Twenty strokes per minute is one stroke every three seconds. That interval is what the clause is protecting, and it was written in an era when a chemical feed pump got its stroke energy from plant air or from a fixed-energy solenoid, and got its turndown by shortening the stroke. Both of those mechanisms fail quietly at the bottom of the range. The number in the standard is a stall-and-slip margin dressed up as a percentage.
Fixes That Fail the Reviewer
Four workarounds come up first on every fluoridation permit that lands below the 20 percent floor, and each one trades a real problem for a worse one.
Downsizing the pump until the design point lands above 20 percent. This works on the day you size it and fails on the seasonal extremes. The feed range has to span minimum plant flow at minimum dose through maximum plant flow at maximum dose, plus the make-up feed after a clearwell drawdown. Size the pump so the low end clears 20 percent and the high end no longer covers the design condition; size it for the high end and the low end drops back under the floor. The turndown requirement is a range requirement, not a point requirement.
Diluting the solution to raise the stroke count. Physically sound, administratively expensive. A dilution tank introduces a second concentration that has to be verified, a second inventory to sample, and a mixing accuracy that stacks on top of the pump accuracy. With fluorosilicic acid it is worse than an accuracy problem: dilution drives hydrolysis, off-gassing at the injection point, and silica deposition that fouls the injection quill and the suction check valve. You have converted a metering problem into a maintenance problem the operators will inherit.
Submitting the manufacturer's turndown ratio as the proof. A 1:1000 turndown claim on a cut sheet is a marketing boundary, not a design point. The vendor has an incentive to publish the widest range that the control electronics can command; that number says nothing about repeatability at your discharge pressure, your solution viscosity, or your degassing conditions. A reviewer who has seen a few of these will not accept the spec sheet alone, and should not.
Duty-cycling the pump on a timer to fake a low average rate.It also defeats flow pacing, which is the point of having a variable-speed feeder.
Getting the low end by shortening the stroke. This is the exact mechanism the clause was written against. Details below.
The Two Clauses and the Numbers in Them
| Clause | Requirement | Floor | Relief written into the clause |
|---|---|---|---|
| 4.7.2(e) — Fluoridation, chemical feed equipment and controls | Fluoride solution applied by a positive displacement pump; stroke rate not less than 20 strokes per minute | Feed rate not less than 20% of rated pump capacity | None stated |
| 5.0.3(g) — Chemical application, general equipment design | Chemical feeders and pumps operate no lower than 20% of the feed range | 20% of feed range | 10% of rated maximum where two fully independent adjustment mechanisms are fitted, such as pulse rate and stroke length |
Read 5.0.3(g) carefully, because it contains the engineering argument in plain text. The standard does not care about 20 percent as a magic number. It cares that a single adjustment mechanism run to its extreme becomes inaccurate, and it grants half the floor to any feeder that gets its turndown from two mechanisms operating independently. That sentence is the door.
Stall Margin: The Original Reason for 20 Strokes per Minute
An air-actuated diaphragm or piston pump develops stroke force from a supply pressure acting across a fixed area. Command it slower and you do not get a slower stroke; you get the same stroke separated by longer dwell. At long dwell the pilot spool leaks down, the diaphragm relaxes against the seat, and the pump either skips strokes or completes them partially. The failure is silent. The stroke counter increments, the pump makes noise, and the feed rate collapses.
Solenoid-driven pumps behave the same way for a different reason: stroke energy comes from a fixed capacitor discharge into the coil, so the pump either has enough energy to lift the diaphragm against system pressure or it does not. There is no partial credit. Below some rate the controller extends the interval far enough that a fouled or gas-bound check valve never gets flushed, and the head loses prime.
Twenty strokes per minute is a conservative floor that keeps those pumps above the dwell interval where these effects appear. It is an equipment-class rule, not a physics constant.
Slip and Stroke-Length Turndown
The second half of the intent is volumetric accuracy. A diaphragm pump turned down by shortening the stroke displaces less per stroke, but the parasitic volumes do not shrink with it. Check-valve seating lag, ball lift, tubing compliance, dissolved gas coming out of solution in the head, and backflow through the discharge check are all roughly constant per stroke. At full stroke they are a small percentage of displaced volume. At 20 percent stroke length they are the same absolute volume against one-fifth the displacement, so the error multiplies by five. At 10 percent it multiplies by ten.
That is why the older generation of pumps was worth avoiding below 10 percent of full capacity: the inaccuracy is inherent to the turndown method, not to the build quality. A pump that reduces capacity by shortening stroke length is a different machine at the bottom of its range than at the top.
What a Stepper-Driven Diaphragm Changes
A stepper motor driving the diaphragm changes both failure mechanisms at once. Torque is developed by the motor and gearing rather than by a fixed pneumatic or magnetic impulse, so the drive completes a full-displacement stroke regardless of how slowly it is commanded — there is no dwell interval below which the stroke does not happen. Turndown comes from stretching the stroke over a longer interval at full displacement, not from truncating it, so the parasitic-volume ratio that ruins stroke-length turndown does not degrade in the same way. Spreading the discharge over a longer period also lowers peak velocity through the injection point, which is favorable for check-valve seating and for mixing.
The argument to the reviewer is therefore not "the pump is better." It is that the two physical mechanisms the clause defends against — stroke stall at long dwell, and slip-dominated error at short stroke — are structurally absent in a positive-displacement pump that varies speed at constant full stroke. Bring the pump's dosing accuracy and repeatability figures from the cut sheet to the meeting, and bring the drawdown data described below, because the reviewer will reasonably discount the first without the second.
Keep the design point well inside the published boundary anyway. A pump that can be commanded across three decades should not be designed to run at the last one. Pick the operating minimum with margin, the same way you would not run a coupling at its published maximum misalignment just because the catalog allows it.
Backpressure, the Variable That Sets Accuracy
The number that matters here is differential pressure across the pump head, and it is the single most common reason a metering pump that passed a shop test misses its setpoint in the field. A metering pump needs a consistent minimum discharge pressure to seat the discharge check and to stop the suction side from carrying flow through on its own. At least one manufacturer specifies a minimum of about 25 psig; read the minimum differential pressure figure out of your pump's installation manual rather than assuming a value.
Two conditions have to hold at every operating point:
- Discharge pressure must exceed suction pressure by the manufacturer's minimum differential. Injecting into an open clearwell, a wet well, or a low-head channel with a flooded suction tank almost never satisfies this by itself.
- Static head on the suction side must never exceed discharge pressure, or the tank siphons through a stopped pump and overfeeds until the day tank is empty.
Both are fixed by a spring-loaded back pressure valve on the pump discharge, sized to hold the minimum differential at maximum feed rate, plus an anti-siphon valve at the injection point. Add both before you run any calibration, because a drawdown test performed without a back pressure valve measures the piping, not the pump.
Sizing to a Turndown You Can Defend
Establish the required turndown before arguing about the permitted turndown. The chemical feed arithmetic is unchanged:
lb/day of chemical = Q (MGD) x dose (mg/L) x 8.34 lb/gal
solution feed (gpd) = lb/day / (solution density (lb/gal)
x purity fraction
x available-ion fraction)
required turndown = (Q_max x dose_max) / (Q_min x dose_min)- Compute maximum feed from firm capacity at the maximum permitted dose, including the raw-water background correction.
- Compute minimum feed from minimum sustained plant flow — overnight, off-season, or single-well operation — at the minimum target dose.
- Divide to get required turndown. Most fluoridation installations land in the single digits to low tens, not in the hundreds.
- Select a pump whose rated capacity puts the maximum feed at roughly 70–85 percent of rating, then check where the minimum feed lands as a percentage.
- If the minimum lands between 10 and 20 percent, verify from the cut sheet whether the pump provides two independent adjustments. If it does, 5.0.3(g) grants the 10 percent floor on its face and no deviation is needed.
- If the minimum lands below 10 percent, either split the duty across two differently sized pumps — a common and easily permitted arrangement — or submit the deviation with the verification package below.
Drawdown Verification and Loss-of-Feed Detection
A deviation request carries when it substitutes measurement for the prescriptive limit. Two elements do that: a calibrated capacity curve across the intended range, and an instrument that detects the stall or underfeed the clause was written to prevent.
| Field symptom | Mechanism | Where to look |
|---|---|---|
| Residual drifts low only at minimum plant flow | Pump operating below its repeatable range; slip is a large fraction of displacement | Drawdown at the low setpoint vs commanded rate |
| Stroke indicator active, zero delivery | Gas binding or vapor lock in the head; fouled suction check | Bleed/degassing valve, suction check ball and seat |
| Day tank drains overnight, residual spikes | Siphon: suction static head exceeds discharge pressure | Back pressure valve setting, anti-siphon valve at quill |
| Drawdown repeats poorly trial to trial | Varying discharge pressure; no fixed minimum differential | Discharge gauge during a full stroke cycle |
| Missed or partial strokes at low command | Fixed-energy drive below its stroke-energy threshold | Drive type — expected on air/solenoid pumps, not on a geared stepper drive |
Run the calibration this way:
- Install a graduated calibration column on the suction, sized so that a five-minute run at the lowest intended setpoint produces a readable displacement.
- Set the back pressure valve to hold the manufacturer's minimum differential at maximum rate; confirm the discharge gauge does not fall below it at minimum rate.
- Prime and bleed the head until the degassing port runs solid liquid.
- Draw down at 100, 50, 20, 10 percent and the lowest setpoint you intend to permit — three trials each, long enough that meniscus read error stays under 1 percent of the measured volume.
- Compute error as (measured − commanded) / commanded x 100 and record spread across the three trials as repeatability.
- Repeat the low-end points at minimum and maximum system pressure to prove the curve is not pressure-sensitive.
- Declare as the permitted minimum the lowest setpoint where both error and repeatability stay inside your accuracy band, with margin. That measured number, not the catalog turndown, is what goes in the submittal.
Pair it with continuous detection. A flow-verification device on the discharge — a dosing monitor or flow switch selected to actuate at the minimum intended feed rate, not at some generic threshold — alarms on stall or loss of prime so a failure cannot run undetected between residual samples. Trend it against the on-line fluoride or residual analyzer and against day tank level or scale weight; three independent indications of feed make the low-flow operating point defensible in an inspection.
Frequently Asked Questions
What happens if a metering pump runs below 20 percent of its feed range?
On a pump that turns down by shortening stroke length, check-valve lag, backflow and gas in the head stay constant while displacement shrinks, so percentage error grows in inverse proportion — roughly five times worse at 20 percent stroke, ten times at 10 percent. On an air- or solenoid-actuated pump the additional risk is skipped or partial strokes at long dwell, which produces zero feed while the stroke counter keeps incrementing.
What happens if there is no back pressure valve on the pump discharge?
Delivery becomes a function of system pressure instead of pump setting, drawdown results stop repeating, and if suction static head ever exceeds discharge pressure the day tank siphons through the stopped pump and overfeeds until it empties. Set a spring-loaded back pressure valve to hold the manufacturer's minimum differential — one manufacturer specifies at least 25 psig — and add an anti-siphon valve at the injection point.
What happens if I claim the manufacturer's full turndown ratio in the permit submittal?
Expect it to be rejected, and correctly so: a 1:1000 catalog turndown is a control boundary, not a verified accuracy range at your discharge pressure and solution. Submit measured drawdown data with error and repeatability at each setpoint, and set the permitted minimum well above the point where either metric starts to degrade.
Stop arguing the clause and start a formal deviation request once you have the drawdown curve, the back pressure valve setting, and the loss-of-feed alarm documented — that package is what a primacy agency reviewer can approve, and 5.0.3(g)'s dual-adjustment language gives them the precedent to work from. If the pump's minimum differential pressure, dosing accuracy, or verified minimum stroke behavior is not stated clearly in the installation manual, get it in writing from the manufacturer's technical support before the submittal rather than during review. Where the state supplements the Recommended Standards for Water Works with its own fluoridation rule, that supplement governs; take the interpretation question to the drinking water program directly.