Visual walkthrough — Bode plot — magnitude and phase vs frequency

We will use only these tools, and each is introduced the moment it is needed:

• a sine wave (a wiggle that repeats),
• an arrow that spins (the phasor),
• a right triangle (to read off size and angle),
• the operations $\log$ (to turn multiply into add) and $\arctan$ (to read an angle off a triangle).

---

Step 1 — A sine wave, and the two things you can do to it

WHAT. Look at the black input wave in the figure. Now ask: if I feed this into a machine (an amplifier, a filter, a rocket's steering loop), what can possibly come out?

WHY. We start here because every signal in the real world — wind gusts, sensor hiss, sloshing fuel — can be built by stacking sine waves. If we understand one sine, we understand all of them.

PICTURE. The output (cyan) is also a wiggle at the same pace, but it can differ in only two ways:

1. It can be taller or shorter — that is a change in amplitude (size).
2. It can be shifted left/right in time — that is a change in phase (timing).

That "same pace, only size and shift change" is not obvious. Step 2 shows why it must be true.

---

Step 2 — Why nothing else can change (the spinning-arrow trick)

WHAT. We replace $\cos (\omega t)$ with a spinning arrow of length 1, turning $\omega$ radians every second. Its shadow is the sine wave.

WHY this tool and not another. We could stay with wiggly cosines, but adding and scaling them is painful (trig identities everywhere). A spinning arrow, by contrast, is just multiply to scale, rotate to shift — one clean object that carries both size and timing at once. That is exactly the two-things-that-can-change from Step 1, packed into one arrow. This idea comes from the Laplace transform world, where $e^{st}$ is the natural language of linear systems.

PICTURE. A linear time-invariant system has a golden property: if you send in a pure spinning arrow, out comes a spinning arrow at the same spin-rate $\omega$ — only stretched and rotated. That is why the frequency never changes. The system cannot invent new spin-rates; it can only scale and turn the one you gave it.

---

Step 3 — Name the stretch and the turn: the number $G(j\omega )$

WHAT. The output arrow is the input arrow multiplied by $G(j\omega )$: $y(t)=\underset{\text{stretch \& turn}}{\underbrace{G(j\omega )}}\underset{\text{input spinning arrow}}{\underbrace{e^{j\omega t}}}$ Here $G(j\omega )$ scales the length and adds to the angle; $e^{j\omega t}$ is the incoming unit arrow from Step 2.

WHY. Multiplying two arrows in the complex plane does two things at once: it multiplies lengths and adds angles. That is precisely "scale the amplitude, shift the phase." The two independent effects from Step 1 are the two independent parts of one multiplication.

PICTURE. Watch the green output arrow: it is longer than the input (this system amplifies) and sits behind it by some angle (this system delays). Length = gain, angle-behind = phase lag.

---

Step 4 — Split the arrow into length and angle (polar form + the right triangle)

WHAT. Write $G(j\omega )$ in polar form: $G(j\omega )=\underset{\text{length}=\text{gain}}{\underbrace{|G(j\omega )|}}{e}^{j\underset{\text{angle}=\text{phase}}{\underbrace{\varphi }}}$

• $|G(j\omega )|$ ::: the magnitude — the length of the arrow. Read it as the long side of the right triangle.
• $\varphi =\angle G(j\omega )$ ::: the phase — the angle the arrow makes with the horizontal.

WHY $\arctan$ enters here. To get the angle $\varphi$ from the arrow's horizontal part $a$ and vertical part $b$, we form the ratio $b/a$ (rise over run — the steepness of the arrow). But a ratio is a number, not an angle. The function that answers "which angle has this steepness?" is $\arctan$: $\varphi =\arctan \frac{b}{a},|G|=\sqrt{a^2+b^2}$ $\arctan$ is chosen because it undoes $\tan$ (opposite-over-adjacent), turning the triangle's slope back into the angle that made it. (The full quadrant care — signs of $a,b$ — is the standard Nyquist plot-style bookkeeping; here $a>0$ so we are safe.)

PICTURE. Put back the input's real shadow: $\text{Re}\{y\}=|G(j\omega )|\cos (\omega t+\varphi )$. Same frequency $\omega$, taller by factor $|G|$, shifted by angle $\varphi$ — the two changes from Step 1, now with names.

---

Step 5 — Sweep the frequency: one arrow becomes two curves

WHAT. Steps 1–4 fixed one frequency $\omega$. Now turn a dial and try every $\omega$. At each setting the arrow $G(j\omega )$ has some length and some angle. Record them:

• length $|G(j\omega )|$ vs $\omega$ → the magnitude curve,
• angle $\angle G(j\omega )$ vs $\omega$ → the phase curve.

WHY two graphs. A complex number needs two real numbers to pin it down (length and angle). Splitting them gives two ordinary graphs — the two stacked plots of a Bode diagram.

PICTURE. As $\omega$ climbs, the arrow shrinks and swings further behind. Trace its length onto the top graph, its angle onto the bottom graph. The arrow's motion is the Bode plot.

---

Step 6 — Why $\log$: turn a product of arrows into a stack of ramps

WHAT. Take the magnitude of a product $G=G_1{G}_2$ and hit it with $20{\log }_{10}$: $\underset{\text{dB, total}}{\underbrace{20{\log }_{10}|G|}}=\underset{\text{piece 1}}{\underbrace{20{\log }_{10}|G_1|}}+\underset{\text{piece 2}}{\underbrace{20{\log }_{10}|G_2|}}$

• ${\log }_{10}$ ::: the "how many tens multiplied" function; its key trick is $\log (ab)=\log a+\log b$.
• the factor $20$ ::: gain is an amplitude ratio, and power $\propto$ amplitude², giving $10\log (A^2)=20\log A$.
• unit dB (decibel) ::: the vertical unit after this conversion.

WHY the $x$-axis is also $\log$. Plot against ${\log }_{10}\omega$ so that a factor like $1/(j\omega )$ becomes a straight line (constant slope in dB per decade), and so decades from 0.01 to 1000 rad/s all fit. Angles need no log — they already add: $\angle (G_1{G}_2)=\angle G_1+\angle G_2$.

PICTURE. The tangled product curve (top) becomes two straight ramps that you simply add (bottom). This is why a Lead-lag compensator or a Low-pass filter can be sketched by hand, factor by factor — the core skill of Loop shaping.

---

Step 7 — The edge cases (never leave the reader stranded)

WHAT & WHY & PICTURE, one degenerate input at a time:

• $\omega \to 0$ (DC, no wiggle). The arrow settles to $G(0)$, a real number = the steady gain $K$. Magnitude flattens to $20{\log }_{10}K$; phase → $0°$. Picture: the arrow stops spinning and points straight along the axis.
• $\omega \to \infty$ (infinitely fast). For a pole $1/(1+j\omega /{\omega }_c)$ the arrow shrinks toward zero length and swings to $-90°$. Magnitude ramps down at $-20$ dB/dec; phase → $-90°$. Picture: the arrow curls under and vanishes.
• At the corner $\omega ={\omega }_c$. The triangle has equal legs ($1$ and $1$), so it is a $45°$ triangle: phase is exactly $-45°$, and length $=1/\sqrt{2}$, i.e. $-3$ dB. Picture: the arrow at half-past, length $1/\sqrt{2}$. This is the famous −3 dB corner — not $0$ dB, not $0°$.
• Pure integrator $1/(j\omega )$, at any $\omega$. Length $=1/\omega$ (drops forever), angle fixed at $-90°$. Picture: the arrow always points straight down, only its length changes. Two of these stacked → $-180°$, the edge-of-instability plant from the parent's Example 2. See Stability margins for why $-180°$ is the danger line.

---

The one-picture summary

Everything on one board: a sine enters → becomes a spinning arrow → the system multiplies it by $G(j\omega )$ (stretch + turn) → we read the arrow's length and angle off a right triangle → sweeping $\omega$ traces those two numbers into the magnitude and phase curves → $\log$ makes products of pieces stack into straight ramps.

Recall Feynman retelling — say it to a friend

Push a swing. If you push slowly and gently, the swing follows you and goes about as high as your push; if you push faster, it lags behind your hand and doesn't go as high. Now imagine your push is a perfect repeating wiggle. The swing wiggles back at the same speed — it can only end up taller/shorter and earlier/later. Draw an arrow whose length is "how tall" and whose angle is "how late." That single arrow is $G(j\omega )$. Spin through every pushing speed and jot down the arrow's length and angle at each one — length on the top graph, angle on the bottom graph. Because real machines are several simple parts multiplied together, and because $\log$ turns "multiply" into "add," you can just stack the parts' graphs to draw the whole thing by hand. That stack of two graphs is the Bode plot.

Recall Quick checks

Why does frequency never change through an LTI system? ::: $e^{j\omega t}$ is an eigenfunction — the system can only stretch and rotate the arrow, not change its spin-rate. What are the two numbers a Bode plot records? ::: the length $|G(j\omega )|$ (gain, in dB) and the angle $\angle G(j\omega )$ (phase, in degrees). Why plot in dB against $\log \omega$? ::: $\log$ turns products of factors into sums, so factors stack and simple pieces become straight lines. At a first-order corner, what are gain and phase? ::: $-3$ dB and $-45°$ — the equal-legs, 45° right triangle.