⚡ electrical engineering

Compositional small-gain and small-phase stability analysis

This paper presents a compositional framework that extends small-gain and small-phase stability analysis to complex MIMO LTI systems by iteratively bounding key parameters (such as phase sectors and Crawford numbers) across series, parallel, and feedback interconnections, thereby enabling stability verification for systems that are intractable using classical methods.

Original authors: Anton Ponomarev

Published 2026-06-16

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Anton Ponomarev

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to predict whether a giant, complex machine will stay stable or fall apart. This machine is built by snapping together many smaller, simpler machines (like gears, springs, and levers) in different ways: some are lined up in a row (series), some are side-by-side (parallel), and some are connected in a loop where the output feeds back into the input (feedback).

For decades, engineers have had two main "rules of thumb" to check if these machines are safe:

The Small-Gain Rule: If the "strength" (gain) of every part is weak enough, the whole thing won't explode. It's like saying, "If no single person in a tug-of-war pulls too hard, the rope won't snap."
The Small-Phase Rule: This is a bit more abstract. It checks the "direction" or "timing" (phase) of the parts. If the parts are all pointing in roughly the same direction and not spinning wildly out of sync, the system is safe.

The Problem:
The "Small-Gain" rule is easy to use for big machines. If you have 10 parts in a row, you just multiply their strengths. If the total is low, you're safe.

However, the "Small-Phase" rule has a major flaw when you try to use it on big, complex machines. When you combine parts, their "directions" don't just add up neatly. Sometimes, two parts that seem safe individually, when combined, create a chaotic mess that the old rule can't predict. It's like trying to predict the path of a ball by just adding the angles of two bounces; sometimes the math gets messy and the ball goes somewhere unexpected. Because of this, engineers couldn't easily use the "Small-Phase" rule for complex networks.

The Solution: A New "Compositional" Approach
The author of this paper, Anton Ponomarev, introduces a new method that allows us to use the "Small-Phase" rule on these complex machines, just like we do with the "Small-Gain" rule.

Here is the core idea, explained with an analogy:

Imagine each machine part is a compass.

The Old Way: If you try to combine two compasses, the old rule says, "We can't tell where the new compass points because the directions might get confused."
The New Way: The author realizes that if a compass is "well-behaved" (it doesn't wobble too much and doesn't point near the "danger zone" of zero), we can predict its new direction.

To make this work, the author invents a new set of measurements for each part, called "Ring-Sectorial Bounds." Instead of just looking at the compass needle, he looks at:

How far it is from the center (The Crawford number): Is the compass needle pointing strongly away from the center, or is it wobbling near the middle?
How wide its spread is (The Crawford defect): Is the compass needle tight and precise, or is it a fuzzy, wide arc?

The Magic Trick: "Composing" the Bounds
The paper proves that if you know these specific measurements for two parts, you can calculate the measurements for the combined part.

Series (One after another): You can multiply their "strengths" and add their "directions," but you have to add a little "safety margin" based on how much they wobble (the defect).
Parallel (Side by side): You can add their ranges together.
Feedback (Loops): You can check if the loop is safe by seeing if the combined direction stays away from a specific "danger angle" (180 degrees).

Why This Matters (According to the Paper)
The author shows that this new method can solve problems that the old methods couldn't.

Example 1: In some cases, a system might have very high "gain" (it's very strong). The old rules might say, "It's too strong, it will break!" But the new method looks at the direction and realizes, "Actually, because it's so strong and pointing in a specific way, it's actually very stable." It's like realizing a very strong wind is actually pushing a sailboat in the right direction, keeping it steady, rather than capsizing it.
Example 2: It allows engineers to break a huge, scary system down into small, manageable pieces, check each piece, and then "stack" the results to prove the whole thing is safe.

In Summary
This paper provides a new mathematical toolkit. It takes a rule that was previously too messy to use on complex networks (the Small-Phase rule) and gives it a "translation guide" (the Ring-Sectorial bounds). This guide allows engineers to build complex systems out of smaller blocks, check the stability of each block, and confidently say, "The whole machine is safe," even when the parts are arranged in complicated loops and chains.

The paper does not discuss medical applications or future commercial products; it strictly focuses on the mathematical theory of how to analyze the stability of these interconnected systems.

Technical Summary: Compositional Small-Gain and Small-Phase Stability Analysis

Problem Statement
The paper addresses the internal stability analysis of complex Multi-Input Multi-Output (MIMO) Linear Time-Invariant (LTI) systems constructed via hierarchical interconnections of series, parallel, and feedback loops. While the classical small-gain theorem has been successfully extended to such complex networks by exploiting the submultiplicativity of matrix norms ( $\|A_1 A_2\| \leq \|A_1\|\|A_2\|$ ), the small-phase theorem faces a fundamental obstacle. The angular field of values (numerical range) is not generally submultiplicative; specifically, the spectral inclusion $\sigma(A_1 A_2) \subseteq W'(A_1)W'(A_2)$ does not allow for iterative application in loops with three or more subsystems. Consequently, existing small-phase analysis is often limited to single-loop systems or requires restrictive assumptions (e.g., sectorial matrices) that do not scale well to complex compositions.

Methodology
The author proposes a "compositional" approach that establishes an analogue to the submultiplicativity of norms for phase-related parameters. The methodology relies on the following core concepts:

Ring-Sectorial Bounds: Instead of bounding the numerical range $W(A)$ solely by a phase sector $\Phi(A)$ , the method utilizes a set of six functions $S(A)$ that bound both $W(A)$ and $W(A^{-1})$ . This set includes the numerical radius $r(A)$ , the Crawford number $c(A)$ (the distance from the origin to $W(A)$ ), and the phase sector $\Phi(A)$ . These parameters define "ring-sectorial" bounds, effectively constraining the numerical range to a ring sector.
Crawford Defect: To handle the non-submultiplicativity of phase, the paper introduces the "Crawford defect" $\delta(A) = \sqrt{\frac{1}{c(A)c(A^{-1})} - 1}$ . This quantity measures the relative radial spread of the numerical range. The defect is zero if and only if the matrix is a scalar multiple of a unitary matrix.
Composition Rules: The paper derives specific theorems for composing the bounds in $S(A)$ $S (A)$ across elementary interconnections:
- Series: Theorem 3 provides bounds for the product of matrices. It shows that if the product of the Crawford defects of the subsystems is sufficiently small, the resulting system remains sectorial, and its bounds can be calculated iteratively.
- Parallel: Theorem 4 establishes bounds for the sum of matrices using Minkowski sums of their ring-sectorial bounds.
- Feedback: Theorem 5 extends the analysis to feedback loops, providing alternative bounds (phase-gap and high-gain bounds) when the standard small-gain condition is not met.

The analysis proceeds hierarchically: starting from the bottom-level subsystems, bounds are composed upwards through the interconnection structure. At each feedback level, stability is verified using the small-gain-or-phase condition (Theorem 1), and new bounds for the composite system are generated for the next level.

Key Contributions

Theoretical Extension of Small-Phase Analysis: The primary technical contribution is Theorem 3, which establishes that the set of ring-sectorial bounds $S(\cdot)$ is composable under series interconnection, provided the Crawford defects are small. This effectively generalizes the small-phase theorem to arbitrary numbers of subsystems in a loop, a capability previously blocked by the lack of submultiplicativity in phase sectors.
Unified Compositional Framework: The paper integrates small-gain and small-phase analysis into a single iterative algorithm (Section V-A). This allows for the stability verification of decomposable systems (those formed by elementary interconnections) where neither pure gain nor pure phase analysis alone would suffice.
High-Gain Stability Insight: The analysis reveals that the compositional approach can detect stability in high-gain regimes (where the norm is large) that are invisible to classical small-gain analysis. This is achieved by leveraging the Crawford number to ensure the numerical range is "far enough from the origin," which, upon inversion in the feedback formula, translates to a small gain.
Connection to Segmental Phase: The paper discusses the relationship between the proposed ring-sectorial bounds and the recently introduced "segmental phase" [8]. It suggests that the Crawford defect serves as an upper bound for the normalized Crawford defect, which relates directly to the segmental phase radius, offering a potential path to reduce conservatism by combining sectorial and segmental approaches.

Results and Examples
The paper validates the method through three analytical examples:

Single-Loop Perturbation: A system with a near-unity feedback perturbation is analyzed. The compositional approach yields a larger stability domain than the classical small-gain-or-phase theorem, specifically identifying a "high-gain-and-small-phase" region where the system remains stable despite high loop gains.
Subsystems with Small Feedback Perturbations: A two-loop system where each subsystem has a small inner feedback perturbation. The method successfully derives sufficient conditions for internal stability by composing the bounds of the inner loops before analyzing the outer loop.
Multiplicative Perturbations: A system with multiplicative perturbations on subsystems. The results demonstrate how the compositional approach handles the accumulation of phase errors and norm bounds, providing a new small-phase condition that complements the classical small-gain condition.

Significance and Claims
The paper claims that its main conceptual contribution is the ability to apply small-gain-or-phase stability conditions to iterated compositions of series, parallel, and feedback interconnections. The author notes that while the composition formulas can be complex and potentially conservative, they provide analytically manageable and intuitively convincing results for "shallow" systems.

The significance lies in extending the applicability of phase-based stability analysis to complex networks, a task where the classical small-phase theorem previously failed due to the inability to iterate spectral inclusions. The method offers a systematic way to identify the source of instability (e.g., a specific internal loop violating conditions) and suggests that combining sectorial and segmental approaches could further reduce conservatism. The paper does not claim to solve the stability problem for all non-decomposable systems but focuses on the class of systems that can be rendered decomposable via block diagram transformations.

More like this