Balance, and what an unbalanced cloth does

Look at a piece of denim and what is visible is warp. Look at the back and there is more weft than the front had, but still not much. Look at a piece of ordinary shirting and warp and weft are visible in equal measure, which is why the surface reads as a texture rather than as a direction.

That is balance: the share of the face each thread system takes. It is the mean of the weave matrix — count the intersections where the warp is up, divide by the total — and it is one of the only quantities in this subject that a reader can take off a draft without computing anything.

What this essay is about is the gap between how easy balance is to read and how much of the cloth’s behaviour depends on it.

The number, and where it comes from #

For a two-two twill: half the intersections are warp-up, so the balance is a half and the cloth is balanced. For a three-one twill: three quarters, and the cloth is warp-faced. For a one-three twill: a quarter, weft-faced. For an eight-end satin: seven eighths, which is as warp-faced as a common weave gets.

Two refinements are worth having.

Regularity. A weave can average a half and still have some ends showing more warp than others. A cloth is regular when every end shows the same fraction and every pick likewise, which every named weave on this site is and most fancy weaves are not. Balance without regularity is a mean over a distribution, and the distribution matters — an end that never shows is an end that also never wears.

Which face. Balance as computed is the fraction on the face; the back is one minus it. A three-one twill is warp-faced on one side and weft-faced on the other, and calling it warp-faced is a decision about which side is the face — a decision made outside the draft.

What follows: wear #

The consequence that matters most commercially is abrasion, and it is direct.

A rubbing surface finds whatever is on top. In a warp-faced cloth that is warp, so the warp takes essentially all the wear and the weft is protected. In a weft-faced cloth the reverse. In a balanced cloth both wear, at half the rate each.

That is why the two systems in a warp-faced cloth are so often different yarns. Denim’s warp is a stronger, harder-twisted yarn than its weft, and it is the warp that is dyed — because dye is expensive, and dyeing the system nobody sees is waste. The whole economics of indigo denim, and the reason it fades the way it does, follows from the balance of a three-one twill.

The same argument runs the other way in a weft-faced cloth. A tapestry or a rep is weft-faced precisely so that the weft can carry the design and the warp can be a cheap, strong structural thread that never shows. The warp in a tapestry is not decorative in any sense: it is scaffolding.

What follows: colour #

Balance decides which system’s colour is seen, and by how much. That is obvious enough. What is less obvious is that it interacts with colour-and-weave effects in a way that is not a simple weighting.

A balanced cloth with dark warp and light weft shows a fine speckle, because half the intersections are one colour and half the other and they are interleaved. A warp-faced cloth in the same two yarns shows dark, with light spots. Those are not the same pattern at different intensities; they are different patterns, because the arrangement of the minority colour changes with the weave.

What follows: cover, and the count that hides it #

A cover factor is thread spacing times thread diameter, summed over the two systems with the overlap subtracted. It does not mention balance, and it does not need to: cover is about how much of the area is covered by thread, not about which thread.

But the setts do depend on balance, because a warp-faced cloth is warp-faced partly by having more warp in it. Denim has roughly twice as many ends as picks. So the two things a reader might mean by “a warp-faced cloth” — more warp on the surface, and more warp threads present — are different statements that happen to travel together.

Separating them matters when reading a specification. A cloth quoted as 100 ends and 50 picks per inch is unbalanced in its sett; whether it is warp-faced depends on the weave, and a plain weave at those setts is a warp-dense balanced cloth — which is a rib, not a satin.

What the census says #

Balance turns out to have a connection to cloth integrity that nobody would predict, and it comes out of the four-by-four enumeration.

Of the 22,874 four-by-four drafts in which every end and every pick interlaces at least once, 144 describe more than one cloth. Of those 22,874, exactly 90 are balanced and regular — two up and two down in every end and every pick.

Not one of the 90 separates.

That is a measurement, not a theorem. Nothing in it proves that balance forces integrity in general, and the sample is one repeat size. It survived a sample of balanced six-by-six drafts as well, which raises the confidence and settles nothing. Saying otherwise would be exactly the over-claim this site exists to avoid.

What it does suggest is a mechanism, and the mechanism is plausible. A separation needs some threads to be systematically above others, and balance is a constraint that forbids any thread from being above more than half the time. It is not obvious that the constraint is strong enough, and the enumeration says that at this size it is.

The unbalanced cloth’s other price #

Two costs of going far from balance, and the second is the one that gets designers into trouble.

Float length. A warp-faced cloth is warp-faced because the warp stays on the face for long runs, so the float is long by construction. Everything float length brings comes with it: lustre, drape, snagging, poor abrasion resistance at the float, and a cloth that can be set very densely because there are few bends to make room for.

Differential shrinkage and skew. In an unbalanced cloth the two systems take different crimps, so they respond differently to washing. A cloth whose warp crimp is much larger than its weft crimp shrinks much more in length than in width, and if the imbalance is not symmetric about the cloth’s own axes the fabric skews — the reason a pair of jeans’ side seam wanders round the leg after washing. That is a mechanical consequence of an arithmetic property, and it is one the draft predicts qualitatively and does not predict quantitatively.

Why the warp is usually the one on the face #

Across the whole of weaving, warp-faced cloths vastly outnumber weft-faced ones, and the reason is not aesthetic.

The warp is under tension for the whole time the cloth is being woven. It is abraded by the heddles, by the reed, and by every shed that opens and closes — several thousand times for a thread near the back of a long warp. So the warp is spun harder, plied more often, and often sized with starch before weaving, all of which makes it the stronger and harder yarn of the two.

Having gone to that expense, a weaver has a strong yarn and a weaker one, and putting the strong one on the surface is the obvious move: the surface is what wears. Warp-faced construction is therefore the default that falls out of how a loom works, and weft-faced construction is a deliberate departure made when the design has to be in the weft — which is exactly the tapestry case, where the weft carries the picture and is changed colour by colour.

There is a third case where the warp is on the face for a different reason again. In a satin the point is lustre, and lustre comes from long uninterrupted lengths of thread lying flat. The warp is the system that can be laid straightest, because it is under tension, so a warp-faced satin is brighter than a weft-faced one in the same yarn. That is a mechanical reason for an optical property, and it is the reason silk satin is woven the way round it is.

The limit: when one system covers completely #

Push balance far enough and the minority system disappears from the surface entirely, which is a construction rather than a failure and has its own name.

A rep or rib is a plain weave set so densely in one system that the other is completely covered. The weave is still plain — every intersection alternates — so the balance of the matrix is still exactly a half. What has changed is the sett: with four times as many ends as picks, the warp crowds together, the picks are buried, and the cloth shows nothing but warp.

That is worth pausing on, because it separates two things the word “warp-faced” runs together.

Structural imbalance is a property of the weave. A three-one twill is warp-faced in every setting, including a square one, because three quarters of its intersections are warp-up.

Setting imbalance is a property of the cloth. A plain weave is balanced in every setting, and a plain weave set four-to-one still shows only warp.

Both produce a warp-faced surface and they behave quite differently. The structurally warp-faced cloth has long floats and everything that comes with them. The densely set plain weave has the highest interlacing count there is and floats of one — so it is stiff, hard-wearing, and does not drape, which is precisely why grosgrain ribbon and Ottoman upholstery are made that way.

A specification quoting only “warp-faced” does not distinguish them, and the two cloths have almost nothing in common.

Measuring it on a real cloth #

Balance as defined here is a property of a draft. Getting it from a piece of fabric requires either the draft or a measurement, and the two routes disagree in an instructive way.

The analytical route is to pick the cloth apart: unravel enough of it to establish the repeat, write down the matrix, and take the mean. That gives the number in this essay exactly, and it is what a fabric analysis does.

The optical route is to look at the surface and estimate what fraction of it is warp. That gives a different number, and it is systematically further from a half than the analytical one — because the thread on the face is not merely visible, it is raised, and it spreads over its neighbours. A three-one twill measures three quarters analytically and looks more like nine tenths.

Neither number is wrong; they answer different questions. The analytical one predicts the wear and the topology. The optical one predicts the colour, because what a viewer sees is the projected area rather than the intersection count. A cloth described as “almost entirely warp on the face” is usually being described optically, and converting that back to a weave is not possible without knowing the sett.

That gap is a small instance of the site’s standing theme: the quantity that is easy to see is not the quantity the model computes, and they can be a fifth apart without either being in error.

Where the words came from #

The vocabulary is old and, unusually for this subject, consistent.

Warp-faced and weft-faced are straightforward and mean what they say. Balanced in the weaving trade means what it means here — an equal share of the face — but in the spinning trade a balanced yarn is one whose twist is neutralised so that it does not kink, which is an entirely different property, and the two collide constantly in writing that covers both.

The older English terms are more specific and less used. A tabby is a balanced plain weave; a rep is a warp-dense one; a half-linen or union describes the fibre rather than the balance. Weft-faced cloths carry the names of their trades: tapestry, kilim, rep, grosgrain, and in every case the name arrived with the product rather than with the construction.

The one term that repays care is sateen. A satin is warp-faced; a sateen is the same weave turned over, weft-faced, and the distinction is real and routinely ignored. Cotton sheeting sold as “sateen” is usually a weft-faced satin, which is why it wears differently from a silk satin of nominally the same construction: it is the weft doing the work, and the weft in a sheet is not the strong system.

The optical balance, written down #

The gap between the analytical balance and what a viewer sees is left above as a fifth and as a caution. It can be written as an expression, and once it is, the whole of the discrepancy turns out to belong to the sett rather than to the weave — which is the essay’s own distinction, arriving in a place it was not expected.

What a viewer sees is projected area, so each system’s share of the surface is its intersection count weighted by how much of the plan it covers when it is on top — which is its own cover factor. Writing b for the balance and r for the ratio of the two systems’ covers,

visible warp = b·r ÷ (b·r + 1 − b).

A three-one twill at a square sett looks exactly three-quarters warp — the analytical number, with no exaggeration whatever. Denim, at two ends to every pick, looks 86 per cent, which is where the essay’s nine tenths comes from.

So the optical exaggeration is not an artefact of threads standing proud. It is the sett ratio, and nothing else. A cloth whose two covers are equal is seen exactly as its matrix says, and a cloth that looks more warp-faced than its weave is a cloth with more warp in it.

That gives a second field measurement of the sett ratio, and one that needs no protractor: estimate the visible warp fraction, take the weave’s balance off the draft, and invert. It is a poor measurement — an eye judging an area fraction is doing something notoriously unreliable — and it is an independent one, so it can be checked against the twill angle, which measures the same ratio from a completely different property. Two rough readings that agree are worth more than either.

The sett cancels out of the wear #

The same weighting settles a question the wear section raises and leaves qualitative: how much faster does a warp thread wear than a weft thread in an unbalanced cloth?

Each system’s total exposure is its visible share; each system’s exposure per thread is that share divided by how many threads it has. So the ratio of the two is the visible odds divided by the sett ratio — and the sett ratio, which is what put the visible odds up in the first place, cancels straight out:

wear per warp thread ÷ wear per weft thread = b ÷ (1 − b).

The odds of the balance, exactly, with no sett in it at all.

A denim’s warp thread wears three times as fast as its weft thread, whether the cloth is set square or two to one. Setting a cloth warp-dense makes it look more warp-faced and does not make any individual warp thread wear faster — because the extra ends share the extra exposure exactly.

That is a cleaner statement of the trade’s practice than the surface argument gives on its own. Denim’s warp is spun harder, plied, sized and dyed because each of its ends is doing three ends’ worth of wearing, and that ratio is fixed by the twill’s sequence — not by how densely the cloth is set, which a mill may change from one order to the next.

And it says where the trade’s habit breaks. An eight-end satin’s face thread wears seven times as fast as its back thread, which is a far harsher ratio than denim’s and is carried by a cloth chosen for lustre rather than for wear. A satin’s warp is not usually the hard, plied, sized yarn a denim’s is — it is the fine lustrous one — so the system doing seven-eighths of the wearing is the weaker of the two. That is the arithmetic behind satin’s reputation, and it is a construction decision rather than a property of silk.

What balance does not decide #

The list is longer than it looks and is worth having in one place, because balance is easy to read and therefore over-read.

It does not decide firmness. Interlacings per intersection is a separate quantity. A balanced weave can be very firm (plain) or quite loose (a balanced twill of long repeat), and an unbalanced one likewise.

It does not decide layer count. The census above found the two related statistically at one repeat size; the criterion itself has no balance term in it.

It does not decide strength. Which system carries load depends on which is straighter and which is more numerous, not on which is visible. A weft-faced tapestry’s strength is in its hidden warp.

And it says nothing about the yarn. The same balance in two cloths made of different yarns gives entirely different surfaces, and every quantity in this essay is a property of the matrix.

Where the ladder goes next #

The rung below is thread count is not quality, which is the cover ladder’s base and the reason any of these quantities need separating from one another at all.

The natural companions are the float, which balance drags along with it, and the interlacings, which it does not. The construction where balance becomes a design variable in its own right is the backed cloth, where a face and a back can be given opposite balances on purpose.

And the geometry underneath all of it is the sett, which decides how many threads there are before the weave decides which of them show.