The second tone forbids the only join into the centre
Worth reading first: Only one start in six can reach the shallowest shading · An eight-end shading is pinned at three everywhere but its centre · An even shading cannot keep its surface level.
A shaded damask passes from one face of a weave to the other in tones, each a weave with one more raised cell in every pick and every end than the last; the faces are usually a satin’s, though a tone step does not need a satin. On eight ends there are seven tones, and the ones that matter are the five in the middle, held to floats short enough to weave: three at the second, third, fifth and sixth tones and two at the centre. Only one start in six can reach the shallowest shading counted every such chain — 1,001,574,400 of them — and found that the depth each one sinks is decided by its centre alone.
That simplification rested on a pattern nobody had a reason for. In all billion chains, the tones either side of the centre never interlace more than the centre does. They tie it in one chain in eight and otherwise fall below. Since the deepest tone of a ramp is the one that interlaces most, a shoulder that out-interlaced the centre would set the depth, and none ever does.
The census closed by naming a route to a proof and a question about it: whether the rule holds pick by pick, or only on the whole. The answer is that it holds thread by thread on both systems — but not for the reason the route proposed, and not on a single pick taken alone. It holds because of the second tone.
Three things one cell can do
A thread’s interlacings are its changes of face, which round a repeat are twice its number of runs of raised cells — the quantity interlacings and firmness made the first measure of a weave. A shading step raises one cell in that thread, and the cell can land in only three kinds of place.
Beside a run, with an unraised cell on its other side, it lengthens the run. The number of runs is unchanged and so are the interlacings. In a gap of exactly one it fills the gap and two runs become one: two interlacings fewer. In a gap of three or more, away from both edges, it starts a run of its own: two interlacings more.
So whether a tone interlaces more than the one below it is a count, thread by thread, of joins against cells standing alone. A shoulder could beat the centre only if, on the step into the centre, some thread joined more than it stood alone — and the question is what stops that.
The bounds read one face each
The float bounds are where the answer comes from, and they need reading exactly. A shading’s floats are its face floats, as the float decides defined them for every weave here: down an end, the runs of raised cells, where the warp lies on the face; along a pick, the runs of unraised cells, where the weft does. Each system’s bound reads one face. A pick may lie under a long run of raised ends — that is a float on the back, and a float presses on nothing whichever face it is on — and nothing in the bound objects.
That asymmetry is what defeats the obvious proof. A bound on both faces would forbid a join outright in either system, since a join makes a run of raised cells at least three long. With one face each, it forbids a join down an end and says nothing directly about one along a pick.
Down an end, the centre forbids the join
Take the step from the third tone to the centre, down an end. A join there fills a gap of one between two runs of raised cells, each at least one long, so the run it makes is at least three. The centre’s warp floats are held to two. So no end joins on its way into the centre, ever, whatever the end looked like at the third tone.
Every end therefore keeps its interlacings or gains two. That is half the proof, and it costs one sentence.
Along a pick, one shape can join
A pick is harder, because the centre bounds its runs of unraised cells, not its raised ones, and a join creates a long raised run that no bound sees. But the bound on the unraised runs still constrains what a joining pick can look like.
The centre’s pick has four unraised cells and none of its gaps may be longer than two, so it has at least two gaps — and therefore at least two runs of raised cells. A join removes one run. So the third tone’s pick had at least three runs, and with only three raised cells, it had three single cells. Its five unraised cells lie in three gaps; the join filled a gap of one, and the other two gaps must each be two or less after the step, which with four cells between them makes both exactly two. The only third-tone pick that can join into the centre has raised cells two, three and three apart round the repeat — X.X..X.., in some rotation or its mirror.
Over every eight-cell pick there are eight single-cell steps that join into the centre, and every one of them starts from that shape. Taken alone, a pick can lose interlacings into the centre. So the rule does not hold pick by pick, as a statement about picks. It holds pick by pick as a statement about picks of a shading, and the difference is the second tone.
The second tone has already ruled it out
An eight-end shading is pinned at three everywhere but its centre, and the pin at the second tone is the one that matters here. A pick of the second tone has two raised cells and six unraised, and held to weft floats of three, its six unraised cells must split three and three. So its two raised cells sit exactly four apart, opposite each other round the repeat, on every pick.
Every later tone contains the second, so every third-tone pick holds that opposite pair. The joining shape has its raised cells two, three and three apart, and no two of them four apart. It is not a third-tone pick of any shading. The only join the centre’s own bound leaves open along a pick is a pick the second tone’s bound has already excluded.
With both halves, every thread of both systems keeps or gains on the step into the centre. The third tone interlaces no more than the centre — not on average, and not as a tendency over many chains, but on every pick and every end of every chain the family holds.
Every step of the family, thread by thread
The proof uses two bounds and a pin, and it is worth seeing what each step of the chain looks like when the same count is run for all of them.
The first step can only add and the last can only take away, because a single raised cell is a run of its own and a single unraised cell is a gap of one. Between them, the second and fifth steps go both ways on one system or the other: down an end, a third-tone mark can join the second tone’s two, which the third tone’s warp float of three allows; along a pick, a sixth-tone mark can stand alone in the fifth tone’s gaps.
Only the two steps at the centre have a sign fixed on every thread. That is exactly the content of the census’s observation, and no more: the ramp climbs to the centre and falls away from it, and the shoulders’ own neighbours are free to wander. A chain’s interlacings are not monotone along the whole ramp; they are simply never higher beside the centre than at it.
Allow the centre floats of three and the structure goes. A join down an end into the centre is now legal, and so is the pick shape the pin was excluding at the old bound, with others besides; the slider’s second position shows joins arriving on both systems. The centre’s own bound and the second tone’s are the two premises, and neither is decorative.
One chain, counted cell by cell
The argument is about every chain at once. It is also easy to watch on one.
The chain is one of the census’s witnesses. Its step into the centre raises eight cells, one in each pick and each end. Six of the picks and all eight ends keep their interlacings — the raised cell extends a run in every one of them — and two picks gain two. The changes of face go from 76 to 80, an interlacing rate of 0.594 to 0.625, and nothing went the other way.
The step out of the centre, through the cloth’s other face
The step from the centre to the fifth tone needs the opposite statement — that no thread gains — and it comes from the same argument read through the cloth’s back.
Exchange raised for unraised cells and picks for ends. A shading chain read backwards from the seventh tone becomes another shading chain read forwards from the first; the fifth tone becomes a third, the sixth a second; and each system’s face float becomes the other’s, since an end’s raised run on the face is, on the back, a pick’s unraised run turned through a right angle. The step out of the centre is the step into it, seen from the back.
So the two halves change places. Along a pick, out of the centre, a cell cannot stand alone: that needs a gap of three, and the centre’s picks have none. Down an end it can, in one shape, and the shape is excluded by the sixth tone — whose ends, held to warp floats of three with only two cells unraised, must have those two cells exactly opposite. Every later tone’s unraised cells are among the earlier ones’, so every fifth-tone end holds that opposite unraised pair, and the shape that could stand alone holds none.
On the same chain the step out of the centre takes the changes of face from 80 to 70: three picks and two ends lose two each, and nothing gains. The count of possible single-thread steps makes the mirror exact: along a pick into the centre, 80 extend, 8 join and 24 stand alone over every unpinned pick; down an end out of the centre, 80 extend, 24 join and 8 stand alone. The joins of one are the cells standing alone of the other, number for number.
What a centre that wins does to the ramp
An even shading cannot keep its surface level: a float bound is a bound on run length, a bound on run length forces a floor under how often a thread changes face, and a tone that interlaces more is pressed thinner. So a shaded ramp is a valley, and the satin digs it. Its depth is set by whichever tone interlaces most.
The proof makes that tone the centre, always. The valley’s floor is at the middle of the ramp in every even eight-end shading, which is why the census could sort a billion chains by their centres alone and lose nothing: the shoulders, the second and sixth tones and the extremes cannot set the depth, because the centre is at least as deep as each of them, thread by thread. The fourteen depth classes are fourteen kinds of centre.
It also says where a designer’s freedom is. The census found that only one start in six can reach the shallowest ramp, because the first two parts constrain the centre. The proof adds that whatever the shoulders do above that floor is cosmetic to the depth — they can tie the centre or fall below it, and either leaves the valley where the centre put it. A shallow ramp is a shallow centre, and nothing else.
How the count was made
Every claim above is about single threads, and a single eight-cell thread has 256 states, so each step was walked exhaustively: every thread of the lower tone’s raised count within its face-float bound, every cell that could be raised, kept if the upper tone’s thread is within its bound, and classed by what it does to the runs. The face floats are read as they are for every weave here: raised runs down an end, unraised runs along a pick, cyclically.
The pins are the two facts the chain adds to a thread: from the second tone on, a pick’s raised cells hold an opposite pair; up to the sixth tone, an end’s unraised cells do. Both follow from the bounds of three at those tones alone. The chain drawn is the census witness for the class whose worst tone interlaces at 0.625, and its changes of face are counted directly from the tones.
Required of it, and checked: that into the centre no end joins; that a pick joins only from the one shape and its mirror, none of which holds an opposite pair, so that a pinned pick never joins; that out of the centre no pick stands alone, and a pinned end never does; that the two steps are mirrors of each other count for count; and that a centre allowed floats of three can be reached by a join. A repeat too small or too large to walk, a step outside the repeat, a thread that is neither pick nor end, and a cell raised twice are refused.
What the proof does not reach
The shoulders’ own neighbours. The second and fifth steps go both ways, so the proof says nothing about whether the third tone interlaces more than the second. The census does not need it to. A second-tone thread has two raised cells, or along a pick two raised cells opposite, and can change face at most four times; the centre’s floats of two force at least two runs on every thread, so at least four changes. Neither the second tone nor, through the cloth’s back, the sixth can out-interlace the centre — a cruder argument than the one above, and enough for them.
Other repeats. The argument used eight in two places: a second tone pinned with its pair opposite, and a joining shape with no pair opposite. At six ends every middle tone floats two and six was special for that reason; at ten or twelve ends the second tone is not pinned by a float of three and the excluded shape is not the only one. Whether a centre floating two still wins there is a separate count.
The depth itself. Interlacing is what the pressing rule reads, and the depth follows from it through the floor a float bound puts under interlacing and the pressing of each tone. The proof is about interlacings; the depth inherits it only through that rule, which counts a thread’s turns and not where in the repeat they fall.
Who found what, and when
Shaded damasks built tone by tone from a satin are old in figured weaving — the eight-end satin is the damask weaver’s usual base — and the face-float bound as the constraint on them is how any designer of one works. The chain census that found the pattern was a split-at-the-centre count of the whole family. What is added here is the proof: that a step into a tone whose floats are held to two cannot join two runs down an end; that along a pick it can join only from a single shape with no raised cells four apart; that the second tone’s float bound pins every pick’s raised cells four apart and so excludes that shape; and that the step out of the centre is the same argument read through the cloth’s back, with the sixth tone in the second’s place.
Still open: whether a centre floating two wins at ten ends
The proof’s two premises are the centre’s own float of two and a second tone pinned to an opposite pair. At ten ends the second tone cannot float three at all: eight unraised cells in two gaps cannot both be three or less. Held to four it splits them four and four and its raised cells sit five apart, opposite again — so a pin of the same kind exists. But the centre is now the fifth tone, two steps beyond the third rather than one, and the shapes from which a pick could join into it are different shapes, with more cells and more gaps.
What has to be found is whether the opposite pair still excludes them, or whether some joining shape at ten ends holds a pair five apart and slips through. The count is the same thread-by-thread walk over 1,024 states a thread instead of 256, with the ten-end family’s own bounds; it would say whether a ten-end centre wins by the same argument, needs another fact from the chain, or can be beaten by its shoulders.
Shares its objects with
Essays naming at least two of the same things, that neither author linked.
- A damask's edge floats further than its figure — both name census, float, float length, shading, tone
- A six-end shading can be even or have a plain centre, not both — both name census, float length, shading, tone
- A selvedge holds only where its edge end changes face — both name census, float, interlacing
- A shading changes two things at once — both name float, shading, tone
- A weave is a halftone screen with n greys — both name float, shading, tone
- How many layers a draft can have — both name census, float, interlacing
Named objects
A flat tag is an object no other essay names yet.