Control-System Elements: Roles, Signal Paths & Hierarchies

A physical automation system carries two different things around one loop: information moves from the process toward the controller, while energy moves from a supply through a power stage and actuator into the process. Confusing those paths produces bad boundaries: a 3.3 V command is not motor power, and a sensor reading is not the physical quantity itself. Modern systems keep sensing, conditioning, decision, power-conversion, and actuation roles distinct even when several share one enclosure.

The functional chain #

The same box can fill several rows. A smart pressure transmitter contains a sensing element, excitation, analog conditioning, conversion, calibration data, diagnostics, and a communications endpoint. A servo drive contains current sensing, control loops, a power converter, and protection. The roles remain separate because their constraints remain separate.

Signal path and energy path #

The diagram makes the split physical. The information path is approximately

$$x(t) \rightarrow \text{sensor} \rightarrow \text{conditioning} \rightarrow y[k] \rightarrow \text{controller} \rightarrow u[k]$$

while the energy path is

$$\text{supply} \rightarrow \text{converter commanded by }u[k] \rightarrow \text{actuator} \rightarrow \text{process}.$$

The paths meet at two places. At the sensor, measurement extracts information while inevitably loading the process by some amount. At the actuator, the low-power command gates a much larger energy flow. This is why gain alone is not enough to describe an automation element: loading, isolation, saturation, bandwidth, power loss, and failure behavior matter just as much.

Worked trace: a pressure-control disturbance #

Consider a variable-speed pump holding a pipe at 4.0 bar. The pressure transmitter is scaled for 0–10 bar and produces 4–20 mA; a 250 Ω input resistor converts that current to 1–5 V for an isolated ADC. The controller samples the result every 10 ms, compares it with the 4.0 bar reference, and sends a torque-producing command to a variable-frequency drive. The drive draws three-phase electrical power, controls motor current, and the motor-pump set converts that power into pressure and flow.

Now an outlet valve opens:

1. Flow demand rises, so stored pressure in the pipe falls. This is the process disturbance; it enters in litres per second, not volts.
2. At 3.8 bar the ideal transmitter current is $4\text{ mA}+0.38(16\text{ mA})=10.08$ mA, or 2.52 V across 250 Ω. This is a measurement of pressure, not pressure itself.
3. The controller sees a $+0.2$ bar error under the convention $e=r-y$. Its PI law increases the torque or speed request, subject to rate and current limits.
4. The drive uses the low-power command to gate semiconductor switches. The command carries information; energy comes from the mains and DC link.
5. Motor current produces torque, pump speed rises, and delivered flow increases until pipe pressure returns toward 4.0 bar.

The trace exposes failures that a one-line block diagram hides. A plugged impulse line can make the transmitter report stale pressure; an ADC input can saturate; the PI integrator can wind up against the drive current limit; the motor can be thermally unable to sustain the requested torque; and the pump can reach its hydraulic envelope without restoring pressure. Each fault belongs to a different role even though all are observed as “pressure did not recover.”

Local loop versus plant hierarchy #

ISA-95 places the physical process at Level 0, sensing and manipulation at Level 1, and monitoring and supervisory control at Level 2. That is a hierarchy of activities and responsibilities, not a demand that every message climb a pyramid. A networked sensor may publish directly to analytics, while the safety-relevant control action still closes locally through deterministic I/O.

Interfaces are contracts #

At every boundary, specify more than a connector or protocol:

1. Quantity and units: what physical or logical quantity crosses the boundary?
2. Range and direction: what values are valid, and what sign convention is used?
3. Static behavior: gain, offset, nonlinearity, hysteresis, saturation, and loading.
4. Dynamic behavior: bandwidth, delay, sample period, jitter, and settling time.
5. Failure behavior: open circuit, stale value, loss of excitation, overrange, and safe state.
6. Authority: which component may command energy, reset a trip, or change a setpoint?

This contract is the bridge between feedback-control theory and hardware. A mathematically stable controller is not a working system if its measurement arrives late, the converter clips its output, or the actuator cannot produce the requested force.

Active, passive, continuous, and discrete #

The textbook's classifications are still useful if interpreted functionally:

• An active element controls energy drawn from an auxiliary source; an amplifier or power converter belongs here. A self-generating transducer instead converts energy from the measurand into its output, but it can still load the measured process.
• A continuous element has a continuously varying output over its operating region. A discrete element changes state at thresholds; relays and limit switches usually add hysteresis so noise does not chatter them.
• A reversible element encodes input sign or phase in output polarity, phase, or direction. A non-reversible element reports magnitude or one-sided state only.
• A linear model is local. Saturation, dead band, friction, magnetic hysteresis, quantization, and contact bounce remain present even when the nominal transfer law is linear.

Design review checklist #

For each loop, trace one representative disturbance and one representative command through every role. At each transition record units, limits, timing, energy source, loading, and failure state. If a boundary cannot answer those questions, the architecture is not yet a control-system architecture; it is only a block diagram.