Iterative Youla–Kucera Loop Shaping

• Iterative Youla–Kucera loop shaping is a design methodology that iteratively constructs low-order controllers to create precise, narrow-band notches for multi-band disturbance rejection.
• It employs a systematic approach using Youla–Kucera parameterization and order reduction to ensure numerical robustness and guaranteed stability.
• Validated in HDD servo applications, the method achieves over 50 dB disturbance attenuation at targeted frequencies while minimizing sensitivity peaks in untargeted bands.

Iterative Youla–Kucera (YK) loop shaping is a systematic design methodology for multi-band disturbance rejection in motion control. It utilizes the Youla–Kucera parameterization in an iterative fashion to achieve precise shaping of the closed-loop frequency response—specifically, to insert multiple narrow-band sensitivity notches—while maintaining numerical robustness, guaranteed stability at every step, and transparent management of fundamental performance trade-offs. This technique is especially effective for modern high-precision actuation platforms where nanometer-level disturbance attenuation across multiple spectral bands is critical (Hu et al., 19 Aug 2025).

1. Youla–Kucera Parameterization and Inversion-Based Loop Shaping

The approach begins with a standard single-input single-output (SISO) negative-feedback architecture:

• Plant: $P(z)$
• Baseline controller: $C_0(z)$
• Nominal loop gain: $L_0(z) = P(z) C_0(z)$
• Sensitivity: $S_0(z) = \frac{1}{1 + L_0(z)}$

With a right coprime factorization $P(z) = \frac{N(z)}{D(z)}$, all stabilizing controllers admit the Youla–Kucera parameterization: $$C_{\rm all}(z) = \frac{C_0(z) + D(z) Q(z)}{1 - N(z) Q(z)} , \quad Q(z) \in \mathcal S$$ where $\mathcal S$ is the set of stable, proper rational functions. The corresponding sensitivity function becomes $\widetilde S(z) = S_0(z) \bigl(1 - N(z) Q(z)\bigr)$.

For motion-control applications with an existing stable loop, the inversion-based YK reformulation treats $L_0(z)$ as a fictitious plant under unity-feedback, exploiting an approximate stable inverse $\widehat L^{-1}(z)$: $C_0(z)$0 where $C_0(z)$1 is the relative degree. This structure enables direct frequency-domain notch placement at desired disturbance bands and underpins the subsequent iterative shaping process.

2. Iterative Design Algorithm

Rather than computing a single possibly high-order Youla parameter, the iterative YK method constructs the controller as a cascade of low-order augmentations. At each iteration $C_0(z)$2, a new $C_0(z)$3 is synthesized and the controller updated by

$C_0(z)$4

with the corresponding closed-loop sensitivity

$C_0(z)$5

Stability is enforced at each step by requiring $C_0(z)$6 and $C_0(z)$7 for $C_0(z)$8, guaranteeing no unstable closed-loop poles are introduced. The iterative product $C_0(z)$9 determines the aggregate notch structure, and the process converges to a prescribed loop shape when this product achieves the target zeros at the desired frequencies.

To enforce deep notches at specific narrow-band frequencies $L_0(z) = P(z) C_0(z)$0, disturbance-rejection polynomials are introduced: $L_0(z) = P(z) C_0(z)$1 yielding the loop-shape template

$L_0(z) = P(z) C_0(z)$2

where $L_0(z) = P(z) C_0(z)$3 govern notch bandwidth and depth, and $L_0(z) = P(z) C_0(z)$4 is a low-order polynomial ensuring degree balance.

Notch depth and width are independently tunable through scalar scaling of $L_0(z) = P(z) C_0(z)$5 ($L_0(z) = P(z) C_0(z)$6, $L_0(z) = P(z) C_0(z)$7) and choice of $L_0(z) = P(z) C_0(z)$8 (bandwidth via $L_0(z) = P(z) C_0(z)$9), respectively.

3. Numerical Robustness and Waterbed Effect Mitigation

A central motivation for the iterative approach is numerical robustness. When implementing a single $S_0(z) = \frac{1}{1 + L_0(z)}$0 with many closely spaced poles and zeros (as required for multi-band disturbance rejection), finite-precision arithmetic pushes poles outside the unit circle, potentially destabilizing the implemented filter beyond four–five notches.

The iterative YK protocol mitigates this through:

• Iterative decomposition: Each $S_0(z) = \frac{1}{1 + L_0(z)}$1 is low order (corresponding to 1–2 notches), keeping poles inside the unit disk even under quantization.
• Depth scaling: Scaling $S_0(z) = \frac{1}{1 + L_0(z)}$2 ensures $S_0(z) = \frac{1}{1 + L_0(z)}$3 does not deviate excessively outside targeted bands, minimizing unwanted sensitivity amplification ("waterbed effect").
• Order reduction: After each iteration, the controller $S_0(z) = \frac{1}{1 + L_0(z)}$4 is reduced (balanced truncation or Hankel-norm reduction) to a chosen order $S_0(z) = \frac{1}{1 + L_0(z)}$5, preventing error accumulation and excessive complexity.

These mechanisms ensure each stage remains stable and well-conditioned; the cumulative controller inherits these properties throughout the sequence.

4. Empirical Validation: Hard Disk Drive Servo Application

The iterative YK method was validated on a dual-stage hard disk drive (HDD) system, targeting precise multi-band vibration rejection. The plant comprised $S_0(z) = \frac{1}{1 + L_0(z)}$6 with a baseline controller $S_0(z) = \frac{1}{1 + L_0(z)}$7, and loop gain $S_0(z) = \frac{1}{1 + L_0(z)}$8. The iterative sequence produced the aggregate controller $S_0(z) = \frac{1}{1 + L_0(z)}$9 and corresponding sensitivity function $P(z) = \frac{N(z)}{D(z)}$0.

The achieved attenuation at five targeted vibration frequencies (229, 338, 545, 633, 740 Hz) consistently exceeded 50 dB, compared to ≈40 dB using cascaded second-order notch filters (ESPRC) and minimal unintended gain elsewhere ($P(z) = \frac{N(z)}{D(z)}$1 dB outside notched regions). Quantitatively, root-mean-square (RMS) position error under worst-case disturbances improved: from ≈10 nm (baseline) to ≈4 nm (ESPRC), and finally to ≈2 nm under the iterative YK design. Peak sensitivity in untargeted frequency bands remained under 3 dB, confirming the absence of excessive waterbed amplification (Hu et al., 19 Aug 2025).

5. Implementation Guidelines and Generalization

Effective usage of iterative YK loop shaping is governed by the following protocol:

• Frequency grouping: Disturbance frequencies are grouped in sets of one or two per iteration ($P(z) = \frac{N(z)}{D(z)}$2 order ≤ 4), mitigating high-order instabilities.
• Edge parameters ($P(z) = \frac{N(z)}{D(z)}$3): For 3 dB notch bandwidth $P(z) = \frac{N(z)}{D(z)}$4, set $P(z) = \frac{N(z)}{D(z)}$5, and $P(z) = \frac{N(z)}{D(z)}$6 for maximum depth.
• Notch depth scaling ($P(z) = \frac{N(z)}{D(z)}$7): Initialize at $P(z) = \frac{N(z)}{D(z)}$8, increasing in later passes for sharper notches while keeping track of global sensitivity bounds.
• Order reduction ($P(z) = \frac{N(z)}{D(z)}$9): Apply balanced truncation after each iteration, maintaining controller order within 2–4.
• Waterbed management: Deep notches (high $$C_{\rm all}(z) = \frac{C_0(z) + D(z) Q(z)}{1 - N(z) Q(z)} , \quad Q(z) \in \mathcal S$$0) can elevate sensitivity elsewhere; the stepwise method allows iterative tuning of $$C_{\rm all}(z) = \frac{C_0(z) + D(z) Q(z)}{1 - N(z) Q(z)} , \quad Q(z) \in \mathcal S$$1 or notch width to balance attenuation against overall robustness.
• Practical checks: After quantization, confirm that all implemented poles reside within the unit circle. Monitor gain crossover and phase margin at each stage. Widely separated frequency bands should be handled in separate iterations to avoid placement of poles across spectral decades.

The method supports real-time tuning of notch parameters and addresses the controller complexity/waterbed trade-off in a granular and transparent manner. It generalizes straightforwardly to any precision motion-control platform requiring multi-band rejection, beyond storage systems.

6. Significance and Context in Modern Control

Iterative Youla–Kucera loop shaping constitutes an advance in numerically robust, high-precision disturbance-rejection controller synthesis. It enables enforcement of exact frequency-domain templates at multiple targeted bands without prohibitive computational fragility. The methodology maintains stability invariants, offers independent tuning of notch depth and width, and permits integration with classical order-reduction and state-space techniques.

A plausible implication is its deployment in other fields where robust multi-band attenuation is critical and actuator dynamic range is tightly constrained. The method's capacity for nanometer-scale disturbance rejection, as validated in HDD systems, situates it as a reference approach for future ultra-high-density storage, precision metrology, and electromechanical actuation settings requiring tight sensitivity shaping across multiple spectral regions (Hu et al., 19 Aug 2025).