Lesson 3.3Lesson 3.3 · Building Management Systems
Control Loops & Sequences
Feedback control - setpoint, sensor, controller, actuator - and why tuning and sequences matter
A room that never quite settles, a valve that hunts all day - almost every comfort and energy fault traces back to a control loop that is doing exactly what it was told, badly.
Under every setpoint the BMS chases is a feedback control loop: a sensor measures, a controller compares that measurement to the target, and an actuator acts to close the gap. It is the same loop whether you are holding a room at 22 C or a car at 60 mph.
Get these loops right and the building is quietly comfortable and efficient. Get them wrong - badly tuned, poorly sequenced, running open-loop - and you get hunting valves, fighting plant and wasted energy that no dashboard will explain until you understand the loop underneath.
Setpoint-sensor-controller-actuator, feedback. PID: P now, I erase gap, D damp. Hunting = bad tuning. Sequence = intended behaviour.
The feedback loop: four parts, one circle
A control loop has four parts, and every BMS is thousands of them running at once. The setpoint is the target - say 22 C. The sensor measures the actual value - the process variable, say 20.6 C. The controller computes the error (setpoint minus measurement, here +1.4 C) and decides how to respond. The actuator carries out that response on the physical world - opening a heating valve, speeding a fan, moving a damper. That action changes the process, the sensor measures the new value, and the loop goes round again, continuously.
The magic word is feedback: the output of the process is measured and fed back to the input of the decision. Because the controller keeps watching the result of its own actions, it can correct for anything it did not predict - a door left open, a sudden crowd, the sun coming out. This is why feedback control is so powerful and so ubiquitous: it does not need a perfect model of the building, only a sensor telling it whether it is on target and which way it is off.
This loop is also the concrete mechanism behind everything in Lesson 3.1. When we said the BMS 'measures, compares and acts', this is the measuring, comparing and acting - the sensor, the error calculation, the actuator. Every point you saw in the point list is either the measurement going into a loop or the command coming out of one.
Setpoint -> (error) -> controller -> actuator -> process -> sensor -> back to error. Feedback = watch your own result.
Open versus closed loop
The distinction that separates good control from wishful control is open loop versus closed loop. A closed-loop (feedback) controller uses the sensor measurement to decide what to do - it knows whether it hit the target and adjusts. An open-loop controller acts without checking the result: it runs on assumption. A timer that turns heating on for two hours regardless of the actual temperature is open loop; a thermostat that heats until the room reaches setpoint and then modulates is closed loop.
Open loop is not always wrong - it is simpler, cheaper and fine when the outcome is predictable and consequences are small (a corridor light on a timeclock). But for anything where conditions vary - which is most of a building - open loop drifts. It cannot compensate for a cold snap, extra people, or a fouled coil, because it never looks. A great deal of energy waste and discomfort in real buildings comes from control that is effectively open loop: schedules and fixed outputs that ignore what the space is actually doing. A digital twin, notably, is fundamentally a closed-loop idea - it exists precisely to feed real measurements back into decisions. A building running open loop is a building making decisions with its eyes shut.
There is a useful middle idea, too: feedforward. A smart controller can act on a disturbance it can measure before the error appears - opening the heating a little as soon as the outside temperature drops, rather than waiting for the room to get cold. Feedforward anticipates; feedback corrects. The best building control blends them: feedforward for the disturbances you can see coming, feedback to catch everything you cannot. Hold on to that distinction, because the predictive and model-based control of Module 7 is essentially feedforward taken to its sophisticated limit, layered on top of the humble feedback loops running underneath.
PID, conceptually - and why tuning matters
How hard should the controller push for a given error? The workhorse answer is PID control - Proportional, Integral, Derivative - and you can understand it without a single equation. The proportional term pushes in proportion to the error right now: far from setpoint, push hard; close, ease off. Alone, it tends to leave a small permanent gap (the room settles slightly cool). The integral term accumulates error over time and keeps nudging until that lingering offset is erased - it provides the patience proportional lacks. The derivative term reacts to how fast the error is changing, damping the response to head off overshoot - it provides the caution. Most building loops use P and I; D is added where overshoot is a problem.
The reason to care is tuning. The three terms are weighted by gains, and those weights must suit the specific process - a fast electric heater and a slow underfloor slab need very different settings. Tune them well and the value rises smoothly to setpoint and holds. Tune them badly and the loop hunts - overshooting, undershooting, oscillating around the target forever. Hunting is not harmless: it swings temperatures occupants feel, cycles valves and dampers until they wear out, and burns energy driving the plant back and forth. Poorly tuned loops are one of the most common and most invisible faults in real buildings - invisible, that is, until you look at the trend and see the sawtooth. This is a place a twin shines: oscillation that no operator would notice jumps out of a time-series, flagging a loop that needs retuning.
P = now. I = erase the lingering gap. D = damp the swings. Bad gains -> the loop hunts forever.
Sequences of operation: loops assembled into behaviour
A single loop holds one variable. A real building coordinates many, and the recipe that says how is the sequence of operation - the written, then programmed, logic describing how a system behaves across all conditions. A good AHU sequence specifies: on the occupancy schedule, start the fan; modulate the heating and cooling valves to hold supply-air temperature; use free cooling from outside air when it is cool enough (an economiser); modulate the fresh-air damper to hold CO2 below its setpoint; stage the plant up and down as load changes; and shut down and set back when unoccupied. It also defines alarms, safeties and interlocks - what must never happen, like heating and cooling at once.
Sequences matter because most building faults are not broken hardware; they are bad or misapplied logic. The classic example is simultaneous heating and cooling - a heating valve and a cooling valve both open in the same air stream, each fighting the other while the meter spins. It looks fine on a walk-round; only the sequence, or the trend data, reveals it. The industry has responded with standardised, high-performance sequences (notably ASHRAE's Guideline 36) precisely because ad-hoc sequences waste so much energy. For a digital twin, the sequence of operation is the ground truth of intended behaviour: compare what the building should do (the sequence) with what the trends show it doing, and the gaps are your faults. That comparison - intended versus actual - is the doorway from control into the fault detection and analytics of Module 6.
A last point ties the module together. Loops and sequences are where a twin stops being a pretty picture and starts paying for itself. A twin that merely mirrors geometry tells you nothing new; a twin that ingests loop trends and knows the intended sequence can say something an operator cannot: this valve has hunted for three weeks, that AHU heats and cools at once every afternoon, this zone never reaches setpoint after 15:00. Those are served decisions - the honest test of a twin from Lesson 0.1. Control is therefore not a dry preliminary to the exciting twin; it is the very substance the twin reasons about. Master the loop and the sequence, and the later analytics, fault detection and predictive control all have something real to stand on.
PID controller
Proportional-Integral-Derivative feedback control
The workhorse control law of building loops; tuning its gains well is what makes a loop settle instead of hunt.
Closed-loop control
Control that feeds the measurement back into the decision
The essence of feedback - and, at building scale, the essence of a digital twin. Open loop acts without checking.
Sequence of operation
The written logic for how a system behaves in all conditions
The ground truth of intended behaviour; comparing it to trends is where fault detection begins.
ASHRAE Guideline 36
Standardised high-performance HVAC control sequences
Codifies efficient, proven sequences so buildings do not reinvent (and misapply) control logic each time.
Workshop — trace a loop and read it for tuning
This exercise makes feedback control concrete: you will identify the four parts of one real loop and, if you can reach a trend, judge whether it is tuned or hunting. It is the core diagnostic skill of the whole module.
A controlled space and, ideally, any trend graph (BMS history or a smart-thermostat app). A notebook. No specialist software required.
Goal: identify a loop's four parts and assess its behaviour from a trend Inputs: one controlled space (a thermostat-controlled room will do) and, ideally, a BMS trend or a smart-thermostat history graph Time: ~30 minutes
- 1Pick one controlled variable in a real space - a room temperature is easiest. Name its setpoint, its sensor, its controller and its actuator (for a home system: thermostat setting, thermostat sensor, thermostat logic, boiler or AC).
- 2Decide whether it is open or closed loop: does it measure the actual temperature and adjust, or just run on a timer? Justify your answer in one line.
- 3Get a trend if you can - a BMS history or a smart-thermostat app graph over a day. Sketch the temperature curve against the setpoint line.
- 4Read the curve for tuning: does the value rise and settle smoothly, or does it overshoot and hunt (a sawtooth around the setpoint)? Note how often the plant switches - frequent cycling is a tuning clue.
- 5Look for simultaneous conflict if data allows: any sign of heating and cooling both active, or one system fighting another? Describe what you would check next.
- 6Write three sentences: the loop's four parts, whether it is tuned or hunting from the trend, and one thing a digital twin watching this loop could flag that a person in the room would miss.
You’ll walk away with
A one-page loop analysis: the four parts named for one real loop, an open/closed-loop verdict, a sketched trend with a tuned-or-hunting judgement, and one fault a twin could catch from the time-series.
Three altitudes on the same idea
Read the band that fits you — or all three.
You may not tune loops, but you commission them - or fail to. Insisting that sequences of operation are written, reviewed and properly commissioned (not just installed) is the difference between a building that performs and one that quietly wastes energy from day one. Reference a recognised standard such as ASHRAE Guideline 36, and require trend-based commissioning proof, so loops are shown to settle rather than assumed to.
Comfort complaints are usually control problems, not design failures. A room that swings hot and cold, or never quite settles, is often a badly tuned loop or a poor sequence - not your material or layout choices. Knowing this lets you push for the real fix (retuning, resequencing) instead of second-guessing the design, and to set realistic expectations with occupants about how responsive a well-controlled space can be.
Understanding feedback control puts you ahead of most building professionals. The loop - setpoint, sensor, controller, actuator - and the idea of PID and sequences appear in controls, commissioning, analytics and every autonomous-building topic later in this course. You do not need the maths to be valuable; you need to read a trend, spot a hunting loop, and know what a sequence of operation is. That literacy is rare and hireable.
“If the room reaches the right temperature, the control loop is working fine.”
Do it yourself
Reason it through - trace the loop.
- 1Name the four parts of a feedback control loop and what each does.
- 2In one sentence, what is the difference between open-loop and closed-loop control?
- 3Explain the P, I and D terms of PID in plain language - one line each.
- 4What does it mean for a loop to 'hunt', and why is hunting harmful even if the setpoint is met?
- 5What is a sequence of operation, and how would a twin use it to find faults?
The one line to carry out
Peer-reviewed journals & authoritative standards
- 01PID controller — Wikipedia, 2026.
- 02Control theory — Wikipedia, 2026.
- 03HVAC control system — Wikipedia, 2026.
- 04Model predictive control — Wikipedia, 2026.
We have seen how a single BMS controls a building through loops and sequences. Next we ask a harder question: who owns that BMS and its data? Open versus proprietary systems decides whether a twin - and the building owner - can actually get in.
The author
Amogh N P
Architect, interior designer, and creative polymath. Studio Matrx began in his notebooks — his vision of design made honest, useful, and open to everyone. Its Academy is written and taught in his memory, and free, forever.
More about Amogh →