Frequency Modulation (FM)

Frequency Modulation is an angle modulation technique where the instantaneous frequency of the carrier is varied linearly with the message signal amplitude. FM offers superior noise performance compared to AM by trading bandwidth for signal quality, making it the standard for high-fidelity audio broadcasting, two-way radio, and many wireless communication systems.

Mathematical Representation

Instantaneous Frequency

f(t) = fc + kf · m(t)

Where kf is the frequency sensitivity (Hz/V).

Instantaneous Phase

φ(t) = 2π ∫ f(τ)dτ = 2πfc·t + 2πkf ∫ m(τ)dτ

FM Signal

s(t) = Ac · cos[2πfc·t + 2πkf ∫₀ᵗ m(τ)dτ]

For single-tone modulation m(t) = Am·cos(2πfm·t):

s(t) = Ac · cos[2πfc·t + β·sin(2πfm·t)]

Where β = Δf/fm = kf·Am/fm is the modulation index.

FM Waveform Visualization

Key Parameters

Carson's Rule for FM Bandwidth

BT ≈ 2(Δf + fm) = 2fm(β + 1)

This approximation contains approximately 98% of the signal power.

Exact Bandwidth (using Bessel functions)

The FM signal can be expanded as:

s(t) = Ac Σ Jn(β) · cos[2π(fc + n·fm)·t]

Where Jn(β) is the Bessel function of the first kind, order n.

Significant sidebands exist where |Jn(β)| > 0.01, approximately (β + 1) pairs.

Narrowband FM (NBFM) vs Wideband FM (WBFM)

Bessel Function Table (Selected Values)

Note: At β = 2.4, J0 = 0, meaning the carrier component vanishes completely!

FM Power Analysis

Total power of FM signal is constant regardless of modulation:

PT = Ac²/(2R) = Pc (always)

This is because FM only changes frequency, not amplitude. The power is distributed among the carrier and sidebands, but the total remains constant.

Power distribution: PT = Ac²/2 × [J0²(β) + 2·Σ Jn²(β)] = Ac²/2

FM Spectrum

Noise Performance of FM

SNR Improvement

For WBFM with threshold above (input SNR > 10 dB):

SNRo = 3β²(β+1) × SNRi (for single-tone)

Or more precisely:

(S/N)out = (3/2) × β² × (BT/fm) × (S/N)baseband

FM Threshold Effect

Below a certain input SNR (typically ~10 dB), FM performance degrades catastrophically. Impulse noise spikes exceed the signal, causing "clicks" that cannot be removed by the limiter.

Pre-emphasis and De-emphasis

FM noise has a parabolic (f²) power spectral density — more noise at higher frequencies:

Pre-emphasis: Boost high-frequency message components before modulation (τ = 75μs in USA, 50μs in Europe)

De-emphasis: Attenuate high frequencies after demodulation, reducing noise

Combined improvement: approximately 13 dB additional SNR gain.

Solved Example 1

Problem: An FM broadcast station operates with Δf = 75 kHz and maximum audio frequency of 15 kHz. Calculate (a) modulation index, (b) bandwidth by Carson's rule, (c) number of significant sidebands.

Solution:

(a) β = Δf/fm = 75,000/15,000 = 5

(b) BT = 2(Δf + fm) = 2(75,000 + 15,000) = 180 kHz (This fits within the 200 kHz FM channel spacing)

(c) Significant sidebands ≈ β + 1 = 5 + 1 = 6 pairs (12 sideband components)

Solved Example 2

Problem: A carrier of 100 MHz is frequency modulated by a 5 kHz sinusoid causing a peak deviation of 40 kHz. Find (a) the modulation index, (b) approximate bandwidth, (c) if the modulating frequency is doubled to 10 kHz with the same amplitude, find new β and bandwidth.

Solution:

(a) β = Δf/fm = 40,000/5,000 = 8

(b) BT = 2(40,000 + 5,000) = 90 kHz

(c) When fm doubles (amplitude same → deviation same): New β = 40,000/10,000 = 4 New BT = 2(40,000 + 10,000) = 100 kHz

Note: β decreased but bandwidth increased! This illustrates that bandwidth depends on both β and fm.

Solved Example 3

Problem: Calculate the FM signal power at the carrier frequency for β = 2.4 and total signal amplitude Ac = 10V across 50Ω.

Solution:

Total power: PT = Ac²/(2R) = 100/100 = 1 W (constant for all β)

Carrier component power: Pc = [Ac·J0(β)]²/(2R) = [10 × 0]²/100 = 0 W

At β = 2.4, the carrier completely vanishes! All power is in the sidebands.

First sideband pair power: 2 × [Ac·J1(2.4)]²/(2R) = 2 × [10×0.52]²/100 = 0.54 W

FM Generation Methods

Direct FM (Voltage-Controlled Oscillator)

The message directly controls the oscillator frequency:

Problem: Carrier frequency stability (solved with AFC or PLL)

Indirect FM (Armstrong Method)

Advantage: Crystal-controlled carrier stability with wideband FM output.

Interview Questions

Q1: Why does FM provide better noise performance than AM?

FM encodes information in frequency variations while maintaining constant amplitude. Most noise affects signal amplitude, which can be removed by a limiter circuit before the FM discriminator without losing information. Additionally, WBFM provides an SNR improvement proportional to β², meaning wider deviation gives exponentially better noise immunity — trading bandwidth for quality. AM cannot use limiters because amplitude carries the information.

Q2: Explain Carson's rule and when it fails.

Carson's rule approximates FM bandwidth as BT = 2(Δf + fm), containing about 98% of signal power. It works well for both narrowband (reduces to 2fm when β<<1) and wideband FM. It can underestimate bandwidth for very large β (>10) where more Bessel sidebands become significant, and it assumes single-tone modulation. For complex multi-tone signals, use the deviation ratio D = Δf/fm(max) in place of β.

Q3: What is the FM threshold effect and how can it be mitigated?

Below approximately 10 dB input SNR, FM performance collapses suddenly rather than degrading gradually like AM. This happens when noise spikes momentarily exceed the signal, causing 2π phase jumps that appear as impulsive "clicks" in the demodulated output. Mitigation techniques include: threshold extension demodulators (FMFB - FM with feedback), PLL demodulators that have lower threshold, and reducing bandwidth (lower β) at the cost of noise improvement above threshold.

Q4: What happens to the FM carrier component at β = 2.4?

At β = 2.4, the zero-order Bessel function J0(2.4) = 0, meaning the carrier component completely disappears from the spectrum. All signal power is transferred to the sidebands. This can be used experimentally to measure exact deviation: increase modulation until the carrier null is observed on a spectrum analyzer, confirming β = 2.4.