FSK Modulation

Frequency Shift Keying (FSK): Encoding Data in Frequency

Frequency Shift Keying is one of the oldest and most reliable digital modulation techniques. Instead of changing the amplitude or phase of a carrier, FSK changes its frequency to represent different digital symbols. If you have ever heard a dial-up modem's distinctive screech, you were listening to FSK in action — those alternating tones were representing ones and zeros.

The Fundamental Concept

Think of it this way: imagine two tuning forks, one pitched slightly higher than the other. To send a "1," you strike the higher fork. To send a "0," you strike the lower fork. The receiver simply listens to which pitch arrives and decodes the data. This frequency-based encoding makes FSK naturally resistant to amplitude variations — even if the signal gets weaker (amplitude drops), the frequency remains unchanged.

Binary FSK (BFSK) signal:

s₁(t) = A × cos(2πf₁ × t) for bit "1" s₀(t) = A × cos(2πf₂ × t) for bit "0"

Where f₁ and f₂ are two distinct frequencies separated by Δf = f₁ - f₂.

The general expression: s(t) = A × cos(2π(fc + mΔf) × t)

Where m = +1 for bit "1" and m = -1 for bit "0", and fc is the center frequency.

Continuous-Phase FSK (CPFSK)

In basic FSK, abruptly switching between frequencies creates phase discontinuities at bit boundaries, which generate spectral splatter (wideband interference). Continuous-Phase FSK solves this by ensuring the phase is continuous across bit transitions:

s(t) = A × cos(2πfct + 2πΔf ∫m(τ)dτ)

The most important CPFSK variant is Minimum Shift Keying (MSK), where:

• Frequency deviation: Δf = 1/(4Tb) = Rb/4
• Modulation index: h = 2ΔfTb = 0.5 (the minimum that ensures orthogonality)

MSK achieves the narrowest bandwidth possible while maintaining orthogonality between the two frequencies. It is used in GSM cellular systems (as GMSK — Gaussian-filtered MSK).

Bandwidth Analysis

FSK bandwidth depends on the frequency deviation and data rate:

Carson's rule for FSK bandwidth: BW ≈ 2(Δf + Rb) = 2Rb(1 + β)

Where β = Δf/Rb is the modulation index.

• Wideband FSK (β >> 1): BW ≈ 2Δf — noise immunity improves but bandwidth is large
• Narrowband FSK (β << 1): BW ≈ 2Rb — bandwidth-efficient but less noise-resistant
• MSK (β = 0.5): BW ≈ 1.5Rb with 99% power containment

Comparison of bandwidth requirements:

FSK Modulator Circuits

Voltage-Controlled Oscillator (VCO) Method:

1. Binary data → level converter (maps 0/1 to voltage levels V₁, V₂)
2. Voltage levels → VCO input (frequency varies linearly with input voltage)
3. VCO output = FSK signal

This naturally produces continuous-phase FSK since the VCO phase is inherently continuous.

Switched Oscillator Method:

1. Two separate oscillators at f₁ and f₂
2. Data bit selects which oscillator's output passes through
3. Creates discontinuous-phase FSK (phase jumps at transitions)

Digital Synthesis (DDS) Method:

1. Numerically Controlled Oscillator (NCO)
2. Data bit changes frequency control word
3. DAC converts digital samples to analog waveform
4. Provides precise, stable frequency control

FSK Demodulation Techniques

Coherent Detection:

1. Two matched filters (or correlators) tuned to f₁ and f₂
2. Sample outputs at bit-time intervals
3. Compare: larger output indicates the transmitted frequency
4. BER = Q(√(Eb/N₀)) for orthogonal BFSK

Non-Coherent Detection:

1. Two bandpass filters centered at f₁ and f₂
2. Envelope detectors on each filter output
3. Compare envelopes: larger indicates transmitted frequency
4. BER = (1/2) × exp(-Eb/(2N₀)) — about 1 dB worse than coherent

Discriminator Detection:

1. A frequency-to-voltage converter (FM discriminator)
2. Output voltage proportional to instantaneous frequency
3. Threshold comparison recovers binary data
4. Simple but suboptimal — used in low-cost receivers

Error Performance

For BFSK in AWGN:

Coherent detection: BER = Q(√(Eb/N₀)) Non-coherent detection: BER = (1/2) × exp(-Eb/(2N₀))

At BER = 10⁻⁵:

• Coherent BFSK needs Eb/N₀ ≈ 12.6 dB
• Non-coherent BFSK needs Eb/N₀ ≈ 13.5 dB
• BPSK needs Eb/N₀ ≈ 9.6 dB

FSK requires about 3 dB more power than BPSK, but offers the crucial advantage of non-coherent detection — no carrier phase recovery is needed, greatly simplifying the receiver.

M-ary FSK: Trading Bandwidth for Power

M-FSK uses M different frequencies to transmit log₂(M) bits per symbol:

Symbol error probability (coherent): Ps ≈ (M-1) × Q(√(Eb × log₂M / N₀))

As M increases:

• Power efficiency improves (less Eb/N₀ needed for same BER)
• Bandwidth increases (M orthogonal frequencies needed)
• Approaches Shannon capacity as M → ∞

This is the bandwidth-power trade-off: M-FSK trades bandwidth for power efficiency, opposite to M-QAM which trades power for bandwidth.

Example: 64-FSK transmits 6 bits per symbol and needs only Eb/N₀ ≈ 4 dB for BER = 10⁻⁵, but requires 64 orthogonal frequency slots.

Practical Applications

FSK is chosen for applications valuing reliability and simplicity over bandwidth efficiency:

1. Caller ID — 1200 baud FSK (Bell 202 standard) between ring signals
2. Pagers — POCSAG protocol uses 512/1200/2400 baud FSK
3. LoRa/IoT — Chirp Spread Spectrum (a variant of FSK) for long-range IoT
4. Bluetooth — GFSK (Gaussian FSK) at 1 Mbps
5. RFID — Many RFID standards use FSK for robustness
6. AIS (ship tracking) — GMSK at 9600 baud
7. Amateur radio — AFSK (Audio FSK) for packet radio and APRS

Why FSK Survives in the Modern Era

Despite being less spectrally efficient than PSK or QAM, FSK thrives because:

• Constant envelope allows Class C amplifiers (highest efficiency)
• Non-coherent detection simplifies receivers dramatically
• Excellent performance in fading channels (amplitude insensitive)
• Simple implementation — a VCO and a comparator suffice
• Robust to non-linear amplification (no AM/PM conversion)

Key Takeaways

1. FSK encodes digital data by switching carrier frequency between discrete values, providing inherent amplitude noise immunity.
2. Continuous-phase FSK (CPFSK/MSK) eliminates spectral splatter from phase discontinuities and achieves excellent spectral efficiency.
3. Non-coherent FSK detection requires about 1 dB more SNR than coherent but eliminates carrier synchronization complexity.
4. M-FSK trades bandwidth for power efficiency — increasing M improves power efficiency while consuming more spectrum.
5. FSK's constant envelope enables use of highly efficient non-linear amplifiers (Class C), crucial for battery-powered devices.
6. Applications span from century-old telegraphy principles to modern IoT (LoRa) and Bluetooth, proving FSK's enduring utility.