The Trickling Filter

Take a 200 litre food-grade barrel and cut the top out. The bed depth you want is at least a metre; deeper is better for treatment and worse for draught, and 1.8 m is the classic figure for a rock filter. The media is the part people get wrong. Use washed stone of a narrow size range, 40 to 60 mm. Two properties fight each other here and you are choosing a point between them. Small stone gives more surface area per cubic metre, which is where the biofilm lives, so it treats more. Small stone also gives smaller voids, and the voids are what the air moves through and what the sloughed film has to fall out through. Go too fine and the bed ponds: the surface seals with film, water sits on top, air stops moving, and the whole thing goes anaerobic and smells. A 50 mm stone gives roughly 45 to 55 percent void space and around 40 to 60 square metres of surface per cubic metre. Graded 25 mm stone would roughly double the surface and roughly halve the tolerance for error. Wash the stone until the water runs clear. Fines are what block a bed, and they arrive with the delivery, not with the sewage. Now the underdrain, which is the actual air inlet and is why this step is not just 'put stones in a barrel'. Build a false floor: a grid of 32 mm PVC pipe drilled with 8 mm holes on its underside, sitting on 100 mm spacer blocks so there is a continuous open plenum beneath the media. Bring at least two 50 mm vent pipes from that plenum out through the barrel wall and turn them up, open to atmosphere. These are not overflow pipes. They are how air gets in, and a filter whose underdrain is flooded or sealed cannot breathe and will not work no matter how well the top is dosed. Lay 150 mm of 75 mm stone directly on the false floor to carry the load, then fill with the 40 to 60 mm media. Leave 200 mm of freeboard above the media.

The feed must arrive as a flush and then stop. Continuous dribble finds one channel through the stone and polishes it forever while the rest of the bed stays dry; a pulse floods the whole surface, spreads sideways, and then drains so air can follow the water down. Intermittent dosing is not a convenience, it is the mechanism. The 1893 installations used fixed spray nozzles and, shortly after, tipping troughs and dosing siphons that filled slowly and emptied all at once. Rotary distributors — the four-arm sprinkler most people picture — came later and do the same job continuously by moving the pulse around the bed instead of turning it on and off. The bench version: feed from a small submersible pump on a repeat-cycle timer, two minutes on and eight minutes off, through a length of 32 mm pipe drilled with 2 mm holes and capped, laid across the media. Aim for a hydraulic loading in the region of 1 to 4 cubic metres per square metre of bed surface per day. On a 560 mm diameter barrel that is about 0.25 square metres, so 250 to 1000 litres per day, which is 10 to 40 litres an hour. Start at the bottom of that range. Feed a defined synthetic wastewater, not real sewage. Per 10 litres of tap water: 1.5 g glucose, 1.5 g peptone, 0.5 g ammonium chloride, 0.2 g potassium dihydrogen phosphate, 0.5 g sodium bicarbonate for alkalinity. That gives roughly 250 to 300 mg/l of biochemical oxygen demand, which is ordinary domestic strength, and it is safe to handle. Seed the bed once with a litre of settled effluent or ditch water to introduce the organisms; after that it colonises itself. Now wait. For the first week nothing visible happens. Somewhere in the second or third week the stones near the top go slippery and take on a brown-grey cast, and the outlet turns from cloudy to clear. That is the film. In cold weather it can take six weeks. You cannot dose your way past this and a bed that is starved or dried during ripening starts over.

The film grows. It has to — that is the process — and growth is the problem, because a film thicker than about two millimetres cannot get oxygen to its base. The inner layer goes anaerobic, loses its grip on the stone, and the whole patch lets go and washes out. This is sloughing, and it is normal and necessary. What you do not want is all of the bed sloughing at once, which shows up as a sudden slug of solids in the effluent and a filter that treats badly for days afterwards while it regrows. Measure it. Take a settled sample of the effluent daily and read its turbidity, and once a week measure suspended solids properly by filtering a known volume through a dried, pre-weighed glass-fibre filter and drying it at 105 degrees Celsius to constant mass. A ripened bed running well gives a low, boring baseline with small peaks. A bed about to pond gives a rising baseline. The control you have is recirculation: return some of the settled effluent to the head of the filter and mix it with the incoming feed. Three things happen at once. The total flow over the bed rises, and higher flow scours the film continuously so it sheds in small amounts rather than in sheets. The incoming feed is diluted, so the top of the bed sees a lower organic strength and grows a thinner film. And the bed never goes dry between doses at low flow. Recirculation ratios of 0.5:1 to 3:1 against the raw flow are ordinary. The cost is real: you are pumping the same water repeatedly, and the hydraulic loading rises while the organic loading does not, so the tank downstream sees more flow. Run yours at 1:1 for a fortnight and then at 2:1, and compare effluent solids and how often you have to rake the surface. That comparison is the whole engineering judgement of the device. One more failure to know by name: ponding. Water standing on the surface means the top voids are blocked, by film or by fines. It is fixed by raking or by resting the bed, and prevented by washing the stone properly at the start and not overloading the top.