---
title: Tunable Lead CgLp Structure
url: https://www.emergentmind.com/topics/tunable-lead-cglp-structure
type: topic
---

# Tunable Lead CgLp Structure

Tunable Lead CgLp Structure denotes a class of reset-based compensator architectures in which a Constant in gain, Lead in phase (CgLp) element is parameterized or structurally modified so that its first-harmonic response provides phase lead around a prescribed frequency region while maintaining approximately constant magnitude. The topic emerged from reset-control formulations that cascade a reset lag with a linear lead, with the explicit aim of circumventing the linear gain–phase limitation associated with the waterbed effect. In the subsequent literature, “tunable lead” came to include not only adjustment of the lead amount and band through parameters such as reset ratio, corner frequencies, and damping, but also architectural mechanisms that localize or suppress reset nonlinearity, including single-state reset, fractional-order reset, feedthrough augmentation, split-lead sequencing, and reset-trigger shaping [1805.12406][2004.03529][2005.02898][2009.11131][2507.01491][2507.04707][2602.10724].

## 1. Origin and defining idea

The original CgLp element was introduced as a reset lag in series with a linear lead of the same nominal order. In its first-order form, the reset lag is paired with
\[
L(s)=\frac{s/\omega_r+1}{s/\omega_f+1},
\]
while in the second-order form a GSORE-based reset lag is paired with
\[
L(s)=\frac{(s/\omega_r)^2+2\beta_r s/\omega_r+1}{(s/\omega_f)^2+2s/\omega_f+1}.
\]
The design intent is that the lead cancels the lag’s magnitude effect, whereas reset reduces the lag’s phase lag relative to the corresponding linear low-pass; the series combination therefore exhibits nearly constant gain with positive phase over a band typically identified as \([\omega_r,\omega_f]\) [1805.12406].

This construction was motivated by the observation that conventional linear lead compensation cannot increase phase around crossover without modifying gain elsewhere. CgLp was therefore positioned as a nonlinear replacement for part of the derivative/lead action in PID-like controllers, particularly in precision motion systems where bandwidth, robustness, tracking, and noise attenuation are simultaneously constrained by Bode’s gain–phase relationship [1805.12406][2005.02898].

From the outset, however, CgLp was a first-harmonic concept: “constant in gain, lead in phase” refers to the describing-function view of the nonlinear element. Later work on tunable lead CgLp structures was driven largely by the gap between that first-harmonic design picture and the actual closed-loop behavior once higher-order harmonics are included [2004.03529][2009.05116].

## 2. Canonical reset formulation and frequency-domain description

The common mathematical model is the standard reset system
\[
\Sigma_R :=
\begin{cases}
\dot{x}_r(t)=A x_r(t)+B e(t), & e(t)\neq 0,\\
x_r(t^+)=A_\rho x_r(t), & e(t)=0,\\
u(t)=C x_r(t)+D e(t),
\end{cases}
\]
where \(A_\rho\) determines which states reset and by what factor. For scalar or uniform reset, the reset ratio is commonly denoted \(\gamma\), with \(\gamma=1\) recovering the base linear system and \(|\gamma|<1\) producing partial reset [2005.02898][2009.05116].

For sinusoidal input \(e(t)=E\sin(\omega t)\), the first-harmonic describing function of the reset element is written as
\[
G_1(\omega)=C(j\omega I-A)^{-1}(I+j\Theta_D(\omega))B + D,
\]
and higher-order sinusoidal input describing functions satisfy
\[
G_n(\omega)=C(jn\omega I-A)^{-1}j\Theta_D(\omega)B,\qquad \text{odd }n>1,
\]
with even harmonics equal to zero. In the CgLp context, the total first-harmonic response is the product of the reset part and the linear lead filter, so design is ordinarily performed on \(G_1\), while validation increasingly relies on \(G_3,G_5,\dots\) or related closed-loop harmonic transfer functions [2004.03529][2009.05116][2507.04707].

This formalism establishes the central technical distinction of the subject. A tunable lead CgLp structure is not merely a phase-lead network with adjustable corner frequencies; it is a hybrid system whose usable phase advantage depends simultaneously on the first-harmonic response and on the architecture’s ability to limit or place the resulting higher-order harmonics. The later literature treats these two layers—fundamental shaping and harmonic shaping—as separate but coupled design problems [2004.03529][2005.02898].

## 3. Principal tuning variables and design logic

In first-order CgLp formulations embedded in PID-like controllers, the tunable lead portion is commonly parameterized by \(K_p,\omega_r,\omega_d,\omega_t,\gamma\), with \(\omega_i=\omega_c/10\) and \(\omega_f=8\omega_c\) fixed in one systematic framework. In that setting, \(\omega_r\) and \(\omega_d\) place the lead around crossover, \(\omega_t\) limits the high-frequency extent of the lead, and \(\gamma\) controls reset strength: lower \(\gamma\) generally increases phase advantage but also raises high-order harmonic content and control effort [2005.02898].

HOSIDF-based tuning work sharpened this view by introducing explicit third-harmonic indicators. For groups of CgLp designs that provide the same first-harmonic phase lead at the same crossover, the third-harmonic peak frequency \(\omega_p\) and peak magnitude \(M_p\) were used as selection criteria. The stated rule is that tracking-oriented tuning should prefer the largest \(\omega_p\), whereas noise-oriented tuning should prefer the smallest \(M_p\), which usually corresponds to the largest \(\gamma\) within the admissible group [2009.05116].

A second harmonic-aware rule was developed for GFORE- and GSORE-based CgLp through the scalar measure
\[
\sigma=
\frac{1-\gamma}{(\alpha\omega_r)^2}
\quad\text{or}\quad
\sigma=
\frac{1-\gamma}{(\kappa\omega_r)^2},
\]
depending on the order. Within a set of designs that all satisfy the same phase-lead requirement at bandwidth, the preferred design is the one with minimum \(\sigma\), because \(\sigma\) scales the low-frequency higher-order harmonics. For CgLp-SORE, setting
\[
\beta=\frac{1}{2\kappa}
\]
cancels the low-frequency higher-order term in the stated approximation, which turns damping selection into an explicit harmonic-suppression device rather than only a phase-shaping parameter [2005.03944].

More elaborate tuning frameworks combined loop-shaping constraints with nonlinear performance metrics. One such method imposes the crossover condition, a phase-margin condition, an iso-damping condition,
\[
\frac{d}{d\omega}\angle\big(\mathcal{N}_{\text{CgLp}}(j\omega)\,\text{PID}(j\omega)\,G(j\omega)\big)\Big|_{\omega=\omega_c}=0,
\]
the \(H_\beta\) stability condition, and a modulus-margin bound on the pseudo-sensitivity \(S_\infty\), and then minimizes
\[
J=\max_{\omega\le \omega_l}\left|\frac{S_\infty(j\omega)}{\omega}\right|_{\text{dB}}.
\]
Here \(S_\infty,T_\infty,CS_\infty,\) and \(PS_\infty\) are computed from the actual periodic steady state rather than from a fundamental-only approximation, so the tuning process explicitly internalizes harmonic distortion [2005.02898].

## 4. Localized linearization: single-state and fractional-order variants

A major branch of tunable lead CgLp research replaced uniform reset with partial or structured reset so that the element becomes effectively linear at a chosen frequency or over a chosen range. The second-order single-state reset element (SOSRE) is the canonical example. In that structure, only one state of the second-order lag is reset, with
\[
A_\rho=\operatorname{diag}(1,\gamma_2,1,1),
\]
rather than resetting both lag states. The key result is that the SOSRE-based CgLp behaves exactly as its corresponding linear second-order lag at
\[
\omega^\star=\omega_{r\alpha}=\frac{\omega_r}{\alpha},
\]
because the reset state reaches zero at reset instants. At that frequency, all higher harmonics vanish and the first harmonic coincides with the linear transfer function. The design implication stated in the paper is explicit:
\[
\text{to suppress the third harmonic at plant resonance }\omega_n,\ \text{choose }\omega_{r\alpha}=\omega_n/3.
\]
In the reported example, a plant with resonance at \(30\,\text{rad/s}\) led to \(\omega_{r\alpha}=10\,\text{rad/s}\), and with reference \(r(t)=\sin(10t)\) the SOSRE CgLp achieved \(L_2(e)=0.099\) and \(L_\infty(e)=0.069\), compared with \(0.171\) and \(0.123\) for a conventional PID, \(0.368\) and \(0.105\) for FORE CgLp, and \(1.214\) and \(0.672\) for SORE CgLp [2004.03529].

The fractional-order single-state reset element (FOSRE) generalizes the same idea. It resets one linear integrator state while leaving a fractional-order component linear, implemented through a CRONE approximation. Its tuning centers on the phase
\[
\psi(\omega)=\angle \frac{X_2(j\omega)}{E(j\omega)}\Big|_{A_\rho=I}.
\]
If \(\psi(\omega_{lb})=0\), the reset state is zero at reset instants, so the element behaves linearly at \(\omega_{lb}\), and all higher-order HOSIDFs vanish there. The published tuning guideline starts with \(\lambda=-0.1\) and \(\beta=1\), then optimizes \(\omega_l\) and \(\omega_{r\alpha}\) subject to \(\psi(\omega_{lb})=0\) and minimum \(|\psi(\omega)|\) on \([0,\omega_{lb}]\), while adjusting \(\beta\) and then \(\lambda\) if the crossover-phase requirement is not met. In the reported precision-stage example, FOSRE-based CgLp designed for \(150\,\text{Hz}\) bandwidth and \(45^\circ\) phase margin yielded much smaller low-frequency harmonics than SOSRE CgLp, significantly smaller tracking errors than PID and SOSRE at \(0.5,0.8,2,\) and \(4\,\text{Hz}\) for most cases, and smaller control peaks than SOSRE [2009.11131].

| Variant | Structural change | Tunable effect |
|---|---|---|
| SOSRE-based CgLp | Reset only one state of second-order lag | Exact linear behavior at \(\omega_{r\alpha}\) |
| FOSRE-based CgLp | Single-state reset plus fractional-order linear block | Suppression of nonlinearity on selected low-frequency range |
| Feedthrough-modified CgLp | Add \(D_r\) to proportional GFORE | Nearly constant first-harmonic gain across all frequencies |
| Split-lead CgLp | Divide lead into pre- and post-reset parts | Balance noise amplification against higher harmonics |

These variants share a common interpretation: the tunability is not limited to moving a phase bump. It includes deliberate placement of frequencies at which reset action becomes ineffective or attenuated, thereby reconciling CgLp’s nonlinear phase advantage with the need for quasi-linear behavior at disturbance, resonance, or operating frequencies where harmonics are especially harmful [2004.03529][2009.11131].

## 5. Feedthrough augmentation, sequencing, and reset-trigger shaping

A later modification introduced a proportional GFORE with nonzero feedthrough,
\[
A_r=-\omega_r,\quad B_r=1,\quad C_r=\omega_r,\quad D_r=\frac{\omega_l}{\omega_f-\omega_l},\quad A_\rho\in(-1,1),
\]
combined with
\[
k_c=\frac{\omega_f-\omega_l}{\omega_f},\qquad
C_{\mathfrak c}(s)=\frac{1+s/\omega_l}{1+s/\omega_f}.
\]
The stated purpose is to make the first-harmonic CgLp magnitude exactly \(1\) at \(\omega\to 0\) and \(\omega\to\infty\), and approximately constant in between, while reducing \(|\mathfrak{C}_3|/|\mathfrak{C}_1|\) at high frequency. That same work proposed a backward calculation of \(\omega_f\) from a desired phase \(\theta_{\mathrm{CgLp}}(\omega)\) and defined a sensitivity improvement indicator
\[
\delta_s(\omega)=\frac{|S_\infty(\omega)|-|S(j\omega)|}{|S(j\omega)|}\,\%.
\]
In an industrial wire bonder application, the CgLp-based add-on with inverse notch shaping reduced the settling interval by \(80.4\%\) in forward motion and \(50.5\%\) in backward motion relative to the original LTI controller, and reduced stationary RMS error by \(58.9\%\) and \(46.4\%\), respectively; a second add-on design delivered average improvements of about \(28\%\) in settling period and \(32\%\) in RMS error across three reference classes [2507.01491].

Another line of work focused on a long-standing sequencing debate. If the lead is placed before the reset element, the reset input is noise-amplified; if the lead is placed after reset, higher-order harmonics generated by the reset element are magnified. The proposed tunable lead CgLp resolves this by splitting the lead as
\[
L_1(s)=\frac{1+s/\omega_r}{1+s/\omega_x},\qquad
L_2(s)=\frac{1+s/\omega_x}{1+s/\omega_f},
\]
so that
\[
C_{\mathrm{CgLp}}(s)=L_2(s)\,\mathcal{R}(s)\,L_1(s),
\qquad \omega_r\le \omega_x\le \omega_f.
\]
The two conventional orderings appear as limits: \(\omega_x=\omega_r\) yields reset–lead, and \(\omega_x=\omega_f\) yields lead–reset. An additional \(N\)–\(N^{-1}\) filtering technique then targets specific higher-order sensitivity components, notably the third harmonic around \(\omega_c/3\), without increasing the reset-path gain above unity [2507.04707].

A related but distinct architecture regulates the reset trigger rather than the control path. In a dual-loop piezo-nanopositioning system, aggressively tuned CgLp blocks with \(5^\circ,10^\circ,15^\circ,\) and \(20^\circ\) phase lead increased crossover but also produced pronounced higher-order harmonics and multiple-reset behavior. The remedy was a shaping filter in the reset-trigger path,
\[
C_s(s)=\mathcal{R}_{bl}(s)\,N_{s_1}(s)\,N_{s_2}(s)\,C_L(s),
\]
with a fractional lead–lag term
\[
C_L(s)=\left(\frac{1+s/\omega_L}{1+s/\omega_H}\right)^\lambda.
\]
For the \(20^\circ\) case, the reported shaping parameters were \(\omega_L=200\,\text{Hz}\), \(\omega_H=800\,\text{Hz}\), \(\lambda=-0.9\), and \(Q=1.15\). This reduced third-harmonic magnitude below \(200\,\text{Hz}\), removed multiple resets at \(80\,\text{Hz}\), and preserved the crossover increase. Relative to a well-tuned linear baseline, the experimental open-loop crossover increased by approximately \(55\,\text{Hz}\) and the closed-loop bandwidth by about \(34\,\text{Hz}\) [2602.10724].

## 6. Integration into motion-control loops and reported performance

In practical applications, tunable lead CgLp structures are almost always embedded in broader controller architectures rather than used in isolation. One systematic design on a precision “Spider” stage combined CgLp with PI, derivative lead, and low-pass filtering, with design targets \(\omega_c=100\,\text{Hz}\), \(\varphi_m=30^\circ\), and modulus margin \(M_m\le 6.5\,\text{dB}\). The tuned CgLp+PID controller achieved better tracking for frequencies below \(10\,\text{Hz}\), better disturbance rejection, and lower high-frequency noise sensitivity than a tuned PID; the reported improvements were approximately \(30\%\) in tracking, \(30\%\) in disturbance rejection, and \(40\%\) in noise rejection, at the cost of higher control sensitivity because reset-induced jumps increase control effort [2005.02898].

The original precision-motion study compared linear PID, reset integrator, GFORE-based CgLp, and GSORE-based CgLp under matched bandwidth and phase-margin conditions. With fixed bandwidth \(100\,\text{Hz}\) and \(30^\circ\) phase margin, the best results occurred around \(\gamma\approx 0.4\). In that regime, GSORE-based CgLp improved tracking RMS from \(162.9\) for linear PID to \(126.6\), and precision RMS from \(814.1\) to \(387.3\). In a bandwidth-increase scenario with preserved nominal precision and phase margin, GFORE-based CgLp pushed crossover from \(100\,\text{Hz}\) to \(127\,\text{Hz}\), an increase of about \(27\%\) [1805.12406].

These experimental results established two recurring facts that later tunable-lead research retained. First, the best design is rarely the most aggressive reset design. Second, structural control of higher-order harmonics can matter more than further first-harmonic phase advantage once the crossover specification is already met. This helps explain why single-state, fractional-order, feedthrough, split-lead, and shaping-filter variants all focus on reducing or relocating nonlinearity rather than simply maximizing raw reset strength [1805.12406][2004.03529][2009.11131].

## 7. Conceptual boundaries, debates, and recurring misconceptions

A central misconception is that “constant in gain” means exact linear unity gain. In the literature, the phrase is a first-harmonic property. Several papers state explicitly that describing-function design can be misleading when higher-order harmonics are non-negligible. The conventional SORE-based CgLp is the canonical warning case: its third harmonic can become so large that \(|G_3(\omega)|>|G_1(\omega)|\) over a broad frequency range, making the first-harmonic design picture practically invalid [2004.03529].

A second misconception is that stronger reset always improves performance. The experimental and HOSIDF-based tuning studies do not support that claim. Lower \(\gamma\) can increase phase advantage, but it also increases \(M_p\), control effort, and harmonic distortion; both the original motion-control experiments and later harmonic-aware tuning papers report performance optima at intermediate reset levels rather than at the strongest reset admissible by the model [1805.12406][2009.05116][2005.03944].

A third issue concerns architectural universality. The sequencing dispute—lead before reset or after reset—does not admit a single best answer independent of context. Lead before reset reduces higher-order harmonics in the closed loop but amplifies noise at the reset input; lead after reset avoids that noise amplification but magnifies the reset-generated harmonics. Split-lead and trigger-shaping approaches were introduced precisely because the preferred arrangement depends on the noise floor, the disturbance spectrum, and where the plant is most sensitive to harmonic excitation [2507.04707][2602.10724].

Finally, localized linearization should not be confused with global linearity. SOSRE-based CgLp is exactly linear only at \(\omega_{r\alpha}\); FOSRE-based CgLp suppresses nonlinearity on a designed range but retains it elsewhere; feedthrough-modified and shaping-filtered designs reduce undesirable nonlinear effects without eliminating reset dynamics altogether. The literature also emphasizes that state-space realization matters for nonlinear reset systems: changing the realization can change the behavior, so implementation must preserve the intended reset state and reset surface. Stability, correspondingly, is not inferred from classical linear margins alone; the reported methods rely on quadratic stability or \(H_\beta\)-type conditions, together with time-domain and harmonic-domain validation [2004.03529][2005.02898][2009.11131].

In that sense, the modern tunable lead CgLp structure is best understood not as a single filter topology but as a design family. Its unifying principle is constant-gain phase recovery via reset; its technical substance lies in how that phase recovery is tuned, localized, and regularized so that the nonlinear advantages survive contact with harmonics, resonances, sampling, and noise in high-performance control systems [2507.01491][2507.04707][2602.10724].

Source: https://www.emergentmind.com/topics/tunable-lead-cglp-structure