Heater PID Derivative Action: Avoid a Setpoint Kick

System boundary

The temperature-command generator, sampled PID algorithm, sensor input and heater actuator. The printed heater is the controlled load; this review does not establish equipment safety or a universal set of controller gains.

Integration interfaces

Integration risks

Find the signal that is actually differentiated

Write error as e=r−y, where r is requested temperature and y is measured temperature. Differentiating error gives dr/dt−dy/dt. A setpoint step therefore enters the derivative term even if the measured temperature is momentarily unchanged. Differentiating measurement instead uses −dy/dt for the usual heating sign convention, removing that direct setpoint contribution.

This is an algorithm choice, not a property of thick-film heater material. Check the controller documentation or implementation rather than assuming all PID labels mean the same equation. A two-degree-of-freedom controller may apply a separate setpoint weight to derivative action; record that weight explicitly if it exists.

Calculate a command-induced derivative pulse

Consider a deliberately simple discrete derivative without filtering: D=Kd×Δe/Δt. Use an illustrative Kd of 0.4 output-percent-seconds per kelvin, a sample interval of 0.1 second and a 20-kelvin setpoint step while the measured temperature is unchanged. The error derivative contributes 80 output percent in that update. This is a software arithmetic example, not a recommended tuning value or measured heater response.

With derivative on measurement, the same instant contributes zero derivative output if the measured temperature truly did not change. Proportional action can still jump, and a filter or limiter changes the actual derivative waveform. Preserve those distinctions when interpreting a saturated output; a 100 percent command alone does not reveal how much came from each PID term.

Use a measurement disturbance to verify the other sign

For a fixed setpoint, error changes are the negative of measurement changes. The two derivative forms then agree mathematically under the same filter and timing. In the preceding illustrative configuration, a measured rise of 0.5 kelvin over 0.1 second contributes minus 2 output percent in either form. A wiring or software sign error that makes the derivative increase heat during a rapid temperature rise must be corrected before tuning.

Do not create an unsafe physical disturbance to demonstrate the equation. Algorithm verification can start with controlled recorded or simulated input sequences, followed by an approved equipment test. Record which checks were numerical and which involved the installed heater. A simulated step confirms arithmetic, not sensor integrity or thermal protection.

Review noise amplification and derivative filtering

Differentiation responds to rapid measurement changes, including noise. A small difference between adjacent readings divided by a short interval can become a large command term. Derivative on measurement removes the direct setpoint kick but does not remove that sensitivity. Specify the filter form and time constant together with derivative gain; a gain value alone is not a reproducible configuration.

More filtering reduces rapid fluctuations but introduces lag into the derivative path. The useful choice depends on sensor bandwidth, sampling and process dynamics. Avoid choosing a filter solely because the power trace looks smooth. Compare its response to a relevant physical temperature change and retain the protection system's response requirements independently of control-loop filtering.

Separate four causes of a sharp power command

Inspect the internal contributions before changing the heater or tuning all gains at once. A command step, an input glitch and a mode transition can look similar at the limited actuator output. Their corrective actions are not interchangeable, so retain raw target, measurement and PID-term traces around the event.

Keep time units and sample handling explicit

Some implementations store a continuous-time derivative gain and divide by the elapsed sample interval. Others store a coefficient already adjusted for a fixed update period. Moving the same numerical setting between them can change derivative strength. Document the actual equation and units instead of relying on a parameter name such as D or derivative time.

Define behavior for delayed or missing samples. Using a nominal interval for a measurement difference accumulated over a longer time exaggerates the inferred rate. Conversely, treating repeated stale samples as fresh can distort a later update. Retain timestamps and test these cases with controlled data before they are exercised on a powered thermal assembly.

Verify steps and ramps without conflating their purpose

A setpoint ramp can reduce abrupt demand, but it is not proof that the derivative input is correctly selected. Test the mathematical command path with an intentional step and a known ramp while holding the measurement input controlled. Then test the installed system within approved limits using the actual command profile. Record calculated derivative output, limited actuator command and measured heater power where required.

For a ramp, derivative on error includes the target ramp rate as well as the measured rate. That contribution may be intended in a particular design, but it must be deliberate. Do not erase it by changing settings without considering the complete control strategy. The acceptance question is whether the chosen algorithm implements the specified command response.

Release the derivative configuration with its test vectors

Keep the algorithm form, gain units, filter, update interval and setpoint weighting in the controller handoff. Attach test vectors showing a target step, a measured-temperature ramp, a constant input and a delayed sample. Expected sign and magnitude are more useful for future firmware checks than a screenshot of one successful warm-up. Save the independent numerical calculation so settings can be transferred without ambiguous time scaling.

For the heater integration review, provide the resulting actuator waveform and the thermal load rather than claiming a smoother PID trace establishes a higher component rating. Derivative behavior, integral anti-windup, power-source limits and independent overtemperature shutdown remain separate checks. Each should have its own observable result before the assembled equipment is accepted.