The straight weft is the one a shear empties
Worth reading first: A frame squared again keeps less than it had · A cloth held both ways settles higher · Every crossing is a force.
A frame squared again keeps less than it had found that a cloth settled on a frame loses load when a rail slides along itself and is put back. The rail stretches the threads clamped across it by , they settle back to their floor at the longer length, and when the frame is squared they come back short. The other system loses too, through the crossings, by a fixed share of what the stretched one lost, because on a frame each system’s floor is held up by the other’s pressing. That account stretched only the ends, because only the warp-way curve had been computed, and it closed on a prediction about the other direction.
The prediction came from the share. The system that is not stretched loses of the stretched system’s loss, where is the stretched system’s own crimp angle, because that angle is how hard the stretched threads press on the crossings the others slide through. A poplin’s ends are crimped at 40.7 degrees and its picks at 10.7. At a friction of 0.3 the ends pass 0.39 of their loss to the picks, and the picks pass only 0.11 of theirs to the ends. So a warp-dense cloth should be robust to one direction of shear and fragile to the other, and the earlier essay suggested that the weft-way shear would be the gentle one, since so little of it would reach the warp.
Half of that is right, and the half that is wrong is the larger.
Exchanging the two systems is the whole calculation
Every model of a woven cloth here is written along the warp. The locus a cloth extends along, the bending energy that picks its least-energy state, the load–extension curve that gives the slope , and the two frame-floor equations all take the ends as the system being pulled and the picks as the one that crimps in exchange. None of those calculations knows which thread went through a heddle and which went through a shuttle. Peirce’s geometry is a statement about two families of threads crossing one another, and it is symmetric under swapping them.
So a weft-way question is a warp-way question about a different cloth: the same cloth with its two systems exchanged, counts and setts together. The poplin here is 15 tex at 32 ends a centimetre in the warp and 20 tex at 22 picks a centimetre in the weft. Asked about its weft, the model is handed a cloth of 20 tex at 22 ends against 15 tex at 32 picks, and everything it returns about “the ends” is a statement about the poplin’s picks.
That is the kind of step that should be checked rather than trusted, and the checks are short. Exchanging twice gives back the cloth it started from. A cloth that is genuinely the same both ways, a muslin set square at 24 by 24, loses exactly the same share to a weft-way shear as to a warp-way one, to a billionth. And whichever way the rail slides, the system that is not stretched loses precisely of what the stretched one does, with read from the stretched system. If any of those failed, the exchange would be hiding an asymmetry in the arithmetic rather than revealing one in the cloth.
The prediction was right about the warp
Slide the rail so that it stretches the picks by three degrees’ worth, = 0.137 per cent, and hold it until the cloth settles. The picks drop back to their floor at the longer length. Square the frame, and they come back short. The ends were never stretched, but their floor depends on how hard the picks press the crossings, and the picks now press less, so the ends settle too.
They settle very little. Repeated until it stops, the weft-way shear costs the ends 5.1 per cent of their floor. The warp-way shear, which stretches the ends directly, costs them 6.5. For the warp, the two directions of shear are nearly the same event, and the coupling factor did exactly what it was predicted to do: of every newton the picks lost, eleven hundredths reached the ends.
That near-equality is itself a little surprising, and the reason for it is in the next section. In absolute terms the ends lose 0.0055 newtons when they are stretched and 0.0043 when the picks are. The first is a direct loss on a shallow curve. The second is a small share of a very large loss.
And wrong about the cloth
The picks lose 68.9 per cent of their floor. The warp-way shear of the same size took 6.5 per cent from the ends it stretched, so the weft-way shear is more than ten times as destructive to the system it stretches. At 3.6 degrees, held and repeated, the picks come back with no load at all, hanging slack across a square frame whose warp is still carrying more than nine tenths of what it held. The ends need 11.7 degrees for the same thing.
The cause is in the two curves. What a shear takes is , the slope of the stretched system’s load–extension curve at its clamp times the stretch, and is the same whichever system the rail stretches. The poplin’s warp is heavily crimped, so pulling along it has a long way to go before anything resists: its curve stays shallow all the way out to 12.54 per cent, where its frame floor of 0.0844 newtons is reached, and climbs there at 3.8 newtons per unit strain. The weft has almost no crimp to give. It reaches its smaller floor, 0.0553 newtons, at 0.87 per cent, and at that point its curve is climbing at 26.6 newtons per unit strain. The same 0.137 per cent of stretch is 0.0052 newtons on the warp and 0.036 on the weft, and the weft had less to lose from the start.
So the one crimp angle that decides how much a shear passes on also decides how much it takes. A thread with little crimp presses its crossings lightly, which is why it passes on only 0.11 of a loss. It also has little interchange left to spend before it stops extending by straightening and has to be stretched along its own length, which is why it loses so much. The poplin’s weft is at 10.7 degrees, and both of its properties follow from that one number: it is a poor transmitter of loss and a very good receiver of it. The earlier prediction read the first property and not the second.
This is the same asymmetry that an unbalanced cloth wrinkles one way and not the other found in folding: a poplin keeps a deep crimp budget along its warp and almost none along its weft, so a demand along the warp is met from the crimp and a demand along the weft goes into the threads. A fold demands length and a sheared frame demands length, and in both a warp-dense cloth pays for the weft-way demand out of thread rather than out of geometry.
Seven cloths, both ways
The poplin is the extreme case, and the other cloths show what decides how extreme.
The close, nearly balanced cloths behave as balance says they should, which is what a cloth with one budget for two directions would lead anyone to expect. The duck loses 12 per cent either way, the filter cloth 12 and 13, the muslin 11 and 14, the batiste 12 and 15, the sheeting 18 and 17. A small difference in sett produces a small difference in loss and in both directions the system that is not stretched loses a few per cent through the crossings.
The voile is the one that would not have been guessed. Its counts are equal, 12 tex both ways, and its setts nearly so, 24 ends against 22 picks a centimetre. Yet its picks lose 46 per cent to a three-degree weft-way shear and its ends only 20 to a warp-way one. The voile is open, and an open cloth’s crimp is fixed by how far apart the threads are rather than by how they jam, so two fewer picks a centimetre leave its weft at 15.1 degrees against its warp’s 21.1. That small difference in angle turns out to decide which system runs out of crimp first.
How much extension the cloth has left
Plotting the losses against the crimp angle orders most of the points and leaves the sheeting out of place, since the sheeting’s warp is crimped at 39 degrees and still loses 18 per cent. The quantity that orders all fourteen cloth-directions is a different one: how much further the cloth can extend past the strain where its floor is reached before its locus ends.
A locus ends in one of two ways when a cloth is pulled. Either the stretched system runs out of crimp and goes straight, or the other system, crimping ever more deeply in exchange, runs out of room between its neighbours and jams. Either way, the curve turns sharply upward as the end approaches, and a floor that sits close to the end sits on the steep part.
The poplin’s weft has 0.71 per cent left, and its locus ends with the picks straight. The voile’s weft has 0.59 per cent and ends the same way. Those two are the large losses. The sheeting has 1.07 and 1.21 per cent left in its two directions and ends by jamming both times, which is why a well-crimped warp can still lose a fifth of its floor: it is close to the point where its weft locks against itself, and a jammed cloth has nowhere further to go. Every cloth with more than one and a half per cent of room loses between 7 and 15 per cent, and the poplin’s warp, with 2.38 per cent of room, loses least of all.
So the crimp angle is a proxy and the room is the thing itself. A low crimp angle usually means little room, because a straight thread is close to straight, but a close set can take the room away from a crimped thread too. The crimp interchange a cloth has left is what lets it absorb a stretch geometrically, and once it is spent the yarn takes the load.
The repeats are the same series both ways
Repeating the shear takes a further amount each time, and the earlier essay showed the losses form a geometric series with ratio . That ratio is symmetric in the two crimp angles, so it is the same whichever system is stretched: 0.044 for the poplin. Each repeat of a weft-way shear takes four and a half per cent of the one before, exactly as a warp-way one does.
So the first shear does nearly all the damage in either direction, and the frame remembers only its worst shear, not how many it has had. What differs is how bad the worst day is. A one-degree shear along the weft costs the picks 7.6 per cent of their floor; along the warp it costs the ends 0.7. At two degrees the figures are 31 and 2.9 per cent. At four the picks are empty and the ends, having lost 7.3 per cent through the crossings, are still close to what a warp-way shear of the same size would have cost them directly, 11.6. A tilt a person would struggle to see is the difference between a weft that holds a third of its load and one that holds none.
A tenth of a per cent of clamp is the difference
All of those figures put each system two tenths of a per cent past the strain at which its floor is reached, which was the earlier essay’s convention and keeps the comparison fair between cloths. For the poplin’s weft that convention turns out to matter a great deal, because two tenths of a per cent is more than a quarter of the room the weft has.
Clamped exactly at its floor, the poplin’s weft loses 37 per cent to a three-degree shear. One tenth of a per cent further, it loses 48. Two tenths, 69. At 0.29 per cent past its floor it loses everything. The warp, over the same range, goes from six per cent to seven and stays there, because its curve barely changes slope in the space of a few tenths.
Below the floor both systems have the allowance the earlier essay found, the same margin that lets a tensioned cloth keep a load it was never pulled past: a load less than friction can hold, which a shear can raise and lower again without anything rearranging. At two tenths of a per cent below its floor the weft survives a three-degree shear untouched. For a warp-dense cloth on a frame that can slide, the weft should be tensioned under its floor and the warp may be tensioned past it, and the direction of the asymmetry is the opposite of what a tensioner pulling hardest on the heavier, denser warp might expect. The warp is the system that can take an extra pull. The weft is the one that cannot take any.
Above a friction of 0.355 the warp loses more to the shear that misses it
Friction lifts both floors, and with them where each system has to be clamped. The earlier essay found that past a friction of about 0.37 a sheeting keeps less after a shear the more grip it has. The poplin does something stranger.
The weft-way loss to the picks is not monotone in friction. At a friction of 0.12 it is 70 per cent, at 0.2 it bottoms out at 59, at 0.3 it is 69, and from about 0.39 a three-degree weft-way shear leaves the picks with nothing. A low friction gives a low floor, so even a small is a large share of it; a high friction gives a higher floor reached further along a curve that is already steep. Somewhere between, the two effects balance.
The ends are where the surprise is. Under a warp-way shear they lose 6.5 per cent at a friction of 0.3, 8.1 at 0.36 and 9.5 at 0.4. Under a weft-way shear, which never stretches them, they lose 5.1, 8.5 and 11.5. Above a friction of about 0.355 a poplin’s warp loses more to the shear that stretches its weft than to the one that stretches the warp itself. The share passed on is still only 0.11, but the picks’ loss it is a share of has grown so fast that a ninth of it outweighs everything the warp’s own shallow curve would surrender.
That is every crossing being a force at its least intuitive. A system can be damaged most by a load it never carried, provided the other system carries that load badly enough and passes any of it on.
What this means for a sliding frame
A rail that slides along itself stretches the threads clamped across it, not the threads running along it. So for a warp-dense cloth on a frame with one sliding member, which way the cloth is mounted decides whether a shear is an inconvenience or a failure. The sliding rail should hold the ends, so that its movement stretches the warp, which has room and a shallow curve. Mounted the other way, the same three-degree movement takes seven tenths of the weft’s load.
The same reasoning extends to any warp-dense fabric held in tension on a structure that can rack imperfectly — a blind, an awning, a printing screen, a tensioned display panel. It says nothing about which way such fabrics are mounted in practice, which is a matter of widths and selvedges and where a seam can go. It says what the choice costs.
How the shear was computed
Each direction is the same calculation as the earlier essay’s, on a cloth whose systems have been exchanged for the weft-way case. The two floors come from the framed cloth’s two linear equations, each system’s frictional floor raised by the other’s pressing at the crossings, with the crimp angles of the least-energy state on the cloth’s own locus and a resting contact force of 0.2 newtons, the order a relaxed cloth’s contact force implies. The slope is the tangent of the stretched system’s free-to-narrow load–extension curve at its clamp.
Each shear stretches the clamped system by , lets it settle to its floor given the other system, brings it back down by when the frame is squared, and lets the other system settle to its floor given the new load. Forty repeats are run and the result is checked against the geometric series’ closed form.
The locus is sampled four times as finely as before, at 2,561 states, because the weft-way curves of the warp-dense cloths end within a per cent of their floors and the slope there changes quickly. The poplin’s weft-way loss moves 2.5 per cent between the coarser and the finer sampling and under 0.3 per cent beyond it. The earlier essay’s warp-way figures move in their third digit, which is why the sheeting appears here at 18 per cent rather than the 18.7 quoted there.
Required of it, and not shown on the page: that exchanging the systems twice give back the cloth; that a square cloth lose the same both ways; that the unstretched system lose exactly of the stretched one’s loss for three cloths in both directions; that the poplin’s weft-way slope be more than five times its warp-way slope and its loss more than five times as large; and that the refusals hold — a rail sliding in a direction that is neither, a shear past forty-five degrees, a cloth the table does not carry, and the cheesecloth, whose weft-way curve does not reach its floor.
What the exchanged cloth cannot see
The exchange assumes the cloth is the same object turned through a right angle. In the geometry it is. In a real loom-state cloth the warp was woven under tension and the weft laid in slack, the two yarns are usually spun differently, and finishing works on them differently; none of that is in a model that knows a thread only by its count, its sett and its crimp. A poplin’s weft is often the softer, lower-twisted yarn, and that would lower its slope.
Nothing extends a straight thread. The poplin’s weft-way locus ends when the picks go straight, and past that point a real pick would stretch along its own length at the yarn’s modulus. The clamps here stay inside the locus, at most 0.29 per cent past the floor, but the steepness of the last stretch of curve is the model’s steepness, and how a nearly straight spun yarn behaves in its last tenth of a per cent of crimp is exactly where a geometric model is least reliable.
The unloading slope is still borrowed. When the frame is squared, each system is assumed to unload along the slope it loaded on. And a shear along a diagonal stretches both systems at once by different amounts, which neither direction computed here describes.
One friction and one contact force stand for every crossing, though a fabric is a population of contacts and would lose its weakest first. Friction is really two coefficients, and a weft yarn with a different finish might grip differently; the crossing of the ends’ curves at 0.355 would move with it.
Whose results these rest on
That a woven fabric’s two directions respond differently to load is old knowledge, measured since biaxial testing began and central to how coated fabrics for tensioned structures are specified, by warp and fill separately. That the difference comes largely from crimp interchange is Peirce’s geometry and the biaxial work that followed it. What is added here is the consequence on a settled frame: that the loss from a sliding-rail shear is decided by how much extension the stretched system has left past its floor; that a warp-dense poplin’s weft, with 0.71 per cent of room, loses ten times what its warp does to the same shear; that the low crimp angle which keeps a weft-way loss from reaching the warp is the same angle that makes the loss large; that a nearly balanced but open voile shows the same asymmetry by a factor of two; and that above a friction near 0.355 a poplin’s warp loses more to a shear that stretches only its weft.
Still open: what a diagonal shear does to a cloth with one fragile direction
A frame rarely goes out of square along one of its own rails, and a racking frame, which the bias accommodates by rotating its threads, costs nothing at all. A gust or a settling post pushes a corner, and a corner moving diagonally stretches both systems at once, each by an amount set by how far the corner travelled along it. For a balanced cloth that only adds two losses of the same kind. For a poplin it pairs a cheap loss with a ruinous one, and the coupling then passes each system’s loss to the other.
The question is whether there is a direction of corner movement a warp-dense cloth tolerates far better than the others, and whether it lies close to the warp or somewhere between. With both systems stretched the two floor equations no longer separate, so the calculation is one step beyond the one done here. The test on a frame would be the one already proposed for this cloth: a square panel, a load cell on a few ends and a few picks, a corner moved by a known amount in a series of directions and returned. The prediction to beat is that the picks’ load falls off a cliff as soon as the movement has any component along the weft at all.
Shares its objects with
Essays naming at least two of the same things, that neither author linked.
- A cloth extends by moving its crimp — both name crimp, jamming
- A cloth gives back less than it took — both name crimp, jamming
- A cloth shrinks most the first time — both name frictional floor, yarn friction
- A cloth's Poisson ratio is not a material's — both name crimp, shear
- A jersey gets taller before it gets shorter — both name jamming, load-extension
- A knee is a dome imposed a thousand times — both name shear, yarn friction
Named objects
A flat tag is an object no other essay names yet.
CrimpFrictional floorJammingLoad-extensionLocusShearYarn friction