The stitch that weakens the seam

A seam is stronger with more stitches in it, up to a point, and past that point it is weaker. Everybody who sews knows the first half. The second half is worth computing, because the point is not a matter of taste and its position moves with something the sewing machine knows nothing about: how closely the cloth was woven.

There are two limits, they run opposite ways, and the seam is whichever is smaller.

The rising line: sewing thread across the seam #

The first limit is the easy one. A seam carries its load through the sewing thread that crosses it, so the strength available goes as the number of stitches per unit length times the strength of the thread in each.

That is a straight line through the origin, and both of its inputs are stated rather than computed: a sewing thread’s break load is a measurement, and how many of its strands cross the seam depends on the stitch type. A lockstitch puts two threads through every hole — a needle thread and a bobbin thread interlocking in the middle of the cloth — where a chainstitch puts one thread looping through itself. The arithmetic here uses one stated figure per stitch and lets the reader move it, because the point being made is about the shape rather than about the number.

Nothing in that line is a fabric property. A seam that is thread-limited is a seam whose strength has nothing to do with the cloth, which is why the interesting half is the other one.

The falling line: what the needle does on the way through #

The second limit is the fabric, and this is where the geometry enters.

A needle coming down between two warp ends does not necessarily damage anything. It needs room, and the room available is the clear gap between the threads — p − d, the same length a filter cloth is specified by — plus whatever it can borrow by crowding its neighbours. Threads shift: a needle can push two ends apart and take some of the neighbouring gaps as well.

So the model has a regime boundary rather than a fitted curve. If the needle is narrower than the room it can find, it displaces threads and severs nothing, and the fabric line is flat. If it is wider, it has to go through a thread, and the fabric loses that thread’s share of the load.

How much room can be borrowed is a stated number — three neighbouring gaps’ worth, by default — and everything below is quoted at it. What is not fitted is the shape: there is a sett at which the needle starts to bite, it is where the room runs out, and it is

sett = 10 ÷ (d + needle ÷ (1 + borrow))

which for a 0.2 mm yarn and a 0.8 mm needle is twenty-five threads per centimetre. Below that the cloth is undamaged; above it every strike costs something. The closed form is checked against a search over four thousand setts and they agree to within a per cent.

The optimum, and the two regimes #

Where the two lines cross is the optimum, and it is arithmetic:

n* = F ÷ (T + F·P/sett)

with F the fabric’s strength, T the thread’s, and P the chance a strike severs. For the cloth in the hero figure — thirty threads per centimetre, a 0.8 mm needle, and a needle that severs four strikes in five — the crossing is at fifteen stitches per centimetre — and since a machine is set in whole stitches, the strongest row on the ladder is sixteen, where the seam carries 172 newtons against the fabric’s own 300.

The surprise: a closer sett does not always want fewer stitches #

The obvious rule — the closer the cloth, the fewer stitches it wants — is false, and finding that out was the model earning its keep.

The damage per stitch is P/sett: the chance of severing a thread, divided by how many threads there are to share the load. Both halves move with the sett. P rises as the gap closes, which pushes the optimum down. But once every strike severs a thread — P saturated at one — the denominator keeps rising, so each severed thread is a smaller share of the total and the optimum edges back up.

So the column of optima is not monotone: twenty-four stitches per centimetre at an open sett (the machine’s limit), eighteen at twenty-eight threads per centimetre, fourteen at thirty-two and thirty-six, sixteen at forty. The assertion originally written against this arithmetic said “a closer sett wants no more stitches than an open one”, and it failed — correctly, and for a reason that was in the model all along.

What is asserted instead is the regime statement, which is true and is the useful half: a cloth the needle damages has an optimum below the machine’s limit, and a cloth it does not is sewn as densely as the machine will go. That is the pattern this site has now met three times — an assertion written against the numbers in front of it rather than against its claim — and the third time is what makes it a rule of the house rather than an anecdote.

The same gap, a fourth time #

The clear gap between two threads is doing its fourth job on this site, and it is worth collecting them.

It is the filter’s pore, where it must be small enough to hold a soil back — and where the hole a rating quotes is not the hole a grain finds. It is the resin channel of a reinforcement, where it must be large enough to carry a flow. It is the slack a tear pulls out of a weave before its threads take the load. And it is the needle’s room.

One expression, four inequalities, four trades that do not read each other’s standards. The reason it recurs is not coincidence: a woven fabric has exactly two lengths in it — a spacing and a diameter — and every question about what can get through it or between it is a question about their difference.

The two failures pull the design in opposite directions #

Putting this rung beside the previous one gives the specification a shape, and it is a familiar one.

Slippage wants a closely set cloth and a generous seam allowance: more crossings, more grip, threads that break rather than slide.

Needle damage wants an openly set cloth, or a finer needle, or fewer stitches: room for the needle to pass, no threads severed along the stitch line.

Both are properties of the same sett, moving in opposite directions, so a seam is another interval — and, as in every other case in this field, the interval can be empty. A very closely set cloth sewn with an ordinary needle is a cloth whose seam is damaged wherever it is strong, and the answer is not a stitch density: it is a finer needle, a ball-point needle that displaces rather than pierces, or a different fabric.

The needle is a purchase decision, and the arithmetic says which one #

The threshold above is a ratio between a needle’s width and a yarn’s diameter, so the lever a sewing room actually pulls is the needle.

Needle sizes are quoted in the metric system as hundredths of a millimetre of blade diameter: a size 70 is 0.7 mm, a size 90 is 0.9, a size 110 is 1.1. Put those through the closed form at a 0.2 mm yarn and the biting setts are 29, 23 and 19 threads per centimetre — so changing from a 110 to a 70 moves the onset of damage by ten threads per centimetre, which is the difference between a fabric that sews cleanly and one that does not.

The second lever is the needle’s point rather than its diameter, and it is the same arithmetic with the borrowing changed. A ball point is rounded so that it pushes threads aside instead of piercing them, which is exactly a larger borrow: more room found, damage postponed. A sharp or a cutting point does the opposite deliberately, for leather and coated fabrics where displacement is not available.

Both of those are in the model as stated parameters rather than as mechanisms, which is honest about what geometry can supply. What it can supply is the ordering and the threshold’s shape: the onset sett goes as 1/(d + needle/(1 + borrow)), so a finer needle and a rounder point both push it up, and a coarser yarn pushes it down.

The sett past which no needle is fine enough #

The closed form is written above as a sett at which a given needle begins to bite. Inverted, it answers the question a sewing room actually asks — which needle will this cloth take? — and inverting it turns up a limit the table of needle sizes does not.

Solving for the needle’s width gives

needle ≤ (1 + borrow) × (10 ÷ sett − d),

with the sett in threads per centimetre and everything else in millimetres. For a 0.2 mm yarn at the default borrowing that is a straightforward table:

The last two rows have no entry, and that is the finding. Sewing needles are not made below about 0.6 mm — a size 60 is the finest in ordinary industrial use — so past

sett = 10 ÷ (d + 0.6/(1 + borrow)), which is 28.6 threads per centimetre in a 0.2 mm yarn,

there is no needle fine enough to pass without severing something. The cloth is on the falling side of the curve whatever the sewing room does with its needle rack.

Three consequences, and the second is the one that explains a practice.

The threshold moves with the yarn and not with the fabric’s weight. It is a sett against a diameter, so a closely set cloth of coarse yarn can be easier to sew than an openly set one of fine yarn — which is the opposite of the trade’s proxy, where a needle is chosen by fabric weight. The proxy works because weight correlates with both, and it fails exactly where the two come apart: a fine, densely set shirting is light and hard to sew.

Past the threshold the answer is a ball point rather than a finer needle. A rounded point displaces threads instead of piercing them, which in this arithmetic is a larger borrow — and borrowing is the term the threshold is most sensitive to, since it divides the needle’s width. Going from three gaps’ worth of borrowing to five moves the limit from 28.6 to 32.3 threads per centimetre. That is why a ball point is specified for the cloths a finer needle cannot save, and it is a different lever rather than more of the same one.

And past that, the answer is the cloth. A construction beyond what any needle and any point will take is a construction whose seams are perforated by definition, and the remedies left are outside the sewing room: a bonded seam, a welded one, or a different sett. That is the same shape of conclusion the applied field keeps reaching — an interval of two inequalities, which can be empty — and here the empty case is reached at a sett a shirting mill would not think unusual.

What was counted, and how #

Two of the three inputs to the falling line are stated and one is computed. The fabric’s strip strength and the sewing thread’s break load are measurements; the chance that a strike severs a thread is computed from the clear gap, the needle’s width and the room it can borrow.

The optimum is found by evaluating both lines over a ladder of densities and taking the larger of the two minima — and then asserted to sit within one ladder step of the closed-form crossing, which is the check that the arithmetic and the search are describing the same figure. The regime statement is asserted at every sett in the range, and the range is asserted to span both regimes, because a threshold whose two sides are not both exercised is a threshold nothing has tested.

The biting sett is computed twice: once in closed form and once by walking a fine ladder of setts and finding the first at which the sever chance leaves zero. They agree to within a per cent, which is the ladder’s own step.

What the picture cannot show #

The trade-off figure draws two straight lines and their crossing, and there are two things it cannot draw.

It cannot show the scatter. Both limits are statistical — threads vary, strikes vary, and a seam fails at its weakest centimetre rather than at its average one — so the real relationship is a cloud with a rounded top and the figure is its skeleton. A reader who takes the crossing as a sharp optimum is taking more from the picture than the model supplies.

And it cannot show the needle. Whether a strike severs a thread happens inside the cloth, at a scale the figure does not draw, and the whole of it enters as one number on the falling line’s slope. That is why the gap in section is drawn separately: the mechanism and the consequence need two pictures, because no single drawing holds both a needle between two threads and a seam strength in newtons.

Where the model stops #

Two straight lines are a caricature. Real seam strength against stitch density is a curve with a rounded maximum, because both limits are statistical: threads vary in strength, needle strikes vary in what they hit, and the seam fails at its weakest point rather than at its average. The crossing of two lines is the right shape and the wrong smoothness.

Needle damage is not only severance. A needle can cut a filament, abrade it, melt it in a synthetic at high sewing speeds, or simply displace a thread permanently and leave a visible line. Each is a different loss and only the first is counted here — and the melting one is why sewing speed appears in garment specifications at all.

Nothing here knows about the stitch’s geometry. A lockstitch’s interlacing sits inside the cloth’s thickness and clamps it; a chainstitch’s loops sit on the underside and can be unravelled; an overlock wraps the edge entirely. Those change both lines, and the arithmetic treats a stitch as a strand across a line.

And the fabric’s own strength is a measurement standing in for a distribution. The step from “a thread was severed” to “the seam lost that fraction of its strength” assumes the load redistributes evenly among the survivors, which is the same assumption the net-section arithmetic of the next rung makes explicit and defends.

There is one more asymmetry between the two limits, and it decides which way a sewing room errs. The thread-limited line is a design quantity: a machine can be set to any density within its range, and a supervisor can change it in seconds. The fabric line is a purchase: the sett arrived with the cloth. So the practical form of the arithmetic is not “what is the optimum” but “given this cloth and this needle, is the optimum inside the machine’s range” — and when it is not, the fabric is being sewn on the falling side of a curve nobody has drawn.

Who found it, and when #

Stitch density as a specified quantity is as old as industrial sewing, and the optimum is folklore in every sewing room: too few stitches and the seam pulls open, too many and the cloth has been perforated. Needle damage has been a measured property since at least the mid twentieth century, with standard methods counting severed threads along a stitch line.

What is not usually written down is that the threshold is geometric. The trade’s rule is expressed as a needle size for a fabric weight, which is a sensible proxy and hides the mechanism; the arithmetic above says the mechanism is the clear gap against the needle’s width, and produces a sett at which the damage begins. It is the same substitution this field keeps making: a table indexed by a proxy, replaced by an inequality on the two lengths a fabric actually has.

Where the ladder goes next #

A seam is one kind of interruption in a cloth. A hole is another, and a cloth’s response to one is nothing like a film’s: the threads carry their own load and hand almost nothing to their neighbours, so a hole costs exactly the threads it removes. The next rung computes what that means for a buttonhole, an eyelet and a nick in the selvedge.