The Cartridge Blocked in Four Days. The Bed in Front of It Had Been Channelling Since About Month Six.

A water plant I was asked to look at was going through its five-micron cartridge every four or five days. The owner had made his peace with it. He bought them by the box, the changeover took a few minutes, the water coming off the machine tested fine, and in his accounting this was a consumable rather than a fault. He wanted to talk about the membrane. I wanted to talk about the vessel two steps before it, because the cartridge was already telling us what was wrong with that vessel and nobody had read the message.

We have come at this plant from other angles before, and I want to be clear what is new here. The note on refilling station pre-treatment works through the feed-flow arithmetic and says in passing that a cartridge brown and tight in four days is telling you the filter upstream has channelled. It stops there. The note on the iodine number is about the carbon, which fails in a different way and for different reasons. The tertiary filtration note deals with a filter polishing sewage rather than one protecting a membrane. This one is about the granular bed itself: what actually happens inside it, roughly how long it takes, and the one chemical trap in it that I did not know about for years.

A sand bed does not wear out. It rearranges itself

The picture most owners carry is that a filter bed gradually clogs and that backwashing unclogs it. That picture is close enough to be useful and wrong in the way that matters. A granular bed is not a sieve with a fixed set of holes. It is a loose pile of particles, and the water finds its way between them by whatever route offers least resistance. As the bed ages, that route stops being evenly spread across the whole surface.

The description our supplier uses for it is biodynamic instability leading to transient wormhole channelling, and the plain version is this: biological growth and fine deposits make parts of the bed stickier and tighter than the rest, the water abandons those parts, and it concentrates into a smaller number of preferred paths. Those paths are not stable — they open and close — which is why the word transient is in there and why the fault is so hard to catch in the act. Water going down a channel is not being filtered. It is passing through a vessel full of media without ever meeting much of it.

So the bed has not lost capacity in the sense of being full. It has lost contact. That distinction keeps turning up on the wastewater side too: a fixed-film media bed is judged on surface actually in contact with water rather than on the volume of plastic in the tank, and a clarifier carrying over is very often a short-circuit rather than an overload. The machine looks right from outside in all three cases.

About six months, and what to do with a number like that

The figure our supplier publishes is that this takes around six months before it begins to affect performance. I want to label that properly, because this blog is not worth much if I blur the line: that is the manufacturer's published figure, not a measurement made by us on a Philippine plant. We have not run a controlled comparison here, and I would rather say so than let a brochure figure harden into a fact.

What makes it worth repeating anyway is that six months is a usable order of magnitude, and most of this industry talks about the problem with no timescale attached at all. It is not six weeks and it is not six years. It is long enough that the plant runs beautifully through its acceptance period and through the first stretch when the owner is still paying attention, and short enough that the thing is quietly degraded well inside the first year. Fine at handover, wrong by the time anybody looks again — that is the same shape we ran into on a sewage plant three weeks old, and the same reason we argue about where the commissioning period actually ends.

Treat it the way you would treat any supplier's figure: as a planning number that tells you what kind of thing you are dealing with, not as a date in a calendar. Our own habit on money is to use rates we have proven on jobs we actually built rather than rates from a brochure, and the honest position here is that the mechanism is well established and the clock on it is somebody else's.

Channelling does not show on the gauge

This is the part that makes it expensive. If a bed blinds evenly, differential pressure across the vessel rises, and a gauge at the inlet and the outlet will tell you so. If a bed channels, the water has found an easier path, so the differential does not rise. It can fall. The vessel passes its rated flow at a pressure drop that looks entirely healthy while doing a fraction of the filtration it was bought to do.

So the instrument on the vessel reports the opposite of the fault. That note on the iodine number describes a carbon bed failing the same silent way, with no gauge movement and no change in appearance, and it is the recurring shape on this blog: a figure that is real, printed and authoritative and is measuring something else. A seat count read as a population. A trap's flow rating read as a size. A clause quoting a regulation that is no longer in force.

The cartridge is the instrument you actually have. It sits downstream, it costs almost nothing, and it integrates everything that got past the bed. A cartridge that lasts weeks says the bed is working. A cartridge loaded in days says it is not, and the discipline that makes that useful is to write the changeover date down rather than carry it in your head. Four changes in a month against one is not a judgement call once it is on paper. Nobody needs a laboratory to see it, which is the same argument we make for reading a settled cone instead of chasing a solids figure.

Backwash is what is meant to prevent this, and it is also what can set it up

Backwashing is supposed to fluidise the bed, lift it, let the fines wash out and let the grains resettle by size. Done at the right rate for the right duration, it restores the grading and breaks up the beginnings of a channel. Done at too low a rate it does something worse than nothing: it disturbs the bed enough to move material around but not enough to fluidise and re-grade it, and the bed resettles mixed.

Too low a rate is the common case, and it is usually a drawing decision made long before any operator arrives. That tertiary filtration note works through the arithmetic of it on the sewage side — backwash is quoted per square metre of bed, it is a large multiple of the service flow, and it sizes the pump, the pipework, the valves and the drain. If any one of those was sized for the service flow instead, the bed cannot be washed properly and no amount of attention will fix it. That is a design problem presented as a maintenance problem, which is the whole argument of why design decides the cost, of what a good operator can and cannot carry, and of why a plant that only works for one particular man is a design finding rather than a staffing one.

There is also media loss. Every backwash carries a little of the bed out with the fines, and over years the bed gets shallower, which shortens the path the water takes through it. We keep a loading table by vessel size for exactly this reason, and a note on retrofitting existing gravity filters, where bed height and underdrains are usually what has gone wrong. A bed that has lost depth is not proportionally worse. It channels sooner.

The trap I did not know about for years: sand contains free silica

This is the part of the subject I would most want an owner to read, because it is not intuitive and it is not visible anywhere in the plant.

Silica sand contains free silica, and a bed of it in moving water gives some up. On most plants that is of no consequence whatsoever. On a plant with a membrane behind it, it matters, and our supplier's position — again, their published claim and not our measurement — is that free silica blocks membranes, and that it does so particularly where aluminium is present in the water, or where aluminium is used as the flocculant.

Read that second clause slowly. An aluminium-based coagulant is the standard, cheap answer to a filter that is not holding fine solids. It is dosed ahead of the bed precisely to improve what the bed does. So on a plant with a membrane, the chemical added to help the filter can combine with what the filter medium itself releases and scale the very thing the filter exists to protect. The remedy and the damage arrive down the same pipe.

I am not claiming this is the usual cause of a fouled membrane. It is not. Most fouling is biological, or it is plain hardness on a plant whose softener cannot deliver its capacity at the flow it is being asked for. But it belongs on the list of things to rule out, and almost nobody has it on their list. It also lands on a point that tertiary filtration note already made: a coagulant is never only a solids chemical. It takes phosphate down with it as a side effect, it consumes alkalinity on a plant already working near its buffer limit, and here it may be feeding a scale. Three effects, one dosing pump, one line in the quotation.

It is also part of why we are uncomfortable with any design where compliance or performance leans on a chemical arriving every month. That is a standing position of ours rather than a reaction to this one mechanism, and it is the reason we would rather pay once for a medium than forever for a drum.

Finer is not the same as cleaner

The usual answer to a filtration problem is to buy a finer filter, and ultrafiltration is where that instinct ends up. A membrane down to around three hundredths of a micron sounds like the end of the argument.

It is not, and the reason is worth holding onto: ultrafiltration is a purely mechanical barrier. It removes particles by size and removes essentially nothing that is dissolved. The dissolved organic matter — the food — passes straight through and arrives at whatever is downstream, which means a UF membrane biofouls like anything else, and so does the reverse osmosis behind it. Buying a finer pore does not change the diet of the biology growing on the other side of it. The reverse osmosis note makes the same point from the membrane's end.

We make the identical argument about sewage on the other side of this blog. A tertiary filter removes suspended material and does almost nothing for what is dissolved, which is why ammonia, nitrate and phosphate pass through sand, glass and disc filters alike, and why a nutrient failure is never a filter problem. Our nutrient field guide sets that out at length, and the parameters that matter by industry are on the site as well. Different industry, same misunderstanding: people reach for a mechanical answer to a chemical question, because the mechanical answer is the one you can buy.

What we specify instead, and what I can and cannot prove

We use a glass-based activated medium in place of sand and gravel, and I will give its figures with the same label as before: these are the manufacturer's published figures together with an independent test house's verification, not measurements from our own plants. Particle removal independently verified at about 97 per cent down to five microns on the coarser grade and about 99.7 per cent down to one micron on the finer. With coagulation, a nominal filtration better than a tenth of a micron. A surface area stated as up to three hundred times that of sand. And a silt density index on the treated water put at somewhere between a half and a quarter of what the same duty gives on sand.

That last one is the number that would settle the argument if anybody measured it. Silt density index is a field test rather than a laboratory one: you pass the water through a small membrane disc at a fixed pressure and time how fast it plugs. It is the standard way of asking whether water is fit to put in front of a membrane, it needs no chemistry, and in all the water plants I have been shown I can count on one hand the owners who have ever had one run. Membrane manufacturers commonly state a maximum. Almost nobody checks it again after commissioning — which is exactly the pattern in asking for an influent result and finding that almost nobody has one.

Where the decision breaks, honestly. If the plant has no membrane, no aluminium dosing and an attentive operator who backwashes properly, a well-built sand and anthracite bed is not bad engineering and I will not pretend it is. If there is a membrane behind it, the sums move, because the thing being protected is the expensive item and the failure is slow and invisible. That is the same way we argued the fixed-film media question: I can tell you which way the decision breaks, I cannot hand you side-by-side Philippine trials, and I would rather say so than dress a preference up as proof. If you want to pull any of it apart, run it past an AI or a second engineer — we mean that, and we say it on every design we issue.

Six questions before the filter is bought

In the order you would actually ask them.

  1. What is the real feed flow this vessel has to pass, as against the product flow printed on the system? On a membrane plant those are very different numbers, and that refilling-station note works the arithmetic through in full.
  2. What is the bed area, and what backwash rate does that area need? Multiply it out before anybody sizes the backwash pump, the pipework, the valves or the drain.
  3. Can the drain actually take that flow, and where does it go? Backwash is a concentrated return, and like everything else you take out of a plant it has to end up somewhere.
  4. Is a coagulant in the design, and if so which one? If it is aluminium-based and there is a membrane downstream, ask the supplier about silica directly.
  5. Who logs the cartridge changes, and where? That log is your only instrument on the bed in front of it.
  6. What is the media replacement interval, and is it priced in the quotation? A washed medium and a replaced medium are two different maintenance obligations, which is the kind of thing we look for when we read two quotations line by line and the reason we put three options side by side rather than one.

None of those six needs a laboratory and most of them are answered on paper, before anybody spends anything. That is deliberate — it is the same test we apply when somebody asks us what to ask before signing.

What getting it wrong costs, stated plainly

No alarm in this, just the arithmetic. A channelled bed does not fail. It passes unfiltered water to a cartridge, which loads faster and gets changed more often, and that is a small recurring cost and the one everybody sees. What it also does is pass fines, and on a membrane plant possibly a scale-former, to the element behind it, which shortens the life of the most expensive consumable in the plant. The bill arrives months later, in a different line of the accounts, on a plant whose owner believes the pre-treatment is fine because the water tastes right and nothing has broken. It is the same reason we treat how a thing fails after the sale as part of specifying it.

Against that, re-bedding a vessel is a day's work at a known price. We would rather have the conversation at the drawing stage, which is why our price estimator and design generator both put the filtration stage in front of people early, and why we publish operating cost per cubic metre rather than a capital figure standing on its own. The cheapest plant to buy and the cheapest plant to own are routinely not the same plant, which is the whole of the reuse arithmetic as well.

Where this does not transfer

Everything above comes off one class of plant and one supplier's published data, so take it as a way of thinking rather than as a specification.

If your feed is a clean deep well with low organics, biological growth in the bed is slow and the six-month figure is pessimistic — the mechanism needs something to grow on. If your feed is surface water, or storage that goes warm and stagnant, it will be optimistic. A plant running a few hours a day is a different case again, because a bed that sits wet and still between runs is a better reactor than one in continuous service, and small plants are harder to run than large ones for closely related reasons. The silica point only bites where there is a membrane and aluminium in the picture; remove either and it falls away entirely. And every figure in the paragraph about media performance is somebody else's measurement in a generic duty, not ours on your water.

The one part I would carry anywhere is the first. A filter bed can be full of media, pass its rated flow, show a healthy pressure drop and be doing very little, and the cheapest item downstream of it is the only honest witness you have. We keep notes on tertiary filtration, on refilling station work, on the full process flow and on the oxidation window metals need before any filter can hold them, and the figures on those pages are the supplier's. We keep a separate note on side-stream filtration on cooling towers, where the same bed problem shows up as a chemical bill. The package comparison and the media's published specification are both on the site if you want the underlying numbers.

If you are reading this because a cartridge in your own plant is loading faster than it used to, that is a real finding and worth acting on before the membrane tells you the same thing more expensively. We do the same work on the wastewater side across Cebu and the rest of the country, from new plants to maintenance on plants somebody else built; the completed projects page shows the range, and our consultancy work is set out separately. If you want somebody to look at your filtration train with you, book a free 15-minute consultation.