Pattern and colour

A round thread does not quiet a bobbinet's null

The faint fringe family a bobbinet lights beats through its net's fifth harmonic, and a sharp-edged thread's fifth harmonic swings through a null as the twisted sides widen. A round thread was expected to damp the swing. It cannot, because a round section's transform has zeros of its own: drawn with translucent round threads the null survives, and moves to a wider twist by at most 1.22, the same ratio that sets the size of a star's image in a telescope. Translucency also gives the other family a null it never had, by letting the twisted sides cancel the single threads. The across family stays the stronger, by more than before, partly because the along family collapses.

Worth reading first: A bobbinet's twists light the faint fringe family · A net of three directions beats a voile one way at a time · A net over a voile beats through a harmonic.

A bobbinet’s twists light the faint fringe family. A tulle hung in front of a voile and seen from across a room makes two families of fringes, one along the voile’s threads and one across them. Drawn as three sets of straight bars, the net leaves the across family a fifteenth as strong as the other, because that family needs two of the net’s directions at once. Drawn as the hexagonal mesh a bobbinet really has, with single threads on four sides of each hole and twisted pairs on the other two, the across family is as strong as the other or stronger almost everywhere.

That account left one thing it did not trust. Both families beat through the net’s fifth harmonic, and the strength of a sharp-edged band at a high harmonic swings through nulls as the band widens. The across family passed through a null near a twist of two, where the twisted sides are about twice as wide as the single threads. The essay put that down to drawing its threads as bands with clean edges, and predicted that a round thread, whose light falls off towards its edges, would damp the swings and keep the across family the stronger over the whole range, at a few times lower strength.

This essay draws the threads round. The across family is indeed the stronger at every twist, and the strengths fall. But the null does not go away, and the reason is a piece of optics far older than tulle.

Why everything happens at the fifth harmonic

The two families are slow for a reason that has nothing to do with the thread. A net over a voile beats through a harmonic: a net of 1.55-millimetre pitch and a voile of 0.3 millimetres have no first harmonics anywhere near each other, so the eye sees no beat between them, but the net’s fifth harmonic lies at 0.31 millimetres and beats against the voile’s first. Moving back from the curtain changes how the far layer is projected onto the near one, and two sheers make a moiré that walks with the viewer for exactly that reason.

A net of three directions beats a voile one way at a time then found the across family’s slow beat at a point two of the net’s directions make together, at 3\sqrt{3} over the pitch, and in the honeycomb’s own spectrum the along family’s point sits at five times the first ring’s radius and the across family’s at three times 3\sqrt{3}, 5.2 times. High harmonics are where the shape of a thread’s section matters most, because a harmonic samples the section on a scale of a fifth of the pitch, little more than the twisted sides’ own width. At the first harmonic every one of the profiles drawn here would give nearly the same amplitude, since a thread a tenth of the pitch wide is too narrow to show its shape to so coarse a wave. At the fifth and beyond, a band and a half-ellipse of the same width are visibly different objects.

That is also why this is a different question from the one watered silk is a beat answers. A moiré in pressed silk is two copies of one rib beating at their first harmonic, and the profile of the rib hardly enters. A tulle over a voile beats at its fifth, and the profile is most of the answer.

What a round thread looks like from in front

A thread seen against the light is not a band. An opaque thread is, near enough, since a round section casts the same sharp shadow as a flat one of the same width. A thread that lets some light through, as a fine white nylon does, passes more of it near its edges, where the light crosses less of the thread, than at its centre, where it crosses a whole diameter.

The amount of thread a ray crosses is a chord of the circle, so across a round thread the absorbance is a half-ellipse: greatest at the centre line, falling to nothing at the edges. Write the centre’s absorbance as the thread’s optical depth. A depth of one passes 37 per cent of the light at the centre and all of it at the edges; a depth of four passes 2 per cent at the centre and is very nearly a band.

The bobbinet’s twisted sides are two threads twisted round one another, wider than a single thread by the twist ratio. They are drawn here as a round section of that width, so their centre’s depth is the single thread’s times the twist, and where two sides meet at a corner their absorbances add. Everything else is the earlier essay’s mesh: the same honeycomb, the same pitch of 1.55 millimetres, single threads a tenth of the pitch wide, the same voile 50 millimetres behind and the eye three metres away.

The two slow fringe families' strengths against a bobbinet's twist width. The strength of the two slow fringe families a 1.55 mm net makes over a 0.3 mm voile 50 mm behind it, seen from 3.0 m — the one beating along the voile's threads parallel to a net direction and the one beating across them through the point at √3 — for a bobbinet of round translucent threads passing 37% of the light at a single thread's centre (solid) and a sharp-edged straight-bar net (dashed), against the twisted sides' width as a multiple of the single threads', at a cover of 0.1. twist 1: bobbinet 0.0017 along and 0.0062 across, bars 0.0193 and 0.0013; twist 1.5: bobbinet 0.0013 along and 0.0060 across, bars 0.0193 and 0.0013; twist 2: bobbinet 0.0007 along and 0.0008 across, bars 0.0193 and 0.0013; twist 2.5: bobbinet 0.0002 along and 0.0064 across, bars 0.0193 and 0.0013; twist 3: bobbinet 0.0003 along and 0.0104 across, bars 0.0193 and 0.0013; twist 3.5: bobbinet 0.0005 along and 0.0081 across, bars 0.0193 and 0.0013. The bobbinet's across family is the stronger of its two at 1, 1.25, 1.5, 1.75, 2, 2.25, 2.5, 2.75, 3, 3.25, 3.5. A round thread's section is a half-ellipse in absorbance, and its transform has zeros of its own, so the curves still swing.
Fig. 1 The two slow families against the twist, for a bobbinet of round threads passing 37% of the light at a single thread’s centre, with the sharp straight-bar net dashed. The across family falls to 0.0008 at a twist of two and rises to 0.0104 at three; the along family falls from 0.0017 to almost nothing at 2.75.

At a depth of one the across family is the stronger of the two at every twist from one to three and a half. The prediction’s first half holds. Its strengths are lower than the sharp mesh’s, 0.0062 against 0.0092 at a twist of one, and the along family’s fall further, 0.0017 against 0.0036, which is the prediction’s second half. But the across family still drops to almost nothing near a twist of two, 0.0008, and rises again past it. The swing is still there.

A half-ellipse rings too

The expectation that a soft edge damps a high harmonic is a good rule with a condition attached. A shape whose edges are truly smooth, a Gaussian blur for instance, has a transform that falls away steadily with no zeros at all. A half-ellipse is not that kind of shape. It is soft at its centre and still has a corner at each edge, where its slope runs to infinity as it meets the background.

The transform of a band is a sine over its argument, and its first zero is at π\pi. The transform of a half-ellipse is a Bessel function, J1(x)/xJ_1(x)/x, and its first zero is at 3.83. That is the same function that describes the light from a round lens, and the ratio of those two zeros, 3.83 over π\pi, is the 1.22 in the rule every astronomer knows: a telescope of aperture DD resolves an angle of 1.22 λ/D1.22\,\lambda/D, not λ/D\lambda/D, because its aperture is round rather than square.

So a round thread’s high harmonic still swings through nulls. They come at slightly different widths from a band’s, and that is the whole difference.

The across null is the twisted sides’ own

Which part of the mesh passes through nought can be read off directly. The honeycomb is symmetric about the centre of each hole, so each kind of side’s contribution to an amplitude is a real number with a sign, and the twisted sides’ part and the single threads’ part can be drawn separately.

The two slow families' net points, split into what each kind of side contributes. For a bobbinet of single threads a tenth of the pitch wide, sharp-edged and opaque, the signed amplitude at the net point each slow family beats through, split into the twisted sides' part and the obliques' part, against the twist. the across family's net point: at a twist of 1, -33.5 and 3.1 thousandths, total -30.4; at 3, 42.5 and -3.0, total 39.5; the along family's net point: at a twist of 1, 1.8 and -13.7 thousandths, total -11.9; at 3, 7.4 and -11.0, total -3.7. A family vanishes where its total crosses nought. The across family's crossing is the twisted sides' own part changing sign; the along family's, where there is one, is the two parts cancelling. The figure cannot show the voile's side of the product.
Fig. 2 The signed amplitude at each slow family’s net point for the sharp, opaque mesh, split into the twisted sides’ part, the single obliques’ part and the whole. At the across point the twisted sides’ part runs from −0.033 at a twist of one to +0.042 at three and carries the whole; the obliques add a few thousandths. At the along point the obliques dominate throughout and the whole never reaches nought.

For the opaque mesh, the across family’s null is the twisted sides passing through nought on their own. Their part runs from −0.033 at a twist of one to +0.042 at three, the obliques contribute only a few thousandths of either sign, and the whole follows the twisted sides across zero at a twist of 1.87. That is a band’s own null: at the across family’s net point, the twisted sides are a row of bars whose width has just reached a whole cycle of the harmonic.

The two slow families' net points, split into what each kind of side contributes, round threads. For a bobbinet of single threads a tenth of the pitch wide, round and translucent, passing 37% of the light at a single thread's centre, the signed amplitude at the net point each slow family beats through, split into the twisted sides' part and the obliques' part, against the twist. the across family's net point: at a twist of 1, -21.5 and -0.8 thousandths, total -20.4; at 3, 37.9 and -0.8, total 34.5; the along family's net point: at a twist of 1, 2.8 and -6.5 thousandths, total -5.7; at 3, 7.9 and -6.5, total 0.8. A family vanishes where its total crosses nought. The across family's crossing is the twisted sides' own part changing sign; the along family's, where there is one, is the two parts cancelling. The figure cannot show the voile's side of the product.
Fig. 3 The same split for round threads passing 37% of the light at a single thread’s centre. At the across point the twisted sides’ part still changes sign, now near 2.07. At the along point the obliques’ part has shrunk to −0.0065 and the twisted sides’ grown to +0.008, and the whole crosses nought near 2.7.

Drawn round at a depth of one, the twisted sides’ part at the across point still changes sign, now at a twist of about 2.07 rather than 1.87. The null has moved outwards and not gone. It has moved because the twisted sides are now half-ellipses, whose first zero lies 1.22 times further out than a band’s, and it has moved by less than 1.22 because a depth of one is only partly translucent: the darkness of a thread is one less its transmittance, which saturates towards a band as the depth grows.

How far out the null can go

The move can be followed from an opaque thread to a nearly invisible one.

Where each slow family vanishes, against how translucent a bobbinet's thread is. The twist, as the twisted sides' width over a single thread's, at which each of a bobbinet's two slow fringe families passes through a null, against the share of light a single round thread passes at its centre, at a single-thread width of 0.1 of the pitch. The across family's null sits at 1.87 for an opaque thread and moves out to 2.27 at 95%, approaching 2.28, the opaque value times 1.22, the ratio of the first zero of a half-ellipse's transform to a band's. The along family has no null below a twist of 3.5 while a thread passes less than about 14% of the light, and then gains one that moves in from 3.39 to 1.66. The figure cannot say how translucent a real tulle thread is.
Fig. 4 The twist at which each slow family vanishes, against the share of light a single thread passes at its centre. The across family’s null moves from 1.87 for an opaque thread to 2.27 at 95%, approaching 2.28, the opaque value times 1.22. The along family has no null until a thread passes about a seventh of the light, then gains one that moves in from 3.39 to 1.66.

The across family’s null runs from a twist of 1.87 for an opaque thread to 2.27 for one that passes 95 per cent of the light at its centre, and the limit it approaches is 1.87 times 1.22, a twist of 2.28. At very small depths the darkness of a thread is very nearly its absorbance, which is exactly a half-ellipse, and the ratio of the zeros is the Bessel ratio. At large depths the darkness is a band with slightly rounded shoulders, and the null sits where a band’s does. Every real tulle lies somewhere between, and wherever it lies, its across family vanishes at some twist between 1.87 and 2.28.

That narrows rather than removes the earlier essay’s caution. A designer who wanted the across family strong would avoid a twist near two whatever the thread; one who wanted it gone, to keep a curtain from beating against a voile behind it, would aim for the same twist. The window is a little wide and moves with the yarn’s translucency, and its middle is fixed.

The along family gains a null of its own

The figure above also has a second line, and it is the larger surprise. The along family, which never vanished in the opaque mesh, has a null for any thread that passes more than about a seventh of the light at its centre.

The split shows why. At the along family’s net point the two kinds of side pull in opposite directions: the obliques’ part is negative and the twisted sides’ positive. In the opaque mesh the obliques win at every twist, −0.014 against at most +0.007, so the whole never reaches zero. Make the threads translucent and the obliques’ part shrinks, since a single thread is the narrower and thinner of the two and loses most of its darkness, while the twisted sides, twice as wide and twice as thick at the centre, keep more of theirs. At a depth of one the obliques fall to −0.0065 and the twisted sides reach +0.008, and the two cancel near a twist of 2.7.

So the along family’s null is a different kind of null from the across family’s. It is not a band’s own zero but a cancellation between two kinds of side, and it depends entirely on how differently the two kinds of side lose their darkness. At 14 per cent through a centre it sits at a twist of 3.39, at the edge of what a bobbinet machine makes; at 37 per cent at 2.69; at 95 per cent at 1.66, below the across family’s null. The more translucent the thread, the narrower the twist at which the along family disappears.

The two slow fringe families' strengths against a bobbinet's twist width. The strength of the two slow fringe families a 1.55 mm net makes over a 0.3 mm voile 50 mm behind it, seen from 3.0 m — the one beating along the voile's threads parallel to a net direction and the one beating across them through the point at √3 — for a bobbinet of round translucent threads passing 61% of the light at a single thread's centre (solid) and a sharp-edged straight-bar net (dashed), against the twisted sides' width as a multiple of the single threads', at a cover of 0.1. twist 1: bobbinet 0.0009 along and 0.0040 across, bars 0.0193 and 0.0013; twist 1.5: bobbinet 0.0006 along and 0.0044 across, bars 0.0193 and 0.0013; twist 2: bobbinet 0.0001 along and 0.0014 across, bars 0.0193 and 0.0013; twist 2.5: bobbinet 0.0003 along and 0.0037 across, bars 0.0193 and 0.0013; twist 3: bobbinet 0.0006 along and 0.0074 across, bars 0.0193 and 0.0013; twist 3.5: bobbinet 0.0007 along and 0.0068 across, bars 0.0193 and 0.0013. The bobbinet's across family is the stronger of its two at 1, 1.25, 1.5, 1.75, 2, 2.25, 2.5, 2.75, 3, 3.25, 3.5. A round thread's section is a half-ellipse in absorbance, and its transform has zeros of its own, so the curves still swing.
Fig. 5 The two families against the twist for round threads passing 61% of the light at a single thread’s centre. The across family runs from 0.0040 at a twist of one to 0.0074 at three, with its dip near two; the along family is never above 0.0009 and passes through nought between two and two and a half.

At a depth of a half, 61 per cent through a centre, both families are far weaker than the opaque mesh’s and the across family is the stronger everywhere, by four times or more. The prediction that the across family wins over the whole range comes true, and it comes true partly for a reason it did not foresee: not only because the across family stays strong but because the along family collapses through a null of its own.

The thread’s width matters as much as its twist

The earlier essay checked its finding at every single-thread width from a twentieth of the pitch to a quarter, with the twisted sides twice as wide. The same sweep with round threads shows how close a twist of two sits to the across family’s null at every width.

The two slow fringe families' strengths against a bobbinet's thread width. The strength of the two slow fringe families a 1.55 mm net makes over a 0.3 mm voile 50 mm behind it, seen from 3.0 m — the one beating along the voile's threads parallel to a net direction and the one beating across them through the point at √3 — for a bobbinet of round translucent threads passing 37% of the light at a single thread's centre (solid) and a sharp-edged straight-bar net (dashed), against the single threads' width, the twisted sides 2 times as wide. cover 0.050: bobbinet 0.0004 along and 0.0088 across, bars 0.0136 and 0.0006; cover 0.100: bobbinet 0.0007 along and 0.0008 across, bars 0.0193 and 0.0013; cover 0.150: bobbinet 0.0009 along and 0.0081 across, bars 0.0136 and 0.0005; cover 0.200: bobbinet 0.0010 along and 0.0008 across, bars 0.0000 and 0.0010; cover 0.250: bobbinet 0.0009 along and 0.0076 across, bars 0.0136 and 0.0012. The bobbinet's across family is the stronger of its two at 0.050, 0.075, 0.100, 0.125, 0.150, 0.175, 0.225, 0.250. A round thread's section is a half-ellipse in absorbance, and its transform has zeros of its own, so the curves still swing.
Fig. 6 The two families against the single threads’ width, twisted sides twice as wide, for round threads passing 37% at a centre. The across family swings from 0.0088 at a twentieth of the pitch to 0.0008 at a tenth and back, and again at a fifth; the along family stays under 0.0011 throughout, and is the stronger only near a fifth.

The across family swings between about 0.008 and nearly nothing with the single thread’s width, low at a tenth of the pitch and again at a fifth, high at a twentieth, three twentieths and a quarter. Only near a fifth does the along family win, by 0.0010 to 0.0008, and only because the across family is at its null. The swing is the width of the twisted sides passing through whole cycles of the harmonic, and with the twist fixed at two the single thread’s width moves both.

Where the round thread is not the end of it

Translucency is a guess. The optical depth of a tulle’s threads has not been measured here, and it depends on the fibre, its delustrant, its dye and how the light falls. A white nylon tulle against a window is more translucent than a black one; a heavily delustred yarn scatters as much as it absorbs, and scattered light is not removed from what the eye sees but spread. The sweep from 2 to 95 per cent through a centre covers what any thread could be, and the conclusions are stated over the whole sweep for that reason.

The twisted sides are drawn as one round section. A real twist is two threads wound round each other, whose outline varies along the side, and whose absorbance across it is two overlapping half-ellipses rather than one wide one. Two narrower half-ellipses would move the twisted sides’ zero, since each is narrower than the pair, and blur it, since their spacing varies along the twist; which way the first effect goes has not been computed.

The voile is still the sharp one. Its threads were left as the earlier essays drew them, so that this comparison is the net’s alone; the voile’s first harmonic, which is all the slow families use of it, is little changed by a round profile.

Strength is still not visibility. Each family’s strength is the product of the net’s amplitude and the voile’s, as before. Whether a fringe of that strength is seen depends on the eye’s sensitivity at its spacing, which the reed leaves its own mark and a slub finds the width of the cloth handled for their own patterns, and on whatever glass stands in front, which glass hides a black net and not a white voile priced.

How the amplitudes were computed

The mesh is the earlier essay’s honeycomb, sampled at 240 points across its unit cell. At each point every side within half its own width contributes an absorbance of the thread’s depth, times its width over a single thread’s, times the square root of one less the square of its distance from the side’s centre line over its half-width; the absorbances add, and the point’s transmittance is ee to minus their sum. At infinite depth this is the sharp mesh exactly. Each amplitude is a direct sum of transmittance times e−2πi G⋅re^{-2\pi i\,G\cdot r} over the cell, divided by the number of points.

The parts use the cell’s symmetry about a hole’s centre, about which every transform is real. Each kind of side’s darkness, one less the transmittance it would have alone, is transformed separately, and an amplitude off the origin is minus the transform of the darkness. A null is found where an amplitude changes sign, on a grid of twists a twentieth apart, with the crossing interpolated.

At infinite depth the sum is the sharp mesh’s exactly, by construction, and a depth of nought or less is refused. The across family’s null at the smallest depth was set against 1.22 times the opaque one, 2.27 against 2.28, as the check that the half-ellipse is what moves it.

Who found what

The diffraction pattern of a circular aperture is Airy’s, from 1835, and its first dark ring at 1.22 wavelengths over the aperture is the textbook limit of a telescope’s resolution; the transform of a half-ellipse is the same Bessel function. Moiré in nets and sheers is ordinary experience and has been used to measure fabric structure, and bobbinet tulle has been made on the machines descended from Heathcoat’s since 1809.

What is added here is the bobbinet drawn with translucent round threads; the finding that the across family’s null survives, because it is the twisted sides’ own zero and a half-ellipse has zeros too, and moves out by at most the Airy ratio; the along family’s new null, a cancellation between the twisted and the single sides that only translucency allows; and the confirmation, by a route the prediction did not foresee, that the across family is the stronger at every twist.

Still open: the twist a tulle maker actually runs

The whole of this places a bobbinet’s slow fringes against its twist, and every result says a twist near two is special: the across family vanishes somewhere between 1.87 and 2.28, and for a translucent thread the along family’s null is nearby. What twist a bobbinet machine gives its twisted sides is a fact about the machine, how many times the bobbin threads are traversed round the warp threads before they move on, and a fine tulle and a coarse one are made with different numbers of traverses.

A ratio of the twisted sides’ width to the single threads’, read under a low-power microscope on a few commercial tulles, would place each on these curves. If tulle makers’ ordinary settings put the ratio near two, then the curtains most people hang over voiles are the ones whose across family is weakest, and the earlier essay’s walk-back test, the wide check’s bars swelling and vanishing near nine metres, would be hard to see in practice for the reason found here rather than for the reason the straight-bar model gave.

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.

BeatHarmonicMagnificationMoirePlane group