Two Degree-of-Freedom Controllers for Hard Disk Drives with Novel Reference Trajectory Design

Susumu Hara, T. Hara, Yi Li, Masayoshi TomizukaThe Proceedings of the Conference on Information Intelligence and Precision Equipment IIP 2000 · DOI

TL;DR

This paper proposes a novel two-stage reference trajectory design for Hard Disk Drive (HDD) head positioning that explicitly accounts for back electromotive force (e.m.f.). By combining this optimized trajectory with Adaptive Robust Control (ARC), the researchers achieved a 5–15% reduction in seek times while maintaining high precision and robustness against external disturbances.

What Problem Does This Paper Address?

In the world of Hard Disk Drives, speed is king. The "seek time"—the time it takes for the read/write head to move from one data track to another—is a critical performance bottleneck. Traditionally, HDDs use Mode Switching Control (MSC), which flips between a high-speed "seek" mode and a high-precision "settling/following" mode.

However, MSC often struggles with the transition between these modes, leading to vibrations or overshoots that delay data access. Furthermore, as seek speeds increase, physical effects like back electromotive force (e.m.f.) from the voice coil motor (VCM) and structural resonances become significant hurdles. The authors aim to replace clunky mode-switching with a unified Two Degree-of-Freedom (2-DOF) controller that uses a smarter, vibration-minimized trajectory.

Methods

The authors introduce a 2-DOF control architecture consisting of a feedforward reference generator and a feedback controller. The core innovation lies in the two-stage trajectory design:

  1. Off-line Simulation: Using a mathematical model of the VCM that includes back e.m.f., the system pre-calculates optimal switching times for different seek distances. This creates a lookup table for the fastest possible acceleration and deceleration profiles.
  2. On-line Generation & SMART: During actual operation, the system generates the trajectory in real-time. To handle the "errors" caused by the fact that digital controllers work in discrete time steps, they use the SMART (Structural vibration Minimized AcceleRation Trajectory) algorithm. SMART adjusts the end of the trajectory to ensure the head arrives at the target track with zero velocity and acceleration, effectively "canceling" potential vibrations.

For the feedback portion, they propose Adaptive Robust Control (ARC). Unlike standard controllers, ARC can adapt to changing parameters (like friction or torque constant fluctuations) while remaining "robust" against unpredictable external shocks.

Datasets

The study does not use a "dataset" in the machine learning sense. Instead, it utilizes the physical specifications and performance parameters of a commercial hard drive:

  • Hardware Model: Fujitsu M2954 HDD.
  • Parameters: 7200 rpm spindle speed, 3.945 μm track pitch, and a 14.88 kHz sampling rate.
  • Evaluation Scenarios: The researchers tested seek distances of 300 tracks (short), 2000 tracks (medium), and 5000 tracks (long) to ensure the method scales across different operations.

Key Insights and Ideas

  • Modeling Physics Matters: Most older models ignored back e.m.f. By including it, the authors could design a deceleration curve that pushes the motor to its physical limit without causing voltage saturation, shaving precious milliseconds off the seek time.
  • The Discrete-Time Correction: In high-speed control, the tiny gap between digital samples can lead to a "miss" at the target. Using SMART as a real-time "patch" for these discrete-time errors allows for a much smoother transition into the tracking phase.
  • ARC vs. Disturbance Observers: The paper highlights that while many engineers use "Disturbance Observers" to handle noise, ARC is superior because it handles "anti-windup" (preventing the controller from over-reacting when the motor hits its power limit) and provides better stability during parameter adaptation.

Key Results

  • Faster Seeking: Experimental results showed a 5% to 15% reduction in seek time compared to conventional Mode Switching Control.
  • Precision: The system consistently settled within a 0.5 μm error margin without significant overshoot, even during long 5000-track seeks.
  • Robustness: Numerical simulations of the ARC component demonstrated that it could suppress tracking errors caused by bias forces (like cable tension) and a 10% variation in the motor's torque constant—common issues in mass-produced hardware.

Limitations and Future Work

  • Short Seek Limitations: The authors acknowledge that for extremely short distances, the two-stage algorithm might be overkill or inapplicable, suggesting that SMART alone might be better suited for "micro-seeks."
  • Experimental Gap for ARC: While the trajectory design was verified with physical experiments on a Fujitsu drive, the Adaptive Robust Control (ARC) section was primarily validated through numerical simulations. The authors note that full experimental implementation of the ARC feedback loop is a target for future research.
  • Computational Overhead: Implementing complex real-time trajectory generation and adaptive algorithms requires more processing power than simple MSC, which may be a constraint for low-cost HDD controllers.

Summary updated

Abstract

This paper considers the determination of reference trajectories for use with a two-degree-of-freedom (2-DOF) controller for the head positioning of Hard Disk Drives (HDDs). The reference HDD model used for the design includes the back e.m.f. (electromotive force) of the voice coil motor. In the second stage, the idea of SMART (Structure vibration Minimized AcceleRation Trajectory) is utilized to adjust the errors at the end of the trajectories. 2-DOF control with the proposed reference trajectory is compared to conventional mode switching control, and its effectiveness is verified. To minimize performance degradation due to disturbances and uncertainties, the use of Adaptive Robust Control (ARC) is suggested.
BibTeX
@inproceedings{hara2000two,
  author = {Susumu Hara and T. Hara and Yi Li and Masayoshi Tomizuka},
  title = {{Two Degree-of-Freedom Controllers for Hard Disk Drives with Novel Reference Trajectory Design}},
  booktitle = {The Proceedings of the Conference on Information Intelligence and Precision Equipment IIP},
  year = {2000},
  doi = {10.1299/jsmeiip.2000.168},
}
Figure