A chenille is a yarn that is already a fabric

Everything in this collection has been arranged in two levels. There is the thread, which has a count, a diameter and a twist and is otherwise a line; and there is the cloth, which has a matrix, a sett, a crimp and a structure. Structure lives at the second level and the first level supplies numbers to it.

That arrangement is a convenience and the trade does not respect it. A whole class of yarns has structure inside the thread — chenille, bouclé, gimp, slub, snarl, loop — and for those the two levels are not separate at all.

The clearest of them is the chenille, and it is clearest because it is not a metaphor. A chenille yarn is a piece of woven cloth cut into strips.

The claim #

A fancy yarn is a structure built at the thread level rather than at the cloth level, and every structural question this collection asks of a cloth can be asked of it.

• Does it hang together? A chenille’s pile is held by its core threads exactly as a tuft is held by its ground, and the criterion is the same capstan argument. A chenille that sheds is a chenille whose core twist is too soft.
• How many systems has it? Three — two core threads and a pile — which makes a chenille yarn a structure a binary matrix cannot express, for exactly the reason a pile fabric is.
• What is its count? Mostly pile, and the pile carries no load, so a chenille’s strength per tex is far below any ordinary yarn’s. That is a design consequence rather than a defect.
• And a slub yarn is a periodic fault made on purpose, so everything this collection computes about periodic faults applies to it — including the beat against the cloth’s width.

How a chenille is made, which is the argument #

Weave a narrow gauze: two, four or six warp ends in tightly spaced pairs, with wide gaps between the pairs, and a weft laid across all of them. Then cut the weft, lengthwise, exactly midway between each pair of warp ends.

What comes off the cutter is a set of strips, each one a pair of warp ends holding a fringe of cut weft on both sides. Twist the pair and the fringe stands out all round: a caterpillar of pile on a two-thread core, which is what chenille means.

Every quantity in the finished yarn is a quantity of the gauze it was cut from.

• The pile length is half the gap between warp pairs, less the take-up.
• The pile density along the yarn is the gauze’s pick density, exactly.
• The core is the two warp ends, and how hard they are twisted after cutting decides how tightly the pile is gripped.
• The count is the sum: two core ends plus two pile lengths per pick, per unit length of yarn.

That last one is worth doing, because it produces the number that decides what a chenille is good for. A chenille with 8 picks per centimetre of gauze and a 4 mm pile carries 8 mm of pile per centimetre from each side — 16 mm of pile yarn for 10 mm of yarn length, per side, which is to say the pile is several times the core in mass. A chenille yarn is mostly pile by weight, and the pile is attached at its middle and free at both ends.

What moving a structure between levels costs #

Building the pile in the yarn rather than in the cloth is a real choice with real consequences, and they are not all in one direction.

What it buys. A chenille can be woven or knitted on ordinary machinery: the pile arrives already attached, so the loom needs no second warp beam, no wires, no cutting motion, and none of the apparatus a velvet or a terry demands. The pile can also be put exactly where the designer wants it, by using the chenille as one thread among others, which a woven pile cannot do without a second warp.

What it costs. The pile is now inside a thread that has to be handled, wound, threaded through heddles and beaten up, and every one of those operations abrades a structure whose pile is held only by twist. Chenille sheds, and it sheds most in the loom.

And what it changes about the arithmetic. A chenille’s diameter is not a mass diameter in any useful sense: the pile stands off in every direction, so the yarn’s contact diameter is several times the diameter its mass implies. Everything this collection says about the two diameters of an ordinary yarn applies here with the gap enormously widened — a chenille is a hair layer with a yarn inside it.

The consequence is that none of this collection’s cloth arithmetic works on a chenille cloth. The cover factor computed from the mass diameter is a fraction of what the cloth shows; the sett a warp will take is set by the pile rather than by the core; and the weight is right, because a weight never went through a diameter.

The slub, which is the variation ladder run backwards #

A chenille is the extreme case. The commoner fancy yarns change the thread less drastically, and one of them is worth following because it inverts an argument this collection has spent several rungs building.

A slub yarn is spun with deliberate thick places — a controlled variation in the drafting, at an amplitude and a spacing the spinner chooses.

Everything in this collection about yarn irregularity has been about accidental variation: a floor set by counting fibres, an index measuring how far a process exceeds it, and an averaging that beats the result down by the square root of the threads in view. All of that assumes the variation is random.

A slub is not random. It has a period, and a periodic variation is not averaged: it beats a random one of the same size by √(2n/π) at its own frequency, which over a few hundred threads is a factor of ten or more.

So a slub yarn is a designed defect, and the design problem is the same as the defect problem with the sign changed. The spinner picks the period; the weaver’s cloth width picks what pattern it makes.

That is a genuinely practical statement and it is the reason slub yarns are specified with a random or variable slub distribution. Making the period irregular deliberately puts the variation back into the regime where the cloth’s own averaging handles it — which is a designer choosing to give up the visibility the periodicity would have bought.

The count arithmetic, which is the whole design #

The pile-to-core ratio is estimated above from one example. Written as a formula it becomes the chenille designer’s only real lever, because every property anybody buys a chenille for and every property they complain about is on one side of it or the other.

Per unit length of finished yarn, the core is two warp ends of the gauze and the pile is two arms per pick. So

pile mass ÷ core mass = P × L × (weft tex ÷ core tex),

with P the gauze’s picks per unit length and L the pile length. And the fraction of the finished count that is core — which is the fraction carrying any load at all — is the reciprocal of one plus that:

core fraction = 1 ÷ (1 + P·L·t_w/t_c).

For the worked example — eight picks per centimetre, a four-millimetre pile, the same count in both systems — the ratio is 3.2 and the core fraction is 0.238. So the yarn’s strength per tex is under a quarter of the core yarn’s own, before any allowance for what the cutting and the twisting did to it.

That single expression prices every decision a chenille maker makes.

A longer pile is a softer, fuller yarn and a proportionally weaker one. Doubling L from four millimetres to eight takes the core fraction from 0.238 to 0.135 — nearly halving the strength per tex for a yarn that looks twice as good.

A denser gauze is the same trade. P and L enter identically, so eight picks at four millimetres and four picks at eight millimetres give the same strength and quite different surfaces — the first a close even pile, the second a sparse coarse one.

And the only lever that improves both at once is the count ratio. A heavier core against a finer weft raises the core fraction without touching the pile’s length or its spacing, so the yarn keeps its handle and gains its strength. That is why the core of a good chenille is markedly coarser than its pile, and it is a design decision the finished yarn does not advertise.

The awkward corollary is the one a weaver meets. A chenille cannot be made strong and full at the same time, because the two quantities are the two ends of one ratio, and there is no construction that escapes it — the pile is attached at its middle and carries nothing whatever it is made of. So a chenille used as a warp has to be either weak or thin, which is why it almost never is one.

The structural questions, asked at the thread level #

The rest of the collection’s apparatus transfers directly and it is worth listing what it says.

Integrity. The criterion for a cloth being one cloth is that the above-and-below relation between threads is strongly connected. A chenille’s core and pile form exactly such a relation, and the pile is a set of threads each held at one point. It is not connected, in the criterion’s sense — a pile end is attached and is not part of a two-way relation with anything — which is the same verdict the criterion gives about a woven pile and for the same reason. The criterion is topological and says a tuft bound at one point is attached; what decides whether it stays in is the friction the criterion cannot see.

Counting. A chenille’s count is a sum of its parts and its fibre count is a sum of theirs, so the evenness floor applies to it — but only to the core, since the pile’s mass is not carried along the yarn in the way the floor’s argument assumes. A chenille’s measured irregularity is dominated by the pile’s own placement, which is the gauze’s pick density and is a machine setting.

Strength. The pile contributes nothing, so a chenille’s strength per tex is the core’s strength divided by the whole count. For the example above, with the pile several times the core, that is a small fraction of an ordinary yarn’s — and it is why chenille is used for wefts and furnishings and not for warps or for anything structural.

The class, and what each member does to a thread #

Set out together, because the family is usually presented as a catalogue of effects and it is a small set of mechanisms.

Slub — a periodic change of count, made in the drafting. The thread stays one thread; only its mass per unit length varies. Everything above about periodicity applies.

Bouclé and loop — a periodic change of length. An effect yarn is fed faster than a core, so the surplus forms loops, and a binder yarn twisted the other way traps them. Three yarns, one of which does nothing but hold.

Gimp and spiral — a differential twist. Two yarns twisted together at rates that leave one longer than the other, so the longer one wraps the shorter in a helix of visible pitch. This is the folding argument run deliberately out of balance.

Snarl — a deliberate over-twist. The effect yarn is fed with more twist than it can hold and allowed to snarl at intervals, which is the crepe mechanism made local instead of general.

Chenille — a third system, made by cutting a woven gauze.

And knop or nep — a local accumulation, made by holding the effect yarn stationary while the core runs, so that mass piles up at a point.

Six mechanisms, and only the last two add anything to the thread that a single spinning frame could not. What the list makes visible is that a fancy yarn is nearly always a timing device: two or more strands delivered at different rates, with the difference showing as loop, snarl, spiral or knop. The structure is in the delivery, not in the fibre.

What a fancy yarn does to this collection’s arithmetic #

Worth being blunt, because the answer is mostly “breaks it”.

The count is right. A fancy yarn’s tex is a mass per unit length like any other, so areal weight survives.

The diameter is meaningless. A yarn whose effect is a loop standing off the core has no single diameter, so every cover, jam, hole and thickness computation in this collection is inapplicable rather than merely inaccurate.

The evenness floor applies only to the base yarn. The deliberate variation is not a departure from a Poisson floor; it is a different signal added on top, and the two are measured together by an instrument that cannot distinguish them. This is a real and practical difficulty: an evenness tester run on a slub yarn reports an enormous coefficient of variation and says nothing about how well the yarn was spun.

And the strength is the base yarn’s, or the binder’s, or the core’s — whichever is continuous and load-bearing. In every member of the class above there is exactly one such component, and the fancy part contributes nothing.

That last observation is the useful one for anybody buying such a yarn. Ask which strand is continuous, and every mechanical property of the yarn is that strand’s, divided by the whole count.

What was counted, and how #

Nothing in this essay is a new computation, and that is deliberate. Every quantity in it is one this collection already computes at the cloth level, asked of a smaller object — the capstan grip, the third-system argument, the integrity criterion, the periodic-visibility ratio and the beat against a cloth width.

What is asserted is the transfer: that the same functions, at the same tolerances, apply. The wrap-angle and capstan machinery is the pile family’s own and is exercised at the tuft parameters this collection already gates; the periodic-visibility ratio is checked against a simulation at four widths and required to match its closed form to three per cent.

Where the model stops #

A chenille’s own geometry is not modelled. Nothing here computes the pile’s angle to the core, how the two core threads twist about each other, or how the pile distributes round the yarn. The claim is that the questions transfer, not that the answers have been computed.

The shedding is argued rather than predicted. The capstan grip on a chenille’s pile depends on the core twist, the friction and the pile’s own diameter, all of which are known in principle; what is not known is how much abrasion a pass through a heddle applies.

The slub arithmetic assumes a strictly periodic slub, and commercial slub yarns are deliberately not. The intermediate case — a distribution of periods — is where the design work actually happens and this collection has nothing for it.

And “a yarn that is already a fabric” is a description rather than a theorem. A chenille is a woven structure by construction; a bouclé is not, and a slub is certainly not. What they share is that their structure is at the thread level, and the strength of the analogy varies a good deal across the class.

Where the ladder goes next #

Into the other members of the class, each of which has a different structural question attached: a bouclé’s loops are held by a binder thread and its failure is the binder slipping; a gimp’s core and effect yarn are twisted at different rates so that one buckles, which is the same mechanism a seersucker uses at the cloth level.

And outward, to the general point the class exhibits. Structure can be built at any level, and moving it changes what can be computed rather than what exists. A pile in the cloth is countable from a matrix; the same pile in the yarn is not, and the collection’s arithmetic goes quiet — not because the fabric is more complicated but because the machinery was built at one level and the structure has moved to another.