A woven filter beats its own rating

A filter is rated by the hole it does not show established the arithmetic of a woven filter’s rating. The retention is decided by the largest channel through the cloth rather than by the average one, the largest channel is a property of the distribution rather than of the construction, and a cloth whose mean opening is comfortably below a particle size can still pass it through the widest hole it happens to have.

That is a statement about geometric sieving and it is complete. A particle smaller than every channel passes through every channel, and the cloth stops none of it.

The cloth #

Everybody who has run a woven filter knows that it does not behave as the sieving arithmetic says. A new cloth passes fines for a few minutes and then stops passing them, and the fines it stops are far smaller than any hole it has. The explanation is the cake: the particles that have been retained build a layer on the surface, and the layer is a much finer filter than the cloth.

That explanation is right and it has a hole in it. A cake has to start from something. If the cloth retains nothing at all in its first moments, no cake forms and the cloth passes fines forever.

The claim #

A woven filter cloth’s hair layer intercepts a small fraction of particles far below its rated opening — a fraction of a per cent to a few per cent for a bare cloth. That is not filtration and it is exactly what a cake needs: a small retention seeds a layer, and the layer does the rest. Raising the same cloth takes the ten-micrometre capture from a fraction of a per cent to a fifth, which is why a napped filter cloth exists as a product.

The mechanism is interception, and — unlike the fourth-power channel arithmetic that decides a cloth’s flow — this essay takes it as a purely geometric statement: a particle whose path passes within its own radius of a fibre touches it. No flow model, so every number is a lower bound.

Interception, taken as geometry and nothing more #

The single-fibre efficiency for interception is the ratio of the particle’s diameter to the fibre’s, and the standard log-penetration form for a bed of fibres gives the total:

E = 1 − exp( −4 α t η / (π d_f) )

with α the bed’s solid fraction, t its depth, and η the single-fibre efficiency.

The boundary matters here more than anywhere in this ladder. Whether a streamline bends around a fibre — and therefore whether a particle following it is carried past — is a flow question and belongs to somebody else’s subject. So η is taken as the bare geometric ratio, which is a lower bound at every particle size, and the essay says so at every use.

What this collection can supply is α and t, because both come from the hair population: the solid fraction is the total protruding fibre length per unit area times a fibre’s cross-section, divided by the canopy depth.

The numbers, which are small and are supposed to be #

For a bare filter cloth, the canopy is not closed and the layer is very dilute — of the order of seventy parts per million solid over a depth of a millimetre.

The captures come out at a fraction of a per cent for a one-micrometre particle and a few per cent for fifty. Those are useless as filtration and are not zero, which is the whole of the argument.

A cake needs a seeding rate rather than an efficiency. A stream at any realistic solids loading, passing through a cloth at a few per cent capture, deposits a monolayer on the surface in seconds — and the moment a monolayer exists the interstices between the deposited particles are the filter, and they are a hundred times finer than anything the cloth has.

So the cloth’s job in the first seconds is not to filter but to be slightly sticky, and a few per cent is ample.

Why a napped filter cloth is a real product #

Raising a filter cloth multiplies the population, so α rises in proportion and the canopy closes and acquires a depth. Both effects raise the capture.

At thirty-twofold the ten-micrometre capture goes from under a per cent to about a fifth, and the fifty-micrometre capture to nearly three quarters. That is filtration, and it happens on a cloth whose largest opening is unchanged.

Napped and needled filter media — and raising is only available to a cloth with float — are used exactly where a cake cannot be relied on — intermittent duty, low solids loading, or a duty where the cake is blown off between cycles. In those cases the cloth has to do the work itself, and the only way a woven cloth can is with a layer that is not part of the weave.

The rating is then a lie in the useful direction. A napped filter cloth’s stated opening is still its largest channel, and its actual retention is far finer, and the specification has no way to say so.

The two jobs, and which one this is #

A filter cloth has two jobs — to retain and to release — and they pull in opposite directions. A cloth that retains finely holds its cake and blinds; a cloth that releases cleanly passes fines.

The hair layer is squarely on the retention side and it is on the wrong side of the release question. A napped cloth blinds, because particles caught in a canopy are caught in a three-dimensional structure and cannot be dislodged by a pressure pulse the way a surface cake can.

So the choice between a bare and a napped filter cloth is the retain–release trade-off with the hair layer as its dial, and the arithmetic here prices only one side of it.

Why the air permeability does not move #

The pairing with permeability is the same one the opacity essay makes and it is worth restating because it is what makes the hair layer invisible to every specification a filter cloth carries.

A flow resistance is a series with a bottleneck and half the air goes through a tenth of the holes: the resistance is set almost entirely by the narrowest place in the widest channel, because the dependence is a fourth power. A layer that is seventy parts per million solid narrows nothing.

A capture is an integral along a path and every fibre in the way counts.

So a filter cloth’s permeability specification, its rated opening and its measured air flow can all be identical across three fabrics with a factor of thirty between their retentions. Every number on the datasheet is blind to the thing doing the filtering.

What the same layer does to the release side #

There is a second and less welcome consequence, and it belongs to the cloth’s other job.

A cake is released by reversing the flow or by flexing the cloth, and it comes away cleanly when it is a coherent layer sitting on a smooth surface. A hair layer is neither smooth nor a surface: it is a three-dimensional structure that the first particles have grown into, so the base of the cake is interlocked with the cloth rather than resting on it.

So the same interception that seeds the cake anchors it. A cloth that starts filtering quickly is a cloth that cleans badly, and the two are not independent choices but one choice seen twice.

That is the retain–release trade-off in its sharpest form — the same exchange a filter cloth’s two jobs sets out, and it says something the datasheet cannot: two cloths with identical openings, identical permeability and identical construction will differ in both halves of their performance according to a surface property nothing on the specification records.

What was counted, and how #

Three assertions and one of them is the essay’s central and deliberately modest claim.

That a coarser particle is caught more readily than a finer one, over the whole size range — which is what interception says and which would be broken by a sign error anywhere in the bed arithmetic.

That a particle far below the cloth’s rating is still caught by some of them, strictly greater than zero. That is the seeding claim and it is stated as an inequality rather than as a value, because the value is a lower bound with a flow model missing from it.

And that a napped cloth catches more than a bare one at every size.

Nothing is fitted to a filtration measurement. The bed comes from the hair population and the efficiency is a geometric ratio.

How long the seeding takes #

The claim is that a few per cent is enough to start a cake, and “enough” is worth converting into a time, because the time is what decides whether a duty needs a napped cloth or a bare one.

A monolayer of particles of diameter a at an ordinary packing weighs about 0.6·ρ·a per unit area — for a ten-micrometre mineral at two thousand kilograms a cubic metre, twelve grams a square metre. The rate at which the cloth collects is the capture times the solids concentration times the flux through the cloth. Divide:

seeding time = 0.6 ρ a ÷ (E c v).

At one per cent solids in water, a flux of a millimetre a second and the bare cloth’s one per cent capture, that is about two minutes — which is what a filtration engineer means by a cloth conditioning, arrived at from a hair population and an interception ratio with nothing fitted to a filtration measurement anywhere.

The scaling is where the argument earns its keep, because every term is a duty parameter.

The last row is the case the essay names and could not price. A dilute stream through a bare woven cloth spends hours passing fines before it has collected enough to filter — and if the duty is intermittent, or the cake is blown off between cycles, it never gets there at all. Three and a half hours of unconditioned running is not a slow start; it is a filter that does not work.

And the napped cloth’s six seconds is why the product exists. Raising multiplies the capture by twenty, which divides the seeding time by twenty, and a cloth that conditions in seconds conditions inside every cycle however the duty is run.

So the choice between the two media is not a choice about retention at all. Both cloths end up filtering through a cake; the napped one gets there quickly enough for the cake to exist, and the bare one is relying on a duty steady enough to let it. That is a much sharper statement of when to nap than “low solids loading”, and it is one line of arithmetic on the capture the previous section computed.

The caution is that every term but the capture is a duty figure with an order of magnitude of latitude, and the capture is a lower bound with no flow model in it. So the two minutes is an upper bound on a number known to a factor of ten, and what survives is the ratios between the rows — which is the part the decision turns on.

Where the model stops #

There is no flow model, which is the largest omission and is deliberate: whether a streamline carries a particle past a fibre is a question about flow around a cylinder and is not this site’s subject. Every efficiency here is therefore a lower bound, and at small particle sizes it is a very loose one, because diffusion — which this arithmetic has no term for at all — dominates below about a micrometre.

The bed is uniform and it is not. The canopy’s solid fraction falls exponentially with height, so the top of the bed is far more open than the bottom, and a log-penetration form with a mean fraction is an approximation.

Nothing accumulates. The whole point of the essay is that capture seeds a cake, and the model computes only the first instant. The cake’s own arithmetic is a different subject and this collection does not have it.

And the interception ratio exceeds one for the larger particles, where the geometric form stops meaning anything — a particle wider than the fibre is not intercepted, it is struck. Those points are drawn because the trend is the argument and they should be read as an upper end rather than as a calculation.

What a specification would have to carry #

The essay’s practical consequence is the same one this ladder keeps arriving at, and here it is unusually concrete because filter cloths are specified more tightly than most textiles.

A filter cloth datasheet carries a rated opening, an air permeability, a construction and a weight. None of the four responds to the hair layer, and the hair layer decides how fast the cloth conditions and how well it cleans.

What would carry it is a surface state — as woven, singed, calendered, raised or needled — which is in fact how filter media are sold, as a family of surface treatments over a common base cloth. So the trade has the vocabulary and treats it as a product range rather than as a measured property.

The measurement that would turn it into one is the same one this ladder keeps proposing: a fabric hair count rather than a yarn hair count, which is a scan of a specimen edge-on and gives the population directly. It would let a datasheet say what the surface treatment actually did, in the units the arithmetic above needs.

The generalisation #

A small efficiency at the start of a self-reinforcing process is worth far more than its size suggests.

The transferable shape is that whenever a process builds its own machinery — a cake that filters, a seed that nucleates, a first crack that concentrates stress — the quantity that matters is not the steady-state performance but whether the process can start at all. A retention of one per cent and a retention of zero are separated by an infinite factor in outcome and by one per cent in measurement.

The diagnostic is to ask whether the number being reported is a rate or a threshold. Filtration efficiency is reported as a rate; what a woven cloth needs from its own surface is a threshold, and the threshold is any.

The one case where the seed is not needed #

There is a duty in which the whole argument is irrelevant and it is worth naming, because it marks the boundary of the essay.

A cloth filtering a stream with essentially no solids in it — a polishing filter, a final guard — never builds a cake, because there is nothing to build one from. Such a cloth has to retain by itself, and a woven cloth cannot: its rating is its largest hole and its hair layer catches a few per cent.

That is why polishing duties use nonwovens and membranes rather than woven cloth, and the reason is structural rather than a matter of degree. A nonwoven is a depth filter in its own right — a mat of fibres many diameters deep, with the same interception arithmetic applied to a solid fraction three orders of magnitude higher — and it does not need a cake because it is one.

So this ladder’s contribution to filtration has a clear edge on it. The hair layer explains how a woven cloth gets started, and it explains nothing at all about the media that never needed to.

Who found it, and when #

That a woven filter’s real retention comes from its cake rather than from its cloth is standard practice and is why filtration engineers speak of a cloth’s “conditioning”. That napped and needled media retain more finely than their openings suggest is likewise standard.

What is added here is where the seeding comes from. A bare woven cloth’s own surface has no mechanism for retaining a particle below its rating, and the hair layer supplies one — computed from a fibre population rather than assumed, and small enough that it could only ever have been a seed.

Where the ladder goes next #

To the far end of the same continuum, where the protruding fibre is put there on purpose, at a chosen length, held by a chosen anchor. Hair, nap and pile are one construction sets this site’s three protruding surfaces beside one another and finds that what separates them is not what they are but how much of them was decided.