The Nyquist plot is a cornerstone of classical control theory, offering engineers an elegant way to assess system stability from frequency-response data. By mapping the open-loop transfer function G(s)H(s) along the Nyquist contour, the Nyquist stability criterion provides a clear graphical condition: the number of closed-loop poles in the right-half plane equals the number of open-loop poles in the right-half plane plus the number of clockwise encirclements of the critical point (-1, 0). This method is powerful for minimum-phase systems, where all zeros lie in the left-half plane. However, when the system contains right-half-plane (RHP) zeros—making it non-minimum phase—the Nyquist plot’s reliability diminishes. Engineers must understand these limitations to avoid misinterpretation and ensure robust control design. This article examines the specific challenges Nyquist plots pose for non-minimum phase systems, explains why they occur, and recommends complementary analysis techniques.

Understanding Non-Minimum Phase Systems

A system is called non-minimum phase if its transfer function contains zeros in the right half of the complex plane. These RHP zeros introduce a distinctive phase lag greater than 90° per zero at high frequencies, and they cause the time-domain step response to exhibit an initial “undershoot” or inverse response—the output first moves in the opposite direction of the steady-state value before reversing and settling. Common examples include:

  • Inverted pendulums and balancing systems: The initial motion of the pivot away from the desired direction.
  • Aircraft pitch dynamics with canard configurations: A downward elevator deflection can produce an initial upward pitch.
  • Chemical process control (e.g., CSTRs): The concentration response to a feed change often shows an inverse transient.
  • Driveline oscillations in vehicles: Gear backlash and compliance produce non-minimum phase behavior in torque transfer.

From a mathematical standpoint, a transfer function with a zero at s = +a (a > 0) contributes a phase angle that decreases further with frequency, unbounded phase lag. In contrast, minimum-phase zeros produce phase lead that asymptotically cancels pole lag. This phase asymmetry is the root cause of the complications in Nyquist analysis.

Nyquist Plots: Fundamentals and Assumptions

The Nyquist plot is constructed by evaluating G(s)H(s) along the imaginary axis (s = jω) and then closing the contour around the right-half plane via an infinite semicircle. The resulting curve shows the real and imaginary parts of the open-loop frequency response. The Nyquist stability criterion states that the number of closed-loop unstable poles Z equals the number of open-loop unstable poles P plus the number of clockwise encirclements N of the point -1: Z = P + N.

This criterion assumes that the Nyquist plot accurately reflects the phase shift contributed by poles and zeros and that the shape of the plot near the critical point directly reveals stability margins. For minimum-phase systems, the plot’s proximity to -1 corresponds well to gain and phase margins derived from Bode plots. But when RHP zeros are present, the phase contribution is “excessively” lagging, distorting the Nyquist contour in ways that can mislead.

Key Limitations of Nyquist Plots for Non-Minimum Phase Systems

Ambiguity in Interpreting Inverse Response

The inverse response of non-minimum phase systems does not manifest directly as a distinct feature in the Nyquist plot. The plot is purely frequency-domain, whereas the inverse response is a time-domain phenomenon. Engineers may see a looping or “nose” in the Nyquist curve at low frequencies, but that shape can arise from various pole-zero configurations. Without additional time-domain simulation, one cannot confidently attribute such features to non-minimum phase behavior. For example, a system with a RHP zero at s = 1 and a stable pole at s = -2 produces a Nyquist plot that begins at Re = -0.5 on the real axis (low frequency) and then bends into the third quadrant—a pattern that might be mistakenly interpreted as a simple phase lag from a slow pole. Only by checking the step response does the initial undershoot become apparent.

Complexity in Encirclement Analysis

The Nyquist criterion relies on the number of encirclements of -1. For non-minimum phase systems, the “excess” phase lag from RHP zeros can cause the Nyquist contour to wrap around the critical point multiple times or in a counter-intuitive direction. Specifically:

  • Additional phase lag: Each RHP zero adds 180° of phase lag, which can make the Nyquist curve cross the negative real axis several times, creating multiple potential critical points.
  • Distorted mapping of the infinite semicircle: For systems with more zeros than poles, the Nyquist plot’s behavior at infinite frequency becomes ambiguous. In non-minimum phase systems, the net number of poles minus zeros (relative degree) still determines the high-frequency asymptote, but the phase at infinity may be offset.
  • Encircling the point -1 without instability: A non-minimum phase open-loop system may have a Nyquist plot that encircles -1 even when the closed-loop is stable, if the open-loop has RHP zeros but no RHP poles. The criterion still holds (Z = P + N), but the visual interpretation is easily misapplied.

For instance, consider an open-loop transfer function with a RHP zero: G(s) = (s - 1)/(s^2 + 2s + 2). Its Nyquist plot crosses the negative real axis near -0.3 and again near -1.2, but does it encircle -1? The answer depends on the exact path, and a single glance can be misleading. Manual inspection of the net encirclements requires careful counting of crossings.

Misleading Stability Margin Calculations

Gain margin and phase margin are typically derived from the Nyquist plot’s distance to the critical point. For non-minimum phase systems, these margins may be overly optimistic or pessimistic:

  • Gain margin ambiguity: The gain margin is defined as the reciprocal of the magnitude at the phase crossover frequency (ω_π where phase = -180°). In non-minimum phase systems, there may be multiple phase crossover frequencies, and the smallest gain margin may not correspond to the most critical instability. A system could have a large gain margin based on the first crossing but become unstable if the gain is increased only moderately, because a later crossing is closer to -1.
  • Phase margin distortion: The phase margin is measured at the gain crossover frequency (ω_c where magnitude = 0 dB). Due to the extra phase lag, the gain crossover frequency might be much lower than expected, yielding a phase margin that seems comfortable but belies a fragile closed-loop response to disturbances. The system might exhibit poor damping or large overshoot in the time domain despite a high phase margin.
  • No single “margin” captures the full risk: Because non-minimum phase zeros impose an inherent bandwidth limitation, traditional margins can fail to predict phenomena like waterbed effect or sensitivity peaking. The peak of the complementary sensitivity function (T_max) is often a better indicator of robustness for non-minimum phase systems.

An illustrative example is a system with a RHP zero near s = +0.5 and lightly damped poles. The Nyquist plot may pass close to -1 at a low frequency, indicating a small gain margin, but the actual closed-loop may be stable with acceptable response. Conversely, another system may show a large margin yet be impossible to control robustly because the RHP zero limits the achievable bandwidth (as predicted by the Bode gain-phase relation).

Sensitivity to Modeling Errors and Parameter Variations

Non-minimum phase zeros often arise from physical effects that are difficult to model accurately—such as flexibility, delay, or non-collocated sensors and actuators. A Nyquist plot based on an idealized model may show a clear stability margin, but small changes in the zero location (or even the pole location) can drastically alter the encirclement pattern. For example, moving a RHP zero slightly toward the imaginary axis brings it closer to destabilizing the system. The Nyquist plot is a snapshot; it does not easily reveal how sensitive the stability margins are to parameter drift. Engineers must therefore supplement Nyquist analysis with parametric uncertainty analysis (e.g., robust control methods like mu-synthesis or Monte Carlo simulations).

Implications for Control System Design

The limitations above have direct consequences for designers working with non-minimum phase plants:

  • Controller bandwidth is fundamentally limited by the RHP zero location. The achievable closed-loop bandwidth cannot exceed approximately 0.5 times the RHP zero magnitude without causing severe phase loss. Nyquist plots can obscure this fundamental constraint, leading designers to push for higher bandwidth than is feasible.
  • Lead-lag compensation may not suffice. While a phase lead compensator can partially offset phase lag from poles, it cannot cancel phase lag from RHP zeros because such cancellation would place the controller zero in the right-half plane, making the controller unstable. Nyquist analysis might incorrectly suggest that a high-gain lead compensator can recover the phase margin.
  • Time-domain specifications (overshoot, settling time) cannot be reliably deduced from Nyquist margins alone. The inverse response adds an extra delay-like behavior that increases the rise time and often leads to larger overshoot than predicted by standard margin-based rules.
  • Design iteration becomes essential. Because Nyquist plots can be ambiguous, engineers must validate designs with step-response simulations, Bode plots (which clearly show the phase lag), and root-locus plots (which show how RHP zeros attract closed-loop poles).

Alternative and Complementary Analysis Techniques

To overcome these limitations, control engineers should use a multi-tool approach:

Bode Plots

Bode magnitude and phase plots make the phase lag from RHP zeros explicit: each zero contributes an extra -20 dB/decade in magnitude (like a pole) and a phase lag of -90° per zero at high frequencies. More importantly, the Bode gain-phase relationship (the fact that phase is determined by the slope of the magnitude curve) imposes a bandwidth limitation for non-minimum phase systems. Bode plots also allow easy identification of multiple gain crossover frequencies and the phase at those points. External resource: For a detailed explanation, see the University of Toronto notes on Bode plots and non-minimum phase systems.

Root Locus

The root-locus method directly shows how closed-loop poles move as gain varies, revealing how RHP zeros attract poles toward the right-half plane. It is especially helpful for understanding the trade-off between stability and performance. For example, a RHP zero at s = +1 will “pull” one branch of the root locus toward the right, limiting the maximum stable gain. Nyquist plots cannot provide this intuitive visualization. External resource: A concise review of root locus and RHP zeros is available at Colorado State University’s mechatronics course.

Time-Domain Simulation

Given a validated model, simulate the closed-loop step response. Inverse response appears immediately, and metrics like overshoot, rise time, and settling time can be computed. This is the only method that directly reveals the transient behavior that Nyquist plots miss. Modern tools like MATLAB/Simulink, Python control library, or even simple Python scripts can run hundreds of simulations under parameter variations.

Gain and Phase Margins from Nyquist (with caution)

One can still compute margins from the Nyquist plot, but only after verifying that the plot is unambiguous. Use the crossing points of the real axis to compute gain margins at each crossing, then take the smallest one. For phase margin, check at all gain crossover frequencies. However, rely more on the distance to the critical point (the modulus margin) as a single robustness indicator. The minimum distance from the Nyquist curve to -1 is directly related to the peak of the sensitivity function (S_max). For non-minimum phase systems, a small modulus margin (< 0.5) often indicates poor robustness even if gain and phase margins look acceptable.

Robust Control Techniques

For systems with significant uncertainty, H-infinity or mu-synthesis methods explicitly account for RHP zeros and their impact on achievable performance. These techniques derive bounds on the sensitivity and complementary sensitivity functions, providing robust stability guarantees that Nyquist plots alone cannot offer. A good reference is the textbook by Skogestad and Postlethwaite, “Multivariable Feedback Control,” which covers the waterbed effect arising from RHP zeros. External resource: The MIT 6.241 Dynamic Systems and Control course materials include discussions on non-minimum phase zeros and their implications.

Practical Considerations and Best Practices

Based on the limitations discussed, here are actionable recommendations for engineers encountering non-minimum phase systems:

  1. Always check the time-domain response before finalizing a design. Nyquist plots are a starting point, not the finish line.
  2. Use the Nichols chart: It combines the magnitude and phase information of Bode plots in a single view, making it easier to see how phase lag from RHP zeros pushes the curve away from the 0 dB / -180° point.
  3. Perform sensitivity analysis: Vary the RHP zero location and other uncertain parameters, and observe how the Nyquist plot and step response change.
  4. Prefer loop-shaping in the frequency domain rather than relying on margins from Nyquist. Shape the open-loop transfer function to meet gain and phase at specific frequencies, and verify that the resulting Nyquist plot does not encircle -1 incorrectly.
  5. For highly non-minimum phase plants, consider feedforward or two-degree-of-freedom control to compensate for the inverse response. For example, a pre-filter that delays the reference signal can reduce the undershoot.
  6. If possible, physically reduce the effect of RHP zeros by redesigning the sensor/actuator placement (avoid non-collocation) or by adding passive damping.

Conclusion

Nyquist plots are an indispensable tool for control system analysis, but their reliability diminishes when applied to non-minimum phase systems. The presence of right-half-plane zeros introduces additional phase lag, inverse response, and multiple potential critical points, all of which can mislead engineers who rely solely on encirclement counts and traditional margin definitions. To overcome these limitations, one must complement Nyquist analysis with Bode plots, root-locus, time-domain simulations, and robust control methods. Only by embracing a multi-faceted approach can engineers design stable, performant controllers for systems that exhibit non-minimum phase behavior. Understanding these limitations not only prevents incorrect conclusions but also enriches one’s grasp of fundamental trade-offs in feedback control.