[Advanced] — how information rides on radio waves, and the shift that turned radio from circuits into code. The bridge from DSP to RF.

The kid version first

A radio wave is a fast sinusoid — say 100 million cycles a second. On its own it carries nothing. Modulation is nudging that wave — its height, its speed, or its timing — in step with your data, so the wobbles encode your voice, your Wi-Fi packet, your GPS signal. The receiver watches the wobbles and reconstructs the data.

Software-defined radio is the realisation that all of this nudging and un-nudging is just arithmetic on samples — so a general-purpose computer with a simple antenna front-end can become any radio, in software.

Why modulate at all

Two reasons you can’t just broadcast your signal directly:

  1. Antenna size. An efficient antenna is roughly the size of the wavelength. Your voice (~3 kHz) has a wavelength of ~100 km — an impossible antenna. Ride it on a 100 MHz carrier (~3 m wavelength) and the antenna is practical. Modulation shifts your low-frequency information up to a high carrier frequency where it can radiate
  2. Sharing the spectrum. Everyone transmitting at once would collide. Assigning each service a different carrier frequency (radio stations, Wi-Fi channels, phone bands) lets many signals coexist — you tune to the carrier you want → channel capacity

The analog modulations

A carrier A·cos(2πf_c·t + φ) has three things you can vary in step with your message:

  • AM (Amplitude Modulation) — vary the height A. Simple, cheap receivers, but noise adds to amplitude so it’s noisy. AM radio, and the video part of old TV
  • FM (Frequency Modulation) — vary the frequency f_c. More robust to noise (noise hits amplitude, which FM ignores), better fidelity. FM radio
  • PM (Phase Modulation) — vary the phase φ. Closely related to FM, and the basis of the digital schemes below

The noise trade is the point: AM is simple but noise-prone; FM/PM trade bandwidth for noise immunity. That trade — bandwidth for robustness — recurs throughout communications.

Digital modulation

Digital data (bits) is carried by switching the carrier between discrete states:

  • ASK — amplitude shift keying: on/off or between levels (a digital AM)
  • FSK — frequency shift keying: bit 0 = one frequency, bit 1 = another. Old modems’ sound
  • PSK — phase shift keying: bits map to carrier phases. BPSK (2 phases = 1 bit), QPSK (4 phases = 2 bits)
  • QAM — quadrature amplitude modulation: vary amplitude and phase together, so each symbol carries many bits. 16-QAM = 4 bits/symbol, 256-QAM = 8. This is how Wi-Fi, cable and LTE pack high data rates into limited spectrum

The trade-off you can feel in your daily life: higher-order QAM sends more bits per symbol but the constellation points sit closer together, so noise flips them more easily. This is why your Wi-Fi drops to a lower rate when the signal is weak — it falls back to a lower-order modulation with more noise margin. Same channel, fewer bits per symbol, more reliable → Shannon’s limit.

I/Q — the representation that makes it all computable

Modern radio represents a signal as two components: In-phase (I) and Quadrature (Q) — the signal multiplied by a cosine and by a sine. Together they’re the real and imaginary parts of a complex signal, and they capture both amplitude and phase at once → why complex numbers.

Why I/Q is the whole game: every modulation scheme becomes simple arithmetic on I/Q samples. QAM is literally a grid of (I, Q) points (the constellation diagram you see in RF tools). Demodulation is measuring which grid point each received sample is nearest. Once a signal is I/Q samples, modulation and demodulation are just DSP — filtering, FFTs, and geometry on complex numbers.

Software-Defined Radio

Traditionally, radio was fixed-function circuits — a specific chip for FM, another for your car key, another for Wi-Fi, each mixing and filtering in hardware. SDR replaces the circuits with an ADC/DAC and a computer: the antenna front-end digitises a chunk of spectrum into I/Q samples, and all the demodulation happens in software.

   antenna ──► [ RF front-end + ADC ] ──► I/Q samples ──► [ SOFTWARE: any radio ]
                digitise a band              (numbers)      FM, ADS-B, GPS, Wi-Fi, ...

Why this is transformative:

  • One device becomes any radio. Reprogram it to decode aircraft transponders, then weather satellites, then your car’s key fob — no hardware change
  • New protocols ship as software updates, not new chips
  • It democratised RF. An RTL-SDR dongle costs ~$30 and turns DSP knowledge into a working radio receiver on your laptop → tooling

GNU Radio is the standard framework — you build a radio as a flowgraph of DSP blocks (filters, FFTs, demodulators). Learning SDR is applied DSP: every block is a concept from this folder.

The honest caveats: SDR trades hardware efficiency for flexibility — a purpose-built chip is far more power-efficient, so phones still use dedicated silicon for the heavy lifting. And transmitting is legally restricted (you need a licence and the right band); receiving on an RTL-SDR is the safe, legal playground → RF.

Key insight

Modulation shifts low-frequency information onto a high-frequency carrier by varying its amplitude, frequency or phase — and once you represent the signal as I/Q (complex) samples, every modulation scheme becomes simple DSP arithmetic. That realisation is what makes software-defined radio possible: digitise a band into I/Q, and a computer can be any radio in software. A $30 dongle plus GNU Radio turns this whole folder into a working receiver, which is the most tangible payoff DSP offers.

Source: [reference] — Aug 2026. Transmitting is licence-restricted; receiving on an RTL-SDR is the legal playground.

↑ Back to Digital Signal Processing