1 Introduction to Nyquist Plots

1.1 Basic concept and geometry

A Nyquist plot represents a complex-valued function by drawing its trajectory in the complex plane as an external parameter—typically angular frequency—varies. For a frequency-domain response \(G(j\omega)\), the plot places the real part on the horizontal axis and the imaginary part on the vertical axis. As \(\omega\) sweeps across a range, the response traces a characteristic curve. The resulting geometry encodes both magnitude-like and phase-like information simultaneously, without requiring separate plots.

1.2 Complex plane conventions

Most Nyquist plots adopt the convention that the horizontal axis is \(\Re\{G(j\omega)\}\) and the vertical axis is \(\Im\{G(j\omega)\}\). The curve’s orientation depends on the chosen frequency sweep direction. In control settings, additional conventions may be used for loop gain (for example, sign choices in feedback), so the plotted function is often a specific combination such as open-loop transfer including the feedback sign.

1.3 Relationship to frequency response

When \(G(j\omega)\) is a frequency response, the Nyquist curve is essentially a re-expression of the same underlying data used in other frequency-response tools. Magnitude and phase can be recovered from the distance to the origin and the angular position of each point, but the Nyquist depiction tends to highlight how the response approaches, crosses, or encircles particular locations in the complex plane.

1.4 Typical use cases in engineering

Nyquist plots are used to assess how systems react across frequencies, particularly in loop analysis. Common applications include evaluating stability margins, identifying resonant behavior, checking sensitivity to gain and phase variations, and diagnosing modeling or numerical issues. The method is also used beyond classical control, for example in signal-processing contexts where complex transfer functions or frequency responses are studied as curves in the complex plane.

2 Mathematical Foundations

2.1 Complex frequency response representation

2.1.1 Transfer function and loop gain

A standard setting in control engineering defines an open-loop transfer function \(L(s)\) and studies its behavior on the imaginary axis, \(L(j\omega)\). In unity-feedback architectures, stability properties are linked to how \(1+L(j\omega)\) behaves, which is closely related to the geometry of \(L(j\omega)\) about critical points. In other formulations, the plotted function may include controller, plant, and sign conventions so that the resulting curve can be used directly for margin interpretation.

2.1.2 Real and imaginary components

Let \(L(j\omega)=a(\omega)+jb(\omega)\). The Nyquist plot is the parametric curve \[ x(\omega)=a(\omega), \quad y(\omega)=b(\omega), \] as \(\omega\) ranges over its sweep interval. Each frequency corresponds to one point, and continuous variation in \(\omega\) produces a smooth trajectory except at singularities or points where numerical evaluation becomes unreliable.

2.1.3 Common sign and scaling conventions

Different texts use different feedback sign and scaling choices. A common decision is whether the Nyquist plot is drawn for \(L(j\omega)\) or for a loop expression that directly yields \(1+L(j\omega)\). Similarly, gain scaling may be applied so the curve aligns with engineering margin definitions. Because critical-point references depend on these conventions, correct interpretation requires knowing exactly what function was plotted.

2.2 Frequency sweep parameterization

2.2.1 Mapping ω to the complex plane

The standard mapping uses \(\omega\in[0,\infty)\) with evaluation at \(s=j\omega\). The curve for positive frequencies often determines the full plot when symmetry holds (for real-coefficient systems, \(L(-j\omega)=\overline{L(j\omega)}\)). In practice, engineers choose a finite band \([\omega_{\min},\omega_{\max}]\) and sample densely enough to resolve features such as rapid phase changes or near-singular behavior.

2.2.2 Handling negative frequencies

For systems with real coefficients, plotting only \(\omega\ge 0\) may omit the mirrored lower-half contribution that would otherwise appear for \(\omega<0\). Some workflows therefore use a one-sided representation for visualization while retaining the theoretical content through conjugate symmetry. In contexts where symmetry assumptions do not hold, negative-frequency evaluation may be required.

2.2.3 One-sided vs two-sided plots

A two-sided Nyquist plot explicitly includes both positive and negative frequency trajectories, yielding a closed curve more directly comparable to theoretical encirclement arguments. A one-sided plot typically draws only \(\omega\ge 0\) and relies on the conjugate symmetry of \(L(j\omega)\) for real systems to infer what happens in the mirrored half. The chosen style affects how encirclements are counted in practice, so the plotting convention should be stated or understood.

2.3 Pole-zero implications on the curve

Poles and zeros of \(L(s)\) strongly shape the Nyquist trajectory. Poles near the imaginary axis tend to produce large excursions or rapid changes in phase across certain frequency bands. Zeros can cause the curve to pass close to the origin or alter its winding behavior. Qualitatively, as frequency increases, the net number of encirclements and the approach to asymptotic regions depend on pole-zero counts and the system’s relative degree.

3 Constructing a Nyquist Plot

3.1 Data preparation from measured or simulated responses

3.1.1 Sampling frequency grid

Construction starts by selecting a frequency grid. Engineers often use logarithmic spacing to capture behavior across decades efficiently, but linear spacing may be preferred when fine resolution around narrow resonances is needed. A good grid ensures that the curve does not “jump” over critical geometrical features between samples.

3.1.2 Computing real/imaginary values

For each \(\omega_k\), the complex response is evaluated as \(L(j\omega_k)\). If the response is obtained from measurements, the data may come as complex frequency response estimates (magnitude and phase or directly complex values). The plot coordinates are then \((\Re\{L(j\omega_k)\}, \Im\{L(j\omega_k)\})\). Care is required to use phase unwrapping or consistent phase conventions when converting from magnitude/phase form to complex form.

3.1.3 Normalization choices (e.g., gain factors)

Sometimes the raw response is scaled to match a loop-gain definition used for stability margins. Normalization might involve dividing by a nominal gain, incorporating controller gain, or expressing the curve for a specific feedback configuration. Since the Nyquist curve’s position relative to critical points is sensitive to such scaling, consistent normalization is essential for meaningful interpretation.

3.2 Plotting procedures

3.2.1 Axis and orientation guidelines

Typical plots use equal aspect considerations only loosely, but consistent axis scaling helps avoid visual misjudgment of distances. The curve direction is determined by the ordering of \(\omega\). Many engineering reports annotate the direction (e.g., arrows) so that readers interpret the traversal and the corresponding winding behavior correctly.

3.2.2 Marker points and direction of traversal

Using markers at selected frequencies can improve readability, especially when the curve spirals or crosses itself. Arrows or a legend specifying the frequency ordering help prevent confusion about which portion corresponds to low versus high frequency.

3.2.3 Dealing with discontinuities and branch cuts

Some complex functions used in modeling involve branch choices (for example, square roots or delays expressed through exponential terms). While the response itself is typically continuous in frequency, numerical evaluation may introduce apparent discontinuities when phase wrapping or branch selection is inconsistent. A practical approach is to compute directly in the complex plane (rather than reconstructing from wrapped phase) and to use phase-unwrapping logic when necessary.

3.3 Practical measurement considerations

3.3.1 Noise and numerical stability

Measurement noise can scatter points and obscure the curve’s true geometry near critical regions. Filtering or smoothing may help but can also distort encirclement-related geometry if overused. Numerical stability issues arise when computing very small or very large values; using appropriate data scaling and floating-point precision reduces these risks.

3.3.2 Windowing and spectral leakage

When frequency response is estimated from time-domain data, windowing affects leakage and thus the reliability of the complex estimates. Poor window selection can bias magnitude and phase, particularly near resonances or where the system exhibits sharp transitions. Proper excitation and averaging improve curve quality.

3.3.3 Units and calibration checks

Nyquist plots are dimensionally consistent in the complex plane, but interpretation depends on what the function represents. Ensuring that the transfer function or loop gain is correctly normalized and that frequency axes use consistent units (e.g., rad/s vs Hz) prevents systematic plotting errors.

4 Stability Analysis with Nyquist Criterion

4.1 Nyquist criterion overview

The Nyquist criterion links closed-loop stability to the behavior of an open-loop complex function around a critical point in the complex plane. In common feedback configurations, the key question is whether the mapped curve of the loop gain encircles the point \((-1,0)\) and how many times, relative to the number of unstable poles in the open-loop system.

4.2 Critical point and encirclement concept

4.2.1 The (−1, 0) point in loop analysis

In unity-feedback systems, the characteristic equation is often written as \(1+L(s)=0\). The critical point \((-1,0)\) corresponds to values where \(L(j\omega)\approx -1\), causing \(1+L(j\omega)\) to approach zero. The Nyquist plot’s proximity and crossing behavior around this location provide direct insight into potential instability.

4.2.2 Meaning of encirclements

Encirclement count is tied to the winding of the Nyquist curve about the critical point. The direction of encirclement matters. Conceptually, encirclements reflect how the complex argument of \(1+L(j\omega)\) changes as frequency sweeps, which in turn indicates whether the closed-loop system’s zeros move into or out of the unstable region.

4.3 Interpreting gain and phase margins

4.3.1 Distance to critical point

Gain and phase margins can be interpreted geometrically as how far the Nyquist curve is from the critical point (and along specific directions or lines in the complex plane). For instance, the gain margin is related to the reciprocal of the distance to the \(-1\) point along an appropriate radial relationship for the phase crossover, while the phase margin relates to the argument margin at the gain crossover condition.

4.3.2 Robustness interpretation

Larger margins generally correspond to a curve that stays farther from \((-1,0)\), meaning moderate perturbations in gain or phase are less likely to drive the system into instability. However, robustness interpretation should consider model uncertainty and the possibility that unmodeled dynamics become significant at frequencies not captured well in the plotted data.

Nyquist analysis is inherently open-loop: it uses the open-loop response to infer closed-loop stability. Once the encirclement and unstable pole count conditions are applied, the criterion yields whether the closed-loop poles reside in the desired half-plane. This linkage allows designers to reason about stability without directly computing all closed-loop poles for every parameter change.

5 Special Cases and Variants

5.1 Open-loop with time delay

Time delays introduce factors like \(e^{-sT}\) into the loop transfer, producing oscillatory phase behavior across frequency. The Nyquist curve can develop additional winding and “spiral-like” patterns, increasing the likelihood of multiple crossings near critical regions. Practical delays can also be approximated numerically, so designers often verify the resulting curve against a more accurate delay representation.

5.2 Non-minimum phase behavior

Non-minimum phase systems have zeros in positions that affect phase evolution strongly. In Nyquist terms, such zeros can push the curve to approach sensitive regions in ways that do not directly align with simple magnitude intuition. This frequently leads to reduced phase margin even when gain response seems acceptable, reflecting the inherent coupling between phase behavior and stability.

5.3 Multi-loop or MIMO extensions (overview)

For multi-loop systems or multiple-input multiple-output (MIMO) systems, classical Nyquist criterion does not extend in a straightforward scalar form. Various generalizations exist that use matrix frequency response and specialized stability tests. In general terms, the “critical encirclement” picture becomes more complex, often requiring additional assumptions or more elaborate computations.

5.4 Discrete-time Nyquist plots (high-level)

In discrete-time control, frequency response is evaluated over the unit circle, typically using \(z=e^{j\omega}\) rather than \(s=j\omega\). The resulting Nyquist-like locus reflects sampled-data dynamics and may display different periodic features due to aliasing and discrete-time stability regions. High-level discrete-time variants map the complex response in a way that aligns with the criterion’s discrete-time stability requirements.

6 Reading and Interpretation

6.1 Shape diagnostics (loops, crossings, and spirals)

The curve’s form serves as a diagnostic tool. Crossings of certain axes or near the critical point indicate potential gain crossover or phase crossover behavior. Spirals often suggest multiple phase wraps or repeated winding introduced by delays or higher-order dynamics. Loops and self-intersections can occur when the response is sufficiently dynamic across frequency, and they must be interpreted with attention to curve direction and encirclement counting.

6.2 Resonance and damping indicators

Resonant systems often show frequency bands where the response magnitude peaks and the phase changes rapidly. On a Nyquist plot, this may manifest as tight curvature, close approach to specific regions, or pronounced excursions away from the origin. Damping tends to spread out and reduce the sharpness of these features, yielding smoother trajectories with less extreme turns.

6.3 Nonlinearities and approximate linearization

Nyquist plots are fundamentally linear tools, but engineers may use them for systems that are locally linearized around an operating point. Nonlinear effects can cause the effective frequency response to vary with amplitude or operating conditions, meaning the plotted curve represents an approximation. Large discrepancies between predicted and observed behavior may appear as unexpected shifts in shape, crossover points, or margin estimates.

6.4 Common pitfalls and misread plots

Common errors include interpreting one-sided plots as though they fully capture encirclement behavior, mixing sign conventions for loop gain, using inconsistent phase conventions, or extrapolating beyond the frequency range where data are reliable. Another pitfall is assuming the curve’s nearest approach to the critical point always determines margin without considering how perturbations scale the loop gain and affect the crossover geometry.

7 Nyquist Plot Examples

7.1 First-order systems

A first-order open-loop transfer often produces a relatively simple curve that transitions monotonically in phase from low to high frequency. The Nyquist trajectory typically moves along a smooth path, rarely forming loops. As a result, gain and phase margins can often be inferred by the curve’s single crossing behavior relative to the critical region.

7.2 Second-order systems

Second-order dynamics frequently lead to more structured trajectories due to conjugate pole pairs. The Nyquist plot can show noticeable curvature and approach patterns associated with oscillatory modes.

7.2.1 Underdamped resonance curves

Underdamped systems have low damping and thus stronger oscillatory behavior. In Nyquist form, this may appear as a pronounced excursion or “bulge” where the response traces a larger loop-like region and phase changes faster around the resonant frequency.

7.2.2 Critically damped and overdamped cases

Critical damping tends to produce a smoother, less dramatic curve than the underdamped case. Overdamped systems often yield trajectories that are more gradual, with reduced sharpness in the transition from low-frequency behavior to high-frequency asymptotics.

7.3 Systems with zeros and poles

Adding zeros changes the route the Nyquist plot takes through the complex plane. A zero can pull the curve closer to the origin at certain frequencies and alter the phase response, which can shift the location of crossovers and affect the curve’s ability to approach or encircle the critical point.

7.4 Comparing designs via overlay plots

Overlaying Nyquist curves from different controller designs provides a direct visual comparison. Designers commonly look for improvements such as increased separation from the critical point, reduced sensitivity to gain changes, or reduced likelihood of additional encirclement events. Consistent plotting conventions and normalization are crucial when comparing curves.

8 Computational and Software Implementation

8.1 MATLAB/Python workflow (conceptual)

8.1.1 Generating frequency response data

A typical workflow evaluates the transfer function at a selected set of frequencies. In MATLAB or Python, this may be done using built-in frequency-response utilities or by directly computing the transfer function expression \(L(j\omega)\). For measured data, the complex response may be imported from a vector of frequency points and corresponding real and imaginary parts.

8.1.2 Transforming to complex-plane coordinates

Once complex values are available, coordinates are computed as real and imaginary components. The plotting step then draws \(x=\Re\{L(j\omega)\}\) versus \(y=\Im\{L(j\omega)\}\). If the curve direction must reflect frequency increase, the order of \(\omega\) in the dataset must be maintained throughout.

8.2 Automation of margin estimation

Software can estimate crossover frequencies where \(\Re\{L(j\omega)\}=-1\) or \(L(j\omega)=1\), depending on the margin definition being implemented. Encirclement counts or winding numbers can be approximated by tracking the curve’s phase relative to the critical point or by detecting crossings of a chosen contour in the complex plane. Numerical robustness often requires dense sampling near critical regions.

8.3 Visualization best practices

8.3.1 Plot aesthetics for engineering reports

Good plots use clear line styles, readable axes labels, and sufficient resolution. Adding grid lines and marking the critical point \((-1,0)\) improves interpretability. When multiple curves are overlaid, distinct colors and legends should distinguish the corresponding designs without ambiguity.

8.3.2 Consistent legends, markers, and scales

Consistency across plots is important for comparisons: the same axes limits, scaling, and normalization should be used when evaluating alternative controllers or model revisions. Markers at selected frequencies can reduce misinterpretation when the trajectory twists or folds.

9 Relationship to Other Frequency-Response Tools

9.1 Bode plot vs Nyquist plot

Bode plots present magnitude and phase separately as functions of frequency, typically on logarithmic scales. Nyquist plots combine these effects into a single geometric locus, which can make stability-encirclement reasoning more direct in feedback analysis. However, Bode plots may be easier for quickly spotting gain roll-off patterns and phase trends.

9.2 Nichols chart connections

Nichols charts plot gain (in dB) versus phase, providing another compact view of frequency response useful for loop shaping and margin visualization. Like Nyquist plots, Nichols charts relate directly to stability margins, but they emphasize different coordinate choices and are often better suited for manual controller tuning workflows.

9.3 Polar plots and Smith charts (conceptual comparison)

Polar plots depict complex values as radius and angle, offering an alternative way to represent phase and magnitude together. Smith charts, commonly used in transmission-line contexts, map impedance to a normalized complex plane. These tools share a “complex-plane geometry” spirit with Nyquist plots, though the specific mapping and stability interpretation differ by application.

10 Glossary

10.1 Key terms and notation

  • Nyquist plot: A parametric plot of a complex function’s real and imaginary parts versus frequency.
  • Loop gain \(L(s)\): The open-loop transfer function used in feedback stability analysis.
  • Critical point \((-1,0)\): The reference point associated with the characteristic equation \(1+L(s)=0\) in unity-feedback analysis.
  • Encirclement: The winding of the Nyquist curve around the critical point, counted with direction.
  • Gain margin: A measure of allowable gain variation before losing stability, related to crossover geometry.
  • Phase margin: A measure of allowable phase variation before instability, related to crossover geometry.

10.2 Common abbreviations and symbols

  • \(\Re\{\cdot\}\): Real part
  • \(\Im\{\cdot\}\): Imaginary part
  • \(j\): Imaginary unit
  • \(\omega\): Angular frequency
  • MIMO: Multiple-input multiple-output
  • \(s\): Laplace-domain complex variable
  • \(z\): Z-transform complex variable for discrete-time analysis