A single PID loop works when your controlled variable responds directly to your manipulated variable. But many real processes have a slow outer dynamic driven by a faster inner one — and trying to control both with one loop produces sluggish response and poor disturbance rejection. Cascade control nests two PID loops: an outer "primary" loop sets the setpoint for an inner "secondary" loop.

The classic example: temperature control via steam. You want to hold a reactor at 80°C by injecting steam through a control valve. A single loop measures reactor temperature and modulates the valve. Problem: if steam supply pressure fluctuates, the valve position no longer corresponds to a predictable flow rate, and the reactor temperature wanders before the controller notices. Cascade fixes this:

• Outer loop (slow): measures reactor temperature, outputs a desired steam flow as its setpoint.
• Inner loop (fast): measures actual steam flow, modulates the valve to hit that setpoint.

Now a supply pressure dip gets corrected in seconds by the inner flow loop, long before the reactor temperature drifts.

When cascade pays off:

• There's a measurable intermediate variable between your final output and your actuator.
• The inner loop disturbances are faster and larger than outer loop disturbances.
• The inner loop dynamics are at least 3–5× faster than the outer loop. This is the rule of thumb: if your inner loop time constant is 10 seconds, your outer loop should have a time constant of 30–50 seconds or more. Violate this and the loops fight each other, causing oscillation.

Tuning order matters. Always tune the inner loop first with the outer loop in manual. Then close the outer loop and tune it while the inner loop runs in auto. Reversing this order leaves the outer loop chasing a moving target.

Other common cascades:

• Position → velocity → current in servo drives (three nested loops, each ~5× faster than the one outside it).
• Level → flow in tank control, where downstream flow disturbances would otherwise upset level.
• Temperature → jacket temperature → coolant flow in jacketed reactors.

Watch the failure modes. If the inner loop saturates (valve fully open or closed), the outer loop's integral term can wind up. Most modern controllers provide "external reset feedback" or "back-calculation" anti-windup that uses the inner loop's actual output to prevent this. Without it, recovery from a saturation event can take many minutes — exactly when you can least afford it.