This is part 2 of my series on control loop delays. All examples come from real projects I've worked on.

Last time we learned a sobering truth: max bandwidth ≈ 1 / (10 × delay).

In digital control, you always have at least one sample of delay—so faster sampling helps. But there's more you can do.

1️⃣ The Simple Wins: Reduce Delay Directly

Start with the obvious:

  • ⚡ Faster communication bus — CAN-FD, EtherCAT, or direct SPI instead of slow serial
  • 🖥️ Faster processor — less computation time = less delay
  • 🔧 Optimize your code — sometimes a few microseconds matter

These are low-hanging fruit. Grab them first.

2️⃣ Loop Cascading: The Secret Weapon

When your sensor is slow, don't fight it—work around it.

Real example: An air pressure controller with a sluggish pressure sensor.

Cascaded Control Loop with Feed-Forward

The solution: 1. Inner loop — controls torque and speed of the fan (fast!) 2. Outer loop — controls pressure (slow sensor is fine here)

But here's where it gets clever: add feed-forward.

Based on operating conditions and desired pressure, a lookup table calculates the required torque directly. The response becomes nearly instantaneous. The slow pressure sensor? It only needs to clean up the residual error.

💡 Key insight: Cascading lets you hide slow sensors behind fast inner loops.

3️⃣ Better Sensors (Or Smarter Sensor Fusion)

Sometimes you need a faster sensor. But what if faster sensors don't exist—or cost too much?

Combine sensors.

Real example: Indoor drone tracking using optical image recognition.

The problem:

  • ❌ Optical tracking was slow
  • ❌ Sometimes produced outliers (lost track, wrong detection)

The solution: Fuse optical data with a fast IMU using a Kalman filter.

Results:

  • ✅ Much lower effective delay
  • ✅ Built-in outlier rejection using Mahalanobis distance
  • ✅ Smooth tracking even when optical updates are sparse

💡 Bonus: Sensor fusion often improves accuracy and speed simultaneously.

🎯 The Takeaway

When you hit the bandwidth wall, you have options:

Strategy When to use
Reduce delay Always try this first
Loop cascading Slow outer sensor, fast inner dynamics available
Sensor fusion Fast sensor exists, but lacks accuracy alone

Which of these have you tried? What worked (or didn't) in your systems?

Next up in this series: Reduce delay with a lead-lag filter.

#ControlSystems #Embedded #Engineering #DSP #RealTimeControl #Kalman