned - less tensile stress in bending.

Wow

it seems to me that some kind of pee is usually involved when making anything medieval

Josh is right, urea is used to soften lignin

I have added a 50/50 wood alcohol/ glycerine mix to the list of marinades to be tested during this second phase about a month from now. I am beginning to suspect that pretty well any moisturising fluid that can be soaked into the wood to reconstitute the dried out cell matter will have some softening effect.
The attached image shows the test pieces from phase 1 of this experiment - all successfully bent longitudinally as well as in cross section. The image of the edge of Maple test piece #8 shows one disadvantage of using string as a clamping device. String cuts into and damages the edges of the softened wood. It does illustrate how soft the wood had become in this case however. To overcome this problem use of strong fabric tape would distribute the pressure more evenly and would also better support the outer surfaces of a piece against failure due to tensile stress in bending.
An improvement in mold design might be to construct it from guided articulated segments. This way the bending would be accomplished in two stages - the first to bend the flute curve with the mold segments laid straight and the rib blank strapped to the mold segments with fabric tape, the second step being to then immediately force the mold, with rib blank in position, into the required curve. Food for thought at this stage.

Yea, well when you live in rural farming areas, as I do, you get used to this sort of thing!
The Scots, like the Romans also had communal collecting pots that they would contribute to on their erratic journey home after consuming a few 'malts' during the course of an evening. Perhaps the liquid intake makes a difference to the final product?
As for the products used for tanning leather by the ancients, urine is only the first step. What follows in the process - although very effective - gave the leather tanning industry an unpopular reputation in polite society and cast it to the nether regions outside of town and city boundaries - and banished the employees of this essential industry to pariah status . I won't say what it was but suffice to say I do not intend to go that route and join the numbers of the great unwashed in modern society - even in the interests of science or whatever!

Today I looked again at the notes that came with the museum drawing of the Dias guitar made by London luthier Stephen Barber.
Barber is convinced that the ribs are made from Kingwood (Dalbergia cearensis) - a member of the rosewood family very similar to Brazilian rosewood but which, he says, is more elastic and easier to bend than the latter. He informs us that Kingwood comes from a tree of small dimension - 15 to 18 cm in diameter (6 to 7 inches). This would have been the size of timber generally available to luthiers of the late 16th C with connections to the 'New World".
The small log size of this timber implies that at least partial slab cut sections of rib stock may have been used?
I have chosen Black Walnut for continuation of these trials because I have some in stock and because it has proven in stage 1 to be easier to bend - and quicker tp prepare - than Maple (we must learn to crawl before we can walk).
Also, Black Walnut has a similar open cell structure to the true Rosewood genus (see attached macro images Walnut to left - Indian Rosewood to right, made from samples that I have in stock) - not that the woods are equivalent - but the similar open cell structure should allow for better absorption of softening fluids for both types of wood? I also have some Brazilian Rosewood lute pegs (made from larger billets purchased in the 1970's) so will macro photo the grain structure of those for comparison.

For comparison - here is a macro image of one of the lute pegs, mentioned in the previous post, made from Brazilian Rosewood (Jacaranda), Dalbergia nigra. I should still have the 'off cuts' left over from the peg turning operation - from which I could make images covering a wider area - but I cannot find them. A better set up for making the macro images needs to be made - with a reference scale to aid comparison - but I haven't got around to it yet. The diameter of the end of the lute peg is
6 mm. Note that the macro images have to be compressed (reduced in size) for efficient posting on the forum which results in a significant loss of definition, or clarity, compared to the original images.
The cell size and distribution of Jacaranda is quite similar to another species of rosewood that I have to hand - Cocobolo (or Grenadillo), Dalbergia retusa.
As Indian Rosewood, Dalbergia latifolia (at least as far as these samples are concerned) seems to have a greater proportion of randomly distributed, large diameter cells than either Jacaranda or Cocobolo I would guess that it would be quicker to soften and easier to bend - but that is just a theory at this point in time, yet to be tested.
Note that all of the rosewoods have material - resin and mineral deposits etc. - packed within a proportion of the cells (visible in the macro images) that may be supporting the cell walls against distortion under compressive loads (experienced during bending). These are 'extractives', which, if they can be dissolved and removed by the pre-soak marinade fluid - might render the wood more pliable? Just a theory yet to be tested
Once the best marinade and soak time has been determined experimentally for the 'easy' wood - Black Walnut - Indian Rosewood will be next in line for testing.

What an great investigation...
I encourage you to continue to both explore the practical methods and to keep us informed of your progress!
I am engrossed!

on a note... not more than obliquely related...

I was told:

...that if you add urea to hide glue, you get a room temperature hide glue... just food for thought and grounds for further research!

I hadn't realised before that urea was used to make liquid hide glue or that liquid hide glue was as popular among woodworkers as it seems to be these days. I made the mistake many years ago of purchasing a half gallon tub of prepared liquid hide glue. As I recall, this stuff was of a rubbery consistency so still had to be heated before use so I suppose was a bit more convenient to use than glue granules. Anyway, by the time I got around to using it, the whole lot had decomposed so was useless. Just as well, I suppose, as it is not recommended to use liquid hide glue for instrument making.
Urea, in white powder form, it seems, is readily available for use as a fertilizer from gardening stores so would be a convenient way to make up into a marinade. Not sure if the locals will stock it this time of year so that will have to wait until Spring next year.
While phase 2 samples are being marinated, here for information and comparison is a macro of the cell structure of the Cocobolo that I have in stock. Very similar in structure to Brazilian Rosewood.

..... and, to complete the picture here for comparison is a macro of the cell structure of the maple used in these trials. Unfortunately a reference scale was not used to enable direct comparison but it can be seen that the maple has smaller cells so absorption of the marinating fluid may be a lot slower than it might be in wood species with larger cells.

Phase 2 bending started today, the samples having marinated for about a month.

Sample #9 Walnut in Sake - marinated for 28 days. The marinating fluid had a light brown tinge to it and the sample was quite flexible when cold. It had curved slightly across its width and - in the same direction - along its length. The sample was boiled in water for 5 minutes and then tied to the mold as for the Phase 1 samples. The sample was placed on the mold with the curvature across its width following the curvature of the mold (i.e. the curve of the fluting).
The sample split at both ends as it was being tied to the mold - the most severe of the splits being 4 cm (1- 5/8 inches) long.
So this is considered to be a failure unlike sample #1 (walnut in sake, phase 1 tests) that bent without splitting after only 4 days marinating. Sample #1 is quarter sawn, sample #9 is part slab sawn which might be part of the problem

For Phase 2 tests, I am using an elastic (stretchy) fabric tape (ice hockey boot laces - cheap, strong, wide and long) instead of string which cuts into the softened wood. The wide tape distributes the pressure more uniformly so should reduce the tendency of a sample to split. Not in this case, however!

The sample will be left overnight to dry and sample #10 will be bent tomorrow - hopefully with better results.

Sample #11 Walnut in plain water - marinated for 30 days. There was bacterial growth in the fluid and a slime on the surface of the sample. The fluid was tinged a medium brown colour.
After boiling for 5 minutes, the sample was tied to the mold. It cracked at the ends like sample #9 and #10. The cracking of sample #9 is shown in the attached image - similar to sample #10 after air drying for 24 hours.
A close inspection of sample #10 indicated that the cracks are being initiated in the late wood cells exposed on the surfaces of each sample. It can be seen that - as there is very little 'run out' of the longitudinal grain in the Phase 2 samples - the open cell structures exposed at the surfaces of the samples are relatively long and so constitute a weakness causing localised thinning of the sample as well as acting as stress raisers. The exposed latewood cells appear in parallel groups on the surface because the samples are slab cut. Slab cut wood is also stiffer across the width of a sample so may be more difficult to bend into a fluted section.

Contrast this with sample #7 from the Phase 1 tests - successfully bent without cracking. It so happens that the longitudinal run out of the Phase 1 samples is about 5 - 10 degrees so that the exposed cell structures are much shorter and more random in distribution. Also, these samples, being quarter sawn, can flex more readily across the width of a sample so should be easier to bend into a fluted section.

It is anticipated that the remainder of the phase 2 samples will also crack when bent for the above reasons - but time will tell.
Longitudinal grain run out may turn out to be an important factor in order to successfully bend fluted ribs.

Sample #12 Walnut/50% glycerine- 50% alcohol marinade. Soaking for 31 days. Fluid tinged pale brown.
The sample cracked at one end during bending - length of crack is 3 cm (1.2 inches).
The attached image is the end grain of sample #12 after 24 hours air drying. The measured thickness is 0.94 mm (0.037 inch) so there may have been some reduction in thickness prior to marinating of around 15%. The camera was hand held (not recommended for macro photos!) but the image is clear enough and illustrates the problem with all of the Phase 2 test samples. It can be seen that the early-wood cells are of significant diameter compared to the thickness of the sample (looking like Swiss cheese - or worm eaten wood!). As there is no longitudinal grain 'run out' in the Phase 2 samples, the exposed cells, on both sides of the sample, are quite long, creating a weakness and points of high stress when subject to bending under tension- equivalent to scoring the samples with a knife.

Sample #13 Walnut/Household Ammonia solution. Soaking for 32 days. The marinade fluid was a deep mahogany brown. The sample was very pliable and after boiling for 5 minutes was tied to the mold without cracking. This is an encouraging result - particularly in view of the problem with Phase 2 samples (although the sample may still crack after air drying for 24 hours). An ammonia solution in water of 5% to 10% is not supposed to be effective in softening wood (unlike liquid ammonia) but both Phase 1 and Phase 2 tests involving household ammonia have both produced promising results so far.

The last test in Phase 2 will be Walnut/Wood Alcohol.

Sample #13 has been released from the mold after air drying at a room temperature of 65 F (18 C) and relative humidity of 50% for 24 hours. The sample has not cracked. However it had shrunk and distorted significantly (shriveled) as can be seen in the attached image where the sample is shown along side sample #11 for comparison. Compared to sample #11, maximum shrinkage across the width of the sample (measured in the middle) is about 20% but there is no measurable shrinkage in the length. The Ammonia solution has softened the cell walls of the wood to the point where they have shrunk in size during drying. The ends of the sample that were not restrained by the binding tape have become 'bell shaped'.
Had the sample been quarter sawn, it is expected that the degree of shrinkage and distortion would have been less.
So - no matter what the experts say - a 5% to 10% solution of Ammonia in water will work in softening the cell wall - like liquid Ammonia does - it just needs more time. Liquid Ammonia presumably softens the wood very quickly - so is commercially more viable. No doubt if wood is left in liquid Ammonia for an extended period, it would also suffer the same degree of shrinkage and distortion? Clearly, in this test, the marinating time has been excessive (so the results are a failure as far as bending is concerned) but it does demonstrate the softening power of Ammonia. Shorter marinating times and/or a more dilute Ammonia solution would likely be more successful. More testing is required in order to find the optimum set of conditions for bending using a dilute Ammonia solution.

Sample #14 Walnut in Wood Alcohol. This sample has been marinating for the same period as the other samples in
Phase 2 but - due to a leaking Ziploc bag - was found to have dried out when inspected on the 7th Dec. Since then the sample has been re-marinating in a fresh solution of wood alcohol but it is now difficult to know if the marinating period should count as 14 days or 33 days. Nevertheless, it was decided to go ahead with bending the sample. After boiling in water for 5 minutes, the sample was tied to the mold without cracking and left to dry for 24 hours.
Another promising result!

Sample #14 cracked at both ends after drying for 24 hours with one crack measuring 2 cm long and the other 1 cm. Nevertheless, this is a positive result as allowance can be made to trim a rib after bending.
The attached image shows a cross section of sample #13 (cut in the middle of the sample) compared to sample #11 to illustrate the degree of shrinkage after marinating in the Ammonia solution. Note that the radius of the fluting is also reduced.

The attached image shows the cell structure of sample #13 (at the mid section) after marinating in Ammonia and drying compared to sample #12 end grain. It can be seen that the larger diameter early wood cells have been reduced in size after drying and the structure of the wood has become compressed and more compact.
Had the sample been quarter sawn there would have been less reduction in the width of the sample but a greater reduction in thickness. This will be worth another test as, after Ammonia treatment, wood becomes quite plastic in nature so should have the advantage of allowing greater scope for more extreme bending of the wood - perhaps without the need for heat.

In order to establish the length of time it takes for Household Ammonia solution to plasticise wood, Walnut samples are to be subject to yield testing. The marinating fluids will be Household Ammonia - undiluted; 50% Ammonia solution/ 50% water; 50% Ammonia solution/ 50% Wood Alcohol.
The marinated samples will be tested every 24 hours by clamping each sample in a test rig, applying a standard, fixed load to the end of each sample and measuring the deflection at the load point on the sample. The attached images show the rig set up.
Samples in an untreated, (i.e. unsoftened) state will retain their elasticity under the (briefly applied) test loading - that is when the load is removed the sample will spring back to its original position on the rig. At some point in time, however, the sample is expected to soften and so will lose its elasticity - i.e. it will not return to its original unloaded position on the rig. This will be the initial yield point of the sample and will be an indication of how long a sample must be marinated to become plasticised. At least, that is how I see the test working so we will see how it goes.

Also, in order to get a better idea of what happens to the cell structures of different wood species when plasticised by Ammonia, several test samples - cubes measuring about 12mm (0.5 inches) on each side - have been left to soak in Household Ammonia solution for 30 days in a sealed jar. The wood samples are Vine, White Ash, Black Ash, White Elm, Black Walnut, Beech, and Indian Rosewood.
There was an immediate action noted on adding the Ammonia solution - the fluid quickly becoming a dark brown colour. So something is definitely going on there. After 30 days the samples will be boiled in fresh water for 10 minutes and the softened end grain sliced with a razor in preparation for taking digital micro images of the cell structures after drying.

The wood samples - cubes measuring 0.5 inch (13 mm) on all sides - have been marinating in Ammonia solution in a sealed jar for 10 days after which time they were removed for examination. Each sample was boiled in water for about 5 minutes and allowed to dry. The duration of the test was reduced as the softening effect of the Ammonia solution appears to be much more rapid than previously thought.

After drying, each sample was observed to have shrunk to a degree - shrinkage being greater along the growth rings than radially across the rings as it would be if the wood had been dried in the usual manner from being freshly cut. The least amount of shrinkage was found in the sample of Indian Rosewood (closely followed by American Elm) and the greatest amount in a sample of White Ash.

Images of the cell structures of each sample - before and after marinating - will be posted separately for information.

The most dramatic example of wood shrinkage after marinating in Ammonia solution is seen in the sample of White Ash. The sample, after drying is now calculated to be only 56% of its original volume. This would imply an increase in Specific Gravity from about 0.6 to 1.0 assuming no loss of weight in the marinating process only shrinkage. There is also likely to be an increase in hardness of the wood - although I have no means of measuring relative hardness to verify this assumption.

The attached image shows the appearance of the sample after marinating and drying compared to its original appearance - both being represented to the same scale before and after. The reference scale is in mm.
The dark appearance of the sample is because all of the samples were marinated together in the same sealed vessel. The Ammonia has extracted wood dyes from the Indian Rosewood and Walnut samples which in turn has stained all of the samples a dark brown colour.

It so happens that the Ash sample has been quarter sawn with the growth rings running horizontal in the image. The greatest amount of shrinkage is along the growth rings (transverse) with a lesser degree across the growth rings (radial) as expected in normal seasoning of wood. However, normal tangential shrinkage in Ash is about 1.6 times greater than radial shrinkage when drying the wood from green to oven dry moisture content whereas, in this case, the tangential shrinkage is 4 times radial.
As the latewood cells of Ash are small and thick walled compared to the large, thin walled, early wood cells there is differential shrinkage between early wood and latewood cells in each growth ring producing the unusual rippled effect observed at the edge of the sample.

In contrast to the 'extreme' shrinkage seen in the White Ash sample the attached images show the American Elm sample after marinating for 10 days together with an untreated sample of Elm - both to the same scale. The reference scale is in mm.
In this case the volume of the marinated sample is calculated to be about 92% of its original size. This shrinkage implies an increase in Specific Gravity from 0.5 to about 0.54.

Both White Ash and American Elm, outwardly, have similar ring porous cell structures so it is not clear why the shrinkages should be so significantly different between the two woods. One clue may be in the amount of darkening observed in the marinated samples - due to absorption of the dark brown dye in the marinade. The colour of the Ash sample is dark whereas that of the Elm is quite pale - indicating that the Ash cell structure is more porous and might, therefore, be expected to react more quickly to the marinade.

The attached images show the marinated sample of Vine wood together with an untreated sample. Vine wood also has a ring porous cell structure but, unlike Ash and Elm, large early wood cells predominate. The reference scale is in mm.
The calculated volume of the sample after marinating is 71% of its original volume and the sample is a dark colour indicating a relatively heavy absorption of the marinating fluid. The change in Specific Gravity implied by this degree of shrinkage is an increase from 0.55 to 0.79 and corresponding increase in hardness.
It is perhaps surprising that the Vine sample did not show a greater amount of shrinkage because of its open cell structure. Perhaps the wide rays running radially on each side of the rows of cells support the cell walls against shrinkage - both in a transverse and radial direction - behaving a bit like reinforcing rods in compression?

To summarise, the least affected sample is Indian Rosewood that, after 10 days marinating, showed no measurable shrinkage.
American Elm is next in decreasing order, shrinking in volume to 92 % of original size. S.G increase from 0.5 to 0.54.
Vine is 71% original size. S.G. increase from 0.55 to 0.78.
Black Walnut is 70% original size. S.G. increase from 0.55 to 0.79.
European Beech is 70% original size. S.G. increase from 0.64 to 0.91.
Black Ash is 59% original size. S.G. increase from 0.49 to 0.83.
White Ash is 56% original size. S.G increase from 0.6 to 1.0.

Presumably, the more resistant a wood is to absorption of a marinade the longer the time it will take for the wood to be softened - although this would require further testing to verify. Perhaps the Rosewoods, for example, are too oily and 'waterproof' to be significantly affected regardless of the time soaking in the marinade?

Note that all of the test samples were fully seasoned and dried prior to marinating. The Black Ash and Black Walnut samples are both over a century old and the Beech about 60 years old.

In the meantime the yield tests for Walnut samples marinating in Ammonia solution, 50/50 Ammonia solution and water and 50/50 Ammonia in Wood Alcohol have been completed.

The purpose of the yield test is to determine the length of time it takes for a marinated sample to become plastic.

Samples were slab cut from the same piece of Black Walnut (over 100 years old so well seasoned) used in the bending trials previously posted. Each sample measured 7 inches long X 1.5 inches wide X 0.045 inches thick (178 mm X 38 mm X 1.14 mm) that allows an inch (25 mm) to be clamped in the test rig leaving a 6 inch (152 mm overhang).
Each marinating sample was removed from the marinade each day (every 24 hours) and loaded as a cantilever beam in the test rig - firmly clamped at one end with a fixed weight attached to the free end. The deflection at the free end was measured using dividers.
A sample was also tested under load to establish that the chosen loading did not produce yielding in a dry untreated sample.
The results are shown on the attached graph.

The yield point (point A) for each sample was achieved for the Ammonia solution marinade in about 3 days, for the 50/50 Ammonia - water marinade in about 4 days and for the 50/50 Ammonia - Alcohol marinade in about 8 days. All of the deflection/time curves are of a similar general shape.
The test was continued after the yield points had been reached - just to see what happened - but was terminated after 13 days as it became increasingly difficult to measure the more extreme deflection. (This difficulty is apparent in the more erratic data points measured after day 10 for the sample marinated in the Ammonia solution).

A further test is in hand to find out how the untreated sample soaked in water responds. These results will be posted later for comparison and information.

This simple yield test method should be useful in establishing the optimum time to marinate a particular species of wood prior to boiling and hot bending. Clearly, from the results of the soaking tests previously posted, Indian Rosewood or Elm will take much longer to become plasticised than Black Walnut.
For some unknown reason, Wood Alcohol has a (beneficial?) modifying effect and significantly slows down the action of the Ammonia solution on the wood fibres.

It should be noted that the force producing bending of the samples is not constant but reduces with increased deflection. This force - or more precisely bending moment - is the product of the applied weight and its horizontal distance (W multiplied by L) from point X the location of the maximum bending stress and will account in part for the gradual flattening of each curve after the yield point has been passed. So. for example the sample soaked in Ammonia solution on day 13 experienced a bending moment of about 70% of that on the first day. This does not affect the final results of the test.

It should be emphasised that the marinated wood only remains plastic when it is wet becoming hard (yet retaining some flexibility) after drying.
The attached image shows the sample that had been marinating in Ammonia solution for 13 days (and subject to 13 loading cycles) after being allowed to dry for 24 hours - compared to a dry untreated sample.
It is interesting to note that the sample has retained a 'memory' reflecting the internal stress levels experienced during the test. The dried sample is heavily fluted at the point of maximum bending stress (point X) as well as being slightly curved longitudinally - the degree of fluting diminishing towards the point of loading W. The fluting occurs naturally due to the balancing of the transverse and longitudinal internal stresses - all without use of a mold!
After drying the sample was found to have shrunk about 8 % in width and 11 % in thickness the length being unaffected.

For the past two weeks, some additional yield tests (using the test rig) have been under taken with the following results.

Test Sample #1 - Walnut sample, slab cut, marinated in water. The sample measured 0.045" thick X 1.5 " wide X 6" long (cantilever) - (1.15 mm X 38 mm X 153 mm) - as for the previous samples tested.
Test Sample #2 - Indian Rosewood marinated in Ammonia solution. Test sample dimensions as before.

The attached image shows a plot of deflection against time (14 days) for the two samples.
Sample #1 showed a slight increase in softening (2 mm deflection) after 1 day marinating in water, but no further softening thereafter.
Sample #2 showed a steady softening (10 mm deflection) after marinating for 5 days, but no further softening thereafter.
This result correlates the previous soaking tests posted that proved Indian rosewood to be largely unaffected by immersion in Ammonia solution.
On conclusion of the marinating/yield testing after two weeks, both samples were subjected to a boiling test. Each sample was immersed in boiling water and subject to yield testing. As the test samples were thin and quickly lost heat they were repeatedly subjected to immersion in boiling water for 1, 2, 3, 4, 5 and 10 minute periods - yield deflections measured after each immersion period (so the total immersion - in boiling water - of each sample was 25 minutes).
The attached plot demonstrates the significant relative softening effect of immersion in boiling water for both samples.

An additional test was undertaken on an un-marinated sample of Indian Rosewood - for comparison purposes. This sample was subjected only to boiling in water with deflections measured at intervals of 1, 2, 3, 4, 5 and 10 minutes - as for samples #1 and #2. The deflection curve is represented as A on the attached graph.
It would appear from this data that Indian Rosewood is only slightly softened by immersion in Ammonia solution - an increase of about 25% after 5 days of cold immersion. Boiling the sample then further increased the softening affect by about 250%. Therefore, the difference in softening effect of immersing a sample in ammonia solution for 5 days and then boiling the sample for 3 minutes compared to just boiling a sample in water for 3 minutes is about an 18% increase.

Thanks for the link. Unfortunately scientists can't really explain anything where practical experience is all that is necessary. I looked another day again at Alexander Batov's web site who seem to be able to bend fluted ribs from the most demanding exotic wood species: king wood, rosewood, ebony, cocobolo, satin wood. So I would imagine walnut would be just a breeze for him Just have a look if you've never seen it:

http://www.vihuelademano.com/index.html

In one of the previous discussions on fluted ribs on this forum I've already mentioned that I saw Alexander's vihuelas on the exhibition in London several years ago and they look absolutely stunning. I've never seen the work like that before. The very thought that he soaks the ribs in order to bend them would seem to be highly unlikely. Would you really want to soak king wood or cocobolo for example? Wouldn't it spoil the wood? I surely appreciate the work published on these pages although I'm not actually sure that the chosen method is the right one. Thanks for sharing your thoughts and ideas anyway.

An update.

While searching for some pieces of Persian boxwood in my stock of woods, I came across the Brazilian rosewood that I was looking for last November but was unable to find. So here - for information - is a better image of the cell structure than the one posted 0n 11-20-2008. The scratches running across the image were made by the knife blade used to prepare the section - my methods are still pretty rough and ready!.

On another more recent thread Peyman noted that an early Persian text recommended soaking instrument (carved) bowl blanks in milk to make the wood easier to carve. Moistening wood with water is a technique used by wood carvers to soften wood prior to carving so milk presumably is a better solution than just plain water - although milk is mostly comprised of water. Perhaps the solids or fats in the milk make a difference? No mention of the animal the milk comes from so perhaps that is not important.
Milk as a potential marinating fluid will, therefore, be added to the list for future testing. Thanks Peyman.

The next phase of testing will be to bend a full sized fluted rib in preparation for making a replica of the Diaz guitar. I know now that this can easily be accomplished using Black Walnut so will run some tests on the more difficult to bend Indian Rosewood. As is apparent, from previous trials, a brief marinating period in Ammonia solution (or Ammonia/Methanol) may provide some marginal advantage in softening Indian Rosewood but use of higher temperatures (100 C to 400C) will be more significant. Also previous trials indicate that the rib blanks should be 'cut on the quarter' with longitudinal grain 'run out' to reduce the tendency to split in the longitudinal direction. The grain run out introduces a potential longitudinal structural weakness that should, however, be more than compensated for by the stiffening effect of the rib fluting.

Use of higher bending temperatures will introduce a handling difficulty so the mold will have to be designed to facilitate a speedy, two stage, bending procedure.

360 degree I mean a circle, with the rib side parrallel, not tapered like an Oud.
My idea was to stack them into a cylinder to make a drum, which I will eventually

Decided to learn to play an Oud a wee bit before I build one.
So I cancelled that project for awhile.

But heres some ribs...I could use for an Oud.





I had them left from the Dias, they are rough 1.2mm thick.

I beleive I can get more bending out of them by reworking. Hand bending over a hot pipe takes practice.

For this the wood must be completely dry. The rib is held under the tap just until it is covered with water, then laid flat for severel minutes until it starts to bend on its own.

If it doesnt do this right away it may be placed under a heat lamp or...wht I use...the sun.

After this is seen it is merely helped on its way with heat.

In order to do this only certain cuts can be used easily.
The thickness to use will vary amongst the different woods and even different stock of the same species but not too much.
Around 1.2 is a good ballpark.
For example honduras would possibley have to be 1.1.

Using too much water or soaking the wood could cause it to crack.

The use of a steambox I have not tried.

I do not use a mold but sometimes "pinch" shape with a clamp for some of the more stubborn stuff.

Thank you for posting the images and description of your bending method AFBustard. Interesting that we are in agreement over the wood thickness (or rather thinness) to be used.
You do not explain what you mean by "certain cuts" that you indicate are necessary for success?

When I first opened this thread in November 2008, I mentioned that - when bending thin lute ribs (lute ribs are generally thin, - under 1.5 mm) - a natural fluting occurs due to the balancing of the longitudinal and transverse stresses in the rib. This natural curvature across the width of a hot bent rib may be slight or quite pronounced, dependant upon a number of factors.
The attached image is an example of this natural fluting in a rib blank hot bent (dry) in the conventional way on a propane heated bending iron to a lute rib profile. The rib material is figured maple, quarter sawn. No clamps or other devices were used to locally enhance the fluting diameter.
For comparison, the second image is of Walnut test sample #7, previously mentioned on this thread, - after marinating and forming on a mold to the longitudinal curvature and fluting diameter of the Dias guitar rib. After several months, this sample is stable having maintained its longitudinal curve and a uniform fluting diameter along its entire length.

The final objective of the experiments of this thread, is to try to produce oud ribs that have at least the same fluting curvature as the Dias guitar ribs - not the more shallow fluting that is the natural result of hot bending. Not an easy objective I would reckon. To be of practical use in making an oud bowl, the formed ribs will have to be precise and consistent in their geometry so will need to be formed using some kind of mold rather than being made 'free hand'.
Precise jointing of these fluted, 'extreme' curvature oud ribs will likely be a challenge in itself for the luthier.
Nobody said it was going to be easy!

This might be a good time to review the Dias rib geometry prior to constructing the experimental rib mold.

The RCM drawings are on paper - drawn to an accuracy of
0.5 mm. Each drawing has a reference scale of 1000mm along the length of the paper but not across. Paper expands and contracts along the 'grain' with changing humidity so, today, the paper has expanded to 1002.5 mm. This is such a small dimensional change that it can be ignored. Note, however, that the original instrument has suffered some distortion over time as represented by the drawing.
The drawings show three cross sections of the ribs at the top bout, sound hole centre and lower bout locations. From this it can be seen that the five fluted ribs not only all taper towards the neck but also reduce in the radius of their fluting. This radius reduction is consistent along the length of each rib - that is, the ribs do not flatten out in the centre.

The image shows a sketch, to scale, of the central rib measured at the bottom bout (with ivory purfling between each rib). The central rib has a longitudinal curve of 8370 mm (34 inches) and a length - measured along the length of the curve - of 365 mm (14.4 inches). The rib width W1 is about 30 mm at the lower bout and about 25 mm at the upper bout. The inner radius of the fluting at the lower bout is about 18 mm reducing to 16.5 mm at the upper bout.
In contrast, the outer ribs of the group of five have a fluting radius of about 18 mm at the lower bout reducing to 11 mm at the upper.
So the ribs should be formed on cone shaped mold(s) or mandrel(s) in order to maintain the consistent geometry necessary for accurately constructing the vaulted back. There will, of course, be some variations as each rib has to be individually fitted to its neighbour.

For the next stage of this investigation, full sized ribs of the Dias guitar geometry will be made (or attempted to be made!). The first trial will be to form a rib using a marinated blank heated to 100 C. This will first be tried with a Walnut blank followed by the more difficult to bend Indian rosewood at higher temperatures using a heated iron.
The first step is to make a mold to the geometry of the central Dias rib using the dimensions taken the RCM drawings.
The mold is made from Ash wood.
The longitudinal curve of the mold has been accurately transferred from the drawings using draughting curves. The curve of the mold has been cut to a radius of 30 inches (760 mm) ,rather than the original 32 inches (internal radius), to allow for some anticipated slight 'spring back' of the formed blank when released from the mold.
The mold is tapered from front to back to conform to the original guitar rib geometry.
At this point the mold has been roughly contoured using a half rounding bit in a router to remove most of the waste. The final fluting profile will be achieved by hand using files etc. with card templates to check the fluting radius at the three positions given by the museum drawings.

A wooden mold will be satisfactory for the higher bending temperatures using this procedure.
Note that the fluted ribs of the so called Chambure vihuela in Paris (also made by Dias?) have scorch marks on the surfaces of the ribs (i.e. the convex surfaces facing inside the instrument).