In this seminar, Process Automation Hall of Fame member Greg McMillan delivers a masterclass on the foundational concepts that drive process control improvement. His central thesis is both practical and unifying: nearly all loop performance problems trace back to a handful of core concepts, and understanding how they interact gives engineers a far more effective framework than fragmented, experience-only knowledge.

Why It Matters

Most process control knowledge is “detailed, fragmented, and experience-driven,” Greg observes. There is no widely shared picture of where improvement opportunities come from or how large they might be. This seminar attempts to fill that gap with a unified set of concepts that apply across industries, loop types, and equipment. Whether a plant is dealing with startup transitions, ambient disturbances, or aging instrumentation, the same small set of principles explains the vast majority of performance limitations.

Key Takeaways

• Delay, speed, and gain account for roughly 90% of process control improvements. Every other issue, including sensitivity, resolution, backlash, noise, and oscillations, matters, but these three dominate.
• Dead time is the single most important limiter. The loop cannot correct what it has not yet seen. Minimizing total loop dead time, from every source, is the highest-leverage design and maintenance activity.
• Disturbance speed is your friend when it’s slow. Real-world upsets are rarely pure step changes. Slowing disturbances down through mixing, volume, and continuous (rather than on/off) control dramatically reduces the error the loop must handle.
• Tuning must match design intent. A plant that invests in reducing dead time through better process design but then tunes controllers sluggishly will never see the benefit. Slow tuning introduces an “implied dead time” that masks the potential for real improvement.

Delay: The Universal Constraint

Greg opens with what he calls the most prevalent limiting concept: delay, or more precisely, total loop dead time. “Without dead time, I’d be out of a job,” he quips. Dead time is the interval before a change completes the full loop and returns to its point of origin. Until that interval passes, the controller is blind to the disturbance and powerless to correct it.

Sources of delay are everywhere. Transportation delays exist in piping, ducts, plug-flow reactors, conveyors, and extruders. Digital devices contribute roughly half their scan or update interval as effective dead time. Analyzers are worse: because results appear at the end of the analysis cycle, the effective delay is approximately 1.5 times the total sample processing and analysis cycle time. Even seemingly minor contributors, such as transmitter damping defaults or aging pH electrodes whose time constants can jump from two seconds to 100 or 200 seconds, can become dominant delays if left unchecked.

Sensitivity and resolution limits also create hidden dead time. Because controller outputs typically ramp rather than step, the time it takes to ramp through a sensitivity, resolution, or backlash band translates directly into additional delay.

Speed: Rate of Change in the Critical Window

The second core concept is speed, which Greg defines as the rate of change of the disturbance, and the process and controller response, particularly in the first four dead-time intervals after a disturbance or set-point change. By the end of that window, the control loop should have completed most of its correction.

High-speed disturbances and slow control systems both degrade performance. Step disturbances, the kind produced by batch operations, on/off control, and manual operator actions, are the worst case. Converting batch operations to fed-batch, replacing level switches with continuous level control, and engaging the controller as early as possible during startups all reduce disturbance speed. Greg notes that variable-speed drives can be slower than expected due to velocity limits imposed by the drive electronics and user-set parameters, a factor often overlooked.

On the process side, slower is better. Larger volumes and better mixing (from agitation, boiling, diffusion, or even simple residence time) act as natural low-pass filters, attenuating disturbances before they reach the measurement point.

Gain: Nothing Works Without It

Gain exists at every point in the loop: controller, valve, process, disturbance, and measurement. Zero gain anywhere is catastrophic, and near-zero gain, such as a rotary valve operating at 80 to 90 percent open on the flat portion of its installed characteristic, is nearly as bad. Greg stresses that inferential measurements, such as temperature-to-composition relationships on distillation columns, must have sufficient gain to be useful. On the measurement side, narrowing the calibration span (100 percent divided by the span) increases effective gain and reduces error.

The Implied Dead Time from Slow Tuning

One of McMillan’s most practical insights is that sluggish tuning introduces an implied dead time that masks the loop’s true dynamics. If a controller is tuned slowly enough, adding filters, delays, or longer scan times produces no visible degradation, giving engineers a false sense that those additions are harmless. Any analysis of the effect of added delays or lags must account for the controller’s tuning aggressiveness. The fastest achievable tuning, where the closed-loop time constant approaches the dead time, serves as a useful benchmark even if the operating tuning is intentionally slower.

Attenuation and the Real World

Greg closes by reinforcing that real disturbances are almost never step changes. By estimating how far a disturbance travels during the dead-time interval using an exponential approximation, engineers can arrive at a much more realistic assessment of the load error the loop actually faces. Volumes and mixing attenuate oscillations; faster controller tuning shifts the break frequency, providing even more attenuation. Understanding these relationships lets engineers size tanks, select equipment, and tune controllers with a coherent, quantitative framework rather than rules of thumb.

Visit Greg’s many posts here on the blog to deepen your process control skills.

Note: The websites referenced at the end of the seminar presentation are no longer available.