The Slow Sand Filter

Two numbers define a filter sand and both are worth measuring rather than trusting. The effective size is the sieve opening that passes the finest ten percent of the sand. For slow sand filtration it should be in the region of a third of a millimetre — fine enough to hold a mat on the surface rather than letting the dirt bury itself in the bed, coarse enough that the bed does not clog in a week. The uniformity coefficient is the ratio between the opening that passes sixty percent and the one that passes ten. A low number means the grains are all much the same size, which is what you want: a wide spread lets the fines migrate down and pack into the gaps, and the bed loses permeability from the inside where you cannot scrape it. Sieve your sand through a nested stack, weigh what is retained on each, and plot the cumulative curve. Read the two sizes off it. If your uniformity coefficient is above about three, wash the fines out and re-measure — this is the step people skip and it is the reason their filter clogs. Wash the sand until the rinse water runs clear regardless. Silt left in the bed at the start behaves exactly like a month of operation you never got the benefit of.

Use a barrel or a tall box, and build it in layers from the bottom: an underdrain of slotted pipe, then coarse gravel, then progressively finer gravel, then at least six hundred millimetres of your graded sand. The gravel grading matters for the same reason as the sand grading — each layer must be fine enough to stop the layer above falling into it. Above the sand leave three hundred millimetres or more of clear space for supernatant water. That standing head is the driving pressure and it is also the buffer that keeps the bed submerged when the inflow stutters. Now the outlet, which is where most home-built filters go wrong. Run the outlet pipe up outside the filter and turn it over at a level above the top of the sand. This means the bed physically cannot be drained below that level, no matter what happens downstream — someone opening a tap, a pump running dry, a leak. An exposed bed dries, the mat dies, and you are three weeks from clean water again without any warning that it has happened. That outlet is a hydraulic interlock made of nothing but pipe routing. It is the same idea as the venous reservoir in a bypass circuit and the vent stack in the last blueprint: arrange the geometry so the dangerous state is impossible rather than merely forbidden.

Fill the filter from the bottom up. Introducing water downward through dry sand traps air in the bed, and trapped air is permanently lost permeability — it does not wash out. Fill slowly through the underdrain until water stands above the sand, then start the feed. Now run it and leave it alone. Feed it real surface water — pond, stream, water butt — because the organisms that build the mat arrive with the water and sterile tap water will never ripen a filter. Measure every day, and measure the right thing. Turbidity is easy and useful: shine a torch through a tall jar of inlet and of outlet and compare, or build the simple photodiode meter and read it as a number. Head loss on the manometer tells you about fouling. And if you can, run a coliform test on inlet and outlet weekly, from a kit, following its instructions exactly. What you will see is not gradual. For the first week or so the outlet is barely better than the inlet — the sand is straining out the coarse material and nothing more. Then over a few days the removal jumps, and keeps improving for another week or two before levelling off. That step is the mat becoming established, and being able to point at the day it happened is the reason to measure daily rather than weekly. Record the whole ripening curve. It is the evidence that the biology, not the sand, is doing the work.

When the head loss reaches your available head, drain the supernatant down to just above the sand and skim the top fifteen to twenty millimetres off with a flat trowel. Take the mat and as little sand as you can. Wash the scrapings if you want the sand back, and keep them separate — that is contaminated material. Refill from the bottom again and restart. Note that recovery is faster than the original ripening, because the organisms are still present deeper in the bed, but it is not instant. Measure the second ripening curve and compare it to the first. After perhaps twenty scrapes the bed has lost enough depth that you must re-sand it, and the correct method is to remove the remaining sand, lay new sand underneath, and return the old ripened sand to the top. Putting new sand on top would throw away everything living in the bed. Where this stops, and read this part twice. A well-run slow sand filter removes a great deal — turbidity, most bacteria, many protozoa including the cyst stages that resist chlorine. It does not remove dissolved chemicals, it does not reliably remove viruses, and its performance depends entirely on a living mat whose condition you cannot see. Every water authority in the world treats filtration as one barrier among several and follows it with disinfection, because a filter that has been disturbed, dried, short-circuited or recently scraped can pass what it normally removes without any visible change in the water. Do not drink the output of this filter on the strength of it looking clear. Clear is not the same as safe, and that sentence is the single most important one in this batch. Build it to understand the process, run it for irrigation or for washing, and if you intend to drink from it, follow it with the chlorination blueprint and verify the residual every single time.