The nodes a drape test throws away

The standard drape test is a hundred years old in outline and seventy in its current form. A circular specimen is laid over a smaller circular pedestal, the overhang falls, a light above casts its shadow onto a ring of paper, and the fraction of the annulus the shadow covers is the drape coefficient.

While that is happening, the specimen does something else that the operator can see and the method does not ask about: it falls into a definite number of folds. Four, or six, or nine, depending on the fabric. Every published photograph of a drape test shows them; no published drape result records them.

They are a buckling mode, and they carry more of the fabric’s stiffness than the number the test reports.

The fold count is the previous rung’s arithmetic, curved round #

A wrinkle’s wavelength comes from a competition: bending resists short waves, tension resists long ones, and the compromise is a quarter power. A draped disc is the same competition with gravity in the place of the tension.

The overhang is a ring of cloth hanging off the edge of a pedestal. Going round the pedestal there is more circumference than the cloth can lie flat over once it has fallen, so the ring is in compression azimuthally — and it does what a compressed sheet does, which is leave the plane. Bending resists many folds; gravity resists deep ones, because a deep fold lifts material.

The length that decides the balance is the one where a fabric’s own weight and its own stiffness come to the same thing, and that length has a name on this site already. It is the bending length: the quantity the cantilever test measures, which is a rigidity with the fabric’s weight already divided out. The scaling that follows is

n ≈ ( R / c )^¾

with R the radius of the overhang and c the bending length — three quarters rather than the wrinkle’s quarter, because a disc’s geometry brings the radius in twice.

A specimen no larger than its own bending length does not drape at all, and the arithmetic refuses rather than returning a fold count of one. That is a real boundary: a stiff enough card laid over a pedestal simply sits there.

What was counted, and how #

The prediction is run over the whole range of ordinary dress fabric — bending lengths from six millimetres to forty-six, which spans a chiffon to a heavy coating — on the standard 150 mm specimen.

The fold count moves by a factor of five and a half; the coefficient moves by fourteen points.

That comparison needs one honest qualification, and it is the site’s own earlier finding. The coefficient in that column is what the specimen would report with its hem at a fixed radius — which isolates the fold count, and is not what happens. In a real test a stiffer fabric’s hem stands further out, and the site’s hem model says the coefficient answers strongly to that: moving the mean hem radius from 9.4 to 14.8 takes the coefficient from 0.08 to 0.96, where changing the fold count from two to twelve at a fixed radius moves it by nineteen points. An earlier essay here asserted that ratio — the coefficient is more than four times as sensitive to the radius as to the folds — and it still holds.

So the honest statement is not that the coefficient is insensitive to stiffness. It is that the coefficient and the fold count are two nearly independent readings of one specimen, the coefficient reads the radius and the fold count reads the bending length, and the standard method takes one of them and discards the other.

Why a fold count is a better measurement than it sounds #

Counting folds looks like the crudest possible observation and it has three properties a shadow area does not.

It is an integer. There is no reading error, no calibration, no lighting dependence and no operator judgement. Two people counting the folds on the same specimen agree exactly or one of them has miscounted.

It is insensitive to the pedestal. The coefficient depends on the ratio of specimen to pedestal, which is why drape coefficients from different standards are not comparable, in the way a thread count from two standards is not and why the literature is full of conversion tables. The fold count depends on the overhang’s radius over the bending length, and the overhang is a length rather than a ratio.

And it is a direct reading of a physical length. Inverting the three-quarter power gives the bending length from the fold count and the radius, so a fold count is a stiffness measurement — a coarse one, because the answer is quantised, but one requiring no instrument at all.

The coarseness is the honest cost. Between four folds and five there is a whole band of fabrics, and the quantisation is worst exactly where most fabrics are. What the count gives cheaply is the order of magnitude and a check on a cantilever reading; what it cannot give is a resolution.

What the fold count would be worth in a mill #

Three uses, and none of them needs an instrument.

A check on a cantilever reading. The cantilever measures one strip in one direction and is easy to get wrong — the strip curls, the edge is not sharp, the operator eyeballs the angle. Counting the folds on a disc of the same cloth gives an independent estimate of the same bending length by an entirely different route, and the two disagreeing is a signal.

A sorting measurement. Fabrics arriving at a cutting room differ in how they will hang, and the quantity that decides it is the bending length. A 150 mm disc over a 90 mm pedestal and a count of folds sorts a delivery into limp, medium and stiff in seconds, with no calibration and no light box.

And a directional reading, which neither the cantilever nor the coefficient supplies cheaply. A cloth cut as a disc and draped shows its stiffest directions as its broadest folds — so where the folds sit relative to the grain is a picture of the fabric’s anisotropy, which the site’s directional drape essay measures with three cantilever readings and a great deal more effort.

That last asymmetry is worth carrying because it is not obvious and it is exploitable. The cantilever’s fractional error is roughly constant across fabrics; the fold count’s is worst where the count is small. The two are therefore complementary rather than redundant, and a mill that took both would have a better estimate of a limp fabric than either alone gives.

Why an integer is a good measurement in one place and a bad one in another #

The fold count’s quantisation is worst where the count is small, and it is worth working out what that costs in the units the measurement is trying to reach, because the answer decides where the reading is usable.

Inverting the three-quarter power, a bending length is the radius divided by the fold count to the four-thirds. So a fold count off by one gives a bending length off by roughly four thirds of one over the count — a third at four folds, an eighth at eleven. At the limp end the reading is good to a tenth, which is comparable with a cantilever’s own repeatability; at the stiff end it is good to a third, which is not a measurement of anything.

That is a steeper degradation than it looks, because the fabrics with low fold counts are also the fabrics whose bending lengths are large, so the absolute error grows twice over. A coating at forty-six millimetres of bending length read to a third is read to fifteen millimetres, which spans most of the stiff range.

So the fold count is a fine instrument for chiffons and a useless one for coatings, and the boundary is somewhere near five folds. That is a sharper statement than “coarse at the stiff end”, and it puts the useful range at bending lengths under about twenty millimetres — which covers dress fabrics, shirtings and linings and excludes everything a tailor would call a cloth.

The complementarity with the cantilever is therefore real rather than rhetorical. A cantilever’s fractional error is roughly constant, so it is the better instrument for stiff fabrics and the equal of the fold count for limp ones; the fold count is free and needs no apparatus. A laboratory that used the count as a screen and the cantilever as the measurement would be using each where it is strongest, and the screen is the half nobody has.

What a fold count would settle that neither test can #

There is one question both instruments are silent about and the fold count answers directly, which is worth setting out because it is the strongest argument for recording it.

Neither the cantilever nor the coefficient can say whether a fabric is behaving as one material. A cantilever reads one strip in one direction; a coefficient reads one area. A cloth that has been finished unevenly, or that has a resin only partly cured, or that has been stretched across its width more than along it, gives a perfectly ordinary reading on both — because both reduce the specimen to a single number and a single number cannot be internally inconsistent.

A fold pattern can. A disc of a uniform fabric falls into evenly spaced folds and a disc of a non-uniform one does not, and the unevenness is visible in the same photograph the coefficient is computed from. Broad folds where the cloth is stiff and narrow ones where it is limp is a picture of the variation, at whatever spatial scale the variation has.

That gives the observation a use nobody has claimed for it. It is not merely a cheap stiffness reading; it is the only routine fabric test that reports a spatial property rather than an average. A drape test on a cloth with a finishing streak in it would show the streak as an irregularity in the fold spacing, and the coefficient computed from the same shadow would be entirely normal.

And the irregularity is easier to see than to measure. A count is an integer and an irregularity is a distribution, so recording it properly needs the outline rather than the count — which the image-based systems already compute and already discard. The information has been available and thrown away twice: once by the operator who did not count, and once by the software that measured the area of an outline it had already traced.

What the standard leaves out and why #

It is worth asking why the fold count was dropped, because it was not an oversight — it was in the earliest descriptions of the test.

The drape test was built to produce a single number that correlates with how a fabric looks made up, and a single number is what a specification can carry. A fold count and a coefficient are two numbers, they do not combine into one, and the coefficient was the one that correlated better with subjective assessments of drape in the trials that established the method.

That is a defensible decision and it has a cost that has been paid for seventy years: the two numbers measure different things, so a method that keeps one of them cannot be inverted for the other. Two fabrics with the same drape coefficient and different fold counts are different fabrics, and every published drape coefficient has thrown that difference away.

There is a modern version of the same loss. Image-based drape testing photographs the shadow and computes not only the area but the whole outline, and from the outline the fold count is immediate — most systems compute it and most reports still quote the coefficient alone.

The two measurements this site now has #

The site’s mechanics has been geometric throughout: a net of inextensible threads says which shapes are available, and no stiffness enters. This anchor has added the other half, and it is worth setting the two measurements side by side because they are the same fabric property read two ways.

The cantilever hangs a strip over an edge and reads a length. It is one direction, one strip, one number, and it is the input to both essays on this anchor.

The draped disc wraps a whole specimen and reads two things: a radius, which the trade records as a coefficient, and a fold count, which it does not. The disc averages over every direction at once, which is a strength when a fabric is isotropic and a problem when it is not — and a woven cloth is never isotropic in bending, so a disc’s fold count is a compromise between the warp’s stiffness, the weft’s and the bias’s.

That last point is where the prediction here is weakest and it is also where it is most interesting. A fabric whose warp and weft stiffnesses differ by a factor of three does not fall into evenly spaced folds; it falls into folds that are broader on the stiff axes and finer on the limp ones, and the count that comes out is neither prediction. The regularity of a drape test’s folds is itself a measurement of anisotropy and nobody records that either.

The same fabric, two shapes, one length #

Both rungs of this anchor compute a shape from one measured length, and setting them beside each other shows how much that one length carries.

The last row is the one worth carrying. A fabric whose rigidity rises by fifty gives a wrinkle spacing under three times as wide, and a fabric whose bending length rises by eight gives a fifth as many folds. The disc is a far more sensitive instrument than the panel, for the same reason a three-quarter power is steeper than a quarter — and the trade’s test is on the disc, which is the right choice, and it reads the wrong number off it.

Where the model stops #

The three-quarter power is a scaling law, not a solution. Its prefactor depends on the boundary — how the specimen is clamped, how much overhang there is, how the ring is loaded — and the arithmetic here takes it as one. So the ratios between fold counts are the trustworthy part and the absolute integers are indicative.

Gravity is the only load. A real specimen has friction against the pedestal at its inner edge, which resists the folds forming and pins their phase, and the standard is careful about the pedestal’s surface for exactly that reason.

The fabric is treated as isotropic, which it is not, as above.

And nothing here is plastic or time-dependent. A specimen left on a drapemeter changes over minutes as the folds settle, and the standard specifies a waiting time. That settling is a viscoelastic process with no term in this arithmetic.

Who found it, and when #

The drape test’s ancestry runs from Peirce’s 1930 paper — which introduced the bending length, the cantilever method and the whole idea of measuring a fabric’s stiffness as a length — through Chu, Cummings and Teixeira in the 1950s, who built the disc-and-shadow apparatus and defined the coefficient, to the national standards that fixed the specimen and pedestal sizes.

The fold count is mentioned in Chu’s own work as an observation. It was not developed, and the reason given at the time was reasonable: nobody had a theory that predicted it, so it was a number with nothing to compare against.

The theory arrived fifty years later from somewhere else entirely. Cerda and Mahadevan’s scaling arguments about wrinkling, written for stretched sheets and hanging curtains, give the elastic–gravity length and the exponents that follow from it — and the draped disc is one of the examples in that literature rather than in the textile one. Reading the fold count as a buckling mode is immediate once the scaling is in hand and was not available before it.

So this is a case of a measurement waiting for a theory, and the theory arriving in a different field and not being carried back. The textile literature has the bending length and the fold count and does not connect them; the physics literature has the connection and does not know that a standard test has been counting folds and discarding them since 1950.

Where the ladder goes next #

This anchor has taken the fabric’s stiffness into two shapes it decides — a wrinkle’s spacing and a disc’s fold count. What neither rung has is the amplitude: how deep a wrinkle goes and how far a fold hangs out, which needs the compression rather than only the geometry. That is the quantity the drape coefficient is actually reading, and computing it rather than measuring it is recorded here as not done.