Chapter Summary

Chapter Summary

Key Points

• 1.

  Use the OO API, not pyplot. Always create figures with fig, ax = plt.subplots() and call methods on ax. This avoids state leaks, enables reusable plotting functions, and composes cleanly into multi-panel figures. Reserve plt.plot() for throwaway REPL sessions only.

• 2.

  Master the core trio: line, scatter, and error bars. ax.plot() for connected curves, ax.scatter() for per-point color/size variation, ax.errorbar() for confidence intervals. Use ax.fill_between() for shaded confidence bands. Always show uncertainty — a BER curve without error bars is scientifically incomplete.

• 3.

  Image plots need correct colormaps and origin. Use viridis or another perceptually uniform colormap — never jet. Set origin='lower' for physical data (spectrograms, channel matrices). Add a colorbar with units. Use pcolormesh for non-uniform grids.

• 4.

  Set figsize to the final print size. For IEEE single-column, use figsize=(3.5, 2.625) with 8pt fonts and 600 DPI. The figure should look correct at 100% zoom — never rely on post-export scaling. Export as PDF or SVG for vector quality.

• 5.

  Complex data needs dual plots or constellations. Use magnitude/phase (Bode) plots with np.unwrap() for frequency responses. Use scatter plots with aspect='equal' for constellation diagrams. HSV coloring maps phase to hue for 2D complex fields.