A little room to wonder

Feedback & Control

Trace a measured value, a target and a correction rule in a fictional system, including what delay or missing information can change.

Why does measuring the result matter after issuing an instruction?

A wind turbine beside a traditional windpump in a rural landscape.
The approved wind-turbine photograph provides an engineering-system context. Sensors and control rules are not visible, and it does not implement the fictional halving model.Doug butler, via Wikimedia Commons. CC0 1.0. · Source · CC0 1.0

Close the loop with an observation

An open-loop instruction applies an action without using the resulting output to choose the next correction. A feedback loop measures an output, compares it with a target and uses that difference in a decision rule. The returning information closes the loop.

A sensor provides a measurement; a controller applies a rule; an actuator or modelled action changes something. Those roles are distinct. A command saying “move to ten” does not itself prove that the measured value reached ten or that a sensor is accurate.

Define error with a sign convention

For this article, error means target minus measured value. A positive error says the measured value is below target; a negative error says it is above. A correction rule needs to interpret that sign consistently with how its action changes the system.

Negative feedback is a structural idea, not a judgement that feedback is unkind. The correction tends to reduce the relevant difference under the model. Delay, excessive correction, uncertain measurements or changing conditions can still make an actual control system perform poorly.

Worked example: halve a paper difference

An invented model has target 10 units and starts at 6. Its rule adds half the current error each step, with an immediate exact measurement. The first error is 4, so the correction is 2 and the new value 8. Then error 2 gives correction 1 and value 9.

The third correction is 0.5, giving 9.5. Under a declared stopping band of absolute error at most 0.5, that is enough to stop the paper procedure. The exact halving rule approaches 10 without reaching it in a finite number of these ideal steps; an acceptable band and exact equality are different requirements.

Feedback depends on trustworthy timing and boundaries

Suppose a controller acts twice using the same old measurement. It can add more correction than intended before seeing the first action’s result. A trace should show when a measurement was taken as well as its value, instead of assuming every reading is current.

This activity uses written numbers only. It does not control water, heat, a motor or an electrical device. A model can clarify error, update rules and stopping conditions while leaving real equipment design, safety limits and verification outside the exercise.

Only if you feel like trying

Trace a feedback ledger

  1. Make columns for target, measured value, error, correction and next value.
  2. Run the 6 to 8 to 9 to 9.5 example and apply its stated stopping band.
  3. Describe what would become uncertain if a correction used an old measurement instead of a fresh one.

Optional. Use paper or talk through the supplied fictional example. No purchase, personal information, account, public sharing or practical experiment is needed.