Digital Modulation Overview
Every radio signal that carries digital data — a text message, an internet packet, a satellite image, a Bluetooth audio stream — is the product of a digital modulation scheme that maps bits onto RF waveform parameters. The engineer or technician who understands digital modulation can read a spectrum analyzer display and identify the mode, predict bandwidth from symbol rate, calculate the link margin needed for reliable communication, and select the right mode for the available channel. This lesson surveys the complete family of digital modulation schemes from the simplest (OOK) to the complex (OFDM and spread spectrum), building the conceptual framework you need to understand and choose any wireless module or data link.
The Three Modulation Parameters
Any sinusoidal carrier has exactly three parameters that can be changed to carry information:
- A(t) — amplitude → ASK, OOK, QAM
- f(t) — frequency → FSK, FM, OFDM
- φ(t) — phase → PSK, PM, QAM
Any digital modulation scheme changes one or more of these parameters in discrete steps synchronized to the symbol clock. The number of discrete states determines how many bits each symbol carries: M states carry log2(M) bits per symbol. A binary scheme (M=2) carries 1 bit/symbol; a 64-state scheme (M=64) carries 6 bits/symbol.
Amplitude-Shift Keying (ASK)
ASK varies the carrier amplitude to represent symbols. Binary ASK with levels {0, A} is on-off keying (OOK) — the simplest digital modulation scheme, used by key fobs, garage-door remotes, wireless doorbells, low-cost 433 MHz modules and infrared remote controls. Multi-level ASK (4-ASK uses {0, A/3, 2A/3, A}; 8-ASK uses eight levels) encodes more bits per symbol at the cost of reduced noise margin.
ASK is very sensitive to amplitude-varying channel impairments: fading, multipath, interference and non-linearities that distort amplitude all directly corrupt ASK symbols. This makes multi-level ASK poorly suited to radio links but acceptable on controlled links such as optical fiber. Simple OOK works well enough for short-range remotes only because the receiver makes a binary (on/off) decision — any signal above the noise threshold is decoded as a “1,” so the absolute amplitude need not be known — and because a lost button press can simply be repeated.
- Simplest implementation — varies only carrier amplitude
- Poor noise and fading immunity for multi-level variants
- OOK (binary ASK) is robust for simple threshold detection
- Occupied bandwidth ≈ symbol rate × (1 + rolloff factor)
Frequency-Shift Keying (FSK)
FSK switches the carrier frequency between M discrete values (tones) to represent M symbols. Binary FSK (2-FSK, BFSK) uses two frequencies f1 and f2 separated by a frequency shift Δf. The ratio of frequency shift to symbol rate is the modulation index h = Δf / fs.
When h = 0.5, the two FSK tones are orthogonal (their cross-correlation is zero) — this is Minimum-Shift Keying (MSK). MSK has the narrowest possible bandwidth for coherent binary FSK and produces a constant-envelope signal (no amplitude variation), making it suitable for nonlinear amplifiers. Gaussian MSK (GMSK) pre-filters the baseband data with a Gaussian filter before modulation, reducing spectral occupancy further at the cost of slight inter-symbol interference. GMSK is the modulation scheme used in GSM (2G cellular).
BW ≈ 2(Δf + fs)
where Δf = frequency deviation (half the total shift), fs = symbol rate
A wireless temperature sensor sends 2-FSK at 9,600 Baud with the two tones 20 kHz apart.
Δf = 10 kHz, fs = 9.6 kBaud
BW ≈ 2(10 + 9.6) = ≈ 39 kHz (Carson's rule applied to FSK)
The receiver's IF filter is typically 50–100 kHz wide to allow for crystal frequency tolerance between transmitter and receiver.
4-FSK uses four tones to carry 2 bits per symbol. Bluetooth Classic and Bluetooth Low Energy use GFSK (two tones, Gaussian filtered) at 1 or 2 Mbaud in a 1–2 MHz channel; many digital two-way radio systems use 4-level FSK; and pagers use 2- and 4-level FSK. Because FSK has a constant envelope, the transmitter can use an efficient saturated amplifier, which is why it is the choice for battery-powered devices.
Phase-Shift Keying (PSK)
PSK switches the carrier phase between M equally spaced values. BPSK (M=2) and QPSK (M=4) were covered in the Phase Modulation lesson. Higher orders:
- 8-PSK: 8 phases, 3 bits/symbol. Used in DVB-S2 satellite broadcasting and some HF modes.
- 16-PSK: 16 phases, 4 bits/symbol. Rarely used alone — 16-QAM achieves the same bits/symbol with better noise margin.
As M increases, adjacent constellation points get closer together (for fixed total power), reducing noise margin. PSK requires careful phase recovery (carrier synchronization) at the receiver — a phase-locked loop or Costas loop that tracks the incoming carrier phase. This complexity limits high-order PSK on channels with phase noise. PSK is constant-envelope (amplitude stays fixed), so it works well through nonlinear amplifiers.
Quadrature Amplitude Modulation (QAM)
QAM simultaneously varies both the amplitude and phase of the carrier, combining ASK and PSK on two orthogonal carriers (I and Q). A 16-QAM signal has 16 points arranged in a 4×4 rectangular grid in the I-Q plane — 4 amplitude levels on the I axis crossed with 4 amplitude levels on the Q axis. Each symbol carries log2(16) = 4 bits.
| Order | Grid | Bits/Symbol | Common Use |
|---|---|---|---|
| 4-QAM (= QPSK) | 2×2 | 2 | Satellite, 802.11 low MCS |
| 16-QAM | 4×4 | 4 | Cable, 802.11, HF ALE |
| 64-QAM | 8×8 | 6 | Cable TV, 802.11n/ac |
| 256-QAM | 16×16 | 8 | DOCSIS cable, 802.11ac/ax |
| 1024-QAM | 32×32 | 10 | Wi-Fi 6 (802.11ax) |
QAM requires both amplitude and phase accuracy — the receiver must know the exact amplitude and phase reference to distinguish adjacent constellation points. This makes QAM sensitive to nonlinearities (which distort amplitude) and phase noise (which rotates the constellation). High-order QAM (256-QAM and above) is only practical on channels with very high SNR (typically >30 dB for 256-QAM).
OFDM: Orthogonal Multi-Carrier Modulation
Orthogonal frequency-division multiplexing (OFDM) divides a wide bandwidth channel into many narrowband subcarriers, each carrying a low-symbol-rate signal. The subcarriers are spaced exactly at intervals of 1/Ts (where Ts is the symbol period) so that each subcarrier is at a zero crossing of every other subcarrier's spectrum — they are orthogonal. The received signal on each subcarrier can be recovered independently despite the overlapping spectra.
OFDM's key advantage: each narrow subcarrier experiences flat fading (nearly constant amplitude and phase across its bandwidth) even when the total channel exhibits frequency-selective fading. A guard interval (cyclic prefix) added to each symbol absorbs multipath echoes without causing inter-symbol interference. This makes OFDM the dominant modulation scheme for broadband wireless: 802.11a/g/n/ac/ax (Wi-Fi), 4G LTE, 5G NR, and digital TV (DVB-T) all use OFDM or its variants.
OFDM is used by Wi-Fi (802.11a/g/n/ac/ax), 4G LTE and 5G, digital audio broadcasting (DAB), digital terrestrial TV (DVB-T) and power-line networking, all of which must cope with strong echoes (multipath) from buildings and walls. Each subcarrier runs slowly enough that the echoes die away within a short guard interval at the start of each symbol.
Spread Spectrum
Spread-spectrum techniques deliberately spread a signal over a bandwidth much wider than the minimum required, trading spectral efficiency for interference immunity and coexistence capability.
Direct-Sequence Spread Spectrum (DSSS)
The data stream is multiplied by a pseudorandom noise (PN) code running at a much higher rate (the chip rate). The transmitted bandwidth equals the chip rate rather than the data rate. At the receiver, correlation with the same PN code de-spreads the signal, recovering the data while spreading any narrowband interferer across the full chip-rate bandwidth. The processing gain is the ratio of chip rate to data rate, typically 10–30 dB. DSSS is used in GPS, CDMA cellular, and 802.11b (Wi-Fi).
Frequency-Hopping Spread Spectrum (FHSS)
The carrier frequency hops among many channels in a pseudorandom sequence synchronized between transmitter and receiver. Each transmission occupies a narrow channel for a short dwell time, then hops. An interferer at a fixed frequency captures only the brief dwell intervals, causing partial-band interference rather than continuous jamming. FHSS is used in Bluetooth, 802.11 FHSS, and military communications.
Comparison Table
Spectral efficiency against required signal-to-noise ratio. Dense modulations such as 256-QAM carry many bits per hertz but need a strong signal; GPS and LoRa trade data rate for the ability to work below the noise.
View Larger| Scheme | Bits/Symbol | Constant Envelope? | Sensitivity to Phase Noise | Sensitivity to Amplitude Variation |
|---|---|---|---|---|
| OOK / BASK | 1 | No | None | High |
| BFSK / MSK | 1 | Yes | Low | None |
| BPSK | 1 | Yes | Low | None |
| QPSK | 2 | Yes | Low-medium | None |
| 8-PSK | 3 | Yes | Medium | None |
| 16-QAM | 4 | No | Medium | Medium |
| 64-QAM | 6 | No | High | High |
| 256-QAM | 8 | No | Very high | Very high |
Digital Modulation in Everyday Devices
Each wireless system chooses its modulation to suit its range, data rate, power budget and cost:
- Key fobs, remotes and cheap 433/315 MHz modules: OOK or simple FSK, a few kbit/s. A single transistor transmitter and a simple receiver keep the cost to cents.
- RFID tags and contactless cards: ASK from the reader to the card; the card answers by load modulation, switching a load on and off to change how much energy it absorbs from the reader's field.
- Bluetooth and Bluetooth Low Energy: GFSK at 1–2 Mbit/s (with PSK modes for higher rates), hopping among channels across 2.4 GHz to avoid interference.
- Zigbee and Thread: O-QPSK with direct-sequence spread spectrum at 250 kbit/s in the 2.4 GHz band.
- LoRa: chirp spread spectrum, a few hundred bit/s to tens of kbit/s, reaching several kilometers on a small battery by trading data rate for sensitivity.
- Wi-Fi, 4G and 5G: OFDM with BPSK, QPSK and 16- to 1024-QAM subcarriers, changing the modulation from moment to moment to match the signal quality.
- GPS: BPSK with direct-sequence spread spectrum, received below the noise floor thanks to its processing gain.
Frequently Asked Questions
Why can't we always use 256-QAM to maximize data throughput?
256-QAM requires extremely high SNR — typically 30 dB or more — because 256 constellation points must be distinguishable in the presence of noise. On a long-distance HF or mobile radio channel, multipath fading routinely causes the SNR to vary by 20–30 dB within seconds. At low SNR, 256-QAM produces catastrophic bit error rates. Adaptive modulation systems (used in Wi-Fi and LTE) solve this by monitoring channel quality and switching modulation order in real time — 256-QAM when SNR is high, BPSK when SNR is low — always operating at the highest sustainable rate. Long-range links such as HF data modems, GPS and LoRa are designed for worst-case conditions, so they use low-order modulation (BPSK, QPSK, BFSK, chirps) for robustness.
What is the difference between symbol rate (baud) and bit rate?
Symbol rate (baud) is the number of symbol transitions per second. Bit rate is the number of bits delivered per second. They are equal only for binary modulation (1 bit/symbol). For M-ary modulation: bit rate = symbol rate × log2(M). A 16-QAM link running at 1,000 baud delivers 4,000 bits/s. The channel bandwidth is determined by the symbol rate, not the bit rate — which is why higher-order modulation improves spectral efficiency (more bits per unit bandwidth).
Why do some radio links prefer FSK over PSK despite its lower spectral efficiency?
PSK requires precise carrier phase recovery at the receiver — the demodulator must know the absolute phase of the received carrier to distinguish phase states. On a fading channel — a long-distance HF link or a moving vehicle — propagation introduces rapid phase variations that are difficult to track, especially at low SNR. FSK detection requires only identifying which of several frequencies is present — a frequency comparison that is much more robust to phase variations. This is why FSK has long been used for HF data links, pagers and low-cost sensor links: it works under conditions where PSK carrier recovery fails. More sophisticated PSK modes use differential encoding (DBPSK) or pilot tones to work around the phase recovery problem.
What is processing gain in spread spectrum and why does it matter?
Processing gain is the ratio of spread bandwidth to data bandwidth: Gp = Bspread / Bdata, expressed in dB as 10×log10(Gp). It represents how much narrowband interference is attenuated when the PN code de-spreads the received signal. A DSSS system with chip rate 10 MHz and data rate 100 kHz has processing gain = 10,000,000 / 100,000 = 100 = 20 dB. This means a narrowband interferer can be 20 dB stronger than the desired signal and still be rejected after despreading. Processing gain allows DSSS systems to operate in the presence of strong interference and enables multiple users to share the same bandwidth (CDMA).
How can GPS and LoRa receive signals that are below the noise floor?
Both use spread spectrum to gain processing gain, trading data rate for sensitivity. A GPS signal spreads a 50 bit/s navigation message over a 2 MHz-wide code; the receiver correlates the incoming signal with its own copy of the code, which collects the signal energy spread across that bandwidth while the noise does not add up in the same way. The processing gain is about 10 × log₁₀(2,000,000 / 50) ≈ 46 dB, so a signal arriving 20 dB below the noise can still be decoded. LoRa uses chirps — tones that sweep across the channel — with spreading factors of 7 to 12; each step up halves the data rate and adds about 2.5 dB of sensitivity, allowing reception at around −20 dB SNR. In both cases the price is a very slow data rate, which is perfectly acceptable for a position fix or a sensor reading.
Test Your Knowledge
Answer the questions below to check your understanding of this lesson. Every answer can be found in the lesson above.