I have worn out three fertilizer injectors in nine years. Two failed the same way: a hardened O-ring in the water motor, a ratio that crept without anyone noticing, and a fortnight of lettuce fed on something closer to tap water than nutrient solution. The third is still running. It is the one I test with a bucket every month.
That experience shapes how I judge dosing hardware. Two families of machine exist, they solve different problems, and neither replaces knowing your reservoir. A proportional injector holds a ratio. An EC and pH controller holds a reading. A grower who confuses the two can feed the wrong solution for six weeks without an alarm.
Two ways to dose, and the physics that separates them
The cheapest dosing device in horticulture is barely a pump at all. A venturi injector, sold for decades as the Hozon and its copies, is a brass mixing valve that screws onto a tap. Water accelerates through a narrow throat, line pressure drops, and the resulting vacuum pulls concentrate up a dip tube from a bucket. Nothing electrical. Nothing to program.
The catch is written into the physics. Suction depends on the pressure difference between the water line and the stock tank, so anything that shifts line pressure shifts the dose. Move the hose, add a second sprinkler, kink the line, and the ratio wanders. The Georgia bulletin on injector selection says these units cannot hold precise control over concentration for that reason, and puts the practical injection ratio near 1:16. A stock tank empties in roughly five minutes of flow, which works out around 750 ml/min of concentrate. The same publication gives a 35 psi minimum for proper operation and notes that excessive back-pressure stops the siphon cold. Venturi units suit small areas and short hose runs. They do not suit anyone who wants two decimal places.
Positive displacement injectors work on a different principle. Water passes through a motor, then that motor drives a small pump. One revolution of the water motor equals one stroke of the pump. The ratio comes from cylinder geometry rather than from pressure. Dosatron, DosMatic, Anderson and Smith all build versions of this, some water-powered and some using a metered electric pump.
Because the mechanism meters water instead of reacting to it, the proportion stays fixed across a broad band of flow and pressure. Published specification bands are wide: injection ratios from 1:50 to 1:1000 on various models, flow capability from under 1 gpm to several hundred gpm on the largest frames. Units with two heads inject two incompatible concentrates separately, which matters more than it sounds.
There is a ceiling on how strong you can make the stock, and it is chemical rather than mechanical. Concentrate has to dissolve completely. Push past roughly 1:200 and some fertiliser salts stop going into solution, so the Georgia authors advise staying at 1:200 or lower for greenhouse work. A 1:100 head needs half the stock strength of a 1:200 head to deliver the same nitrogen, so the dilution ratio sets the recipe for everything upstream.
Where EC and pH controllers earn their money, and where they mislead
An EC and pH controller is a closed loop. Probes sit in a mixing tank. The controller reads them, then runs dosing pumps that pull concentrated fertiliser from one or two stock tanks and acid or base from another. It stops when the reading matches the setpoint. Nothing about that loop is proportional to water flow. It is proportional to a sensor voltage.
That difference cuts both ways. The controller can hold 2.0 mS/cm all afternoon without anyone touching a bucket, which is genuinely useful when you are running a 3,000 L recirculating tank and the plants are drinking 400 litres a day. It can also dose with complete confidence while the probe is wrong. This is the failure mode nobody plans for. A drifted pH electrode does not throw an error. It simply reports a number, the controller believes it, and the acid pump obliges.
Chemistry also constrains the plumbing. Calcium nitrate and magnesium sulfate cannot share a stock tank without forming calcium sulfate, and calcium plus phosphate precipitates just as reliably. Kansas State’s hydroponic systems publication describes the standard workaround: split the concentrate into two tanks, one holding calcium nitrate and iron, the other carrying everything else, and use a twin or triple-head proportioner. If both heads feed one stock tank by mistake, the delivered concentration doubles.
Matching the hardware to the reservoir you actually own
Scale decides this more than budget does. The numbers below assume you are filling from a tank or a line, and they assume you already own decent meters.
| Reservoir volume | Approach that fits | Reasoning |
|---|---|---|
| Up to about 100 L | Hand-mix from two-part stock | No mechanical failure mode exists, because there is no mechanism |
| 100 to 800 L | Water-driven proportional injector on the fill line | Ratio stays fixed at any fill speed; one head is enough for a simple recipe |
| 800 to 5,000 L | Dual-head injector with inline EC and pH monitoring | Two concentrates stay apart, and you watch the mixed solution rather than trust the machine |
| Above 5,000 L, or multiple zones | EC and pH controller driving dosing pumps into a mixing tank | Manual correction stops being feasible once demand varies hour to hour |
The 100 to 800 L band is where a proportional injector beats a controller on cost per litre and on the number of things that can break. A pump head that relies on a sensor needs the sensor to be right. A piston that meters water needs a clean filter and an intact seal. I know which I would rather debug at 6 a.m.
What drift actually costs you
EC is a sum, not a recipe. Bruce Bugbee’s work at Utah State makes the point plainly: because plants take up different nutrients at different rates, conductivity readings mostly reflect the calcium, magnesium and sulfate left behind, while micronutrients contribute less than 0.1% of the signal. Nitrogen, phosphorus and potassium can sit near depletion while the meter still reads a comfortable number.
So the readout can be right and the solution can be wrong. That is the quiet version of drift, and it shows up as pale new growth and stalled fruit set rather than a dramatic collapse.
The loud version is osmotic. Arizona’s controlled environment agriculture chapter gives mature tomatoes a drip EC target of 2.5 to 3.5 mS/cm, with the drain running about one unit higher, and notes the conversion of 1 mS/cm to roughly 640 ppm total dissolved solids. That puts a 2.0 mS/cm lettuce solution near 1,280 ppm on the same scale. Purdue’s lettuce trial found the best growth at 1.3 dS/m, with plants at 2.9 and 3.9 clearly under osmotic stress. Kentucky’s monitoring module splits the difference by stage, suggesting 1.2 to 2.0 mS/cm for vegetative growth and 2.5 to 3.5 mS/cm once fruiting starts.
Every one of those figures assumes pH is behaving. pH decides which nutrients stay soluble enough to be absorbed at all. Below 5.5 and above 6.5 the door starts closing on iron, phosphorus and the micronutrients, and the plant starves in a solution that is technically full of food. Kentucky puts the working window at 5.5 to 6.5. Virginia Tech’s media systems guide narrows it to 5.5 to 6.2 and recommends checking EC and pH two to three times a week.
Alkalinity is the variable most growers blame pH for. Georgia’s bulletin draws the distinction sharply: water at pH 9.0 with alkalinity of 50 will barely move substrate pH, while water at pH 7.0 with alkalinity of 300 will push it up hard. If you dose acid to chase a pH number without knowing your alkalinity, you will spend a season overcorrecting.
Why the meters need calibrating, and what happens when you skip it
A pH electrode is a glass bulb with a hydrated layer that wears and fouls. Calibration measures its slope, and slope is the honest signal. California State University Stanislaus calls 95% to 100% ideal, accepts 90% to 105%, and flags anything outside that as meter error. Expired buffer is the most common cause of a bad calibration. The EPA’s standard operating procedure for pH meters is stricter still, requiring a slope of 100 ± 5% and specifying storage in 3 M KCl. Never in distilled water, because that leaches the reference junction and slowly kills the electrode.
I learned that one the expensive way. A spare probe spent a winter in a beaker of tap water, came back with a slope in the low 80s, and read 0.4 pH high at the calibration point. Two days of dosing against a lying instrument cost me a batch of basil.
EC meters need less ceremony but they still need the water correction. Tennessee’s greenhouse nutrition publication gives the formula directly: subtract the EC of the irrigation water from the EC of the fertiliser solution before comparing with the label value. Skip it on a borehole reading 0.4 mS/cm and you overfeed by around a fifth at a 2.0 target.
Proportional injectors have their own drift, and it is mechanical. Virginia Tech’s fertiliser calculations guide states it without hedging: injector ratios change over time as parts wear, and units should be calibrated monthly through the growing season. Most heads have no calibration screw, so you adjust the stock recipe instead. The test is a bucket and a measuring cup. Collect ten gallons through the injector into a graduated container, watch how much the dip tube removes from the cup, and divide. Ten gallons collected against 6.4 fluid ounces depleted works out at 1:200.
My current head measured 1:178 on that test after nineteen months of service, which meant it had been running about 11% rich. Nothing had failed. It had simply worn. Maryland’s injector pump guide pairs the bucket method with an EC check of the delivered solution, and that combination catches both a worn pump and a stock tank mixed to the wrong strength. The Georgia authors set a threshold: if the measured ratio departs from the setting by more than 5%, the unit needs service rather than a tweak. Cornell’s greenhouse nutrition pages carry a worksheet for the same job if you want a procedure to follow rather than remember.
Controllers fail on a different clock. Peristaltic heads move a few millilitres to a few tens of millilitres per minute, and the tube flattens over time, so delivery falls while the controller keeps counting seconds. My tube lasted fourteen months before flow dropped noticeably at a fixed speed. Fresh silicone restored it. Piston-seal kits are not expensive, and replacing them on a schedule beats discovering the failure through a crop.
The honest case for hand-mixing
Below roughly 100 to 200 litres, a proportional injector is solving a problem you do not have. A two-part stock set, a bucket, a jug and a stirrer will do the job in ten minutes. There is no seal to wear, no filter to clog, no ratio to verify. The failure mode is forgetting, and a calendar fixes that.
I ran a 60 L reservoir twice a week for two seasons of leaf crops with nothing but two measuring jugs and a handheld meter. Mixing took eight minutes. The results were fine. The only reason I moved to an injector was a second bay and a standing commitment that did not wait for me.
Hand-mixing also forces you to look at the solution. You weigh the calcium nitrate, you watch the pH, you notice the colour of the water change. Automation removes that contact, and contact is where most growers learn what a healthy reservoir looks like. Missouri’s hydroponic nutrient solutions guide is blunt that solutions must be measured and adjusted continuously, because both pH and EC shift as plants draw water and nutrients at different rates. Whether a machine does the measuring or you do is a separate question from whether it gets done.
Two safety rules belong in any hand-mixing routine. Keep calcium nitrate out of the same concentrate as sulfates and phosphates, as Georgia and Kansas State both warn, and add acid to water rather than water to acid. A splash of concentrated stock is unpleasant. A splash of phosphoric acid concentrates is a hospital trip.
Failure modes worth designing around
Venturi units clog their suction screens and build salt deposits around the mixing valve. Both are invisible until output drops. Positive displacement units lose suction when back-pressure rises, and a failed check valve can let concentrate migrate back into the potable line, which is why Georgia devotes a section to cross-connection control and local water regulations. Put a filter and pressure gauges upstream, and fit a backflow preventer before you argue with anyone about it.
Controllers concentrate risk in the sensor. A probe that drifts high makes the unit under-dose; a probe that drifts low makes it over-dose, and it will keep dosing into a mixing tank that is already at strength. Set a maximum run time on each dosing pump so a stuck loop cannot empty a stock tank into a small reservoir. On a 3,000 L system that limit is a nicety. On a 60 L system it is the difference between a bad week and a dead crop.
What I would install next
If I were setting up again with a 500 L tank and a single zone, I would buy a dual-head water-driven injector, a filter, two pressure gauges and a backflow preventer. No controller, no wiring, no setpoint to trust. I would test the ratio monthly with a bucket, record the number, and adjust the stock recipe when it moved. Inline EC and pH probes would be a monitor I glance at, not a machine I obey, because a monitor that drifts costs me a reading while a controller that drifts costs me a crop.
Past 5,000 L, or once a second zone runs a different recipe, that logic inverts. The labour of hand-checking a large recirculating system every day outweighs the sensor risk, and a controller with fresh buffers and a scheduled probe replacement earns its keep. The technology is not the decision. The reservoir is.