Automatic control systems
Controlling means measuring, comparing and correcting, over and over. In industrial electronics what gets corrected is almost always power: how much energy is delivered to a furnace, a motor or a heating line, and by what method it is delivered.
01Open loop and closed loop
An action is applied and it is trusted to produce the expected effect. A knob dimmer, an oven with its heating element at 50 % on a timer: nobody measures the result.
It is simple and cheap. It fails as soon as anything changes: the mains voltage, the load, the ambient temperature.
The variable is measured and compared with the setpoint, and the difference —the error— governs the action. The system corrects itself in the face of disturbances.
This is what turns a power circuit into a control system, and what introduces the new problem: stability.
- Setpoint: the value you want.
- Process variable: what is measured.
- Error: setpoint minus measurement.
- Manipulated variable: what the controller adjusts —the firing angle, the duty cycle—.
- Disturbance: anything that affects the process without the controller deciding it: a door was opened, cold material came in, the mains voltage dropped.
- Transient and steady-state behavior: how it gets there and how it stays.
02Types of control
| Type | How it decides | Result |
|---|---|---|
| On-off | Turns on or off at two thresholds | Always oscillates around the setpoint. Simple and sufficient for many things (control with devices). |
| Proportional (P) | Action proportional to the error | Does not oscillate, but leaves a steady-state error: if the error were zero, the action would be too. |
| PI | Accumulates the error over time | Eliminates the steady-state error. It can overshoot and become slow. |
| PID | Adds the rate of change of the error | Anticipates and brakes before arriving. It is the industry standard, and is covered in Year 7. |
| Time-proportional on-off | Turns on for a percentage of each cycle time (for example, 7 s out of every 10) | With a two-state actuator a proportional effect is achieved. It is what many furnace controllers do. |
If the action is output = K · error, for the furnace to deliver 40 % power
—what it needs to make up for the losses— the error has to be nonzero. With zero error,
the output would be zero and the furnace would cool down. That permanent deviation is called offset, and it is
exactly what the integral term resolves.
03Single-phase power control
There are two ways to deliver a percentage of the mains power to a load, and the choice is not cosmetic: it changes the level of interference and which loads can be used.
| Phase-angle control | Integral-cycle control | |
|---|---|---|
| How it regulates | By delaying the firing within each half-cycle | By letting N complete cycles out of every M through |
| Resolution | Continuous | In steps: with a 100-cycle base, 1 % |
| Interference | High: current step 100 times per second | Practically none: it always switches at zero |
| Visible flicker | No | Yes, in lamps |
| Suitable for | Lighting, universal motors, fast loads | Furnaces, heating elements, any load with thermal inertia |
| Power factor | Worsens it: shifts and distorts | Preserves it: each cycle is a whole sine wave |
A 2000 W resistive load at 220 V.
- By phase angle: you have to solve the RMS voltage equation to reach 60 % of the power; it gives a firing angle of approximately 81°. The current has a sudden step one hundred times per second.
- By integral cycles: 6 cycles out of every 10 are let through (120 ms base). The current is a pure sine wave while it conducts.
- In both cases the load receives 1200 W on average. The difference lies in the quality of the mains and in whether flicker matters.
04Three-phase power control
With powers above a few kilowatts, the load is split among the three phases. That improves the balance of the installation but complicates the control: you have to synchronize three firing pulses phase-shifted by 120°.
- A pair of thyristors in antiparallel —or a triac— per phase.
- The load can be wye- or delta-connected (three-phase circuits); the connection changes the currents through each switch.
- With an accessible neutral, each phase can be controlled separately; without a neutral, the three affect each other and the analysis is considerably more complex.
- Synchronization is taken from the mains itself, with a zero-crossing detector per phase.
- Phase sequence: if it is reversed, the firing pulses come out in the wrong order.
- Imbalance: if one phase does not fire, the other two draw more current and the load heats unevenly.
- Harmonics: three simultaneous phase-angle controllers inject a lot of distortion. In large furnaces integral-cycle control is preferred precisely for that reason.
- Isolation: the firing pulses are referenced to different phases, so each one needs its own isolation: pulse transformer or optotriac. The three cannot be driven from a common ground.
- Three heating elements in wye on 380 V: each receives 220 V and consumes 5 kW.
- Current per phase: 5000/220 = 22.7 A. Thyristors rated 50 A and 800 V are chosen, with heatsink and forced-air cooling.
- Integral-cycle control with a 100-cycle base (2 s): 1 % resolution, no harmonics and no perceptible flicker on a thermal load.
- Closed loop with a type K thermocouple and a PID controller, which decides what percentage to request.
- Independent protection: manual-reset safety thermostat and ultrafast fuses per phase.
05The firing circuit
- Synchronization with the mains: the angle is counted from the zero crossing of that phase.
- Isolation between control and power: pulse transformer or optocoupler.
- Sufficient energy in the pulse: gate current above the specified value, for the time needed for the thyristor to latch.
- Pulse train for inductive loads: with a single short pulse, the current may not have reached the holding current when the pulse ends and the thyristor turns off. A series of pulses is sent until the end of the half-cycle.
The classic circuit —a UJT or a dedicated IC such as the TCA785— still works and does not depend on software. A microcontroller is more flexible: it allows start-up ramps, current limits, communication and diagnostics. What does not change is that safety cannot depend on the program: the interlock and the protections are hardwired all the same.
06In the lab
With a lamp and a light sensor, regulate the brightness first in open loop —fixed PWM setting— and then in closed loop with the sensor. Partially cover the lamp: in open loop the measured light drops and stays down; in closed loop, the system raises the power and recovers. Plot the two responses.
Program a pure proportional controller on a small thermal load and note the steady-state error for three gain values. You will see that the error decreases but never reaches zero, and that with very high gain the system starts to oscillate. Add the integral term and compare.
With an isolation transformer and reduced voltage, control a lamp with phase-angle control and a heating element with integral-cycle control at the same power percentage. Compare the line current on the oscilloscope, and bring an AM radio close: the interference from phase-angle control can be heard, that from integral-cycle control cannot.
Fire a thyristor on an inductive load with a short pulse and observe that it sometimes does not latch. Switch to a pulse train and verify that it now always conducts. It is the most direct way to understand the holding current.
07Common mistakes
| Symptom | Usual cause |
|---|---|
| The system oscillates around the setpoint | Gain too high, or process delay greater than expected. |
| It never reaches the setpoint | Pure proportional control: steady-state error. Integral action is missing. |
| It corrects well but takes very long | Integral action too slow, or undersized actuator. |
| One phase of the furnace heats less | A thyristor is not firing: check the pulse, isolation and synchronization of that phase. |
| Interference on the radio and on neighboring equipment | Phase-angle control without a filter. Switch to integral-cycle control if the load allows it. |
| The lamp flickers annoyingly | Integral-cycle control on a load with no inertia. Phase-angle control is the right choice there. |
| The thyristor does not latch with an inductive load | Pulse too short: a pulse train is needed. |
| Everything works on the bench and fails in the plant | Noise on the synchronization or sensor signal. Shield, filter and separate the grounds. |
08Self-assessment
What distinguishes a closed loop from an open one?
The measurement of the result. In closed loop the variable is measured, compared with the setpoint, and the error governs the action; the system corrects itself in the face of disturbances.
Why does a pure proportional controller leave a steady-state error?
Because its output is proportional to the error: to sustain a nonzero action, the error cannot be zero. That deviation is eliminated with integral action.
When is integral-cycle control better, and when phase-angle control?
Integral cycles for loads with thermal inertia —furnaces, heating elements—: it generates no interference. Phase-angle control for lighting and fast loads, where the flicker of integral cycles would be visible.
A 15 kW furnace in wye on 380 V: what is the current per phase?
Each element receives 220 V and consumes 5 kW: I = 5000/220 = 22.7 A per phase.
Why does each phase of a three-phase controller need its own firing isolation?
Because the thyristors of each phase are referenced to different potentials. Driving them from a common ground would short the phases together through the control circuit.
What is a pulse train and what is it used for?
A series of firing pulses repeated until the end of the half-cycle. It is used with inductive loads, where the current rises slowly and with a single pulse it may not reach the holding current before the pulse ends.
What is a disturbance and why does it matter?
Any change that affects the process without the controller deciding it: cold material coming in, an open door, the mains voltage dropping. It is the reason closed loops exist.
The control works well and suddenly starts to oscillate when the load changes. What happened?
The process dynamics changed: with less thermal mass it responds faster and the gain that used to be adequate is now excessive. The parameters must be retuned for the new operating point.
Why does phase-angle control worsen the power factor?
Because the current stops being sinusoidal and is also shifted with respect to the voltage: both a displacement component and a distortion component appear. Capacitors correct the first but not the second.
Can software replace hardwired protections?
No. The program can hang, the sensor can disconnect and the thyristor can fail shorted. Safety is handled with independent devices: manual-reset thermostat, fuses, interlocks.