Mechanics and drape

The quietest shafts belong at the back

Which end of a warp gets the hardest ride is the product of how often it changes sides and how far its shaft is from the fell, and the threading is the one free choice that pairs them. The best threading sorts the shafts by how often they move, busiest in front, and an exhaustive walk of a damask check's 1,820 placements confirms it: the straight draw sits at the worst value, shared by every threading that leaves a busy shaft at the back, and sorting cuts the worst end's work by a fifth. The heddles' own travel wants the same order almost everywhere. The exception is the selvedge, which is small, busy and wants the front, against a body whose weight wants it there too.

Worth reading first: A figure is harder on its warp · The back shaft works hardest · Where the heddles go.

The back shaft works hardest set out a product. An end is strained each time it changes sides, by an amount that grows with its shaft’s distance from the fell, so the work the loom does on it is its flex count times its shaft’s strain. The first factor comes from the weave and the second from the machine, and the only thing that pairs them is the threading, which nobody chooses for this purpose. A figure is harder on its warp found the flex counts of a figured cloth spreading unevenly across its shafts, and ended by proposing a threading chosen to level the product, with a warning that it would probably fight the threading a weaver chooses to share the heddles out.

Both halves can now be settled. The threading that levels the warp is a sort, and on a figured cloth the straight draw a weaver writes first is the worst one available. The heddles’ own work agrees with the sort almost everywhere. The one place it does not is the selvedge, and there the disagreement can be priced.

Every way of placing a damask check's sixteen shafts, by the work its hardest-worked end does. A damask check of an eight-end satin figure on an eight-end sateen ground needs sixteen shafts: twelve whose ends change sides 10 times a repeat and four whose ends change 6 times. Shafts of one count are interchangeable, so there are 1820 distinct placements, and each is drawn by the work on its worst end, flex count times the strain of that end's shaft. They run from 90.83 to 112.97. The straight draw, threading the shafts in the order the draft first uses them, is at 112.97, the worst of all, with 455 placements better. Sorting the shafts by flex count, most-flexed forward, gives 90.83, the best. The chart cannot show what a weaver gains in reading the draft.
Fig. 1 Every way of placing a damask check’s sixteen shafts in its harness, by the work its hardest-worked end does. The straight draw sits at the worst value; sorting the shafts by how often their ends change sides gives the best.

The threading that lowers the worst end

Every end on one shaft has the same column, so it changes sides the same number of times in a repeat. A shaft therefore carries one flex count, and a harness position carries one strain, rising from the front to the back. A threading assigns shafts to positions, and the worst end is the shaft whose flex count times its position’s strain is largest.

The least largest product has a classical answer. Pair the largest flex count with the smallest strain, the next largest with the next smallest, and so on down the harness: sort the shafts by flex count, busiest at the front. That is the rearrangement inequality, and the exchange argument behind it is short. If a busier shaft stands behind a quieter one, swapping them moves the busier one to a smaller strain and the quieter one to a larger, and the larger of the two products cannot go up. Repeat until no such pair remains and the shafts are sorted.

A proof is not a check, so the damask check from the figured-cloth essay was walked whole. Its sixteen shafts fall into two classes, twelve whose ends change sides ten times a repeat and four whose ends change six times, where the ground and the figure meet at the block edges and some columns lose a change. Shafts of one class are interchangeable, so the placements that matter are the ways of choosing four of sixteen positions for the quiet ones: 1,820 of them. The sorted threading has the least worst end of all 1,820, at 90.83 in the units the earlier essays use, and the walk confirms it rather than taking it on trust.

The worst end is the back shaft’s

The first figure has a feature the proof does not predict. Three placements in four sit at exactly the same worst value, 112.97, at the right-hand end of the chart, and the straight draw is one of them.

The reason is that the back shaft has the largest strain of all, and whenever a busy shaft lands there, its product is the worst in the harness regardless of anything else. Of the 1,820 placements, 1,365 put one of the twelve busy shafts at position sixteen, and all 1,365 share the worst value. Only the 455 that put a quiet shaft at the back do better, and how much better depends on what is next to it.

So the practical rule is cruder than a sort and almost as good: look at what goes on the back shaft. The straight draw of a figured cloth threads the shafts in the order the draft first uses them, and for the damask check the last shaft that order reaches is a busy one. It lands at the back because it was used last, not because anybody weighed it.

The work on an end at each harness position of a damask check, threaded straight and sorted by flex. For a damask check on 16 shafts, the work on an end at each harness position — its flex count times that position's strain — under the straight draw and under the threading that sorts shafts by flex count. Straight: 27.6, 49.4, 52.8, 56.4, 60.2, 38.4, 68.1, 72.2, 46.0, 81.1, 85.9, 90.8, 96.0, 60.8, 107.0, 113.0, a spread of 4.09. Sorted: 46.0, 49.4, 52.8, 56.4, 60.2, 64.0, 68.1, 72.2, 76.6, 81.1, 85.9, 90.8, 57.6, 60.8, 64.2, 67.8, a spread of 1.97. Large dots are the most-flexed shafts. The chart cannot show the order in which a weaver would read the threading off the draft.
Fig. 2 The work on an end at each harness position of the damask check, threaded straight and sorted by flex count. The straight draw scatters its four quiet shafts through the harness and leaves busy ones at the back; the sorted threading runs the twelve busy shafts up the front and puts the four quiet ones behind them.

Sorted, the twelve busy shafts take positions one to twelve and the four quiet ones thirteen to sixteen. The worst end is now the twelfth shaft’s, 90.83, a fifth less than the straight draw’s 112.97. The quiet shafts at the back do 57.6 to 67.8, less than the busy ones in front of them, because a six-change column at the back of the harness does less than a ten-change column halfway along. The spread across the warp falls from 4.09 to 1.97.

Nothing about the cloth changes. Every end still makes the same interlacing in the same place; the drawdown, the lifting plan read with the new threading, and the fabric off the loom are the same. A lifting plan says nothing without a threading, and here the converse is the point: the lifting plan is rewritten to match, the pair expresses the same cloth, and only which ends sit where in the harness has moved.

A stripe has nothing to sort

The satin stripe on a plain ground that the back-shaft essay used is a different case, and the difference is instructive.

The work on an end at each harness position of a satin stripe on plain, threaded straight and sorted by flex. For a satin stripe on plain on 10 shafts, the work on an end at each harness position — its flex count times that position's strain — under the straight draw and under the threading that sorts shafts by flex count. Straight: 36.8, 39.5, 10.6, 11.3, 12.0, 12.8, 13.6, 14.4, 15.3, 16.2, a spread of 3.74. Sorted: 36.8, 39.5, 10.6, 11.3, 12.0, 12.8, 13.6, 14.4, 15.3, 16.2, a spread of 3.74. Large dots are the most-flexed shafts. The chart cannot show the order in which a weaver would read the threading off the draft.
Fig. 3 The work at each harness position of a satin stripe on a plain ground, threaded straight and sorted. The two lines lie on each other: the straight draw already has the busy plain shafts in front, and the step between the plain and satin groups sits between two neighbouring shafts.

Its two plain shafts change sides eight times a repeat and its eight satin shafts twice. The straight draw already threads the plain shafts first, because the plain ground is what the draft uses first, so the sorted threading is the straight draw and the spread stays at 3.74. That is not an accident of this stripe. A stripe uses its weaves in blocks, and threading in order of first use keeps each block’s shafts together and puts whichever block comes first at the front.

And no threading could do better here, because the two groups meet at neighbouring positions. Whatever the order, the quietest busy shaft and the busiest quiet shaft must stand next to each other somewhere, and there the product drops by the whole flex ratio, four, against one shaft’s step in strain, about seven per cent. A threading can absorb a flex spread only up to the harness’s own strain ratio — 1.76 from front to back on ten shafts, 2.45 on sixteen — and a stripe whose flex counts differ by four has no threading that levels it. The damask check could be brought under a spread of two because its two counts differ by only five to three, well inside its sixteen shafts’ 2.45.

Where the heddles agree

Where the heddles go treated the threading as an allocation of heddles and asked how to share them out. The lifting has its own cost: every time an end changes sides its heddle travels through its shaft’s lift, which grows towards the back in step with the strain. Summed over a repeat, the harness moves through a total heddle travel of each shaft’s ends times their flex count times their lift.

The least heddle travel is again a sort, by a different key: the shafts with the most ends-times-flexes go to the front. On the damask check every shaft carries two ends, so the key is the flex count and the two sorts are the same; the sorted threading cuts the travel by three per cent as well as the worst end by a fifth. On the satin stripe the plain shafts carry four ends each and flex four times as often, so they lead on both keys. The two measures disagree only when a shaft’s heddle count and its flex count rank opposite ways, and on most warps they do not, because a structure that is used a lot tends to be used across a lot of the cloth.

The figured-cloth essay expected a conflict and the arithmetic finds one only at the margin. There is one shaft on almost every loom where heddle count and flex count do rank opposite ways.

The selvedge disagrees

A selvedge is a narrow band at each edge woven in a structure that holds the turning weft, very often plain weave on two shafts of its own while the body of the cloth is something looser. Its ends are few and they change sides at every pick. The body’s ends are many and, in a twill or a satin, change sides less often.

So the worst end wants the selvedge at the front, where its busy ends meet the least strain, and the heddle travel wants the body at the front, because the body’s shafts carry far more ends.

A selvedge on the front shafts and on the back ones, by three measures. A 2/2 twill body on 4 shafts of 400 ends each, with a plain selvedge on two shafts of eight ends each, threaded with the selvedge in front and then behind. The selvedge ends' work is 19.75 in front and 25.61 behind, 30 per cent more. The body's worst end does 12.81 and 11.29, 12 per cent less with the selvedge behind. Heddle travel a repeat is 193.2 and 176.8 metres, 8.5 per cent less. The chart cannot show which ends actually break.
Fig. 4 A 2/2 twill body on four shafts of 400 ends each and a plain selvedge on two shafts of eight ends each, threaded with the selvedge in front and behind. Behind, the selvedge ends work 30 per cent harder; the body’s worst end does 12 per cent less and the harness travels 8.5 per cent less.

Take a 2/2 twill body on four shafts of 400 ends each, and a plain selvedge of eight ends on each of two shafts. In front, the selvedge ends do 19.75 and the body’s worst end 12.81, the harness travelling 193 metres of heddle a repeat. Behind, the selvedge ends do 25.61, thirty per cent more, the body’s worst end does 11.29 because the body has moved forward two places, and the travel falls to 177 metres, eight and a half per cent less.

That is a real trade between two different things. Heddle travel is energy and wear on the harness, paid by the loom on every pick across the whole warp. The worst end’s work is the strain on sixteen particular ends, paid by those ends. A warp breaks in a tail no tensile test reaches found that an end’s chance of breaking rises steeply with its shaft’s tension, so a thirty per cent rise in strain on the selvedge is not a thirty per cent rise in its breaks but a much larger one, on the ends whose breakage a weaver notices first because a broken selvedge stops the loom.

Deeper harnesses make it worse

The trade moves with the harness. A body on more shafts pushes a selvedge behind it further back, and spreads its own weight over more positions.

What moving the selvedge behind costs its ends and saves the harness, against the depth of the body. For a 2/2 twill body on two to sixteen shafts of 400 ends each and a plain selvedge on two shafts, the extra work on the selvedge ends when their shafts go behind the body instead of in front of it, and the heddle travel saved: 2 shafts, +14% and −8.7%; 3 shafts, +22% and −8.6%; 4 shafts, +30% and −8.5%; 6 shafts, +46% and −8.2%; 8 shafts, +64% and −7.9%; 10 shafts, +84% and −7.6%; 12 shafts, +105% and −7.3%; 14 shafts, +129% and −7.0%; 16 shafts, +155% and −6.8%. The cost grows with the harness, since the selvedge is pushed further back, and the saving shrinks, since the body's own shafts spread over more positions. The chart cannot show how a weaver would value one against the other.
Fig. 5 The extra work on the selvedge ends when their shafts go behind the body instead of in front, and the heddle travel saved, against the number of shafts in the body. The cost rises from 14 per cent on two body shafts to 155 on sixteen; the saving falls from 8.7 per cent to 6.8.

With a body on two shafts the selvedge pays fourteen per cent for a saving of 8.7. On four it pays thirty for 8.5, on eight sixty-four for 7.9 and on sixteen a hundred and fifty-five for 6.8. The cost grows and the saving shrinks as the harness deepens, so a selvedge behind a twill body on four shafts is a defensible choice and one behind a sixteen-shaft satin body is not. On a sixteen-shaft body there is a second consequence: the body’s own back shaft is now the worst end in the warp even with the selvedge in front, at 25.14, so moving the selvedge behind it changes which end is worst as well as by how much.

The saving also depends on how wide the warp is, because the body’s weight is what buys it.

The heddle travel saved by putting the selvedge behind, against the ends on each body shaft. For a 2/2 twill body on four shafts and a plain selvedge of eight ends on each of two shafts, the share of heddle travel saved by threading the selvedge behind the body rather than in front, against the number of ends on each body shaft, on a logarithmic scale: 8 ends, -4.8%, 14 ends, -0.9%, 20 ends, 1.3%, 40 ends, 4.6%, 100 ends, 7.1%, 300 ends, 8.3%, 1000 ends, 8.8%. It changes sign at 16.0 ends a shaft, where the body's ends times their flex count equal the selvedge's, and approaches 8.8% on a wide warp. The chart cannot show a sample warp, where the body is narrow.
Fig. 6 The heddle travel saved by putting the selvedge behind, against the ends on each of four body shafts, on a logarithmic scale. Below sixteen ends a shaft the selvedge outweighs the body and behind costs travel; above it the saving climbs towards nine per cent.

The two orders tie exactly when a body shaft’s ends times its flex count equals a selvedge shaft’s, sixteen ends here, since the body flexes half as often. A sample warp a few centimetres wide is below that and should have its selvedge in front on every count. A production warp of thousands of ends is far above it, and the saving there approaches nine per cent.

The practical reading

Sort the body. Thread the shafts that change sides most at the front and the quietest at the back, and if nothing else is possible, make sure no busy shaft is the last one. On a figured cloth that alone takes a fifth off the worst end, without touching the fabric.

Leave a stripe alone. Its straight draw is already sorted, and its flex counts usually differ by more than the harness can absorb.

Decide the selvedge on purpose. On a shallow harness and a wide warp, putting it behind saves the loom eight per cent of its lifting at the cost of a third more strain on sixteen ends; on a deep harness it costs those ends double or more for less. Which is the better trade depends on what a selvedge break costs against what the lifting costs, and that is a mill’s number, not a geometric one.

How the numbers were reached

The flex count of each end is the number of times its column changes value round the repeat, read off the draft. The strain at each harness position is the shed’s exact extension on an ordinary broad loom: fell to first shaft 300 mm, sixteen millimetres between shafts, fell to back rest 1,200 mm, a 30 mm shed at the reed. The heddle lift is the same geometry’s lift at each position. Work is flex count times strain, in the thousandths the earlier essays used; heddle travel is ends times flex count times lift, summed, in metres a repeat.

The damask check is the eight-end satin figure on an eight-end sateen ground, in blocks of eight, with both weaves started at the same origin. Its placements were enumerated by class, every way of choosing four positions of sixteen for the quiet shafts, and the sorted threading was required to equal the least of them. The selvedge comparison uses two plain shafts and a 2/2 twill body, with the body’s flex count half the selvedge’s, as it is for those two weaves over a common repeat of four picks.

Where the product stops

The work is a strain count, not a force. What the shed costs in newtons turned the strain into a tension, and the flex count multiplies how often an end feels it, but fatigue in a yarn is not linear in either, and the product ranks ends rather than predicting when they fail.

Only the threading moves. A loom with the shed opened by a different amount at each shaft — a staggered or levelled shed, set so the back shafts lift less — changes the strain profile itself, and would change every number here. That is a loom setting rather than a threading and is not computed.

The heddle travel is a proxy for the lifting. It counts distance moved and leaves out the masses, the springs and whatever the dobby or tappet mechanism does, and two harnesses with the same travel can cost different energy.

The selvedge is two plain shafts. Many selvedges are woven on the body’s own shafts, as a weave that holds its selvedge with two of its own columns does, and then there is no separate shaft to place and no trade to make.

Who worked out which part

The shed’s strain and its rise towards the back are standard weaving mechanics, and so is the practice of putting heavily used shafts forward and of choosing a selvedge’s shafts with care. The product of flex count and strain is the earlier essays’ own. The rearrangement inequality is classical, from Hardy, Littlewood and Pólya’s Inequalities of 1934 and older in substance.

What is added here is the sort as the answer to the levelling question, checked by exhaustion on a figured cloth where the straight draw turns out to be the worst threading; the observation that a stripe cannot be levelled because its blocks meet at neighbouring shafts; the agreement of the worst end and the heddle travel everywhere except where heddle count and flex count rank opposite ways; and the price of that exception at the selvedge.

Still open: whether a staggered shed levels what a threading cannot

The satin stripe could not be levelled because its flex counts differ by four and a ten-shaft harness’s strains differ by 1.76 from front to back. The threading can only redistribute what the harness offers. A loom can be set so that each shaft lifts a different amount, the back shafts less, which is how a weaver keeps a clean shed on a deep harness, and that reshapes the strain profile directly. Whether a stagger chosen for the stripe’s own flex counts could flatten its product — lifting the satin shafts’ quiet ends more and the plain shafts’ busy ones less — is a question with the same arithmetic and one more free variable, and the constraint it would have to meet is that the shed still opens wide enough at the reed for the weft to pass.

What links here

Computed from the collection rather than written here: the essays that point at this one.

Shares its objects with

Essays naming at least two of the same things, that neither author linked.

Named objects

A flat tag is an object no other essay names yet.

HarnessHeddleInterlacingSelvedgeShaftShedThreading