3. Digital Communication

Learning Objectives

• Trace a message through the blocks of a digital communication system, from source encoder to decoder
• Explain sampling, quantization, and PCM encoding and apply the Nyquist criterion
• Distinguish ASK, FSK, and PSK and identify which is most bandwidth- and power-efficient
• State Shannon's channel capacity formula and interpret what it means for real systems
• Explain how error-detection and error-correction codes (parity, Hamming, CRC, FEC) protect data
• Identify why digital communication has replaced analog for most high-capacity applications

Quick Answer

Digital communication represents information as a sequence of discrete symbols — usually bits — rather than a continuous waveform. A digital transmitter samples and quantizes the source (if it started analog, like voice), encodes it into bits, adds redundancy for error protection, and modulates the bits onto a carrier using a digital keying scheme such as ASK, FSK, or PSK. The receiver reverses each step. Because the receiver only needs to decide between a small number of discrete signal values, digital systems tolerate far more noise than analog ones and can regenerate a perfect copy of the original bits at each repeater — this is the central reason digital communication now underlies the internet, mobile networks, and virtually all modern high-capacity links.

From Analog Source to Digital Bitstream

Most information in the real world — voice, video, sensor readings — starts out analog. A digital system's first job is converting it faithfully into bits.

Sampling and the Nyquist Criterion

Sampling takes instantaneous snapshots of a continuous signal at regular intervals. The Nyquist-Shannon sampling theorem states that if a signal's highest frequency component is fmax, you must sample at a rate fs ≥ 2fmax to reconstruct it without loss. Standard telephone voice, band-limited to about 3.4 kHz, is sampled at 8 kHz precisely because that comfortably clears the Nyquist rate.

Why it matters: Sample below the Nyquist rate and you get aliasing — high frequencies fold back and masquerade as false low frequencies that cannot be removed afterward. This is why every ADC (analog-to-digital converter) is preceded by an anti-aliasing low-pass filter.

Common misunderstanding: Students often think "more samples per second is always better." Oversampling does improve fidelity margin, but it also increases the data rate and storage needed; real designs pick the minimum rate that satisfies Nyquist with a practical safety margin, not the maximum possible.

Quantization and PCM

After sampling, each sample's amplitude is rounded to the nearest of a fixed set of discrete levels — quantization. Pulse Code Modulation (PCM) then represents each quantized level as a binary code word. An 8-bit PCM system, for example, offers 256 quantization levels. The unavoidable rounding error is called quantization noise, and it sets a fundamental floor on digital audio fidelity that more bits can push down (each extra bit improves SNR by about 6 dB) but never eliminate.

Why it matters: PCM is the encoding used inside CDs (16-bit, 44.1 kHz), digital telephony (8-bit, 8 kHz, μ-law/A-law companding), and as the baseline that more advanced codecs compress further.

Digital Modulation (Keying)

Once the message is a bitstream, it still needs to ride on a carrier for transmission. Digital modulation reuses the same three carrier parameters as analog modulation, but the message signal now takes only discrete values.

• Amplitude Shift Keying (ASK): Carrier amplitude switches between two (or more) levels to represent bits — simplest, but most vulnerable to noise since noise directly changes amplitude.
• Frequency Shift Keying (FSK): Carrier frequency switches between two (or more) discrete tones — more noise-resistant than ASK, used in early modems and some low-rate IoT radios.
• Phase Shift Keying (PSK): Carrier phase switches between discrete values (e.g., BPSK uses 2 phases, QPSK uses 4) — the most bandwidth-efficient and noise-resistant of the three, and the basis of Wi-Fi, LTE, and satellite links.
• Quadrature Amplitude Modulation (QAM): Combines amplitude and phase to pack more bits per symbol (16-QAM, 64-QAM, 256-QAM), trading noise margin for higher data rate — used when the channel SNR is good, as in cable modems and Wi-Fi's fastest modes.

Why it matters: Choosing a modulation scheme is a direct trade between data rate, bandwidth, and required SNR — this is exactly why your Wi-Fi router automatically drops from 256-QAM to a simpler scheme like QPSK when you walk further from it and the signal weakens (adaptive modulation).

Channel Capacity: Shannon's Limit

Shannon's theorem gives the absolute upper bound on error-free data rate for a channel:

C = B log2(1 + S/N)

where C is capacity in bits per second, B is bandwidth in Hz, and S/N is the signal-to-noise power ratio. No coding scheme, however clever, can exceed this limit reliably. Modern codes (turbo codes, LDPC) now operate within a fraction of a decibel of the Shannon limit — a landmark achievement of digital communication theory.

Common misunderstanding: Students sometimes think Shannon's formula gives the actual data rate a system achieves. It only gives the theoretical ceiling; real systems always run somewhat below it due to practical modulation, coding, and synchronization overhead.

Error Detection and Correction

Because channels are noisy, digital systems add controlled redundancy so errors can be caught or fixed:

• Parity check: Adds a single bit so the total number of 1s is even (or odd); detects single-bit errors but cannot correct them.
• Cyclic Redundancy Check (CRC): Computes a checksum via polynomial division; very effective at detecting burst errors, used in Ethernet and storage systems.
• Hamming codes: Add multiple parity bits positioned so the receiver can not only detect but pinpoint and correct a single-bit error.
• Forward Error Correction (FEC): Adds enough redundancy that the receiver can correct errors without asking for retransmission — essential for one-way links like satellite broadcast and deep-space communication, where a retransmission request would be too slow or impossible.

Why it matters: The choice between "detect and retransmit" (used by TCP over the internet) and "correct in place" (FEC, used in satellite and storage) depends on whether a return channel exists and how costly a retransmission would be.

Key Terms

Term	Definition	Related Concept
Sampling	Measuring a continuous signal's amplitude at regular time intervals	Nyquist theorem, ADC
Nyquist rate	Minimum sampling rate (2× highest signal frequency) needed to avoid aliasing	Sampling, anti-aliasing filter
Quantization	Rounding sampled amplitudes to the nearest available discrete level	PCM, quantization noise
PCM (Pulse Code Modulation)	Representing quantized samples as binary code words	Quantization, digital audio
ASK / FSK / PSK	Digital modulation schemes that key a carrier's amplitude, frequency, or phase between discrete states	Digital modulation, QAM
QAM	Modulation combining amplitude and phase to encode multiple bits per symbol	Bandwidth efficiency, adaptive modulation
Shannon capacity	Theoretical maximum error-free data rate of a channel given bandwidth and SNR	C = B log2(1+S/N)
Forward Error Correction (FEC)	Adding redundancy so the receiver can correct errors without retransmission	Hamming code, satellite links
CRC	A checksum-based method for detecting (not correcting) transmission errors	Error detection, Ethernet

Common Mistakes

Misconception: Digital communication is "immune" to noise. Why it's wrong: Digital signals are only more resistant to noise, not immune. If noise is large enough to push a received symbol past the decision threshold, a bit error still occurs. Correct understanding: Digital communication tolerates noise up to a point because the receiver only needs to distinguish between a few discrete levels; beyond that threshold, error correction — not immunity — is what saves the data.

---

Misconception: Higher-order modulation (like 256-QAM) is always better because it carries more bits per symbol. Why it's wrong: Packing more bits per symbol means the discrete symbol points sit closer together in signal space, so a 256-QAM signal needs a much higher SNR to be decoded reliably than BPSK or QPSK. Correct understanding: Modulation order is chosen based on channel quality — high SNR channels can exploit 256-QAM for higher throughput, but weak or noisy channels must fall back to simpler, more robust schemes like QPSK.

---

Misconception: Shannon's capacity formula tells you exactly how fast your Wi-Fi or 4G connection will run. Why it's wrong: Shannon's formula assumes ideal coding with unlimited complexity and delay. Real systems have protocol overhead, imperfect coding, and shared channel access, so achieved throughput is always below the Shannon limit. Correct understanding: Shannon's capacity is an upper bound used for comparing how close real systems get to theoretical optimum, not a prediction of real-world throughput.

Comparison and Connections

Feature	Analog Communication	Digital Communication
Signal representation	Continuous waveform	Discrete bits
Noise handling	Noise accumulates and distorts permanently	Regeneration restores a clean signal below error threshold
Error correction	Not applicable	Built-in via coding (Hamming, CRC, FEC)
Bandwidth efficiency	Fixed by modulation type	Adjustable via modulation order (QPSK to 256-QAM)
Hardware complexity	Simpler	Requires ADC/DAC, coding, synchronization
Typical modern use	Legacy broadcast, simple links	Internet, mobile networks, storage, most new systems

Practice Questions

Recall

1. State the Nyquist sampling theorem and explain why standard telephone speech is sampled at 8 kHz. Answer guidance: Sampling rate must be at least twice the highest frequency component. Telephone speech is band-limited to about 3.4 kHz, so 8 kHz sampling comfortably satisfies fs ≥ 2fmax.

2. Name the six blocks of a typical digital communication system in order. Answer guidance: Source encoder, channel encoder, modulator, channel, demodulator, channel decoder, source decoder (destination). (Reference lists source encoder through decoder as the core chain.)

Understanding

3. Explain why ASK is more vulnerable to noise than FSK or PSK. Answer guidance: ASK encodes information in amplitude, and noise directly adds to or subtracts from amplitude, easily confusing decision thresholds. FSK and PSK encode information in frequency or phase, which are less directly disturbed by additive amplitude noise.

4. Why can a digital repeater regenerate a perfectly clean signal while an analog amplifier cannot? Answer guidance: A digital repeater only needs to decide which of a few discrete symbols was sent, then re-transmit a fresh, ideal version of that symbol. An analog amplifier has no such decision step — it amplifies the signal and any accumulated noise together.

Application

5. A channel has 1 MHz of bandwidth and an SNR of 30 dB (S/N = 1000). Estimate its Shannon capacity. Answer guidance: C = B log2(1+S/N) = 10^6 × log2(1001) ≈ 10^6 × 9.97 ≈ 9.97 Mbps.

6. A satellite broadcast has no return channel to request retransmission. Which error-control approach must it use, and why? Answer guidance: Forward Error Correction (FEC), because the receiver must correct errors using only the redundancy already sent — there is no way to ask the transmitter to resend.

Analysis

7. Compare parity check, CRC, and Hamming codes in terms of what each can do (detect vs. correct) and where each is best applied. Answer guidance: Parity check detects single-bit errors only, cheap but limited. CRC detects burst errors reliably (used in Ethernet, file storage) but does not correct. Hamming codes can detect and correct single-bit errors, useful in memory systems (ECC RAM) where correction without retransmission is valuable.

8. A Wi-Fi router automatically switches from 256-QAM to QPSK as a laptop moves further away. Explain the reasoning in terms of Shannon's theorem and modulation trade-offs. Answer guidance: As distance increases, received SNR drops. High-order 256-QAM needs high SNR to keep symbol points distinguishable; at low SNR it would produce excessive errors. QPSK uses fewer, more widely separated symbol points, tolerating lower SNR at the cost of a lower bit rate — matching Shannon's principle that achievable rate depends on both bandwidth and SNR.

FAQ

Why do we sample and quantize instead of just sending the analog signal digitally as-is? An "analog signal sent digitally" is a contradiction — anything transmitted as bits must first be converted into discrete samples and discrete amplitude levels. Sampling and quantization are exactly this conversion step, turning a continuous voice or video signal into the 0s and 1s that digital hardware, storage, and coding can work with.

What's the practical difference between error detection and error correction? Detection (like CRC) tells you that an error occurred so you can ask for a retransmission — cheap in bits but requires a return channel and adds delay. Correction (like Hamming codes or FEC) uses more redundant bits so the receiver can fix the error itself immediately, which is essential when retransmission is slow, costly, or impossible.

Why does quantization always introduce some noise, no matter how many bits you use? Quantization rounds a continuous value to the nearest of a finite number of levels; the difference between the true and the rounded value is quantization error. Adding bits increases the number of levels (halving the rounding error per extra bit, about 6 dB SNR improvement), but with a finite number of bits some rounding error always remains.

Is QAM a form of digital modulation or analog modulation? QAM is digital modulation — it's used to transmit digital bitstreams — but its underlying mechanism (varying both amplitude and phase of a carrier) is mathematically the same idea used in analog AM and PM combined. The message driving it, however, is a stream of discrete symbols rather than a continuous waveform.

How close are real systems to the Shannon limit today? Modern codes like LDPC (used in Wi-Fi 6, 5G, and DVB-S2 satellite broadcast) and turbo codes (used in 3G/4G and deep-space probes) operate within about 0.5–1 dB of the Shannon limit for their respective channel conditions — remarkably close to the theoretical best possible performance.

Quick Revision

• Digital systems: source encoder → channel encoder → modulator → channel → demodulator → channel decoder → source decoder
• Nyquist rate: sample at fs ≥ 2fmax to avoid aliasing; telephone voice uses 8 kHz sampling
• PCM encodes quantized samples as binary words; more bits reduce quantization noise (~6 dB per bit)
• ASK keys amplitude, FSK keys frequency, PSK keys phase; PSK is most noise-resistant of the three
• QAM combines amplitude and phase to pack more bits per symbol at the cost of needing higher SNR
• Shannon's capacity: C = B log2(1 + S/N) — the theoretical ceiling, not the achieved rate
• Parity check detects only; CRC detects bursts; Hamming codes and FEC can correct errors
• FEC is essential when there is no return channel (satellite broadcast, deep space)
• Digital repeaters regenerate a perfect signal; analog amplifiers amplify noise along with the signal
• Modern codes like LDPC and turbo codes approach the Shannon limit within a fraction of a decibel

Related Topics

Prerequisites: Basics of Communication Systems, Analog Communication, binary number systems, Fourier analysis basics

Related Topics: Modulation Techniques, Signal Processing in Communication, Communication Protocols

Next Topics: Modulation Techniques (digital keying in depth), Wireless Communication, Communication Protocols