Open vs Closed Loop Piezo Stages: Performance and Control

Introduction

Piezoelectric ceramic actuators are widely used in nanopositioning applications due to their sub-nanometer resolution and fast response. However, their inherent nonlinearities — hysteresis and creep — significantly degrade positioning accuracy when operated in open-loop mode. Closed-loop control with high-resolution feedback sensors, particularly capacitive sensors, is essential to achieve linear, repeatable motion. This guide examines the performance differences between open-loop and closed-loop control, explains how capacitive feedback compensates for nonlinear effects, and provides practical engineering guidelines for PID parameter tuning and feedforward compensation algorithms.

Hysteresis and Creep in Open-Loop Operation

In open-loop mode, the displacement of a piezoelectric stack is approximately proportional to the applied voltage, but hysteresis causes up to 10–15% position error depending on the voltage range and excitation history. This path-dependent behavior makes open-loop positioning unsuitable for precision applications requiring better than 1% accuracy. Additionally, creep — a slow drift in position under constant voltage — introduces further instability, especially after rapid voltage changes or during long-duration holds.

For example, a typical open-loop piezo stage may exhibit a positioning error of several hundred nanometers when moving to a target position after a previous movement in the opposite direction. Such nonlinearities are predictable mathematically but cannot be fully eliminated without feedback.

Capacitive Feedback for Closed-Loop Control

Capacitive position sensors offer picometer-level resolution, high bandwidth, and excellent stability over temperature and time. When integrated into a closed-loop system, they measure the actual displacement of the piezo stage and feed the error signal back to the controller. This real-time correction virtually eliminates hysteresis and creep, enabling linear displacement with errors typically below 0.1% of the full range.

The capacitive sensor works by detecting changes in capacitance between a fixed electrode and a moving target attached to the piezo stage. The sensor output is conditioned and digitized, then compared with the desired setpoint. The controller adjusts the drive voltage to minimize the error, effectively forcing the piezo to follow the commanded trajectory regardless of hysteresis or creep.

It is important to note that capacitive sensors are sensitive to stray fields and mechanical misalignment. Proper shielding and differential measurement topologies should be employed to maintain noise levels below 1 nm. In practice, a well-designed capacitive feedback loop can achieve positioning repeatability of 1–2 nm over a 100 µm travel range.

PID Parameter Tuning for Nanopositioning

PID (Proportional-Integral-Derivative) control is the core of most closed-loop piezo controllers. For nanopositioning, the proportional gain (Kp) provides stiffness to counteract hysteresis, but excessive Kp can cause oscillation due to the mechanical resonance of the stage. The integral gain (Ki) eliminates steady-state error but introduces low-frequency phase lag, potentially reducing stability margin. The derivative gain (Kd) adds damping to suppress resonance peaks, but it amplifies high-frequency noise from the sensor.

A practical tuning procedure starts with manual Ziegler-Nichols or relay tuning to find the ultimate gain and period. For piezo stages, a more refined approach is to use a frequency response analyzer to identify the open-loop transfer function, including the resonance mode (typically between 100 Hz and 2 kHz depending on mass and stiffness). Then, design a lead-lag compensator or implement a PID with a low-pass filter on the derivative term (e.g., cut-off at 5–10 times the resonance frequency).

As a rule of thumb, start with Kp set to 10–20% of the value that causes sustained oscillation, then add Ki to reduce settling time (integral action time constant Ti ≈ 1–2 times the period of the dominant resonance). Use Kd sparingly — a derivative gain that provides a damping ratio of 0.7–1.0 is often optimal. Always verify step response settling time and overshoot; for precision positioning, overshoot should be less than 2% and settling to within 1 nm within 50 ms.

Feedforward Compensation Algorithms

Pure feedback control cannot anticipate the rapid nonlinear response of the piezo actuator. Feedforward compensation predicts the required voltage for a given motion profile, reducing the burden on the feedback loop and improving tracking bandwidth. Two common approaches are inverse hysteresis compensation and model predictive feedforward.

Inverse hysteresis compensation uses a mathematical model of the hysteresis loop (e.g., Prandtl-Ishlinskii or Preisach model) to pre-distort the input voltage. For example, if the hysteresis characteristic is known, the feedforward controller applies a voltage that linearly maps to the desired displacement, effectively canceling the nonlinearity. This can reduce tracking error by 80–90% compared to feedback alone.

Implementation steps: (1) Characterize the hysteresis of the piezo stage by applying a triangular voltage sweep and recording the displacement using the capacitive sensor. (2) Fit a rate-independent or rate-dependent hysteresis model to the measured data. (3) Invert the model to compute the voltage required for a given position command. (4) Combine the feedforward signal with the PID output. Tuning the feedforward gain and dynamics typically requires iterative optimization; a useful heuristic is to set the feedforward contribution to 70–90% of the total drive, with PID handling the residual error and disturbances.

Conclusion

Closed-loop control with capacitive sensors transforms the nonlinear, drift-prone behavior of piezoelectric stages into a linear, repeatable positioning platform. The performance advantages — typically an order of magnitude improvement in accuracy and stability — are essential for applications such as scanning probe microscopy, adaptive optics, and semiconductor lithography. Proper PID tuning and the addition of feedforward compensation further push the bandwidth and precision limits. Engineers should carefully characterize their stage dynamics and sensor noise to achieve the best trade-off between speed and accuracy. With these techniques, nanopositioning at the single-nanometer level becomes routine.