Every control loop has a delay between when the controller moves the valve and when the sensor sees the result. Part of that delay is the process lag (the exponential rise of a first-order system), and part is pure dead time — a flat interval where absolutely nothing happens at the measurement. Dead time is the assassin of control loops, and the ratio of dead time to lag determines whether your loop will be easy, hard, or essentially impossible to tune.

Where dead time comes from:

• Transport delay: Fluid traveling down a pipe before reaching the sensor. A 50 m pipe at 1 m/s gives 50 seconds of dead time, period.
• Sensor response: A thermowell with a heavy sheath can add 10–30 seconds before the RTD even begins to move.
• Analyzer cycle time: A gas chromatograph reporting every 4 minutes makes 4 minutes the absolute floor of dead time.
• Computation and scan rate: A PLC scanning every 500 ms adds half a second to every loop.

The controllability rule of thumb uses the dead-time-to-time-constant ratio, θ/τ:

• θ/τ < 0.2 — easy. PI control works beautifully, aggressive tuning is fine.
• 0.2 < θ/τ < 1.0 — moderate. PID helps; tuning needs care.
• θ/τ > 1.0 — dead-time dominant. Standard PID will oscillate unless detuned heavily. Consider a Smith Predictor or model-based control.

Worked example — a sheet dryer: Paper enters a 6 m drying section at 2 m/s. The moisture sensor is at the exit. Dead time θ = 6/2 = 3 seconds before any control action shows up. Suppose the dryer's thermal time constant τ = 4 seconds. Then θ/τ = 0.75 — solidly difficult. A Ziegler-Nichols tuning gives proportional gain Kc ≈ 1.2(τ/θ) = 1.6 and integral time Ti ≈ 2θ = 6 s. Crank Kc higher and the loop will hunt indefinitely, because the controller is reacting to errors that already happened seconds ago.

What engineers actually do about it:

• Move the sensor closer to the actuator. This single change often fixes "untunable" loops.
• Shrink the thermowell or switch to a bare-bulb sensor where safe.
• Use a Smith Predictor: the controller acts on a model-predicted output and corrects against the delayed measurement, effectively tuning out the dead time.
• Detune deliberately. When θ/τ is high, slow controllers are stable controllers.

The single most common tuning mistake is raising the gain to fight sluggishness in a dead-time-dominant loop. You can't outrun a delay — you can only respect it.