Heating is off.
Temperature is still rising.
Heating elements and the lining store heat and continue releasing it into the chamber after power is switched off. A command to switch off at 730 °C does not mean the temperature will stop at 730 °C.
The target temperature is called the setpoint. It stays constant during a hold and changes gradually during a programmed ramp. The controller compares it with the thermocouple reading and controls the heater.
The result depends on heater power, lining mass, load and measurement delay. An empty chamber and a chamber containing heavy workpieces can follow the same profile differently. Opening the door also changes operating conditions. Thermal characteristicsSource summary: heat capacity, heater power, plant dynamics and external disturbances affect the temperature response.Source: OMRON, Technical Guide for Temperature Controllers, Characteristics of the Controlled Object, p. 1.
Hysteresis — for processes that tolerate temperature variation. PID — a starting point for closely following a temperature profile. ADRC — an approach to evaluate when conditions and disturbances change.
Hysteresis
In a simple hysteresis mode, the heater switches between on and off. Different temperature thresholds are used for switching on and switching off.
An illustrative example: switch heating off at 730 °C and back on at 728 °C. The 2 °C difference prevents repeated switching caused by small changes in the reading. The position of the thresholds relative to the setpoint depends on the controller implementation. Why two thresholds are usedSource summary: a temperature band between the on and off thresholds reduces frequent relay switching near the setpoint.Source: OMRON, Technical Guide for Temperature Controllers, Hysteresis, p. 4.
No PID gains are required: set the switching band and check the result. A narrow band does not guarantee accuracy. Thermal inertia can produce an actual overshoot larger than the selected hysteresis band.
When to choose it: the process tolerates variation and a trial cycle confirms the required temperature range. Drying with a broad tolerance is one example.
Hysteresis and thermal inertia
Setpoint 730 °CChange inertia, the switching band and the lead factor: the thresholds and actual temperature respond differently.
Teaching model near the setpoint. The lead factor illustrates inertia compensation; its values are not Kk settings for Nellin firmware.
PID: reduce heating
before reaching the target
A PID controller calculates its control output from the deviation between the measured temperature and the setpoint. The calculation combines three terms.
- P
- Current error. How far the actual temperature is from the target.
- I
- Accumulated error. Accounts for how long the deviation persists and helps remove a steady temperature shortfall.
- D
- Rate of change. Responds to changes in the error or measured temperature, depending on the implementation.
With suitable tuning, PID reduces power before the setpoint is reached. This helps limit overshoot and maintain the temperature during a hold. PID termsSource summary: the proportional term adjusts output according to the error, the integral term removes steady offset, and the derivative term accounts for signal changes.Source: OMRON, Overview of Temperature Controllers, P Action, I Action, D Action and PID Control.
When a solid-state relay is used, “30% power” usually means heating for part of each control period. In an illustrative ten-second cycle, this would be three seconds on and seven seconds off. The average power delivered to the heater changes.
When to choose it: programmed ramps, temperature holds and repeatable cycles. For a muffle furnace performing these tasks, PID is usually the starting point.
Temperature and heating pulses
Setpoint 730 °CAs the temperature approaches the target, the relay spends less time on. Change the illustrative period to see the difference between switching pulses and average power.
An illustrative tuned-PID response, not a calculation using Kp, Ki and Kd. The period is stretched for visibility; temperature is defined separately from the pulses. This is not a recommended PWM period.
PID depends on tuning. Aggressive settings can cause oscillation; conservative settings can delay reaching the target. Prolonged operation at maximum power can cause the integral term to build up. Controllers use anti-windup mechanisms to address this. Integral windupSource summary: when the output is limited, the integrator can continue accumulating error. Anti-windup helps the controller recover more quickly from saturation.Source: MathWorks, Anti-Windup Control Using PID Controller Block, Performance Without Using Anti-Windup.
Autotuning helps select gains, but the result must be checked with the working profile. After changing the heating elements, insulation or load, the previous tuning may need adjustment.
ADRC: estimate
what has changed in the system
Active Disturbance Rejection Control uses an observer: an algorithm that estimates system behaviour from the temperature and the control signal.
The total disturbance estimate includes external effects and plant dynamics that a simplified model does not describe accurately. The controller uses this estimate to adjust its output. The ADRC observerSource summary: an extended state observer estimates the total disturbance, including unknown dynamics and external effects. This estimate is used for compensation.Source: MathWorks, Active Disturbance Rejection Control, Controller Structure.
Changes in load or heat loss can be a reason to evaluate ADRC. For example, tuning may work with a small batch but produce a different transition to a hold with heavier workpieces. This is a scenario for comparing approaches, not a guarantee that ADRC will perform better.
When to evaluate it: changing conditions prevent a tuned PID controller from meeting the process requirements. Assess any advantage through recorded responses and repeatability.
Response to increased heat loss
Setpoint 730 °CHeat loss increases at minute 40. Change the additional loss and observer speed, then follow the temperature and relay activity.
A simplified linear ADRC model with an observer and pulsed output. Parameters and timing are illustrative. This chart does not establish ADRC availability in Nellin controllers.
A detailed mathematical model is usually unnecessary, but an approximate understanding of how the kiln responds to power is required. A common linear implementation uses an estimated input gain, controller bandwidth and observer bandwidth.
Increasing observer speed helps compensate for disturbances, but also increases sensitivity to measurement noise and can reduce stability. ADRC requires tuning too. ADRC tuning and limitationsSource summary: linear ADRC requires the plant order and an estimate of the input gain. A faster observer makes the controller output more sensitive to high-frequency noise.Source: Gernot Herbst and Rafal Madonski, Active Disturbance Rejection Control, Springer, 2024; chapters 3 and 5, sections 3.1.1 and 5.3.3.
Start with the process
requirements
Base the final choice on the acceptable temperature deviation and the results of a trial cycle.
| Operating conditions | Starting point | What to check |
|---|---|---|
| A simple hold with a broad tolerance | Hysteresis | Variation and overshoot after switching off |
| Programmed ramps and precise holds | PID | Profile tracking and the transition to a hold |
| Consistent load, repeatable cycles | Tuned PID | Repeatability over several runs |
| Load or heat loss changes significantly | Compare PID and ADRC, if available | Deviations under different operating conditions |
| A mechanical contactor switches the heater | Check the permitted switching frequency | Contactor life and control settings |
Frequent pulsed control typically uses a solid-state relay, or SSR. Mechanical contacts have a limited switching life. An SSR supports frequent switching, but requires correct selection and heat dissipation. Power switchingSource summary: an SSR has no wearing mechanical contacts and is suitable for frequent switching. Semiconductor losses require heat dissipation.Source: OMRON, Overview of Solid-state Relays, Features and ON/OFF Control.
The charts show different teaching scenarios. Their scales adapt to the data. They are not a comparison of three controllers tested on the same kiln.
Evaluate the full cycle
A short, flat section of a chart is not enough. Check the ramp, the transition to a hold and an extended hold.
- Define acceptable deviations. Specify permitted overshoot, temperature variation and unacceptable undershoot.
- Test the working profile with a representative load. An empty chamber does not always reflect working conditions.
- Repeat the cycle. Compare maximum overshoot, settling time and temperature variation.
- Change one parameter at a time. Keep the previous settings so that the comparison remains clear.
If PID meets the required tolerance with all working loads, a move to ADRC should address a specific, measured problem.
Every controller is limited by the equipment. If the heater stays at maximum power while the kiln falls behind the profile, check heater capacity, heat loss and the requested ramp rate. Without active cooling, switching heating off cannot make the kiln cool at an arbitrary rate.
Control settings in Nellin
The Nellin instructions describe PID and Hyst. PID provides AutoTune to select gains automatically. The hysteresis settings include Kk for thermal inertia compensation; it is adjusted experimentally. Check the available settings against the device model and firmware version.
Controller autotuning and measurement calibration serve different purposes. The former helps control heating; the latter corrects temperature readings. Stable numbers on the display do not establish uniform heating throughout the load.
ADRC is discussed here as a separate control approach. Check the documentation for availability in a particular Nellin model.
Choose a controller based on operating temperature, thermocouple type and the requirements of your kiln.
Choose a temperature controller →