Nyquist stability criterion
The Nyquist stability criterion is a graphical frequency-domain method in control theory for determining the stability of linear time-invariant feedback systems by analyzing the open-loop transfer function's response along the imaginary axis of the complex plane.[1] Introduced by electrical engineer Harry Nyquist in his seminal 1932 paper "Regeneration Theory," the criterion leverages the argument principle from complex analysis to evaluate closed-loop stability without solving for the roots of the characteristic equation. At its core, the method involves constructing the Nyquist plot, a polar representation of the complex-valued open-loop transfer function L(j\omega) = G(j\omega)H(j\omega) (where G(s) is the forward path and H(s) the feedback path) as the frequency \omega varies from -\infty to +\infty.[2] The plot is completed by a semicircular arc in the right-half s-plane at infinity, enclosing the entire right-half plane via a contour known as the Nyquist D-contour; for systems with poles on the imaginary axis, small indentations avoid them.[3] Stability is assessed by counting the net number of counterclockwise encirclements (denoted N) of the critical point -1 + j0 by the Nyquist plot.[1] The number of unstable closed-loop poles Z (right-half plane zeros of $1 + L(s)) is given by Z = P + N, where P is the number of unstable open-loop poles (right-half plane poles of L(s)).[2] For asymptotic stability, Z = 0 must hold, meaning N = -P; if P = 0, the plot must not encircle -1 at all.[3] Beyond binary stability, the criterion yields quantitative measures of relative stability through gain margin (the factor by which gain can increase before instability at the phase crossover frequency, where phase is -180^\circ) and phase margin (the additional phase lag tolerable at the gain crossover frequency, where magnitude is 1).[1] These margins guide controller design in applications ranging from electronics and aerospace to chemical processes, ensuring robust performance against uncertainties.[2] The approach remains influential due to its intuitive visualization and compatibility with experimental frequency data, though it assumes linear time-invariant dynamics and requires careful handling of non-minimum phase systems or delays.[3]Fundamentals of Feedback Systems
Historical Background
The Nyquist stability criterion originated in the work of Swedish-American engineer Harry Nyquist at Bell Laboratories, where he addressed the challenges of ensuring stability in feedback amplifiers for long-distance telephony during the early 1930s. At the time, telephone networks relied on vacuum-tube repeaters with negative feedback to amplify signals over thousands of miles, but these systems were prone to parasitic oscillations that could generate howling noises and degrade transmission quality. Nyquist's motivation stemmed from practical engineering needs at Bell Labs to design reliable amplifiers that avoided such instability without relying on time-consuming transient simulations or root-locus methods, which were not yet developed.[4][5] Nyquist's breakthrough came in his 1932 paper "Regeneration Theory," published in the Bell System Technical Journal, which introduced a frequency-response-based approach to stability analysis for linear feedback systems. In this work, he proposed plotting the complex gain of the open-loop transfer function as a function of frequency to determine if the closed-loop system would exhibit bounded outputs for bounded inputs, effectively providing a contour-based test for encirclements around the critical point -1 in the complex plane. This graphical method marked a shift from time-domain to frequency-domain techniques, offering engineers a direct way to assess amplifier stability using measurable frequency responses.[6] The criterion gained wider influence through subsequent developments at Bell Labs, particularly Hendrik Wade Bode's complementary work on frequency-domain tools in the late 1930s and 1940s. Bode's 1945 book, Network Analysis and Feedback Amplifier Design, built directly on Nyquist's foundation by incorporating logarithmic magnitude and phase plots (now known as Bode plots) to simplify the application of the stability test for amplifier design. By the mid-1940s, the Nyquist criterion had evolved from a telephony-specific tool into a foundational method for evaluating the stability of general linear time-invariant feedback systems in control engineering.[7][8] Later mathematical interpretations linked Nyquist's empirical criterion to Cauchy's argument principle in complex analysis, formalizing its theoretical underpinnings for broader academic adoption.[4]Open-Loop and Closed-Loop Transfer Functions
In feedback control systems, the open-loop transfer function is defined as L(s) = G(s) H(s), where G(s) represents the transfer function of the plant (the process or system being controlled) and H(s) represents the transfer function of the controller or feedback element.[9] This formulation captures the combined dynamics of the forward path in the absence of feedback closure, allowing analysis of the system's inherent response to inputs without loop interaction.[10] The closed-loop transfer function, which describes the overall system behavior from reference input to output under feedback, is given by T(s) = \frac{L(s)}{1 + L(s)} = \frac{G(s) H(s)}{1 + G(s) H(s)}. The denominator $1 + L(s) = 0 forms the characteristic equation, whose roots determine the closed-loop poles.[11] These poles govern the transient response and stability of the system. For continuous-time linear time-invariant (LTI) systems, asymptotic stability requires that all closed-loop poles—solutions to the characteristic equation—lie in the open left-half of the complex s-plane, meaning they have strictly negative real parts.[12] This ensures that system responses decay to zero over time following bounded inputs.[9] Open-loop stability, in contrast, depends on the poles of L(s) itself (the poles of G(s) and H(s)), which may or may not coincide with closed-loop stability; feedback can relocate poles to achieve stability even if the open-loop system has right-half-plane (RHP) poles.[11] Closed-loop poles in the RHP (positive real parts) indicate instability, leading to unbounded responses, while RHP zeros of L(s) (roots of the numerator) do not directly cause instability but can introduce limitations such as undershoot or inverse response in the closed-loop dynamics.[12] Thus, the Nyquist criterion arises to assess closed-loop stability without explicitly solving for these poles.[9]The Nyquist Plot
Construction and Properties
The Nyquist plot is a polar representation of the open-loop transfer function L(j\omega) = G(j\omega) H(j\omega), where G(s) and H(s) are the plant and feedback path transfer functions, respectively, plotted in the complex plane as the frequency \omega varies from $0 to \infty.[1] To construct the plot, the real part \operatorname{Re}[L(j\omega)] and imaginary part \operatorname{Im}[L(j\omega)] are computed for a range of positive frequencies \omega, typically using the substitution s = j\omega into the transfer function and separating the resulting expression into its Cartesian components.[1] These components are then plotted parametrically, with \operatorname{Re}[L(j\omega)] on the horizontal axis and \operatorname{Im}[L(j\omega)] on the vertical axis, tracing a curve as \omega increases; the portion for negative frequencies (\omega from -\infty to $0) is obtained by reflecting the positive-frequency curve across the real axis due to the conjugate symmetry property L(-j\omega) = \overline{L(j\omega)}.[1] For transfer functions that are real-rational (i.e., with real coefficients and rational form), the Nyquist plot possesses symmetry about the real axis, ensuring the full curve from \omega = -\infty to \infty forms a complete, mirror-symmetric locus.[1] At low frequencies (\omega \to 0), the plot approaches the DC gain L(0), which lies on the positive real axis for stable minimum-phase systems, indicating steady-state behavior.[1] As \omega \to \infty, the plot typically rolls off toward the origin for proper transfer functions where the degree of the denominator exceeds that of the numerator, reflecting the attenuation of high-frequency components.[1] In cases where the degree of the denominator exceeds the degree of the numerator by at least 2 (relative degree ≥ 2), the Nyquist plot construction may include the mapping of a large semicircular arc at infinity along the contour, which contracts to the origin in the L-plane, simplifying the overall diagram by confining the high-frequency behavior to a point.[8]Frequency Response Interpretation
In the Nyquist plot, the open-loop frequency response L(j\omega) is visualized in the complex plane, with the magnitude |L(j\omega)| represented as the radial distance from the origin to the curve at each frequency \omega, and the phase \angle L(j\omega) indicated by the angle that the vector from the origin to the point makes with the positive real axis.[13] This polar representation allows engineers to assess how the system's gain and phase vary across the frequency spectrum in a single diagram, contrasting with linear Bode plots that separate magnitude and phase on logarithmic scales.[14] At low frequencies (\omega \approx 0), the plot's position reflects the system's DC gain and type; for instance, a type-0 system starts near a point on the positive real axis proportional to its steady-state gain, while a type-1 system with an integrator pole at the origin extends along the negative imaginary axis toward -\jmath \infty, signifying infinite low-frequency gain essential for tracking constant references.[13] As frequency increases to high values (\omega \to \infty), the plot spirals inward toward the origin for strictly proper systems, where the relative degree (number of poles minus zeros) determines the approach angle—at -90° times the relative degree, which for even relative degree lies along the real axis (positive or negative depending on the value modulo 4)—illustrating the attenuation of high-frequency disturbances.[14] These asymptotic behaviors provide quick qualitative checks on bandwidth and roll-off characteristics without numerical computation.[13] Key features include the gain crossover frequency, defined as the \omega where |L(j\omega)| = 1, marking intersections of the plot with the unit circle centered at the origin, and the phase crossover frequency, where \angle L(j\omega) = -180^\circ, corresponding to crossings of the negative real axis.[14] These points highlight frequencies where gain equals unity or phase inversion occurs, influencing bandwidth and potential phase lag accumulation. The plot's proximity to the critical point -1 offers insights into dynamic robustness; curves passing close to this point suggest high sensitivity to parameter variations, while loops or indentations reveal resonant modes where magnitude peaks and phase shifts abruptly, amplifying specific frequency components.[13] A representative example is the first-order low-pass system L(s) = \frac{K}{\tau s + 1}, whose Nyquist plot traces a semicircle of radius K/2 in the right-half complex plane, starting at K on the real axis for \omega = 0 and ending at the origin as \omega \to \infty, with the phase varying smoothly from $0^\circ to -90^\circ.[13] This simple shape underscores the non-oscillatory damping typical of first-order dynamics, originally analyzed in telephony contexts to avoid amplifier oscillations.[15]Mathematical Foundations
Cauchy's Argument Principle
Cauchy's argument principle is a fundamental theorem in complex analysis that relates the number of zeros and poles of a meromorphic function enclosed by a closed contour to an integral over that contour.[16] The principle states that if f(s) is a meromorphic function in a domain containing the closed contour \Gamma and analytic inside \Gamma except possibly at isolated poles, then \frac{1}{2\pi i} \oint_{\Gamma} \frac{df(s)}{f(s)} = Z - P, where Z denotes the number of zeros of f(s) inside \Gamma (counted with multiplicity) and P denotes the number of poles inside \Gamma (also counted with multiplicity).[16] This formulation arises as a direct consequence of the residue theorem applied to the logarithmic derivative df/f, where the residues at zeros contribute positively and at poles negatively.[16] A key variant of the argument principle expresses this relation in terms of the change in the argument of f(s) along the contour. Specifically, as s traverses the positively oriented (counterclockwise) closed contour \Gamma, the total change in \arg(f(s)) is $2\pi (N_z - N_p), where N_z is the number of zeros enclosed by \Gamma and N_p is the number of poles enclosed, both counted with multiplicity.[16] This change in argument corresponds to the winding number of the image curve f(\Gamma) around the origin in the complex plane.[16] The theorem assumes that f(s) is meromorphic in the region bounded by \Gamma, meaning it is analytic except at isolated poles, and that \Gamma is a simple closed contour oriented counterclockwise with no singularities of f(s) on \Gamma itself.[16] If singularities lie on the contour, the principle requires modifications, such as indentations around those points.[16] A sketch of the proof begins with the residue theorem: the integral \oint_{\Gamma} \frac{df(s)}{f(s)} = 2\pi i \sum \operatorname{Res} \left( \frac{f'(s)}{f(s)}, s_k \right), where the sum is over singularities inside \Gamma. At a zero of order m, the residue is m; at a pole of order m, it is -m. Thus, the integral equals $2\pi i (Z - P), yielding the stated form upon division by $2\pi i. This links directly to the winding number interpretation via the fact that \frac{1}{2\pi i} \oint \frac{df}{f} measures the net encirclements of the origin by f(\Gamma).[16] The argument principle is attributed to Augustin-Louis Cauchy, who developed its core ideas in his 1831 memoir on definite integrals and their applications to celestial mechanics, though the full theorem appeared in later publications.[17] In the context of control theory, Leroy A. MacColl first applied the principle to analyze feedback systems in his 1945 book Fundamental Theory of Servomechanisms.[8]Nyquist Contour and Mapping
The Nyquist D-contour, denoted as Γ, is a closed path in the s-plane designed to enclose the entire right-half plane (RHP), where potential unstable closed-loop poles may reside. It consists of the imaginary axis traversed from s = -jR to s = jR, followed by a large semicircular arc of radius R in the RHP, with the limit taken as R approaches infinity. This contour ensures that any zeros of the characteristic function in the open RHP are captured inside Γ.[4][18] To handle open-loop poles on the imaginary axis, the contour includes small semicircular indentations of radius ε (with ε → 0) detouring to the right of these poles, avoiding their inclusion on the path itself. These indentations ensure the contour remains in the stable region while encircling the RHP. For the mapping, the image of Γ under the function f(s) = 1 + L(s), where L(s) is the open-loop transfer function, forms a closed curve in the complex plane whose winding number around the origin equals the difference between the number of zeros and poles of f(s) inside Γ, per the argument principle.[4][14][18] The mapping of the imaginary axis portion under L(s) produces the standard Nyquist plot, a polar representation of the frequency response as ω varies from -∞ to ∞. The large semicircular arc in the s-plane, for strictly proper L(s) (degree of denominator exceeds numerator by at least one), maps to a point at the origin in the L(s)-plane as R → ∞, due to |L(s)| → 0 for large |s|. Indentations around imaginary-axis poles map to large arcs that may encircle significant portions of the complex plane, reflecting the pole's order and phase contribution.[4][14]Statement of the Criterion
General Formulation
The Nyquist stability criterion is a graphical technique for assessing the stability of a linear time-invariant feedback system by examining the frequency response of the open-loop transfer function L(s). It determines whether the closed-loop system has any poles in the right half-plane (RHP) based on the behavior of the Nyquist plot, which maps the imaginary axis in the s-plane to the complex L-plane.[2] The criterion originates from the work of Harry Nyquist in his seminal 1932 paper on regeneration theory, where he introduced the foundational graphical method for stability analysis in feedback amplifiers.[6] The general formulation of the criterion is given by the equation Z = P + N, where Z is the number of RHP poles of the closed-loop transfer function (i.e., the zeros of $1 + L(s) in the RHP), P is the number of RHP poles of the open-loop transfer function L(s), and N is the net number of clockwise encirclements of the critical point -1 + j0 by the Nyquist plot of L(j\omega) as \omega varies from -\infty to \infty.[2] For closed-loop stability, Z = 0 is required, implying N = -P. The encirclement count N = -\frac{1}{2\pi} \Delta \arg [1 + L(j\omega)], where the change in argument is taken as \omega varies from -\infty to +\infty, reflecting the signed winding number around the critical point (positive for clockwise).[2] This relation holds under the assumption that the Nyquist contour—a large semicircle enclosing the RHP—maps to a plot that encircles -1 + j0 in a manner consistent with the argument principle.[19] Key assumptions for this formulation include that L(s) is strictly proper (degree of numerator less than degree of denominator), ensuring the Nyquist plot approaches the origin as |\omega| \to \infty, and that L(s) has no poles on the imaginary axis, allowing the standard contour without modifications.[2][2] Additionally, the closed-loop system must have no poles on the imaginary axis, so the Nyquist plot does not pass through -1 + j0.[2] Violations of these assumptions require adjustments, such as contour indentations, but the core equation remains applicable.[19] For minimum-phase systems, where P = 0 (no open-loop RHP poles), the criterion simplifies: the closed-loop system is stable if and only if N = 0, meaning the Nyquist plot does not encircle -1 + j0.[2][19] In practice, this often manifests as the plot lying entirely to the right of the line \Re(L(j\omega)) = -1 in the complex plane, avoiding any looping around the critical point. For example, consider an open-loop stable system with L(s) = \frac{1}{s+1}; its Nyquist plot is a circle of radius 1/2 centered at 1/2 + j0, lying entirely in the right half-plane (Re(L) ≥ 0) and thus staying to the right of -1 + j0, confirming closed-loop stability since P = 0 and N = 0.[2]Encirclement and Stability Conditions
In the Nyquist stability criterion, encirclements of the critical point -1 + j0 by the plot of the open-loop transfer function L(j\omega) are counted to assess closed-loop stability. The net number of encirclements N is determined by tracing the direction of loops around this point: clockwise encirclements are counted as positive, while counterclockwise ones are negative. To perform the count, vectors are drawn from -1 + j0 to successive points on the Nyquist plot as \omega varies from -\infty to \infty, and N = - (the total angular change / 2π), since the angular change is positive for counterclockwise direction.[2][4] The stability condition requires that the net encirclements satisfy N = -P, where P is the number of right-half-plane (RHP) poles of the open-loop transfer function L(s); this ensures zero RHP poles in the closed-loop characteristic equation $1 + L(s) = 0, as per the relation Z = P + N with Z = 0 for stability. If N > -P, the closed-loop system has N + P > 0 RHP poles and is unstable, indicating potential oscillations or divergence. Open-loop pole locations to find P are typically determined via Routh-Hurwitz criterion or root locus analysis prior to plotting.[1][2] The Nyquist plot also provides measures of relative stability through gain and phase margins. The gain margin is the factor by which the open-loop gain can be increased before instability, calculated as the reciprocal of the magnitude |L(j\omega_{pc})| at the phase crossover frequency \omega_{pc} where \angle L(j\omega_{pc}) = -180^\circ; in the plot, this corresponds to the distance from the origin to the intersection with the negative real axis beyond -1 + j0. The phase margin is the additional phase lag needed to reach instability, given by $180^\circ + \angle L(j\omega_{gc}) at the gain crossover frequency \omega_{gc} where |L(j\omega_{gc})| = 1; on the plot, it is the angular deviation from the negative real axis to the ray from the origin through the unit-circle intersection point. Larger margins indicate greater robustness to parameter variations.[2][4] To apply the criterion, construct the Nyquist plot of L(j\omega) for \omega from 0 to \infty (mirroring for negative frequencies if the system is real), count the full net encirclements N of -1 + j0, and compare to -P from open-loop analysis; the plot must not pass through the critical point for a defined result.[1][2]Special Cases
Poles on the Imaginary Axis
When the open-loop transfer function L(s) has poles on the imaginary axis, the standard Nyquist contour traverses these singularities along the j\omega-axis, rendering L(s) undefined at those points. This issue is resolved by modifying the contour with small semicircular indentations of radius \epsilon \to 0 detoured to the right into the right-half plane (RHP) around each pole at s = \pm j\omega_k.[2][20] These indentations are traversed clockwise to ensure the modified contour encloses the open RHP while excluding the imaginary-axis poles. In the Nyquist plot, each indentation around a simple pole maps to a large semicircle of radius approaching infinity, encircling the origin in the clockwise direction. This mapping arises because, near the pole, L(s) behaves as L(s) \approx \frac{r}{s - j\omega_k} (where r is the residue), resulting in a phase variation of -180^\circ over the small arc.[2][20][1] The stability criterion remains Z = N + P, where Z is the number of RHP closed-loop poles, N is the net number of clockwise encirclements of the point -1 + j0 by the full Nyquist plot (including the large semicircles from indentations), and P counts only the open-loop RHP poles (excluding those on the imaginary axis, as the indentations leave them outside the contour). The system is stable if Z = 0. The general encirclement rule applies after incorporating these modifications.[2][20] A representative example is the integrator L(s) = 1/s, with a simple pole at the origin. The principal Nyquist plot follows the negative imaginary axis from $0 to -\j\infty (for \omega > 0), while the indentation around the origin connects the ends at infinity via a large clockwise semicircle from +\j\infty (negative frequencies) through the positive real axis to -\j\infty, reflecting the -90^\circ phase of the integrator and ensuring proper encirclement assessment for stability.[1][21]Jordan Poles and Indentations
When the open-loop transfer function L(s) has poles of multiplicity m > 1 located at the same point j\omega_k on the imaginary axis, these higher-order poles necessitate modifications to the standard Nyquist contour to avoid them while preserving the applicability of Cauchy's argument principle.[20] Unlike simple poles, higher-order poles induce larger phase shifts in the mapping of the contour under L(s), requiring careful handling in the indentation to ensure the contour encloses only the right-half plane poles without passing through singularities.[22] The indentation around such a pole is constructed as a small semicircle in the s-plane with radius \epsilon \to 0 into the RHP, traversed clockwise. This ensures that the change in argument of s - j\omega_k along the arc is \pi radians, leading to a mapping under L(s), which near the pole behaves as L(s) \approx c / (s - j\omega_k)^m for some constant c. As \epsilon \to 0, the image of this semicircle under L(s) approaches an arc at infinity with angular extent -m \times 180^\circ, corresponding to m clockwise semicircles or equivalent winding around the origin in the L-plane.[20][23][3] In applying the Nyquist criterion, poles on the imaginary axis are excluded from the open-loop pole count P, which includes only right-half plane poles; the indentations ensure these imaginary poles are outside the enclosed region, but their multiplicity amplifies the argument change in the plot. The number of encirclements N is determined from the full Nyquist plot, including the infinite arc, with the stability condition Z = P + N = 0.[1][2] A representative example occurs in type-2 control systems, where L(s) has a double pole (m = 2) at the origin, such as L(s) = K / s^2. The indentation is a small semicircle around the origin, mapping to a -360^\circ (full clockwise circle) arc at infinity in the Nyquist plot of L(j\omega), which traces the negative real axis for \omega > 0 and connects via the full circular arc, potentially encircling the -1 point depending on K.[21] This configuration highlights how the multiplicity doubles the phase excursion, demanding precise gain tuning to achieve N = -P for closed-loop stability.[20]Derivation
Proof Using Contour Integration
The proof of the Nyquist stability criterion begins by applying Cauchy's argument principle to the function $1 + L(s), where L(s) denotes the open-loop transfer function of the feedback system.[24] The argument principle states that for a meromorphic function f(s) analytic inside and on a simple closed contour \Gamma except for isolated poles, for a counterclockwise \Gamma, the number of zeros Z minus the number of poles P inside \Gamma is given by \frac{1}{2\pi i} \oint_\Gamma \frac{f'(s)}{f(s)} \, ds = Z - P, which equals the net number of counterclockwise encirclements of the origin by f(\Gamma).[13] Consider the Nyquist contour \Gamma, a closed path that encloses the open right half-plane: it traverses the imaginary axis from -jR to jR, followed by a large semicircular arc of radius R in the right half-plane from jR to -jR, with R \to \infty, and includes small semicircular indentations of radius \epsilon \to 0 into the right half-plane around any poles of L(s) on the imaginary axis to maintain analyticity of $1 + L(s) inside \Gamma.[24] Since \Gamma is traversed clockwise, the integral becomes P' - Z = \frac{1}{2\pi i} \oint_\Gamma \frac{f'(s)}{f(s)} \, ds, where Z is the number of zeros of $1 + L(s) (i.e., closed-loop poles) in the right half-plane, and P' is the number of poles of $1 + L(s) in the right half-plane. The poles of $1 + L(s) coincide with those of L(s) except at points where L(s) has a pole but $1 + L(s) = 0; under standard assumptions where such coincidences do not occur, P' = P, the number of open-loop right half-plane poles of L(s).[13] This integral equals \frac{1}{2\pi} \Delta_\Gamma \arg(1 + L(s)) = -N_{1+L}, where N_{1+L} is the net number of counterclockwise encirclements of the origin by the image curve $1 + L(\Gamma) (the negative sign arises because the clockwise contour reverses the winding direction relative to the standard counterclockwise case). Thus, P - Z = -N_{1+L}, or Z = P + N_{1+L}. As R \to \infty, the contribution from the large semicircular arc vanishes because |L(s)| \to 0 along s = Re^{i\theta} for \theta \in [-\pi/2, \pi/2], provided the relative degree of L(s) (denominator degree minus numerator degree) is at least 2, ensuring the argument change on this arc approaches 0. The small indentations around imaginary-axis poles contribute a phase shift of \pi per simple pole (or multiples thereof for higher-order poles), but these are handled separately in the Nyquist plot construction to preserve the total argument change. Thus, the integral over \Gamma reduces to the argument change along the imaginary axis s = j\omega, \omega: -\infty \to \infty.[24] Along the imaginary axis, the mapping s = j\omega traces the Nyquist plot of L(j\omega), and the image under $1 + L(j\omega) encircles the origin precisely when L(j\omega) encircles the critical point -1 + 0j. Specifically, the net encirclements of the origin by $1 + L(j\omega) equal the net encirclements of -1 by L(j\omega), so N_{1+L} = N_{-1}, where N denotes the net counterclockwise encirclements of -1 by the Nyquist plot of L(j\omega).[13] Therefore, Z = P + N. The closed-loop system is stable if and only if Z = 0, requiring N = -P. The indentations ensure the contour avoids singularities while the limit R \to \infty captures all right half-plane dynamics.[24]Relation to Characteristic Equation
The characteristic equation for a single-input single-output (SISO) feedback control system with open-loop transfer function L(s) and unity feedback is given by $1 + L(s) = 0.[13] The roots of this equation, known as the closed-loop poles, determine the system's stability; specifically, any roots in the right-half plane (RHP) of the complex s-plane indicate instability.[2] For multiple-input multiple-output (MIMO) systems, the analogous equation is \det(I + G(s)) = 0, where G(s) is the open-loop transfer matrix, though the Nyquist criterion is most commonly applied to SISO cases.[13] When L(s) is a rational function expressed as L(s) = N(s)/D(s), with N(s) and D(s) as coprime polynomials, the characteristic equation becomes [D(s) + N(s)] / D(s) = 0, so the zeros of the numerator polynomial D(s) + N(s) are the closed-loop poles, provided no unstable cancellations occur between numerator and denominator.[1] These RHP zeros directly correspond to unstable closed-loop behavior, linking the time-domain pole locations to frequency-domain analysis.[2] The Nyquist stability criterion serves as a frequency-domain alternative to directly solving the characteristic polynomial for its roots, which can be computationally intensive for high-order systems.[13] By mapping the Nyquist contour in the s-plane to the L(s)-plane and counting the encirclements of the critical point -1, the criterion determines the number Z of RHP zeros of $1 + L(s), where Z = P + N (P is the number of RHP poles of L(s), and N is the net number of counterclockwise encirclements).[1] Thus, for closed-loop stability, Z = 0 must hold, ensuring no RHP roots of the characteristic equation.[2]Applications and Limitations
Practical Implementation in Design
In practical control system design, the Nyquist stability criterion is applied through a structured workflow to assess and ensure closed-loop stability. The process begins by computing the loop transfer function L(s), which is the product of the plant and controller transfer functions.[13] Next, the Nyquist plot is generated by evaluating L(j\omega) for frequencies \omega from 0 to \infty, along with the necessary indentations around imaginary-axis poles if present.[13] The number of right-half-plane poles P of L(s) is determined from the open-loop system. The number of clockwise encirclements N of the critical point -1 by the Nyquist plot is then counted, and stability is confirmed if the number of unstable closed-loop poles Z = P + N = 0.[13] Software tools facilitate this implementation by automating plotting and stability calculations. In MATLAB, thenyquist() function from the Control System Toolbox generates the Nyquist plot and can display stability margins, such as gain and phase margins, directly on the diagram. Similarly, the Python Control Systems Library provides the nyquist_plot() function, which computes the response along the Nyquist contour and returns data for encirclement analysis, including automated margin evaluation via stability_margins(). These tools enable rapid iteration during design, allowing engineers to simulate adjustments to L(s) and verify Z = 0 without manual computation.
In design applications, the Nyquist plot guides controller shaping to prevent encirclement of the -1 point, thereby ensuring stability. For instance, adding a lead compensator introduces phase lead around the gain crossover frequency, rotating the Nyquist curve counterclockwise to increase the phase margin and distance from -1.[25] This is particularly useful for systems requiring improved transient response, where the compensator C(s) = k \frac{s + z}{s + p} (with z < p) is tuned so the maximum phase lead occurs near the desired crossover, avoiding instability.[25]
A representative case study involves stabilizing a conditionally stable system, where stability depends on gain levels and the Nyquist plot may encircle -1 for certain parameters. Consider a loop transfer function like L(s) = \frac{3(s+1)^2}{s(s+6)^2}, whose Nyquist plot intersects the negative real axis multiple times, potentially leading to instability if gain is too low or high.[26] By adjusting the gain to position the plot such that it does not encircle -1 (e.g., ensuring the curve passes to the right of -1 without net clockwise loops), the system achieves N = 0 and Z = 0, stabilizing it across operating ranges.[26] This approach highlights how Nyquist analysis reveals conditional stability and informs gain scheduling for robust performance.[26]
The Nyquist criterion offers advantages in handling non-minimum phase systems, where right-half-plane zeros complicate other methods like root locus, which may not fully capture stability due to branch asymmetries.[13] In such cases, the frequency-domain encirclement count directly assesses closed-loop poles without requiring pole-zero mapping, making it suitable for systems with delays or non-minimum phase behavior.[13]