Why a hot-water reset schedule and a discharge-air PID loop working together isn't two separate control problems — it's one cascade, and the difference matters for tuning, troubleshooting, and stability.
A single-loop controller measures one process variable, compares it to one setpoint, and drives one final control element. That's the AHU discharge-air-temperature loop most engineers learn first: measure the air leaving the coil, compare to a fixed 55°F setpoint, modulate the valve. Cascade control looks similar at a glance — it also ends in a PID loop driving a valve — but it chains two (or more) loops together, where the output of an outer, slower loop becomes the setpoint of an inner, faster loop. The distinguishing feature isn't how many valves or sensors are on the drawing. It's whether one loop's setpoint is fixed by the engineer, or continuously recalculated by another loop above it.
The outer loop in a cascade (here, outdoor-air reset) typically has no direct feedback path of its own back from the final control element — it only computes a setpoint and hands it to the inner loop, which does the actual fast, tightly-tuned regulation against a real sensor. This separation is why cascades are stable where a single "do everything" loop would not be: the inner loop can be tuned aggressively for a fast disturbance (a sudden supply-fan speed change) because it's only ever chasing a slowly-moving target, and the outer loop can be tuned gently because it only needs to track a slow trend (outdoor temperature) rather than reject fast disturbances directly. In smart-buildings practice, resets are the most common form this takes: OA-reset discharge-air-temperature control, static-pressure reset driving VAV box control, and hot-water-supply-temperature reset are all cascades, even though most BAS graphics never label them that way.
Close, but the precise requirement is stricter than just "one number depends on another." A true cascade requires that the inner loop is itself a closed feedback loop with its own live measurement — the outer loop only ever adjusts that inner loop's setpoint, never touching the final control element directly. A simple linear reset that maps outdoor temperature straight to a chilled-water valve position, with no discharge-air sensor feedback at all, is not a cascade — it's actually an open-loop (feedforward) strategy layered on top of nothing, because there is no inner closed loop being adjusted. The giveaway question: does the "inner" stage have its own sensor and its own error calculation, or does the outer schedule just compute a value that goes straight to the actuator? Only the first case is genuinely cascade control; the second is feedforward control, a related but distinct strategy that reacts to a measured disturbance before it affects the process rather than adjusting an inner loop's target.
Explains how cascade control differs from single-loop control in building automation — an outer, slower loop computes the setpoint fed to an inner, faster closed loop — using a side-by-side comparison of fixed discharge-air-temperature control (single-loop) and outdoor-air reset driving discharge-air setpoint (cascade).
Cascade loops look, at a glance, like an ordinary single loop with an extra input somewhere upstream — both end in a PID block driving a valve or damper. The confusion is compounded because BAS graphics rarely draw the nested-loop structure explicitly; they usually just show a "setpoint" field next to a reset schedule table, with no visual indication that the schedule itself is functioning as an outer control loop. The structural test is whether the inner stage has its own real feedback (a live sensor and its own error calculation) that the outer stage only ever retargets — versus the outer computation feeding straight through to the final control element with no independent inner feedback at all.
In cascade control, two or more control loops are nested. The outer (primary/master) loop measures a slower or more indirect variable and computes a setpoint — not a direct output — for the inner (secondary/slave) loop. The inner loop is a complete, independently closed feedback loop: it measures its own process variable, compares against the setpoint handed down by the outer loop, and drives the final control element. Only the inner loop touches the actuator directly.
This nesting is valuable specifically because the inner loop can be tuned fast and tight against a variable that responds quickly (discharge air temperature reacts to valve position within seconds), while the outer loop only needs to track something that changes slowly (outdoor air temperature changes over minutes to hours) — so it can be tuned gently without hurting overall stability. A textbook non-HVAC example is a boiler drum-level cascade: an outer level loop sets the setpoint for an inner steam-flow/feedwater-flow loop, because feedwater flow responds much faster than drum level itself.
Cascades are everywhere in ASHRAE Guideline 36-style optimized control sequences, even though the term "cascade" rarely appears on the drawings. Discharge-air-temperature reset based on outdoor air temperature or building load, duct static-pressure reset based on the most-open VAV damper, hot-water or chilled-water supply-temperature reset based on outdoor air temperature or valve position, and even simple economizer changeover logic feeding a mixed-air-temperature loop are all cascades in the strict sense: an outer calculation continuously recomputes the inner loop's setpoint. Recognizing a sequence as a cascade matters for troubleshooting — instability or hunting can originate in either loop, and the fix is different depending on which one is misbehaving. An inner loop that hunts needs PID retuning against its own sensor; an outer loop that hunts (rapidly changing the setpoint it hands down) destabilizes the inner loop indirectly and needs its reset rate limited or its own gains reduced, not the inner loop's tuning touched at all.
No. It is a cascade only if the value being reset is itself the setpoint of a real, independently closed inner loop with its own sensor feedback. If the reset computation feeds straight through to a valve or damper position with no inner sensor loop at all, that is feedforward or open-loop scheduling, not cascade control — the giveaway is whether an inner feedback loop with its own error calculation actually exists.
That is exactly what a cascade is — the "elsewhere" calculation is the outer loop. Structuring it explicitly as two loops (rather than one loop with an ad hoc setpoint formula bolted on) is valuable because it lets each loop be tuned and diagnosed independently: PID gains for the fast inner loop don't need to account for the slow outer dynamics, and the outer reset logic can be as simple as a linear schedule without needing its own PID tuning at all.
Yes, though it is uncommon in typical HVAC/BAS applications. Process industries occasionally nest three loops (e.g., a temperature loop setting a flow loop's setpoint, which in turn sets a valve-position loop's setpoint) when the intermediate dynamics genuinely differ enough to benefit from separate tuning at each stage. In smart buildings, two-level cascades (a reset schedule feeding a PID loop) cover the overwhelming majority of real cases.
The inner loop is chasing a moving target rather than a stable one, which can produce oscillation or hunting that looks like a tuning problem in the inner PID loop but actually originates upstream. This is why reset schedules are usually rate-limited (e.g., no more than a few degrees of setpoint change per minute) — slowing the outer loop's effective response protects the inner loop's stability without touching its gains at all.
Try our Smart Buildings Studio
More calculators, simulators, and guides for this discipline.
Building Management System Use Case (Full Access)
✨ Premium ContentA zoomable interactive reader — free preview, then unlock the full set.