DAF Jar and Float Testing: The Bench-to-Pilot Protocol That Fixes Your Design

Why bench and pilot testing decide whether a DAF works

A dissolved air flotation unit is only as good as three numbers: the chemical conditioning that forms strong, floatable flocs, the air-to-solids (A/S) ratio that supplies enough bubble surface to lift them, and the rise rate the resulting float-and-water system can tolerate. None of these can be read reliably from a textbook for a specific effluent. Two food factories on the same trading estate can need coagulant doses that differ threefold and A/S targets that differ by a factor of two. That variability is exactly why bench testing exists.

The protocol splits naturally into two stages that answer different questions. Jar testing is a coagulation-flocculation experiment: it optimises the chemistry — coagulant type, pH, dose, polymer and mixing energy — using conventional gang-stirrer paddles. Float (batch-flotation) testing then takes the best-conditioned water and floats it in a pressurised bench DAF cell, varying recycle and saturator pressure to measure how clarification responds to air. Skipping either stage is the single most common reason a full-scale DAF underperforms its guarantee. If you have not yet fixed your target loading rates, read our companion guide on how to size a DAF system alongside this protocol — testing supplies the inputs that sizing consumes.

The scientific reason bench work translates so well is that DAF performance is governed by bubble-particle collision and attachment physics that are largely scale-independent at the microscale. A 40 µm bubble attaching to a coagulated floc behaves the same in a 1 litre cell as in a 40 m² tank; what changes with scale is the hydraulics — short-circuiting, wall effects and float-blanket stability — which is why a pilot stage still matters for the tank, even when a bench cell has settled the chemistry and air demand.

Jar testing: getting the coagulation chemistry right first

Never run a float test on unconditioned water — you would simply be measuring the wrong variable. Jar testing comes first because the floc you present to the bubbles determines everything downstream. A conventional six-paddle gang stirrer reproduces the two mixing regimes of a full plant: a brief high-shear coagulation phase that destabilises colloids, and a gentle flocculation phase that grows aggregates without shearing them apart.

The destabilisation mechanism is rooted in colloid science. Most industrial suspended solids and emulsified oils carry a negative surface charge; the electrical double layer keeps particles apart. Hydrolysing metal coagulants (aluminium sulphate, polyaluminium chloride, ferric chloride) compress that double layer and, more importantly at typical doses, form positively charged hydroxide precipitates that adsorb and sweep the colloids out of suspension. This is why coagulant dose and pH are coupled: the metal-hydroxide speciation that drives sweep flocculation only forms in a specific pH window.

A defensible jar protocol runs a dose ladder at fixed pH, then repeats the best dose across a pH ladder. Typical mixing: rapid mix at a velocity gradient G of 300–1000 s^-1 for 30–120 s, then slow mix at G of 20–70 s^-1 for 10–20 min to grow flocs, before assessing floc size, settling/floating tendency and supernatant clarity. For DAF specifically you want small, strong, low-density flocs that bubbles can attach to and lift — not the large, dense flocs you would chase for sedimentation. Over-dosing polymer is a classic DAF error: it makes heavy flocs that sink faster than bubbles can raise them.

Record the residual turbidity (NTU) of the supernatant for every jar. The dose that minimises residual turbidity at the correct pH becomes the fixed chemistry you carry into the float test. For a fuller treatment of the underlying chemistry — coagulant selection, charge neutralisation versus sweep, and polymer bridging — see our coagulation and flocculation guide.

The bench DAF (batch flotation) apparatus

A bench flotation unit is a deliberately simple analogue of a full plant. Its purpose is not to look like a DAF — it is to reproduce the one event that matters: releasing a known mass of dissolved air into a known volume of conditioned water and watching what floats.

• Saturator (pressure vessel): a sealed vessel part-filled with clean water, pressurised with compressed air to 4–6 bar gauge and shaken or recirculated so the water approaches air saturation at that pressure.
• Flotation cell: a graduated 1–2 litre transparent column or square jar holding the conditioned sample, into which the pressurised water is injected through a needle valve.
• Needle valve / nozzle: the pressure let-down device. The sudden depressurisation nucleates a cloud of 30–70 µm microbubbles — the same bubble-size regime a full-scale DAF produces.
• Sampling port: a tap at a fixed depth (typically mid-column) to draw clarified sub-natant for turbidity measurement after a set flotation time.

The key discipline is that you know exactly how much air you injected. Because you meter a measured volume of water saturated at a measured pressure and temperature, the dissolved-air mass released is calculable from Henry's law — and that is what lets you compute a real A/S ratio for each test rather than guessing. The physics of that release is covered in depth in our note on the DAF air-to-solids ratio.

Running the A/S series: the heart of the protocol

The single most valuable output of bench work is the A/S versus performance curve. You generate it by holding the chemistry fixed (from the jar test) and stepping the air dose across a series of flotation cells, then measuring residual turbidity in each. To vary the air, you change one of two things: the volume (or ratio) of pressurised recycle water injected, or the saturator pressure. Everything hinges on computing the true A/S for each cell.

The recommended matrix varies R (recycle ratio) at a fixed saturator pressure, giving perhaps six A/S values spanning the design range. For each cell you record residual turbidity after a fixed flotation time (commonly 3–10 min). Plotting residual turbidity against A/S produces the characteristic knee curve: turbidity falls steeply as air increases, then flattens once every floc has enough bubble surface to float. The design A/S is chosen just past the knee — enough air for robust flotation, without wastefully over-pressurising the saturator.

In this illustrative set the knee sits near A/S ≈ 0.020 (cell 4): beyond it, tripling the air from 0.020 to 0.037 buys only a 2 NTU improvement. The engineering choice is therefore an A/S of about 0.020–0.025 — the point of diminishing returns. This is precisely the kind of evidence a supplier performance guarantee should be pinned to.

Worked example 1: computing the bench A/S for one cell

Take cell 3 from the matrix. We floated a 1.0 L conditioned sample at 1,000 mg/L solids, injecting pressurised water at a recycle ratio R = 0.12, from a saturator at 5.0 atm absolute, at 20 °C, with an assumed saturator efficiency f = 0.85.

• Air solubility: S_a = 18.7 mg air/L at 20 °C, 1 atm.
• Air released per litre of recycle: S_a · (f·P − 1) = 18.7 × (0.85 × 5.0 − 1) = 18.7 × (4.25 − 1) = 18.7 × 3.25 = 60.8 mg air per litre of pressurised water.
• Air delivered to the sample: multiply by R = 0.12 → 60.8 × 0.12 = 7.3 mg air per litre of sample.
• Solids present: 1,000 mg/L.
• A/S ratio: 7.3 / 1,000 = 0.0073 ≈ 0.007 kg air / kg solids.

Note this uses the actual injected volume rather than the nominal table value, illustrating why you must record real volumes and temperatures during the test — a 2 °C temperature error shifts S_a by roughly 3–4%, and an unmeasured saturator efficiency can swing the result by 10% or more. The discipline of computing A/S per cell is what converts a set of murky-or-clear beakers into a defensible design curve.

Worked example 2: converting bench results to full-scale recycle and rise rate

Suppose the bench curve selected a design A/S of 0.020 kg/kg for a full plant treating Q = 80 m³/h of effluent at X₀ = 900 mg/L solids, with a saturator running at 5.0 atm absolute, f = 0.85, at 20 °C. We need the full-scale recycle ratio and then the tank rise rate.

• Solids load: 80 m³/h × 900 g/m³ = 72,000 g/h = 72 kg/h TSS.
• Air mass required: A/S × solids = 0.020 × 72 = 1.44 kg air/h = 1,440,000 mg/h.
• Air released per m³ of recycle: S_a·(f·P − 1) = 18.7 × 3.25 = 60.8 mg/L = 60,800 mg/m³.
• Recycle flow required: Q_r = 1,440,000 / 60,800 = 23.7 m³/h.
• Recycle ratio: R = Q_r / Q = 23.7 / 80 = 0.30 (30%).

Now the hydraulics. If the tank is designed for a surface (hydraulic) loading rate of 8 m³/m²·h, the required surface area uses the total flow through the tank:

You can size the full unit around this with our engineering approach to DAF sizing, using the tested A/S as a hard input rather than a textbook assumption. The important conceptual point: the bench test hands you chemistry and A/S; it does not hand you a surface loading rate.

Scale-up rules: why bench SLR is not full-scale SLR

The most dangerous mistake in interpreting bench data is to read a rise rate straight off the cell and apply it to the tank. It does not scale one-to-one, and understanding why keeps designs honest.

In a bench column, flotation is quiescent and batch: microbubbles and flocs rise through still water with no cross-flow, no inlet turbulence, and no float blanket being continuously removed. The effective rise rate a bench cell demonstrates is therefore the intrinsic bubble-floc rise velocity — governed by aggregate density and bubble load — which is often 10–40 m/h. A full-scale tank cannot run anywhere near that, because it has horizontal throughput, inlet momentum, density currents and skimmer disturbance, all of which promote short-circuiting and re-entrainment. Full-scale hydraulic loading is consequently set at a fraction of the bench rise velocity — typically 5–15 m³/m²·h — with a safety margin.

What does transfer cleanly is the microscale physics: bubble size, bubble-particle attachment, the chemical dose, and the A/S ratio. What does not transfer is the tank hydraulics. This is the guiding principle of DAF scale-up — similitude at the bubble scale, not at the tank scale. It is why a bench programme fixes chemistry and air, while a pilot or CFD study fixes hydraulics.

Pilot testing: proving the hydraulics before you buy

Where a bench cell is silent — on short-circuiting, float-blanket behaviour, skimming and steady-state throughput — a continuous-flow pilot skid answers. A pilot DAF is a small but fully continuous unit (typically 0.5–5 m³/h) with a real saturator, recycle pump, contact zone and skimmer, run on a live slipstream of the actual effluent for hours to days.

The pilot confirms three things the bench cannot. First, that the chosen full-scale surface loading rate genuinely holds the float without carry-over under continuous flow. Second, that the float blanket thickens and skims to the expected solids concentration (often 3–6% dry solids), which sets the downstream dewatering duty. Third, that the unit tolerates the plant's real flow and load variability — diurnal peaks, shock loads and temperature swings — rather than a single well-behaved grab sample. A residence-time-distribution (tracer) test on the pilot also quantifies short-circuiting directly, and for a high-value or difficult duty this is where CFD modelling earns its place, cross-checking the pilot hydraulics against a simulated flow field.

Run the pilot at the A/S and chemistry the bench work established, then push the hydraulic loading up in steps until effluent quality degrades — that failure point, discounted by a safety factor, is your defensible full-scale design loading. The combination is powerful: the bench cell tells you whether the water can be floated and at what air dose; the pilot tells you how fast you can push a real tank before it breaks.

Interpreting results and common pitfalls

Good data is easy to misread. The following are the failure modes that most often turn a sound test programme into a bad design decision.

• Reading rise rate off the bench cell. As above, bench flotation is quiescent; its intrinsic rise velocity is 2–4× the loading a continuous tank can sustain. Never specify SLR from a batch column.
• Testing a stale or unrepresentative sample. Effluent chemistry drifts within hours — pH shifts, oils coalesce, biological activity changes solids. Test fresh, on-site where possible, and across the real operating envelope, not one grab.
• Over-dosing polymer. Chasing large flocs makes them dense enough to sink between bubbles. For DAF, target small, strong, buoyant flocs and let the bubbles do the lifting.
• Ignoring temperature. Air solubility falls with rising temperature, so a summer A/S can be 15–20% below a winter A/S at the same pressure and recycle. Compute A/S at the warmest expected condition.
• Not measuring air, only pressure. Saturator efficiency below 100% and unmeasured injected volumes make the real A/S far lower than the nominal figure. Always compute A/S from measured quantities.
• Choosing A/S on the flat of the curve. Past the knee you pay in energy and saturator size for negligible clarity gain. Select the design A/S just past the knee, not at the plateau.

Interpreted with these caveats, a bench-plus-pilot programme is the most cost-effective insurance available on a flotation project: a few days of testing routinely prevents a six-figure error in tank size, saturator capacity or chemical spend. When you are ready to translate the tested numbers into hardware, our overview of what a well-conditioned unit needs — from saturator pressure to float handling — is set out in the MCBA guide to dissolved air flotation systems.

DAF bench-to-pilot testing protocol

1. Optimise chemistry by jar test. Run a coagulant dose ladder at fixed pH, then a pH ladder at the best dose. Record supernatant residual turbidity; fix the coagulant, pH, polymer and dose that minimise it.
2. Prepare the bench flotation cell. Condition a fresh sample with the chosen chemistry using matched rapid/slow mixing. Charge the saturator with clean water and pressurise to 4–6 bar.
3. Run the A/S series. Inject pressurised water at increasing recycle ratios into replicate cells. Record injected volume, pressure and temperature so A/S can be computed for each.
4. Measure and plot performance. After a fixed flotation time, sample sub-natant turbidity from each cell. Plot residual turbidity against computed A/S and identify the knee.
5. Select the design A/S. Choose the A/S just past the knee, then convert to a full-scale recycle ratio and required air mass at the plant flow and solids load.
6. Confirm hydraulics at pilot scale. Run a continuous pilot at the tested chemistry and A/S; step up the surface loading until quality degrades, then apply a safety factor to set the full-scale rise rate.