A second foot saves treadles only on the heavy face of a cloth
Worth reading first: What a dobby stores · How many shafts a draft needs · Three mistakes and the shape each one leaves.
What a dobby stores is lifts — the distinct sets of shafts a draft ever raises — and that essay set a treadle loom beside the dobby in one sentence: a treadle is a lift with a foot attached, the tie-up is the store, and the number of treadles is the chain capacity of a loom that has no chain. A draft that needs eight lifts needs eight treadles. A hand loom with four shafts usually has six treadles under it and one with eight shafts usually has ten, so the store runs out long before the harness does, and the harness has a depth only on looms that have already outgrown their feet.
The sentence is right for one foot. Weavers have two, and on a rising-shed loom — a jack loom, where each treadle pushes up the shafts tied to it and nothing pulls any shaft down — pressing two treadles together lifts every shaft tied to either. That turns the tie-up from a list of lifts into a set of pieces, and every lift is either a piece or two pieces put together. The trade calls it a skeleton tie-up, and it is how hand weavers working eight or more shafts get drafts with many lifts onto ten treadles.
So the question the one-foot count left open has an exact form: given the lifts a draft needs, what is the smallest set of treadles from which each lift is one treadle or the union of two? It can be answered by search, and the answer turns on something the count does not see. A union only adds. Pressing a second treadle can put shafts up; it can never leave one down that the first treadle raised. So the second foot builds large lifts out of small treadles, and it cannot build a small lift out of anything.
Five treadles where one foot needs eight
Take the 7/1 twill: seven up, one down, the down shaft stepping across the harness one place a pick. Its eight lifts are the eight sets of seven shafts, one for each shaft left down — the same eight lifts, in a different order, as an eight-end warp satin, so everything below about the twill is true of the satin too. One foot needs a treadle for each, eight in all.
Two feet need five. The tie-up above is the smallest the search finds, and the treadles in it are not lifts of the cloth at all: each is tied to four, five or six shafts, and no single one of them is ever pressed alone. Every pick presses two, and the pair lifts exactly seven shafts because the two treadles between them leave exactly one shaft untouched, the one the pick is meant to leave down.
That is the whole mechanism, and it is easier to see from the shafts that stay down than from the ones that go up. Each treadle leaves some shafts alone. A pair leaves alone only the shafts that both of them leave alone, so the shaft a pick keeps down has to be the one shaft the two treadles’ gaps have in common. Five treadles have ten pairs among them, and eight of those pairs are made to share exactly one gap each, a different one every time. Four treadles have six pairs and four singles, ten slots in all, which on a count alone would be enough for eight lifts. The search exhausted every family of four and none of them works: there is no way to give eight shafts eight distinct single shared gaps from four sets.
The arrangement was found, not designed, and it is not the only one. What is fixed is the number, five, and the fact that nothing smaller exists.
The same cloth face down needs every treadle
Now turn the cloth over. A 7/1 twill woven face down is a 1/7 twill: each pick lifts one shaft and leaves seven down, and the cloth that comes off the loom is the same fabric, read from the other side. Many hand weavers weave warp-faced cloth exactly this way, because lifting one shaft a pick is a seventh of the work of lifting seven.
Face down it needs all eight treadles, two feet or not. A lift of one shaft can be made from a pair only if both treadles of the pair are tied to nothing but that shaft, and then the pair is the same treadle twice. So a single-shaft lift is always a treadle of its own, and the second foot has nothing to do.
The cloth is identical, the draft is identical up to turning it over, and the treadle count is eight on one side and five on the other. Nothing in the matrix of the weave distinguishes them, which is the point a balanced twill need not have two equal faces made about the cloth’s two surfaces from a different direction: the matrix and its complement are one fabric, and anything that depends on which of them the loom actually lifts is a property of the weaving and not of the cloth. Effort is one such property and has always been the reason for choosing a face. The treadle count is another, and it points the other way.
A balanced twill gets nothing at all
The middle case is the instructive one. A 4/4 twill lifts four shafts of eight on every pick, a run of four stepping across the harness, and it is its own complement, so face up and face down are the same problem.
It needs eight treadles with two feet. A run of four shafts is easily written as two runs of two — shafts one and two with three and four — and the obvious skeleton ties eight treadles to the eight adjacent pairs and presses two at a time. That works, and it saves nothing: eight pairs for eight lifts. The search confirms that nothing better exists. Every lift can be split, but no split is shared widely enough between the lifts to bring the count down, because each lift is a run of four in a different place and the pieces that build one run are not pieces of the others.
So the 4/4 twill, the most even weave on the harness, is exactly where the second foot stops paying. What it needs is lifts with so many shafts in them that a small number of large treadles can cover them all by overlapping, and a lift of half the harness is not yet large enough.
The whole four-shaft catalogue
At four shafts the question can be put to every cloth there is. Every cloth at four by four counted 22,874 four-by-four drafts in which every thread interlaces, and what a dobby stores split them by lifts: 98 need two, 5,184 need three, and the remaining 17,592 need four. One foot needs exactly that many treadles.
Two feet take 5,616 of them down to three treadles, a third of the four-lift drafts, and touch nothing else. No draft with two or three lifts can save. Two lifts need two treadles whatever the feet do, and three lifts from two treadles would need one lift to be the union of the other two. Every shaft is lifted by some pick, so that union would lift every shaft, and the pick under it would pass under every end — which an interlacing draft forbids. No draft at all saves two treadles, since two treadles reach at most three lifts.
Turning the cloth over changes the count for 6,912 drafts: 3,456 need fewer treadles face up than face down and 3,456 the reverse. The two numbers are equal because turning a draft over is its own inverse, so the drafts that are cheaper face up are exactly the turned-over copies of those cheaper face down. A four-shaft weaver who chooses a face for ease of lifting is choosing a treadle count too on almost a third of everything that can be woven.
The four-shaft saving is small, one treadle in four, and it is unlikely to decide anything at a loom with six treadles under four shafts. It is the eight-shaft case where it matters, because there the one-foot count of eight is near the ten treadles a loom has, and a draft with a tabby or a border needs more.
Seven twills of twenty-one, and all on one face
How many twills a repeat admits found that eight ends hold 21 twills once the notation’s redundancies are taken out. Three of them repeat on four ends or fewer and need four lifts or two. The other eighteen need eight lifts and so eight treadles with one foot, and each can be woven face up or face down.
Seven of them save treadles, every one of them face down, and every saving is on a face that lifts five shafts or more. The 3/5, 2/6, 1/7, 1/1/1/5, 1/2/1/4, 1/1/1/1/1/3 and 1/1/1/2/1/2 twills are weft-faced as written, and woven with their warp face up — lifting five, six or seven shafts a pick — they need six treadles, or five for the 1/7. Woven the other way, lifting one to three shafts, every one of them needs eight. Nothing that lifts four shafts or fewer saves a treadle, whichever face and whichever twill.
The heavy face is necessary and not sufficient. Four faces lift five shafts and still need eight: the 1/2/4/1 twill face up, and the 1/1/2/4, 1/2/2/3 and 1/3/2/2 face down. What separates them from the savers is the arrangement of the five. A single run of five, as in the 3/5 face down, splits into overlapping shorter runs that neighbouring lifts can share; five made of a run of one and a run of four, or of two and three, need their pieces in places the other lifts do not use. The census says which; it does not offer a rule that predicts it, and none is claimed here.
The direction of the whole result is what was predicted from the union, and it holds without exception across the eighteen. The second foot can only add, so it pays where the lifts are big.
How the saving grows with the harness
The cleanest family is the lifts that each leave exactly one shaft down — a 7/1 twill’s, an eight-end warp satin’s, a 3/1 twill’s on four shafts — because there is one such set for every harness size, and the question is simply how many treadles n of them need.
Three treadles for three shafts, four from four shafts to six, five from seven to nine. Nine lifts of eight shafts each come from five treadles. The count rises far more slowly than the harness, which is the second foot’s whole value made visible: on a big harness the warp face of a satin or a steep twill becomes cheap in treadles, and the weft face stays at one treadle a shaft. At four shafts it saves nothing, which is why the 3/1 twill is not among the four-shaft savers in the catalogue above even though it is warp-faced: four lifts of three shafts need four treadles either way.
A count of pairs puts a ceiling on all of this. Five treadles have five singles and ten pairs, fifteen lifts at most; ten treadles reach fifty-five. The real minima are nowhere near the ceiling, because a pair has to make exactly a lift the draft needs and not merely some new lift, and the two numbers should not be confused. A ten-treadle loom is not a fifty-five-lift loom for any draft examined here.
What the tie-up is a factorisation of
Four ways to write a weave down described the full draft as a factorisation: threading times tie-up times treadling gives the cloth. A skeleton tie-up factorises once more. The tie-up is now a small set of pieces, the treadling says which one or two pieces each pick takes, and the lifting plan is their union rather than a lookup. Like every factorisation of a draft it has redundancy — the 7/1 twill has more than one five-treadle skeleton — and like the threading it is invisible in the cloth.
It also has a failure the one-foot tie-up does not. Three mistakes and the shape each one leaves priced an error in the tie-up as a lattice across the whole cloth, because a wrong tie on one treadle is wrong on every pick that treadle makes. In a skeleton tie-up most treadles serve several lifts, so one wrong tie lands on every pick that presses that treadle in any pair — between two and five lifts of the eight for the 7/1 above, rather than one. The tie-up gets smaller and each tie in it carries more of the cloth.
The trade between treadles and effort
A weaver at a jack loom has two reasons to prefer a face and they now point in opposite directions. Lifting fewer shafts is lighter, and on a long warp the difference between one shaft a pick and seven is the difference between a comfortable day and a tiring one. Lifting more shafts lets two feet cover the lifts with fewer treadles, and on an eight-shaft loom with ten treadles that is the difference between a draft that fits and one that does not.
For the warp-faced twills and satins the arithmetic is now exact. A 7/1 face down costs eight treadles and one shaft a pick; face up it costs five treadles and seven shafts a pick, with both feet on every pick. A 2/6 or 3/5 costs eight treadles on its light side and six on its heavy side. For the other eleven eight-end twills with eight lifts — the balanced ones among them — there is nothing to trade: eight treadles on either face.
That gives the harness does not grow a counterpart under the loom. A reversal spends threading rather than shafts and a lifting sequence spends order rather than lifts; a skeleton tie-up spends feet rather than treadles, and the exchange rate is set entirely by how many shafts the lifts contain.
What the count leaves out
The loom is a jack loom. On a counterbalance loom the shafts are hung in pairs over rollers, so a treadle that sinks one shaft raises its partner, and pressing two treadles fights the mechanism. On a countermarch loom every treadle is tied to every shaft, up or down, and two treadles pressed together give contradictory instructions. The union arithmetic belongs to the rising shed alone.
Effort and reach are not counted. Pressing two treadles takes both feet, and a pair far apart on the treadle bank is awkward; a real skeleton tie-up is arranged so that the pairs a draft uses most are adjacent and the treadling can be walked. The minimum here is a minimum over all arrangements and says nothing about whether the smallest one is comfortable. Lifting weight rises with the shafts lifted and with the warp tension, and neither is modelled.
Only straight draws and only lift sets are examined. A twill on a point or broken threading has the same lifts as on a straight one, so the counts carry over, but a draft with blocks or with a tabby has lifts beyond the twill’s and needs its own search. Tabby lifts are half the harness each, and whether they can share pieces with a twill’s lifts is a separate count for every draft.
The ten-shaft case is not computed. The search finishes for nine shafts in a quarter of a minute and did not finish for ten in the time it was given; the pattern of three, four and five treadles above is a measurement up to nine, not a rule beyond it.
How the minimum was found
The lifts of a twill on n shafts with a straight draw are the n cyclic shifts of its word, each a set of shafts; woven face down they are the complements. For the four-by-four catalogue they are each draft’s distinct rows, written over its distinct columns as shafts.
The search finds the smallest family of treadles, each a set of shafts, such that every lift equals one treadle or the union of two. It works upward from one treadle, takes the smallest lift not yet made, and branches on making it as a treadle of its own or as the union of two subsets of it, adding whichever are new. It is pruned by a count that cannot be beaten: r new treadles beside t old ones make at most r + rt + r(r − 1)/2 new lifts. The first budget that succeeds is the minimum.
Required of it, not shown: that it agree with a search over every family of treadles drawn from the subsets of the lifts, on four hundred lift sets from the four-shaft catalogue and on the five-shaft single-gap lifts; that every lift of the reported tie-up is made, and made by a treadle or a pair; that two feet never need more treadles than one; and that a set of single-shaft lifts needs one treadle each.
Who found which part
Skeleton tie-ups are long-standing practice among hand weavers working many shafts on jack looms, and the habit of weaving warp-faced cloth face down to lighten the lift is older still; neither is claimed here, and this essay does not know who first tied a skeleton. The counting of drafts and their lifts follows the four-by-four census and the twill count.
What is added here is the observation that a pair of treadles can only add shafts, and its consequences counted exactly: five treadles for the eight lifts of a 7/1 twill or warp satin and eight for the same cloth face down; three to five treadles for single-gap lifts on three to nine shafts; 5,616 four-shaft drafts that save one treadle and none that save two; and seven eight-end twills that save, every one on the face that lifts five shafts or more.
Still open: whether a comfortable skeleton costs a treadle
The minimum is a minimum over every arrangement, and a weaver would not accept most of them: five treadles whose pairs are scattered across the bank would have the feet crossing on every other pick. What the trade actually wants is the smallest tie-up whose pairs are adjacent treadles, or at least within a stride, so the treadling can be walked left to right and back.
That is the same search with one more condition — a pair may only be two neighbouring treadles in some order of the bank — and it has a prediction worth testing. If the adjacency costs nothing on the 7/1 twill, five treadles stay five and the skeleton is free; if it costs one, the warp face’s advantage shrinks from three treadles to two, and the heavy-lifting face stops being worth choosing on a ten-treadle loom for any of the six-treadle twills above. Which of the two is true is a count, and nobody has taken it.
Shares its objects with
Essays naming at least two of the same things, that neither author linked.
- A second wrong digit is silent only by cancelling the first — both name census, harness, lifting plan, shafts
- Where the heddles go — both name census, harness, lifting plan, shafts
- A crepe cannot be structureless — both name census, satin, twill
- A jacquard is every end its own shaft — both name harness, lifting plan, shafts
- A turned block is a moved origin — both name census, satin, twill
- An eight-shaft error hides in the row's mirror — both name lifting plan, satin, twill
Named objects
A flat tag is an object no other essay names yet.