Mechanics and drape

A frame squared again keeps less than it had

A cloth settled on a frame holds its floor, and a frame does not stay square. Whether going out of square costs anything depends on the frame, not the cloth. A frame that racks on its pins keeps every thread's length and loses nothing. A frame whose rail slides along itself stretches the ends by the square of the angle, and squaring it again leaves them short. A sheeting loses a sixth of its warp load to a three-degree shear, the picks lose load they were never asked to carry, and the next shear of the same size takes only an eighth as much again.

Worth reading first: A cloth held both ways settles higher · A tensioned cloth loses its load · The bias is a mechanism.

A cloth stretched on a frame settles. Its crossings rearrange until each system’s load has fallen to what friction can hold, and a cloth held both ways settles higher than one held one way, because each system presses the crossings the other is sliding through. For a sheeting at a friction of 0.3 that pair of floors is 0.115 newtons on every end and 0.113 on every pick, and pulling harder does not change it.

That account held its frame square, and its last question was what happens when the frame is not. Holding a frame square was a strong assumption, because a cloth pulled both ways has only one direction it can give in, and a frame that moves offers it another. Panels are pushed out of square by wind, by a rail settling in its track, by a stretcher whose corners work loose and are knocked true again. Nothing about either floor equation obviously changes under a shear: no length is altered on average and the cloth is not pulled harder. Yet a frame that has been out of square and come back can hold less than it did.

Which of the two happens is not decided by the cloth. It is decided by how the frame goes out of square, and there are two ways for it to do that.

What squaring a sheared frame costs a sheeting's ends. The share of a framed sheeting's settled warp load lost after a shear is held, settled and squared again, against the shear angle, for a cloth clamped at 4.45%, which loses 19% to a 3° shear and goes slack at 6.9°; at 4.65%, which loses 28% to a 3° shear and goes slack at 5.7°; at 4.85%, which loses 48% to a 3° shear and goes slack at 4.3°. A frame that racks on its pins keeps every thread's length and loses nothing at any angle, which is the flat line along the foot. The loss is quadratic in the angle because a sliding rail stretches the ends by 1/cos γ − 1. The picture assumes the settled cloth unloads along the loading curve's own slope, and cannot show how long a shear must be held to settle.
Fig. 1 The share of a framed sheeting’s warp floor that is gone after a shear has been held, allowed to settle, and the frame squared again, against the angle. Each curve is a cloth clamped at a different strain. A frame that racks on its pins loses nothing at any angle; a frame whose rail slides loses load as the square of the angle, and the curves end where the ends come back with no load at all.

Two ways for a frame to go out of square

A pin-jointed frame racks. Its four sides keep their lengths and its corners change their angles, so the rectangle becomes a parallelogram with the same sides. A cloth stretched in it does what the bias is a mechanism says a cloth does under shear: its threads keep their lengths and rotate at the crossings. Each end still runs from its clamp on one rail to its clamp on the other, and that distance is the side of the frame, which has not changed.

A frame square or racked also leaves the cloth where its one budget for two directions put it, since neither mean length has changed. So nothing in the floor equations moves. The crimp angle of an end is set by how far apart the picks are measured along that end, and the crossings are pinned along it. Trellising changes how close neighbouring ends come sideways, which is why it stops at the locking angle, but up to that angle it leaves the path of every thread exactly as it was. The pressing at each crossing is unchanged, the friction it buys is unchanged, and each system sits at the same floor. Racking costs a settled frame nothing.

A frame with a sliding rail shears simply. Think of a canopy whose front bar is pushed sideways while staying parallel to the wall it hangs from, or a panel whose top rail can slide in its track. The rail keeps its distance from the opposite rail and moves along itself. An end clamped at both rails is now a hypotenuse: the rails are as far apart as ever, but the end runs on a slant between them.

That end is longer. If the rails are a distance LL apart and one has slid by Ltan⁡γL\tan\gamma, the end’s new length is L/cos⁡γL/\cos\gamma, so it is stretched by

δ=1cos⁡γ−1≈γ22.\delta = \frac{1}{\cos\gamma} - 1 \approx \frac{\gamma^2}{2}.

The picks, running along the rails, are not stretched at all.

A sliding rail stretches the ends by the square of the angle

The stretch is second order in the angle, and that sounds like a reason to ignore it. At one degree it is 0.015 per cent, and at three degrees 0.137 per cent. At five it is 0.38 per cent.

Those numbers are only small beside the wrong scale. The scale that matters is the width of the region a cloth’s load falls through as it settles. A tensioned cloth loses its load across a band of strain a fraction of a per cent wide, and near the top of that band the load–extension curve is climbing steeply, because the cloth has used up its crimp interchange and is beginning to stretch its yarn. A sheeting clamped at 4.45 per cent carries 0.139 newtons an end before it settles, and its curve there rises by 0.136 newtons for every further per cent of strain. Call that slope kk, per unit strain: 13.6 newtons an end.

A three-degree shear of that cloth by a sliding rail therefore adds kδk\delta = 0.0187 newtons to every end. That is a sixth of the floor, from a tilt a person would struggle to see.

What comes back when the frame is squared

The extra load is above the floor, so the ends settle again. Held long enough, the crossings rearrange exactly as they did the first time, and the ends fall back to their frame floor. That costs nothing visible: the cloth carries the same load as before, at a slightly greater length.

Now square the frame. The rail slides back, the ends return to the distance between the rails, and they are shorter by the stretch they settled at. Their load drops by kδk\delta to 0.0965 newtons. That is below the floor, and below the floor friction does nothing to put load back. A floor is a limit on how much a cloth can hold, not a level it returns to. A cloth below its floor stays where it is put.

So a sheeting at its floor, sheared three degrees by a sliding rail, held, and squared, has lost 16.2 per cent of its warp load, and nothing about the square frame it is now in shows why.

A sheeting's two floors through repeated 3° shearsThe settled load in a framed sheeting's ends and picks after each of 6 shears of 3° by a sliding rail, each held, settled and squared: the ends go from 0.1152 to 0.0965 N on the first and 0.0940 on the second, and the picks, which are never stretched, from 0.1133 to 0.1062 and 0.1053. Each repeat takes 0.132 of the one before, so both settle at the series' limit, 0.0936 and 0.1052 N, drawn as grey rules. The picture holds friction at one value for both systems and cannot show a shear that varies from gust to gust.the first shear takes nearly all of it; the repeats take a series that stopssheeting · friction 0.3 · clamped at 4.45% · each repeat takes ρ = 4μ² sin θ₁ sin θ₂ = 0.132 of the last00.0500.1000246shears of 3° held and squaredsettled load, N per threadendspicksgrey: T − kδ/(1 − ρ), the limit of the seriessheeting
Fig. 2 The settled load in a framed sheeting’s ends and picks through six shears of three degrees, each held and then squared. The first takes the ends from 0.115 to 0.097 newtons and the picks from 0.113 to 0.106. Every later shear takes about an eighth of the one before, and both systems stop at the limits ruled in grey. The slider sets the shear from one degree to six; past five the ends lose more than half their floor.

The picks lose load they were never asked to carry

The picks were not stretched, so on a one-way account they should be untouched. But on a frame the picks’ floor is not a fixed number. It is

T2=μ (N0′+2T1sin⁡θ1),T_2 = \mu\,(N_0' + 2T_1\sin\theta_1),

and T1T_1, the ends’ load, has just fallen by kδk\delta. Each end presses the crossings less, so the picks sliding through those crossings are gripped less, and their floor drops by 2μsin⁡θ1 kδ2\mu\sin\theta_1\,k\delta. The picks were sitting at their old floor, which is now above their new one, so they settle down to it.

For the sheeting, 2μsin⁡θ12\mu\sin\theta_1 is 0.378, so the picks lose 38 per cent of what the ends lose, through nothing but the crossings: 0.113 newtons to 0.106, a fall of 6.2 per cent. The shear never touched them. This is every crossing being a force, run backwards: the same coupling that lifted both floors when the frame was first tensioned now passes a loss from one system to the other.

It also explains why a one-way relaxation test could never find this. A strip with its picks free has no second system to pass the loss to and no first system pressing on it. The coupling exists only on a frame, and so does the loss it carries.

A second shear takes an eighth of the first

Shear the frame again by the same three degrees. The ends are stretched by the same kδk\delta and rise from 0.0965 to 0.1152 newtons, which is exactly their original floor. But their floor is no longer that. It fell when the picks fell, by 2μsin⁡θ22\mu\sin\theta_2 times the picks’ loss, so the ends settle a little again. Square the frame and they come back that little lower; the picks follow.

Each repeat takes a fixed fraction of the loss before it, and the fraction is

ρ=4μ2sin⁡θ1sin⁡θ2,\rho = 4\mu^2\sin\theta_1\sin\theta_2,

which for the sheeting is 0.132. The series of losses is geometric, so it has a sum, and both systems stop at a definite level:

T1−kδ1−ρ,T2−2μsin⁡θ1 kδ1−ρ.T_1 - \frac{k\delta}{1-\rho}, \qquad T_2 - 2\mu\sin\theta_1\,\frac{k\delta}{1-\rho}.

For the sheeting at three degrees that is 0.0936 newtons on the ends and 0.1052 on the picks: 18.7 and 7.2 per cent lost, where the first shear alone took 16.2 and 6.2. The first shear does seven eighths of the damage.

The denominator is the surprise. 1−4μ2sin⁡θ1sin⁡θ21 - 4\mu^2\sin\theta_1\sin\theta_2 is the same expression that appeared in the frame floor’s own solution, where it set the friction at which a cloth would lock and never relax at all. The quantity that decides how much a frame lifts its floor is the same one that decides how much a repeated shear takes from it. A cloth whose two systems grip each other strongly gains most from being framed and passes most of a loss back and forth between its systems before the passing stops.

The largest shear, not the number of them

Once the series has stopped, a further shear of the same size stretches the ends back up to a load their reduced floor already holds. Nothing settles, so squaring the frame takes them back to where they were. A smaller shear does even less. What a frame has lost is set by the largest shear it has been through, and a thousand repeats of a smaller one cost nothing more.

That is the opposite of fatigue, and it is a useful kind of opposite. A panel that has survived its worst storm has paid its whole bill, and its load from then on is fixed until something shears it harder. The loss also goes as the square of the angle, so half the shear costs a quarter of the load: a degree and a half takes 4.7 per cent of a sheeting’s warp floor, where three degrees took 18.7.

The ends do not have far to go to lose everything. A sliding-rail shear of 6.9 degrees, held and repeated, takes the sheeting’s ends from 0.115 newtons to nothing, and they hang slack in a square frame.

Seven cloths and which system pays

The size of the loss depends on how steep the cloth’s curve is where it was clamped and on how strongly its two systems press each other.

What a 3° shear held and squared takes from seven framed cloths. The share of each system's frame floor lost when a sliding rail shears the frame by 3°, the cloth settles and the frame is squared, ends above and picks below, once (filled) and repeated until it stops (outlined): voile, ends 19.3% then 20.0%, picks 4.9% then 5.1%; batiste, ends 11.3% then 11.9%, picks 3.1% then 3.2%; muslin, ends 10.2% then 10.8%, picks 3.0% then 3.2%; duck, ends 10.6% then 11.7%, picks 3.5% then 3.8%; filter, ends 11.2% then 12.5%, picks 3.8% then 4.2%; poplin, ends 6.3% then 6.6%, picks 3.8% then 3.9%; sheeting, ends 16.2% then 18.7%, picks 6.2% then 7.2%. Each cloth is clamped two tenths of a per cent past the strain its frame floor sits at. Only the ends are stretched; the picks lose through the crossings. Thread forces only, and the picture shows only the rail that stretches the ends.
Fig. 3 The share of each system’s frame floor that a three-degree sliding-rail shear takes, for seven cloths each clamped two tenths of a per cent past the strain at which its frame floor is reached. The filled bar is one shear; the outline is the limit of repeating it. The voile and the sheeting lose most from their ends, and the poplin’s picks lose more than half as large a share as its ends.

Each cloth here is clamped at the same distance past its own floor, two tenths of a per cent, so the comparison is fair. The voile and the sheeting lose the most, a fifth or nearly a fifth of their warp floor, because their curves are steep where they settle. The muslin, the batiste, the duck and the filter cloth lose eleven or twelve per cent.

The poplin is the instructive one again. Its ends lose only 6.6 per cent, because its curve is shallow there and it settles far along its locus. But its picks lose 3.9 per cent, more than half as large a share as the ends that were actually stretched. A poplin’s warp is crimped at 41 degrees, so it presses hard on the picks, and its picks start from a low floor of 0.055 newtons. In a warp-dense cloth a shear that stretches the warp costs the weft nearly as much in proportion, because the weft was living on the warp’s pressing in the first place.

The frame that keeps most was pulled least

All of that assumed a frame at its floor: tensioned past it and allowed to settle. A frame tensioned below its floor never settles. Its load is whatever it was pulled to, and that load is below what friction could hold.

Such a frame has an allowance. A sliding-rail shear stretches the ends, and as long as the stretched load stays under the floor, nothing rearranges and squaring the frame takes it straight back. The allowance is the gap between the floor and the load, divided by the slope: (T1−T)/k(T_1 - T)/k in strain, which turns into an angle through 1/cos⁡γ−11/\cos\gamma - 1.

A framed sheeting's settled load before and after a 3° shear, against how tightly it was clamped. The load a framed sheeting's ends hold before a 3° rail shear and after it has been repeated to its limit, against the strain the ends were clamped at. Below 4.25% the cloth never reached its floor and has a shear allowance — 5.3° at 3.85%, 3.5° at 4.05% — so a 3° shear first costs anything at 4.11%. What is left after the shear peaks at 0.102 N for a clamp at 4.10%, the strain whose allowance is exactly 3°. Above the floor the load held is the floor, and what a shear takes grows with the curve's steepness: 19% at 4.45%. The picture uses the loading slope for the unloading, and cannot show a cloth whose crossings are not all at one friction.
Fig. 4 A framed sheeting’s warp load before a three-degree sliding-rail shear and after it has been repeated to its limit, against the strain the ends were clamped at. Below 4.25 per cent the cloth never reached its frame floor and the two curves coincide until the shear exceeds its allowance. The load left afterwards peaks near 4.10 per cent and falls steadily beyond the floor.

A sheeting clamped at 4.00 per cent carries 0.095 newtons an end and survives any sliding-rail shear up to four degrees untouched. At 4.05 per cent the allowance is 3.5 degrees. The best clamping for a frame that will see three degrees is 4.10 per cent, the strain whose allowance is exactly three degrees, which carries 0.102 newtons and keeps 0.102 after the shear. Below it the cloth keeps everything but was given less; above it the shear settles it. Clamped at 4.45 per cent, the same cloth carried 0.139 newtons, settled to 0.115, and after the same shear keeps 0.094.

So the frame that was pulled hardest ends up holding about nine per cent less than one pulled to just under its floor. That is the one-way account’s rule taken one step further. Pulling a cloth past its floor was already known to buy nothing but time. Under shear, it also moves the cloth onto a steeper part of its curve, where each degree of shear costs more, and so a hard pull ends up with less than a moderate one.

The shear that leaves a framed cloth's ends slack, against how hard it was clamped. The rail shear, held and repeated, at which a framed cloth's ends come back with no load at all, against how far past its frame floor's strain the cloth was clamped: sheeting, 8.1° just past its floor and 2.7° at 0.81% past; duck, 9.7° just past its floor and 5.8° at 0.81% past; poplin, 12.4° just past its floor and 9.0° at 0.81% past. The harder the pull, the steeper the curve the cloth sits on and the smaller the shear that empties it. The picture assumes one friction and one contact force for every crossing.
Fig. 5 The sliding-rail shear, held and repeated, at which a framed cloth’s ends come back with no load at all, against how far past its frame floor’s strain it was clamped. The sheeting goes from 8.1 degrees just past its floor to 2.7 degrees at eight tenths of a per cent past; the duck and the poplin, on shallower curves, from 9.7 to 5.8 and 12.4 to 9.0.

The slack angle makes the same point at the other extreme. A sheeting clamped just past its floor needs 8.1 degrees of sliding shear before its ends come back slack. Clamped eight tenths of a per cent further, it needs 2.7. The poplin and the duck sit on shallower curves and keep more room. The tightest frame is the most fragile one, because tightness is measured on the curve, and the curve is steepest where it is tight.

More grip, a higher floor and a deeper fall

Friction is the setting the two-way floor found had the most leverage of all, since it enters both the floor and the gain, and it is really two coefficients rather than one; a single value stands for both here. The shear gives it a further and less welcome role.

A framed sheeting's floor and what a 3° shear leaves of it, against friction. The ends' frame floor of a sheeting and what is left of it after a 3° rail shear repeated to its limit, against the thread friction, each clamped two tenths of a per cent past its own floor's strain: at 0.2, 0.065 N falling to 0.058, a loss of 11%; at 0.3, 0.115 N falling to 0.094, a loss of 19%; at 0.4, 0.188 N falling to 0.107, a loss of 43%. More friction lifts the floor, moves the clamped strain onto a steeper part of the curve, and so raises what a shear takes; the gap between the curves is the loss. The picture holds the crimp angles at rest and cannot show a cloth where friction varies across the piece.
Fig. 6 A framed sheeting’s warp floor, and what is left of it after a three-degree sliding-rail shear repeated to its limit, against the friction between the threads, with the cloth clamped two tenths of a per cent past its own floor’s strain each time. The floor rises steadily. What survives the shear rises with it up to a friction near 0.37 and then falls, reaching nothing by 0.47.

At a friction of 0.2 the floor is 0.065 newtons and a three-degree shear takes eleven per cent of it. At 0.3 the floor is 0.115 and the shear takes nineteen. At 0.4 the floor is 0.188 and the shear takes 43 per cent, leaving 0.107. What is left after the shear peaks at 0.112 newtons near a friction of 0.37, and above that more grip leaves less, until by 0.47 the ends come back slack.

The mechanism is the one before. A higher friction raises the floor, so the cloth has to be pulled further along its curve to reach it, and further along is steeper. At a friction of 0.45 the sheeting, clamped two tenths past its floor, sits at 5.1 per cent strain, where its curve rises eight times as steeply as at 4.45. A finish that raises friction, which the two-way account found buys a framed fabric more than a strip, buys it more only so long as nothing shears the frame. Beyond a stated friction, grip is a liability on a frame that can slide.

How the shear was computed

The floors are the two linear equations of the framed cloth, 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 of magnitude a relaxed cloth’s thickness implies. The slope kk is the tangent of the cloth’s computed warp-way load–extension curve at the strain the ends were clamped at.

Each shear is a sequence of four states. The ends are stretched by kδk\delta; their load, if above their floor given the current picks, settles to that floor; the frame is squared and the ends come back down by kδk\delta, never below nothing; and the picks settle to their own floor given the new ends, if that is below where they stand. The sequence is iterated forty times and checked against the closed form of the geometric series to within a billionth of a newton.

Required of it, and not shown on the page: that a racked frame lose nothing at one, five or twenty degrees; that a repeated shear converge to the closed form for four cloths; that the picks lose exactly 2μsin⁡θ12\mu\sin\theta_1 of what the ends lose; that each repeat take ρ\rho of the one before; that halving the angle quarter the loss; that a cloth clamped inside its band survive a shear inside its allowance and lose past it; and that a shear past the slack angle leave the ends with nothing.

What the stretched-thread picture leaves out

The unloading slope is borrowed. When the frame is squared the ends shorten, and their load is taken to fall along the slope the curve had going up. A settled cloth is not the same object as the one that was pulled, since its crossings have moved, and its unloading curve has not been computed anywhere here. If it is steeper, the losses above are underestimates; if shallower, overestimates. The structure — a one-time loss, quadratic in the angle, shared between the systems by the crossing coupling — does not depend on which.

There is no clock. A shear is taken to be held long enough for the cloth to settle, as the floor itself is. A gust lasting seconds may not settle anything, and how long it takes is a question no relaxation result here has answered. Vibration matters too: agitation lowers the friction a crossing offers, so a flapping panel would settle faster and lower than a still one. The numbers are what a sustained shear does: a rail that has crept in its track over a season and is put back at maintenance, or a stretcher that has been out of true for a winter.

Only the ends are stretched. The warp-way curve is the one computed, so every shear here slides a rail along the picks. A frame sheared the other way stretches the picks, and their curve would be needed. The coupling would then pass the loss from picks to ends with the factor 2μsin⁡θ22\mu\sin\theta_2, which for a poplin is 0.11 where the factor the other way is 0.39.

Past the locking angle, racking is not free. The claim that a racked frame loses nothing rests on the threads rotating freely, and that holds only up to the angle at which neighbours jam. Beyond it, a racked frame is compressing its cloth sideways, and nothing here says what that does to a floor. And one friction and one contact force stand for every crossing. A fabric is a population of contacts, and it would lose its weakest grip first, so a cloth near its slack angle would go slack in patches rather than all at once.

Whose results these rest on

That frames go out of square, and that a stretched fabric panel loses tension when its supports move, is everyday experience in sign-making, canvas work and fabric architecture. The decomposition of a shear into a racking that preserves thread lengths and a sliding that does not is the classical kinematics of the trellis and of simple shear. What is added here is the consequence for a cloth already at its frictional floor: that only the sliding part costs anything, that the cost is kδk\delta on the stretched system and is passed to the other through the crossings, that repeating it sums to a geometric series with the frame floor’s own determinant in it, that the loss is set by the largest shear and not by the number, that a frame tensioned below its floor keeps more after a shear than one tensioned past it, and that above a friction near 0.37 a sheeting keeps less after a shear the more grip it has.

Still open: whether a weft-way shear costs a poplin as little as it should

Everything above stretches the ends, because their curve is the one computed. The coupling says a shear that stretches the picks instead should pass a much smaller share of its loss to the ends in a poplin: 0.11 of it, against the 0.39 its ends pass to its picks, because the poplin’s weft is nearly straight and presses on the warp very little. So a warp-dense cloth should be robust to one shear direction and fragile to the other, and which is which is decided by the crimp angles alone.

Computing it needs the weft-way load–extension curve at the strain the picks were clamped at, which is the same calculation as the warp-way one with the systems exchanged. The test on a real frame is cheap: a square panel of poplin with one sliding rail, a load cell on a few ends and a few picks, sheared three degrees each way in turn and squared. If the arithmetic is right, a rail sliding along the picks should cost the picks about two fifths of what it costs the ends, and a rail sliding along the ends should cost the ends only about a ninth of what it costs the picks. And a second shear of the same size, in either direction, should take only a small fraction of what the first took, a frame’s way of remembering only its worst day.

What links here

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

Reads more easily once this is understood

Essays that name this one as worth reading first.

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.

Frictional floorLoad-extensionRelaxation bandShearTrellisYarn friction