Compound and figured cloths

A figure's streak is plainer from across the room

A figure woven on one beam leaves its ground a little brighter in the columns through the figure and a little darker in the columns beside it, and across a repeat that difference is very nearly one cosine. An eye is ten times less sensitive to a slow gradient than to a fine one, so whether the streak shows depends on how wide the repeat looks. From two metres a fixed two per cent threshold is right only for a repeat about 250 millimetres wide. A narrower one shows at less light, a wider one needs more, the same design woven finer shows its fault in dimmer rooms, and stepping back makes every streak plainer.
18 min read 6 figures Measured, not claimed

Worth reading first: A figure's streak needs a tenth of the room's light from one lamp · A figure woven on one beam shows its columns off the mirror · A random error hides and a periodic one shows.

A figure’s streak needs a tenth of the room’s light from one lamp put a number on when a jacquard’s tension fault shows. A figure woven on one beam pulls its columns to different tensions, a tight column’s crossings stop reflecting a few degrees off the mirror before a slack column’s do, and in a room lit by a lamp and a window the two columns differ by an amount that grows with the lamp’s share of the light. Set a threshold of two per cent between them and a two-degree lamp needs about a tenth of the room’s light to reach it.

The two per cent was a stated number, and the essay said what it was standing in for. An eye does not see a contrast. It sees a contrast at a size, and it sees a soft gradient spread over a hand’s width far less readily than a sharp edge. A tension streak is the softest kind of gradient there is: the tension rises column by column across the figure, with nothing anywhere that steps.

So the question left open was where the slack and the tight columns are, in degrees at the eye, and what an eye does with that. It has an answer with no fitted part beyond a standard observer, and the answer runs against what a mill would expect of a large pattern.

The lamp a figure's streak needs, against the width of its repeat. The share of a room's light that a lamp in the window of tilts must carry for a standard eye 2 m away to see the ground streak of a 5-float figure on a 3-float ground, framed by ground, against the width of the repeat. A 1° lamp: 1.3 per cent at 50 mm, 1.7 per cent at 100 mm, 2.7 per cent at 200 mm, 4.6 per cent at 400 mm, against 3.2 on a fixed two per cent rule (dashed), which it crosses near 253 mm. A 2° lamp: 2.5 per cent at 50 mm, 3.4 per cent at 100 mm, 5.3 per cent at 200 mm, 9.1 per cent at 400 mm, against 6.3 on a fixed two per cent rule (dashed), which it crosses near 253 mm. A 4° lamp: 5.0 per cent at 50 mm, 6.7 per cent at 100 mm, 10.4 per cent at 200 mm, 17.8 per cent at 400 mm, against 12.4 on a fixed two per cent rule (dashed), which it crosses near 254 mm. A narrow repeat shows its streak at less light than the fixed rule allows and a wide one needs more. The chart cannot show colour, or the matte light that dilutes every share.
Fig. 1 The share of a room’s light a lamp in the window of tilts must carry before a standard eye two metres away sees the ground streak of a 5-float figure on a 3-float ground, against the repeat’s width. Solid: through the eye’s sensitivity. Dashed: a fixed two per cent between the darkest and brightest column. The two agree only near 253 mm, for every lamp.

The streak is a cosine across the repeat

Start with what the ground looks like across one repeat, column by column. The design is a round figure framed by ground, a disc 34 blocks across in a square of 48, which is closer to a real figure than a disc filling its square: every column has some ground in it, and the columns at the edge of the repeat are ground from top to bottom. Each column runs at the tension that makes it consume what every other column consumes, which is the regulation a figured warp on one beam performs. The columns through the middle of the disc hold least ground and run slack; the framing columns hold all ground and run tightest, at 0.011 newtons an end above the slack ones.

Light the ground with a two-degree lamp placed in the window of tilts the earlier essay found and a sixteen-degree window at the mirror, and read each column’s luminance against the mean of all of them.

A regulated figure's ground streak across one repeat. The luminance of the ground in each of the 48 columns of one repeat, a disc 34 blocks across framed by ground, 5-float figure on a 3-float ground on one beam, in a room of a 2° lamp in the window of tilts and a 16° window at the mirror, as a contrast about the mean. With 20 per cent of the light from the lamp the ground swings 7.0 per cent from its darkest column to its brightest, and the first harmonic alone has an amplitude of 3.92 per cent, against 0.61 for the third. With 5 per cent of the light from the lamp the ground swings 1.5 per cent from its darkest column to its brightest, and the first harmonic alone has an amplitude of 0.76 per cent, against 0.15 for the third. The columns through the disc's middle hold least ground and run slack and bright; the framing columns hold all ground and run tight. Dashed: each profile's first harmonic. The figure cannot show how the eye weighs a gradient against an edge.
Fig. 2 The ground’s luminance in each of 48 columns across one repeat, about its own mean, with 20 and 5 per cent of the room’s light from a 2° lamp. The columns through the disc run slack and bright, the framing columns tight and dark. The dashed curves are each profile’s first harmonic.

With a fifth of the light from the lamp the ground swings seven per cent from its darkest column to its brightest. With a twentieth it swings one and a half. In both, the profile is very nearly a single cosine with one period to the repeat. The first harmonic has an amplitude of 3.92 per cent at the larger share and carries 96 per cent of the profile’s variance; the third, which is the next largest, has 0.61. At the smaller share the first carries 94 per cent.

There is a reason for the shape. The tension in a column follows its share of ground, a disc’s share of ground follows the chord the disc cuts through the column, and a chord through a circle rises and falls as a half-ellipse. The window of tilts then turns tension into luminance smoothly. Nothing between the design and the eye makes an edge, so the streak arrives as a gradient with one period to the repeat and very little else. Its period is the repeat’s width, and that is the number the eye cares about.

A slow gradient is what an eye sees worst

The eye’s sensitivity to a grating of light and dark bars is not flat in the bars’ width. It rises from the coarsest gratings to a peak at a few cycles a degree and falls away again towards the finest, and that curve has been measured on many observers since the 1950s. This account uses the simplified formula Peter Barten published in 2004 for an average observer, which gives sensitivity as a function of frequency, mean luminance and the size of the field the grating fills. The ground of a figured cloth in an ordinary room is taken at 50 candela a square metre and the field at the width of a 1.4-metre piece as seen from where the viewer stands.

The eye's sensitivity against the streak's spatial frequency. A standard eye's contrast sensitivity, Barten's simplified formula at 50 cd/m² over a field 39° across, against spatial frequency in cycles a degree, on log scales. It peaks at about 2.4 cycles a degree and falls tenfold by 0.1. The streak of a 5-float figure on a 3-float ground, with 5 per cent of the light from a 2° lamp, has a first harmonic of 0.76 per cent, so it is seen wherever the curve stands above 132 (solid line). A 50 mm repeat seen from 2 m puts its fundamental at 0.70 cycles a degree, where the eye's sensitivity is 426. A 100 mm repeat seen from 2 m puts its fundamental at 0.35 cycles a degree, where the eye's sensitivity is 236. A 200 mm repeat seen from 2 m puts its fundamental at 0.17 cycles a degree, where the eye's sensitivity is 121. A 400 mm repeat seen from 2 m puts its fundamental at 0.09 cycles a degree, where the eye's sensitivity is 61. The dashed line is the sensitivity a fixed two per cent threshold implies, 50. The chart cannot show summation across the many repeats of a width, which would lower every threshold a little.
Fig. 3 A standard eye’s contrast sensitivity against spatial frequency, log scales, peaking at about 2.4 cycles a degree. A streak whose first harmonic is 0.76 per cent is seen wherever the curve stands above 132. Four repeats seen from 2 m put their fundamentals at 0.70, 0.35, 0.17 and 0.09 cycles a degree.

The curve peaks at 2.4 cycles a degree, where its sensitivity is about 690: a grating there is seen at a contrast of a seventh of one per cent. At 0.1 cycles a degree the sensitivity is 70, ten times lower. And 0.1 cycles a degree is where a jacquard’s repeat sits. A 150-millimetre repeat seen from two metres subtends 4.3 degrees, so its fundamental is 0.23 cycles a degree. A 400-millimetre repeat from the same place is at 0.09.

A fixed two per cent threshold is a sensitivity of 50 at every frequency, the dashed line in the figure, and it sits below the curve over the whole range a figured cloth occupies at room distances. So for every repeat the earlier essay’s rule was conservative in one sense and wrong in shape: an eye sees less contrast than two per cent at these sizes, and how much less depends on the size.

A harmonic is seen when its amplitude times the eye’s sensitivity at its frequency reaches one. That is the whole of the reading, and the visibility quoted below is the largest such product over the harmonics. At a twentieth of the light from a two-degree lamp the first harmonic is 0.76 per cent, so it needs a sensitivity of 132, and the four repeats in the figure fall on either side of that line: 50 and 100 millimetres above it, 200 just under, 400 well under.

Only the fundamental counts

A streak with sharper edges would have stronger high harmonics, and the eye’s sensitivity rises steeply with frequency over this range, so it is worth checking that the cosine is really what is seen and not some harmonic riding on it.

Which harmonic of a figure's streak the eye actually sees. The first 8 harmonics of the ground streak of a 5-float figure on a 3-float ground, with 5 per cent of the light from a 2° lamp, each as its amplitude times a standard eye's sensitivity at its frequency, for a 150 mm repeat seen from 1 m (first 0.64, third 0.35), 2 m (first 1.22, third 0.62), 5 m (first 2.41, third 0.91). One is the threshold, dashed. The first harmonic carries the streak at every distance; the third reaches about half of it and the rest less, since what they gain in sensitivity they lack in amplitude. The chart cannot show summation across harmonics, which would add a little to the first.
Fig. 4 Each of the first eight harmonics of the streak, as amplitude times the eye’s sensitivity at its frequency, for a 150 mm repeat seen from 1, 2 and 5 m with 5 per cent of the light from a 2° lamp. One is the threshold. The first harmonic is the one that crosses it; the third reaches about half of it.

It is. For a 150-millimetre repeat at a twentieth of the light, the first harmonic stands at 0.64 of threshold from one metre, 1.22 from two and 2.41 from five. The third harmonic is a fifth as large in amplitude and sits at three times the frequency, where the eye is nearly three times as sensitive, so it reaches 0.35, 0.62 and 0.91 at the same distances — never past the first, and never past one on its own. Every other harmonic stays under 0.4. Part of what shows at the sixth and eighth is the staircase of 48 columns, a feature of drawing the disc on blocks and not of the cloth.

So the streak’s visibility is its first harmonic’s amplitude times the eye’s sensitivity at one cycle per repeat, to within the third harmonic’s half share. That reduces the whole question to two numbers: how large the fundamental is, which the lamp decides, and how wide the repeat looks, which the cloth and the viewer decide between them.

Stepping back shows it

The second of those has a consequence that would surprise anyone who inspects cloth at arm’s length. A viewer who steps back makes the repeat look narrower, which moves its fundamental up the sensitivity curve towards the peak.

A figure's streak against the distance it is seen from. How far above threshold the ground streak of a 5-float figure on a 3-float ground stands, with 5 per cent of the light from a 2° lamp, against the viewer's distance on a log scale, for four widths of repeat: 50 mm, 1.84 at 1 m and 4.56 at 4 m, first seen from 0.5 m; 100 mm, 0.95 at 1 m and 2.97 at 4 m, first seen from 1.1 m; 200 mm, 0.48 at 1 m and 1.62 at 4 m, first seen from 2.2 m; 400 mm, 0.24 at 1 m and 0.83 at 4 m, first seen from 5.6 m. One is the threshold. Stepping back moves a repeat's fundamental towards the frequency the eye sees best, so every curve rises before it falls. The chart cannot show what a viewer standing still on a moving fold would see.
Fig. 5 Visibility of the streak against distance, log scale, at 5 per cent of the light from a 2° lamp, for repeats of 50, 100, 200 and 400 mm. One is just seen. A 100 mm repeat is first seen from 1.1 m and a 400 mm repeat from 5.6 m; every curve rises for metres before it falls.

At a twentieth of the light from a two-degree lamp, a 100-millimetre repeat is invisible at arm’s length and seen from 1.1 metres. A 200-millimetre repeat is first seen from 2.2 metres and a 400-millimetre one from 5.6. Every curve keeps rising for metres past the point where it crosses the threshold: the 150-millimetre repeat stands at 1.22 from two metres, 2.11 from four and 2.93 from eight. The curves only turn down where the fundamental nears the peak of the sensitivity curve and the piece itself has shrunk to a few degrees, so that the formula’s field term bites: about six metres for a 50-millimetre repeat and eleven for a 100-millimetre one.

That is the reverse of how a fault is usually looked for. A weaver or an inspector puts the cloth on a table under a lamp and looks closely, and a tension streak is the one fault that looking closely hides. A random error hides and a periodic one shows found the same kind of thing in a reed streak, a fault seen only because it repeats, and noted that it had no sensitivity curve in its arithmetic. The tension streak needs one, because its period is the whole repeat, and it is seen best by the person who sees the curtain from the other side of the room.

The two per cent rule is right at one width

Bisecting on the lamp’s share until the streak is just seen gives the threshold as a function of the repeat’s width, and that is the opening figure.

From two metres, a two-degree lamp makes the streak visible at 2.5 per cent of the light for a 50-millimetre repeat, 3.4 at 100, 5.3 at 200 and 9.1 at 400. A one-degree lamp needs half as much at every width and a four-degree lamp twice as much, so the proportionality to the lamp’s width that the room’s sum predicted survives the eye intact.

The fixed two per cent rule, applied to the same design, asks 3.2, 6.3 and 12.4 per cent of the three lamps at every width. It crosses the eye’s curve near 253 millimetres for all three lamps, and that coincidence is not a coincidence: the lamp’s width scales both thresholds by the same factor, so the width at which they agree belongs to the eye and the distance and not to the light. From two metres a two per cent rule is the right threshold for a 250-millimetre repeat and for nothing else.

So the earlier essay’s numbers stand for a large figure seen from a couple of metres and overstate the light a fine figure needs by a factor of up to two and a half. That is a change of size and not of kind. The proportion to the lamp’s width and the need for a lamp in the window of tilts both hold; what the eye adds is that the threshold slopes with the repeat.

A finer sett shows the same design’s fault

The obvious way to read that slope is that small figures are risky and large ones safe. The less obvious reading is about the sett, and it is the one a mill would get wrong.

One figured design woven at different setts. A design 384 ends wide, a 5-float figure on a 3-float ground on one beam, woven at setts from 12 to 80 ends a centimetre, so its repeat runs from 320 mm down to 48 mm, with yarn and spacing scaled together so every crossing keeps its shape and every column its weave angle. Seen from 2 m, a 2° lamp makes its streak visible at 6.4 per cent of the light at 15 ends, 3.9 at 30 and 2.7 at 60. Seen from 5 m, a 2° lamp makes its streak visible at 3.8 per cent of the light at 15 ends, 2.7 at 30 and 2.2 at 60. The finer cloth shows the same fault at less light. The chart cannot show that a real finer cloth is usually also a different yarn, whose crimp would move every tension.
Fig. 6 One design 384 ends wide, 5 on 3, woven at setts from 12 to 80 ends a centimetre so its repeat runs from 320 down to 48 mm. Seen from 2 m, a 2° lamp makes its streak visible at 6.4 per cent of the light at 15 ends, 3.9 at 30 and 2.7 at 60; from 5 m, at 3.8, 2.7 and 2.2.

Take one design, 384 ends to the repeat, and weave it at different setts with the yarn scaled in proportion. Peirce’s geometry has no length in it once the yarn and the spacing move together, so every crossing has the same shape, every column gives up the same share of its crimp and keeps the same weave angle, and the streak’s profile is identical at every sett. Only its width changes. At 15 ends a centimetre the repeat is 256 millimetres and the streak needs 6.4 per cent of the light from two metres; at 60 ends it is 64 millimetres and needs 2.7. Seen from five metres the same pair is 3.8 and 2.2.

So a fine dress silk and a coarse furnishing cloth woven from one design carry the same tension fault, and the fine one shows it in a room with less than half the directional light. Fine figured cloth is exactly what is bought to be looked at under a showroom’s spotlights, which the earlier essay found to be the room most likely to show a streak, and it is where the margin is thinnest. A large pattern woven coarse hides a fault that the same pattern woven fine shows, which is the opposite of a mill’s instinct that a coarse cloth forgives less.

The figure’s other route to that conclusion, more columns per repeat, needs every end to be its own shaft in any case, so the comparison is between cloths a jacquard could equally weave.

Under a window alone the reversed streak shows too

The earlier essay found that a window at the mirror, with no lamp at all, lights the columns very slightly the wrong way round: the tight columns are a fraction brighter, because they have marginally more face in the same area. It called that difference too small to see. On a fixed two per cent rule it was.

Through the eye it is not, for a fine enough repeat. With the window alone the ground’s first harmonic is 0.29 per cent, with the dark side now in the middle of the figure. A sensitivity of 345 sees it, and the curve stands above 345 between 0.54 and 8.5 cycles a degree. A 50-millimetre repeat lit only by an overcast window shows a reversed streak from two metres, at 1.25 of threshold, and a 150-millimetre repeat from about seven metres. A 400-millimetre repeat never reaches it in a room.

Between the two signs, at 1.5 per cent of the light from a two-degree lamp, the lamp’s streak and the window’s cancel and the first harmonic falls to 0.02 per cent. There, and only there, no distance and no repeat shows anything. The streak does not fade out as a room’s light becomes diffuse. It passes through nothing and comes back reversed, and a fine enough figure is visible on either side of the null.

The pairing matters a little more than the rule said

On the fixed rule the pairing of figure and ground hardly moved the threshold, because what a lamp adds is one crossing’s arc reflecting or not, and the arc has no weave angle in it. Through the eye the pairing moves it by up to a factor of 1.7, and in an order that changes with the repeat.

At 150 millimetres from two metres, an 8-float figure on a 3-float ground is seen at 3.0 per cent of a two-degree lamp’s light, an 8-on-5 at 3.8 and the 5-on-3 at 4.3. At 50 millimetres the 8-on-5 is seen first, at 1.7. At 400 the 8-on-3 is far ahead at 5.5 against 9.1 and 9.4. In every case it is the first harmonic that crosses, so the difference is in how the fundamental grows with the lamp’s share, and it has two parts. A narrow repeat needs only a small fundamental, a quarter of a per cent at 50 millimetres, and a small fundamental is reached just past the null where the lamp and the window cancel — so the pairing whose null comes earliest is seen first on a fine figure. A wide repeat needs a fundamental of one and a half per cent, far past the null, where what counts is how fast the streak grows, and the 8-on-3’s wide range of tensions grows it fastest. The pairing now decides where a streak starts as well as how often a lamp lands in its window, and the second is still the larger effect on what a room shows.

None of this touches the one pairing the mechanism spares. A damask’s two regions share a crimp, every column runs slack, the profile is flat, and every harmonic is nought. A damask is the only figure that costs its beam nothing, and through the eye it is still the only figure that cannot streak.

How the streak was read

The columns and the room are the earlier essays’: a figured warp on one beam regulated column by column along each region’s Peirce locus, and a room whose luminance is each source’s power over its half-width squared times the specular area it finds, the same sum that lit a satin against a plain weave. The design here is a disc 34 blocks across centred in a 48-block repeat. The lamp is placed at the middle of the window of tilts the unframed disc gave, which is a good place and not the best for this design. Because the room is linear in the lamp’s share, each column’s two terms are computed once and every share is a weighted sum.

The streak is each column’s luminance over the mean, and its harmonics come from a discrete Fourier sum over the 48 columns. Halving and doubling the column count moves the first harmonic by under four per cent.

The eye is Barten’s simplified contrast sensitivity at 50 candela a square metre, with its field set to the angle a 1.4-metre piece subtends from the viewer. Visibility is the largest product of a harmonic’s amplitude and the sensitivity at its frequency, and a threshold share is found by bisection on it.

Required of it, not shown: that the standard eye peak between two and five cycles a degree and fall on both sides; that a disc’s profile be symmetric about its middle; that more of the lamp’s light always raise visibility; that a 100-millimetre repeat be plainer from four metres than from half a metre; and that the reading refuse a grating of no frequency, a share over one and a disc wider than its repeat.

What the eye model leaves out

The figure itself is not in the stimulus, and it may hide its own streak. The streak’s fundamental is the figure’s fundamental, at the same frequency and in phase with it, and the figure stands out from the ground by far more than any streak does. A strong pattern at one frequency raises the threshold for a weak one at the same frequency, which is contrast masking. This account reads the ground’s profile as if seen alone, as it is in the plain rows between motifs, and so its thresholds are the least a streak needs. How much the figure raises them is not computed.

The observer is an average one. Barten’s formula is fitted to many observers’ measurements on sine gratings on a screen; individuals differ by a factor of two either way, and a cloth is not a screen. The light level hardly matters at these frequencies: a room five times dimmer, at 10 candela a square metre, moves the 150-millimetre threshold by under one per cent, because the formula’s limit on a slow gradient is its term for lateral inhibition, which has no luminance in it.

Summation is left out, both across harmonics and across the several repeats of a piece’s width, which would make a streak a little easier to see. That too makes these thresholds conservative.

Nothing moves. A figure’s drape sweeps it through every window of tilts, and a band moving along a fold is far more visible than a still one. Matte light, colour and the yarn’s own twist are absent, as before.

Who measured the eye

The contrast sensitivity function was measured with sine gratings by Otto Schade in 1956 and by Fergus Campbell and John Robson in 1968, whose work established that the eye analyses a scene by spatial frequency. James Mannos and David Sakrison gave a widely used closed form in 1974. Peter Barten’s model of 1999 derives the curve from the eye’s optics, its photoreceptor noise and its neural processing, and his simplified formula of 2004 is the one used here. That a regular fault is seen far more readily than a random one is old knowledge in cloth inspection, and a woven cloth’s evenness at the eye’s best scale and a knit’s stripe from a repeating machine have both used the peak of the curve.

What is added here is the streak as a cosine across the repeat; its visibility as one harmonic against one sensitivity; the width near 250 millimetres at which a fixed two per cent rule is right from two metres, whatever the lamp; a streak growing plainer as the viewer steps back; the same design showing its fault at less light when woven finer; and a reversed streak under a window alone, visible on a fine figure.

Still open: whether a figure masks its own streak

Everything above reads the ground as if the figure were not there, and the figure is the strongest thing in the field at exactly the streak’s frequency. Contrast masking has a measured form: the threshold for a weak grating on a strong one of the same frequency rises roughly as the strong one’s contrast to a power a little over a half, and falls away as the two frequencies part.

So the question is how much of the figure’s own contrast falls in the ground’s rows, where the streak is read. In the plain rows between motifs the figure is a few millimetres off and its masking should be weak; in a row that crosses the figure, the ground between two motifs sits beside both. A figure with generous ground between its motifs would then show its streak in those rows and nowhere else, and one packed edge to edge might mask it entirely, which would make the spacing of motifs, not their size, the designer’s protection against a tension fault. A figured warp needs a beam for every share of its figure priced the other protection, and comparing the two is what the masking calculation would allow.

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.

AppearanceJacquardRepeatSettSpecular reflectionWarp tension