Auto-tuning: From system identification to optimized gains
What is auto tuning?
Auto-tuning is the automated process in which a servo drive independently determines the optimal control loop gains (proportional, integral, and derivative components) — without manual trial and error. The process consists of two phases: system identification (the electromechanical system is measured) and gain calculation (an algorithm calculates the optimal parameters).
Why is this important?
Manual tuning of servo control loops requires in-depth knowledge of control engineering and hours of iteration. Auto-tuning solves this problem:
- Faster commissioning — what takes hours manually can be done in seconds
- Lower expertise requirements — the drive handles model extraction and gain calculation
- Better results — frequency-based auto-tuning uses information from the entire frequency range and often delivers better results than manual tuning.
- Compensation of mechanical resonances — identified resonances can be automatically suppressed with notch filters
- Reproducibility — identical results across an entire series production
How does it work?
Phase 1 — System identification:
The drive excites the system with a torque signal (sine sweep/chirp signal) and measures the resulting motion response. The frequency response (Bode diagram) of the system is calculated from the input/output data. The amplitude of the excitation signal increases linearly from 50% to 100% of the configured value during the sweep.
The Bode diagram shows:
- Gain bandwidth — the frequency at which the amplitude drops to −3 dB (70.7%)
- Phase margin — stability reserve; target value 30–60°
- Mechanical resonances — peaks in the amplitude response caused by elastic elements in the drive train
- Moment of inertia and friction — extracted from the low-frequency response
Phase 2 — Gain calculation:
Using the identified system model, the algorithm calculates the controller gains that achieve the desired behavior: target bandwidth, damping ratio, and settling time. Two configurations are available:
- PI-P (standard) — Integrator in the position controller, proportional component in the speed controller. Focus: precise path tracking, fast response, zero steady-state error during ramps
- P-PI — Proportional component in the position controller, integrator in the speed controller. Focus: smooth, overshoot-free step response
Important: Only one integrator at a time (either position OR speed), never both — otherwise instability may occur.
How does SOMANET implement this?
OBLAC Drives offers auto-tuning for all three cascade stages:
Geschwindigkeits-Auto-Tuning (ab OBLAC Drives 19.0): Berechnet PI-Gains für den Geschwindigkeitsregler. Einstellbare Parameter: Dämpfungsgrad (< 1 = schneller mit Überschwingen, > 1 = langsamer ohne Überschwingen) und Bandbreite (max. 100 Hz, begrenzt durch EtherCAT-Zykluszeit).
Position auto-tuning: Calculates the gains of the cascaded position control (PI-P or P-PI). Adjustable parameters: Settling time (ms, 2% criterion) and damping ratio. Four test signals (square wave, ramp, bidirectional ramp, sine wave) are available for validation.
System identification is limited to a bandwidth of 100 Hz (1 kHz EtherCAT communication rate). High-resolution encoders are recommended for precise results—Hall sensors alone are insufficient.
Further links:
