Every dosing pump, bleed valve, or injector in a water treatment system runs on one of four control strategies: on/off, proportional, PID, or fuzzy logic. Most operators never choose one deliberately — they inherit whatever came bundled with the controller and only notice a problem once conductivity starts swinging wildly, a discharge line drifts out of compliance, or a loop that worked fine for years suddenly hunts after a process change.
The four methods are not interchangeable, and using the wrong one for the wrong reason is a common source of chemical waste, off-spec discharge, and unnecessary strain on dosing equipment. Here is what each one actually does, and what happens when it meets a real process — cooling tower bleed control, wastewater pH adjustment, hydroponic nutrient dosing, and a cooling tower’s full biocide, pH, and inhibitor programme — across Malaysia and elsewhere.
On/Off Control: The Simplest Loop You Can Build
On/off control — also called bang-bang control — has a single setpoint and two states. The output is either fully on or fully off. To stop the output from chattering rapidly at the setpoint, a dead band (hysteresis) is added: the output switches on above the setpoint plus a margin, and off below the setpoint minus a margin.
This is the cheapest and most robust control strategy available, and it is genuinely adequate for slow-moving processes with a lot of buffering volume. It is not adequate for anything that needs to hold a tight value continuously, because the process variable will always cycle within the dead band rather than settle at the setpoint.
Cooling tower bleed control (recirculating). A conductivity controller running the bleed valve on/off is the most common setup in the field, and it works passably because of the tower’s own recirculating volume: a full bleed event dumps a slug of concentrated water, and the basin has enough hydraulic mass to absorb that swing without immediate scale or corrosion consequence. The cost shows up in water and chemical consumption rather than water quality. Because the valve doesn’t open until conductivity crosses the upper dead band threshold, the tower is deliberately run below its true maximum cycles of concentration to leave margin — bleeding earlier and more often than the chemistry actually requires, and paying for the make-up water and re-dosed inhibitor that follows every bleed event. Widen the dead band to save water and the swings get large enough to risk scale at the top of the band; narrow it, and bleed frequency — and the water bill — climbs. It’s a pattern common on cooling towers across Malaysia’s manufacturing plants, where make-up water cost and discharge volume are both under increasing scrutiny.
Proportional Control: Output Scaled to the Error
Proportional control scales its output to the size of the error — the difference between setpoint and measured value. A small deviation produces a small correction; a large deviation produces a large one, up to the pump’s maximum output. This removes the on/off cycling problem and gives a much smoother response.
Proportional control has one structural limitation: it needs an error to produce an output. As the process approaches setpoint, the error shrinks, and so does the correction — the loop settles at a value offset from setpoint rather than exactly on it. This offset, sometimes called droop, is permanent unless something else (an operator’s manual bias, or an added integral term) removes it.
Hydroponic nutrient dosing (recirculating). A recirculating nutrient reservoir dosed proportionally against an EC setpoint tends to look fine on the EC meter — the reservoir’s own volume smooths out most of the offset’s practical effect, and growers rarely notice a reservoir running slightly under or over target EC. What proportional control’s offset actually distorts is easy to miss: A and B nutrient stock are almost always dosed through separate proportional channels, and if one channel’s offset differs even slightly from the other’s, the ratio between elements shifts — not just the total ionic concentration the EC probe reports. A reservoir can hold a perfectly acceptable EC reading while quietly running an off-ratio nutrient mix for days, because EC measures total dissolved ions, not which ions they are. Growers running recirculating systems across Malaysia’s hydroponic operations who monitor EC and pH alone typically have no way to see this ratio drift building.
PID Control: Closing the Gap Proportional Leaves Behind
PID adds two terms to proportional control. The integral term accumulates error over time and keeps increasing the output until the error reaches zero — this is what eliminates the permanent offset that proportional control leaves behind. The derivative term looks at how fast the error is changing and adds a correction based on that rate, which dampens overshoot and lets the loop respond ahead of a slow-moving disturbance rather than only after it has fully arrived.
Done well, PID gives you a loop that returns exactly to setpoint and handles disturbances smoothly. Done poorly — with gains set for a different process, or a controller left on factory-default tuning — PID can overshoot, oscillate, or wind up the integral term into a correction that takes a long time to unwind once the disturbance has passed.
pH adjustment in industrial wastewater treatment (once-through). A pH neutralisation skid dosing acid or caustic into a once-through effluent line before discharge is exactly where the integral term earns its place: with no recirculating tank to average conditions out, an offset-prone proportional loop leaves the treated stream sitting near the edge of the discharge limit rather than on the compliance target, with no margin for the next surge of higher-strength effluent. PID closes that gap — but only if it’s tuned around the actual transport lag between the dosing point and the downstream pH probe. Get that dead time wrong, and the derivative term reacts to a disturbance that has already been over-corrected by the integral term, pushing pH past the target on the other side before the effect of the last dose has even reached the sensor. Once tuned correctly, a PID loop can hold tight to a single pH target rather than parking near one edge of the band — a detail that matters on once-through lines across Malaysia handling variable-strength streams like palm oil mill effluent (POME), where a slug of stronger effluent arriving faster than the loop’s tuned response is a discharge exceedance, not a transient blip.
Fuzzy Logic Control: No Model Required
Fuzzy logic control does not work from a fixed mathematical relationship between error and output the way PID does. Instead, it uses rule-based logic built on linguistic terms — “error is large,” “rate of change is small negative” — each with a degree of membership rather than a hard true/false cutoff, and combines multiple rules simultaneously to decide an output. It does not need a precise model of the process to work reasonably well across a wide range of conditions, which is exactly where PID and proportional control start to struggle.
The trade-off is that fuzzy logic is harder to reason about from the outside — there is no single gain value to point at and say “that’s why it overshot.” What it buys in return is stability across conditions that would otherwise each need their own tuning pass, and the ability to weigh several interacting variables at once rather than controlling each in isolation.
Cooling tower biocide, pH, and inhibitor control (recirculating). A cooling tower running a full treatment programme — biocide, pH, and corrosion inhibitor together — is a genuinely multi-variable control problem, especially where inhibitor residual is tracked with an inert fluorescent tracer such as PTSA or a tagged polymer, read by an in-line fluorometer instead of a slow wet-chemistry test. Bleed activity dilutes conductivity and inhibitor/tracer concentration at the same time. pH shifts affect both biocide efficacy and inhibitor solubility. A biocide dose briefly disturbs pH and can interact with the tracer reading. No single fixed loop models all of these relationships correctly, because the interactions aren’t linear and change depending on which variable moved first. A fuzzy controller can weigh several of these inputs at once — tracer-based inhibitor residual, pH deviation, recent bleed activity, time since the last biocide dose — against a rule set (for example: if inhibitor residual is trending low and bleed has been active recently, increase inhibitor dose moderately regardless of current pH) and reach a dosing decision without a unified transfer function tying bleed, pH, biocide, and tracer response together. For larger commercial and industrial cooling towers across Malaysia running full inhibitor-biocide-pH programmes, this kind of simultaneous multi-loop coordination is where fuzzy logic earns its complexity over a bank of separately tuned PID loops.
Matching the Control Method to the Application
None of these four methods is universally “better” — each is a different answer to how tightly a process needs to be held, how many variables interact, and how much the process characteristics change over time. On/off is fine for slow, high-volume-buffer processes where a wide dead band doesn’t matter. Proportional is fine wherever a small permanent offset is acceptable and only one variable is being held. PID is worth the tuning effort wherever the process is reasonably stable, single-loop, and an exact setpoint matters. Fuzzy logic earns its complexity wherever several interacting variables need to be balanced at once, or conditions vary enough that a fixed PID tuning would need constant readjustment.
If your application calls for fuzzy logic, the Aquarius Ultima controller comes equipped with it as standard. If PID is the right fit and you want the flexibility to tune loop response to your own process, both the Nexus 5000 and 5500, as well as Pyxis’ UC80, offer PID control. And where proportional control is genuinely sufficient — plenty of applications don’t need more — the Nexus 2000 and 3000 give you that without paying for capability the process doesn’t use.
Not sure which control strategy fits your process — a cooling tower, a wastewater line, or a hydroponic reservoir anywhere in Malaysia? Contact the Autoflo team at info@autoflotechnology.com and we’ll help you match the controller to the loop.
Frequently Asked Questions
What is the main difference between on/off and proportional control? On/off control drives the output fully open or fully closed around a dead band, causing the process variable to cycle within that band. Proportional control scales the output to the size of the error, giving a smoother response but settling at a small permanent offset from setpoint rather than cycling around it.
Why does proportional control never quite reach the setpoint? Proportional control needs an error to generate an output — as the process approaches setpoint, the error and the correction both shrink toward zero, so the loop settles at an offset (droop) rather than exactly on target. Only adding an integral term, as in PID, removes that offset.
What does the integral term in PID actually fix? The integral term accumulates the error over time and keeps adjusting the output until the error reaches zero, eliminating the steady-state offset that a proportional-only loop leaves behind. It’s the reason a well-tuned PID loop returns to the exact setpoint rather than settling near it.
When does fuzzy logic control make more sense than PID? Fuzzy logic is worth using when several variables interact and need to be balanced simultaneously, or when the process is nonlinear enough that a single PID tuning would need frequent readjustment. It works from rule-based logic rather than a fixed mathematical model, so it can weigh multiple inputs at once without a unified equation tying them together.
Which control method suits cooling tower water treatment? Many cooling towers start with on/off bleed control, which is often adequate for conductivity alone given the tower’s buffering volume, though it runs the tower below its true maximum cycles of concentration to leave margin. Towers running a full biocide, pH, and inhibitor programme — particularly with tracer-based inhibitor monitoring like PTSA or a tagged polymer — benefit from fuzzy logic control, such as the Aquarius Ultima, which can weigh several interacting loops at once instead of needing a separately tuned loop for each.