Skip to content

Complex numbers for signals

Build complex numbers from nothing, as points and arrows in a plane, so 3.4 can describe a sinusoid as a spinning arrow.

Before thisSinusoids (3.2)

Before this3.2
Chapter 3 · Lesson 3 of 4

First, the picture

Drag the point straight left or right and watch only Across change; drag it straight up or down and watch only Up change.

A complex number is a point

Drag the point; two ordinary numbers locate it.

Drag straight left or right and only Across changes. Drag straight up or down and only Up changes: an ordinary (x, y) point, just renamed real and imaginary.

Across
3.00
Up
4.00
Written as
3.00 + 4.00j
Describe this picture

A plane with a horizontal axis labelled real and a vertical axis labelled imaginary, each from −6-6 to 6, and one point you can drag, starting at 3+4j3 + 4j. Three readouts show “Across”, “Up” and “Written as”, the point in the form a+jba + jb.

A complex number is a point

Back in 3.2 you described a sinusoid with two numbers: an amplitude and a phase, a length and an angle. It turns out those two numbers pin down a single point in a flat plane, the same way “3 blocks east, 2 blocks north” pins down one spot on a map. I want to give that point its own name and its own arithmetic, because once you can add and multiply points instead of just plotting them, a spinning sinusoid becomes something you can compute with directly. That’s where we’re headed in 3.4; this page just builds the tool.

Take any ordinary (x,y)(x, y) point and relabel it. Call the sideways position, measured along a horizontal axis, the real part. Call the up-down position, measured along a vertical axis, the imaginary part, and mark that axis with a unit called jj (engineers’ name for “one step up”; you may have seen ii used the same way elsewhere). A point that sits aa across and bb up is written

z=a+jbz = a + jb

and called a complex number. Nothing about it is mysterious or “imaginary” in the everyday sense: it’s an ordinary point, just written as a sum instead of a pair of coordinates.

Go back to the point at the top of the page. It behaves exactly like the (x,y)(x, y) point you already know, with new names on the two axes.

Length and angle: polar form

The same point can be described a second way: instead of “how far across, how far up,” you can give “how far from the centre, and in which direction.” That’s exactly the length-and-angle picture from 3.2’s amplitude and phase, applied to this point instead of to a wave.

The distance from the origin is called the magnitude, written ∣z∣\lvert z\rvert. Across aa and up bb form the two short sides of a right triangle whose long side is that distance, so Pythagoras gives it directly:

∣z∣=a2+b2\lvert z\rvert = \sqrt{a^2+b^2}

The direction is called the argument, written arg⁡z\arg z: the angle the line from the origin to the point makes with the positive real axis, measured in the radians from 3.2. Writing a complex number this way, as a length and an angle, is its polar form.

Drag the point around a circle centred on the origin and watch Length hold still while Angle sweeps. Then drag it straight outward from the origin and watch Angle hold still while Length grows.

Length and angle: polar form

The same point, described by how far and which way.

Drag the point around a circle centred on the origin: Length stays put while Angle sweeps. Drag straight outward from the origin: Angle stays put while Length grows.

Across
3.00
Up
4.00
Written as
3.00 + 4.00j
Length
5.00
Angle
53.1°
Written as (polar)
(5.00, 53.1°)
Describe this picture

The same plane and point, with the line from the origin to the point. Six readouts: “Across”, “Up” and “Written as” for the rectangular form, and “Length”, “Angle” and “Written as (polar)” for the polar form.

Notice the two readouts, rectangular (a+jba+jb) and polar (length, angle), always describe the exact same point.

Adding two complex numbers: nose to tail

Add two complex numbers by adding their real parts and their imaginary parts separately:

(a1+jb1)+(a2+jb2)=(a1+a2)+j(b1+b2)(a_1+jb_1) + (a_2+jb_2) = (a_1+a_2) + j(b_1+b_2)

Picture each number as an arrow from the origin to its point. Adding them is the same as combining two displacement instructions, “3 blocks east, 2 north” then “1 block east, 4 south,” into one net displacement: you walk the first arrow, then walk the second arrow starting from where the first one left off, nose to tail, and the sum is the single arrow from the start straight to where you end up.

Move either point and watch the sum arrow slide by exactly the same amount you moved it.

Adding two complex numbers: nose to tail

Move either point; the sum arrow follows exactly.

Move either point and the sum arrow slides by exactly the same amount: it is the two arrows placed end to end.

Point 1
3.00 + 2.00j
Point 2
-1.00 + 3.00j
Sum
2.00 + 5.00j
Describe this picture

The plane with two points you can drag, each drawn as an arrow from the origin, and their sum, drawn nose to tail. Three readouts show “Point 1”, “Point 2” and “Sum”.

Notice the sum is nothing more than the two arrows placed end to end.

Multiplying two complex numbers: rotate and scale

Multiplying is less obvious from the numbers alone, but it has a clean geometric meaning: it multiplies the two magnitudes together, and it adds the two angles together.

∣z1z2∣=∣z1∣∣z2∣,arg⁡(z1z2)=arg⁡z1+arg⁡z2\lvert z_1 z_2\rvert = \lvert z_1\rvert\lvert z_2\rvert, \qquad \arg(z_1z_2) = \arg z_1 + \arg z_2

So multiplying a point by zz rotates it by zz‘s own angle and stretches it by zz‘s own length. A clean special case: jj itself sits at length 1 and angle 90°90°, so multiplying anything by jj spins it a quarter turn with no change in size at all. Recall z=3+4jz=3+4j from the last two instruments: its length is 32+42=5\sqrt{3^2+4^2}=5 and its angle is about 53.13°53.13° (the angle whose rise over run is 4/34/3; a calculator’s arctan⁡\arctan button finds it) . Multiplying it by jj gives j(3+4j)=3j+4j2=−4+3jj(3+4j) = 3j+4j^2 = -4+3j (using j2=−1j^2=-1, which the next section explains), a point still of length 55 but at angle 53.13°+90°=143.13°53.13°+90°=143.13°, exactly a quarter turn further round.

Move either point and watch the readouts.

Multiplying two complex numbers: rotate and scale

Length multiplies; angle adds.

The product's angle is always the two angles added, and its length is the two lengths multiplied. Multiplying by a point on the unit circle only rotates: it never stretches.

Point 1
3.00 + 4.00j
Point 2
0.00 + 1.00j
Lengths multiply
5.00 × 1.00 = 5.00
Angles add
53.1° + 90.0° = 143.1°
Product
-4.00 + 3.00j
Describe this picture

The plane with two points you can drag and their product. Five readouts show “Point 1”, “Point 2”, “Lengths multiply” (the two lengths and their product), “Angles add” (the two angles and their sum) and “Product”.

The product’s length is always the two lengths multiplied together, and its angle is always the two angles added together. Notice that moving Point 2 onto the unit circle, so Lengths multiply reads ”… × 1.00 = …”, only ever rotates the product; it never stretches it.

Why j×j=−1j \times j = -1

You’ve now seen jj used twice: as the label on the vertical axis, and as a number you can multiply by. Both uses agree, because jj is defined as the point at length 1, angle 90°90°, and multiplying by it is exactly the rotate-and-scale rule above: rotate by 90°90°, don’t stretch at all. Multiply by jj twice in a row and you rotate by 90°90° twice, which is 180°180° total, a half-turn, landing you on the point exactly opposite where you started, at the same distance from the origin. Starting from the point 11 (length 1, angle 0°0°), two quarter-turns land you on −1-1. That’s the whole content of the famous rule

j×j=j2=−1j \times j = j^2 = -1

not a new kind of arithmetic, just two quarter-turns making a half-turn, flipping a point to the opposite side of the origin.

The mirror image

Flip the sign of a complex number’s imaginary part only, leaving the real part alone, and you get its conjugate, written zˉ\bar z:

zˉ=a−jb\bar z = a - jb

Geometrically, that’s a mirror image of the point across the real axis: same distance across, same distance from the axis, just on the opposite side of it. Add a number to its own conjugate and the two imaginary parts are exact opposites, so they cancel completely, leaving only twice the real part:

z+zˉ=(a+jb)+(a−jb)=2az + \bar z = (a+jb)+(a-jb) = 2a

always a point sitting on the real axis. For z=3+4jz=3+4j, zˉ=3−4j\bar z = 3-4j and z+zˉ=6=2(3)z+\bar z = 6 = 2(3) .

Drag the point anywhere and watch its mirror image follow.

The mirror image

One point; its mirror image is always shown too.

The mirror always sits the same distance below the real axis as the point sits above it. Add them together and the Up parts cancel exactly, landing on the real axis at twice Across.

Point
3.00 + 4.00j
Its mirror (the conjugate)
3.00 − 4.00j
Point + mirror
6.00 + 0.00j
Describe this picture

The plane with one point you can drag and its mirror image across the real axis, the conjugate. Three readouts show “Point”, “Its mirror (the conjugate)” and “Point + mirror”.

The mirror image always sits the same distance below the real axis as the point sits above it (or vice versa). Notice Point + mirror always lands on the real axis, at exactly twice Across, no matter where you drag the point.

The maths behind it · rotation matrices

A complex number is exactly a vector in the plane: addition here is ordinary vector addition, and multiplying by a fixed zz turns out to be the same rotate-and-scale operation linear algebra writes as a 2×22\times2 matrix.

Equal steps around the circle

One last picture, because it comes back in Chapter 13, when we break signals into frequencies. Ask: which points, raised to the NN-th power, land exactly back on 11? Raising a point to a power means multiplying it by itself over and over, and by the rotate-and-scale rule each multiplication adds another copy of the point’s own angle. So a point at angle θ\theta lands back at angle 0°0° (up to a full lap) after NN multiplications only if NθN\theta is a whole number of full 360°360° turns. That gives NN possible angles, evenly spaced:

zk=cos⁡ ⁣(2πkN)+jsin⁡ ⁣(2πkN),k=0,1,…,N−1z_k = \cos\!\left(\frac{2\pi k}{N}\right) + j\sin\!\left(\frac{2\pi k}{N}\right), \qquad k = 0, 1, \ldots, N-1

all sitting on the unit circle (length 1), called the NN-th roots of unity. For N=4N=4 they are 1, j, −1, −j1,\ j,\ -1,\ -j, at angles 0°,90°,180°,270°0°, 90°, 180°, 270°; you already checked j2=−1j^2=-1 above, and j4=(j2)2=(−1)2=1j^4=(j^2)^2=(-1)^2=1 confirms jj is one of them.

Drag the N slider and watch the points redistribute themselves evenly around the circle.

Equal steps around the circle

N points, evenly spaced around a circle of length 1.

Every point's angle is exactly 360 degrees divided by N, times its own position number: spaced perfectly evenly, for any N you choose.

4
Angle between neighbours
90.0°
Next point after 1, raised to the Nth power
1.00 + 0.00j
Describe this picture

The unit circle with NN points evenly spaced around it, joined into a regular NN-sided shape. The “Points (N)” slider runs from 3 to 12 and starts at 4. Two readouts show “Angle between neighbours” and “Next point after 1, raised to the Nth power”.

Notice the readout for Angle between neighbours always reads exactly 360° divided by N, and Next point after 1, raised to the Nth power always lands back on 11, whatever N you pick.

Worked example

Take z=3+4jz = 3+4j, the point used as the default above.

  1. Rectangular to polar. ∣z∣=32+42=25=5\lvert z\rvert = \sqrt{3^2+4^2} = \sqrt{25} = 5 (a 3-4-5 triangle), and arg⁡z=arctan⁡(4/3)=53.13°\arg z = \arctan(4/3) = 53.13° (0.9273 rad).
  2. Multiplying by jj. j(3+4j)=3j+4j2=−4+3jj(3+4j) = 3j+4j^2 = -4+3j. Check the length is unchanged: ∣−4+3j∣=16+9=25=5\lvert -4+3j\rvert = \sqrt{16+9} = \sqrt{25} = 5; the new angle is 53.13°+90°=143.13°53.13°+90° = 143.13°, exactly a quarter turn further round.
  3. Conjugate sum. zˉ=3−4j\bar z = 3-4j, so z+zˉ=6=2(3)z+\bar z = 6 = 2(3), real, exactly on the real axis.
  4. 4th roots of unity. 1, j, −1, −j1,\ j,\ -1,\ -j, at angles 0°,90°,180°,270°0°, 90°, 180°, 270°; check j4=(j2)2=(−1)2=1j^4 = (j^2)^2 = (-1)^2 = 1.

Where you’ll meet this

3.4 puts the rotate-only case front and centre: a point of length 1 whose angle grows steadily with time is a phasor, and it traces out exactly the spinning-point picture that built the cosine and sine back in 3.2. Chapter 27 uses the real-plus-imaginary pair again for the two-part signals inside every radio, and Chapter 13 measures a signal’s frequencies at points that sit, quite literally, at the roots of unity you just met.

Reference card

QuantityFormulaNotes
Rectangular formz=a+jbz=a+jbaa: real part, bb: imaginary part
Magnitude∣z∣=a2+b2\lvert z\rvert=\sqrt{a^2+b^2}distance from the origin
Argumentarg⁡z\arg zangle from the positive real axis, in radians
Polar formz=∣z∣(cos⁡θ+jsin⁡θ)z=\lvert z\rvert(\cos\theta+j\sin\theta)
jjj×j=−1j\times j = -1two quarter-turns make a half-turn
Conjugatezˉ=a−jb\bar z = a-jbmirror across the real axis; z+zˉ=2az+\bar z=2a (always real)
Addition(z1+z2)=(a1+a2)+j(b1+b2)(z_1+z_2)=(a_1+a_2)+j(b_1+b_2)nose-to-tail
Multiplication∣z1z2∣=∣z1∣∣z2∣\lvert z_1z_2\rvert=\lvert z_1\rvert\lvert z_2\rvert, arg⁡(z1z2)=arg⁡z1+arg⁡z2\arg(z_1z_2)=\arg z_1+\arg z_2rotate-and-scale
NN-th roots of unityzk=cos⁡(2πk/N)+jsin⁡(2πk/N)z_k=\cos(2\pi k/N)+j\sin(2\pi k/N), k=0,…,N−1k=0,\dots,N-1NN evenly spaced points on the unit circle

End of lesson 3.3

Where to go next.

Phasorium
LibraryEvery lesson, in order

Parts

About Phasorium
Look