Signals and Systems: A New Way to See the World

Everything is a number that moves

Hum a single note. The air in front of your mouth doesn't just sit there — it squeezes and stretches, packing tight then thinning out, thousands of times a second. If you could plot the air pressure at one point against time, you'd draw a wavy line that wiggles up and down. That wavy line *is* your voice. It is a signal: a quantity that carries information by changing as time goes by.

Here is the quiet idea that the whole field is built on: a signal is just a function of an independent variable. For sound, that variable is time, and the function is air pressure: write it as x(t), read aloud as 'x of t', meaning 'the value of x at time t'. The world is overflowing with such functions. The voltage on a wire, the current through a motor, the brightness at a pixel, the altitude of a plane, your body temperature through a fever — each is some quantity that takes on a value at every instant, a number that moves.

A system is anything that transforms a signal

Now put a thing in the path of a signal. Speak into a microphone: sound pressure goes *in*, a voltage waveform comes *out*. Feed that voltage into an amplifier: a small wiggle goes in, a big wiggle comes out. Send it to a speaker: voltage goes in, sound comes back out. Each box in this chain has a job — it takes an input signal and produces an output signal. That, exactly, is a system: anything that transforms an input signal into an output signal.

A thermostat is a system: the input is the room temperature it senses, the output is the on/off command it sends to the furnace. A camera is a system: light intensity in, a grid of pixel values out. Cruise control, noise-cancelling headphones, an electrocardiogram machine, the autofocus in your phone — all systems, all built from the same one-sentence definition. Engineers love to draw this as a block diagram: a labelled box with an arrow going in and an arrow coming out.

      x(t)                                  y(t)
  input signal      +----------------+    output signal
  ---------------->  |    SYSTEM    |  ---------------->
   (e.g. mic        |  (e.g. an    |   (e.g. a louder,
    voltage)        |  amplifier)  |    cleaner wave)
                     +----------------+

        y(t) = T{ x(t) }       "the system T acts on x to make y"

Sorting signals: the four big distinctions

Before we can ask clever questions, we need to sort signals into families, because each family wants different tools. Four distinctions carry most of the weight, and the first one decides almost everything that follows.

1. Continuous-time vs discrete-time. A continuous-time signal x(t) has a value at *every* instant — a smooth, unbroken curve, like real sound pressure. A discrete-time signal x[n] (note the square brackets) exists only at integer steps n = 0, 1, 2, …, like the readings a sensor takes once a millisecond. The wiggly world is continuous; computers can only ever store the discrete kind.
2. Analog vs digital. A close cousin of the first, but about *value* not *time*. An analog signal can take any value in a range (the needle on an old dial can sit anywhere). A digital signal is quantized to a finite set of levels, usually bits — 0 or 1. Your microphone makes analog voltage; an audio file is digital. The bridge between them is sampling, the star of a later rung.
3. Periodic vs aperiodic. A periodic signal repeats itself forever on a fixed beat — the steady 60 Hz hum from a wall socket, a pure musical tone. Its period T is the time for one full cycle. An aperiodic signal never exactly repeats: a spoken word, a heartbeat under stress, a single drumbeat fading away. Periodic signals are wonderfully easy to analyse, which is why we lean on them so hard.
4. Energy vs power. A short burst — a camera flash, a single ping — delivers a finite total amount of *energy*, then it's gone; we call it an energy signal. A signal that keeps going forever — that 60 Hz hum, a radio carrier — has infinite total energy but a well-defined average *power*; we call it a power signal. The distinction tells you which yardstick to measure 'how big' with.

The LEGO bricks: four elementary signals

Here is the move that makes the whole subject *work*. Instead of wrestling with each messy real-world waveform on its own terms, we keep a small box of beautifully simple signals — and then build any complicated signal by stacking copies of those simple ones. Master four bricks and the rest of the track becomes a game of assembly.

  UNIT IMPULSE  δ(t)        UNIT STEP  u(t)        SINUSOID  cos(2πf t)      EXPONENTIAL  e^(-t/τ)
     |                        ____________            /\      /\               |
     |  (spike of            |                       /  \    /  \              | \
  ___|___  area 1)      _____|                  ____/____\__/____\___       ___|__ \______
     0  t              0      t                     \  /    \  /                 0     \___ t
  "a single tap"     "switch turns ON"          "a pure tone, one freq"      "a smooth decay"

1. The unit impulse δ(t) — an idealized instantaneous tap: zero everywhere except a single infinitely thin spike at t = 0, with total area 1. Think of striking a bell once. It seems too extreme to be real, yet it turns out to be the single most important signal in the whole subject — and the next rung will show you why.
2. The unit step u(t) — a switch that flips from 0 to 1 at t = 0 and stays on. Whenever you plug something in or press a power button, you've applied a step. Watching how a system reacts to a step — its step response — is the everyday way engineers test whether a thing settles smoothly or rings and overshoots.
3. The sinusoid cos(2πf t) — a pure tone of a single frequency f. This humble wave is the secret king of the subject: a deep result (you'll meet it as the Fourier idea) says *any* signal, however jagged, is just a sum of sinusoids of different frequencies. Pull a signal apart into its tones and most of its mysteries dissolve.
4. The exponential e^(−t/τ) — a smooth rise or decay set by a time constant τ. A charging capacitor, a cooling cup of coffee, a fading echo: nature relaxes exponentially almost everywhere. Combine an exponential with a sinusoid and you can describe waves that grow, shrink, or ring — the full vocabulary of how real circuits respond.

The one question this whole track answers

Put the two halves together — signals going in, a system in the middle — and a single question lights up, the question this entire track exists to answer: **given a system, how will it respond to *any* input I feed it?** If I knew that, I could predict the exact sound from my speaker, design a thermostat that never overshoots, or build a filter that erases hum and keeps the music. It sounds impossibly ambitious. The breathtaking news is that for an enormous and useful class of systems, the answer is *yes, completely*.

That golden class is the linear, time-invariant (LTI) system. *Linear* means doubling the input doubles the output, and inputs add up without interfering — so if you know how the system treats simple bricks, you know how it treats any stack of them. *Time-invariant* means the system behaves the same today as tomorrow; delay the input and the output just delays to match. Most amplifiers, filters, and circuits you'll meet are LTI to an excellent approximation, which is precisely why we obsess over them.