Loading learning content...
In the early days of radio, listeners accepted static and crackling as inevitable companions to their favorite broadcasts. Electrical storms, automobile ignition systems, and industrial equipment all contributed to a constant background of noise that degraded audio quality. AM radio seemed destined to forever sound like a conversation during a thunderstorm.
Then came FM. When Edwin Armstrong first demonstrated his frequency modulation system in 1935, witnesses were astonished not just by the improved fidelity, but by the silence—the absence of the crackling and hissing that plagued AM. Armstrong had discovered a fundamental principle: by encoding information in frequency rather than amplitude, communication systems could achieve remarkable immunity to the noise that nature and technology conspire to create.
This noise immunity isn't magic—it's mathematics. Understanding why and how FM resists noise illuminates fundamental principles of communication theory that apply far beyond FM radio.
By the end of this page, you will understand the mathematical basis for FM's noise immunity, quantify SNR improvement as a function of modulation index, analyze the FM threshold effect, apply pre-emphasis and de-emphasis techniques, and evaluate noise performance in FSK digital systems.
Before analyzing FM's noise immunity, we must understand the nature of noise itself and how it affects different types of signals.
Types of Noise
Thermal Noise (Johnson-Nyquist Noise):
Shot Noise:
Atmospheric and Man-Made Noise:
Noise Characterization
Noise is typically modeled as a zero-mean Gaussian random process with power spectral density N₀/2 watts/Hz (double-sided spectrum). The noise power in bandwidth B is:
P_n = N₀ × B (for double-sided N₀/2) or N₀ × B (for single-sided N₀)
How Noise Affects Modulated Signals
Noise adds to the received signal:
r(t) = s(t) + n(t)
Where:
For AM (amplitude carries information):
For FM (frequency carries information):
Key Insight: The fundamental reason for FM's noise immunity is that most noise sources produce amplitude variations, and FM encodes information in frequency. By using a limiter to remove amplitude variations, FM receivers reject most noise before it can corrupt the message.
FM receivers include a limiter stage that clips the signal to constant amplitude before frequency detection. This simple circuit is responsible for much of FM's noise advantage—it literally removes amplitude noise from the signal. AM receivers cannot use limiting because their information is in the amplitude.
Before diving into mathematics, let's develop intuition for why FM resists noise.
Phasor Analysis of FM Plus Noise
Consider an FM carrier plus narrowband noise. Using phasor representation:
When signal is strong (A_c >> A_n):
When noise is significant (A_n comparable to A_c):
The Capture Effect Perspective
FM's capture effect extends to noise:
Why FM Noise Spectrum Is Not Flat
A crucial insight: the noise at the output of an FM demodulator is not flat (white) but has a parabolic power spectrum that increases with frequency.
Mathematical basis:
Resulting noise power spectral density:
S_n(f) ∝ f² for |f| ≤ f_m
This means:
Implications:
The parabolic noise spectrum is unique to FM. It arises because frequency is the derivative of phase, and differentiation enhances high frequencies. This characteristic shapes pre-emphasis design and explains why FM audio without de-emphasis sounds excessively bright and hissy.
Let's derive the SNR improvement that FM provides over baseband transmission. This analysis reveals the fundamental bandwidth-noise tradeoff.
Signal Power at FM Detector Output
For single-tone modulation with frequency f_m and frequency deviation Δf:
The message signal power at detector output is: S_o = (A_c × Δf)² / 2 = A_c² × Δf² / 2
Noise Power at FM Detector Output
The noise power in the message bandwidth after FM detection:
N_o = N₀ × f_m³ / (3 × A_c²)
(This accounts for the parabolic noise spectrum.)
Output SNR for FM
SNR_FM = S_o / N_o = (3 × A_c⁴ × Δf²) / (2 × N₀ × f_m³)
Rewriting in terms of modulation index β = Δf/f_m:
SNR_FM = (3/2) × β² × (A_c² / N₀f_m)
The term (A_c² / 2N₀f_m) is the input carrier-to-noise ratio (CNR) in the baseband bandwidth:
SNR_FM = 3β² × CNR_baseband
FM Improvement Factor
Comparing FM output SNR to baseband transmission (AM with same power and bandwidth):
Improvement Factor = SNR_FM / SNR_baseband = 3β²
This is the FM advantage—the factor by which FM provides better noise performance than AM for the same carrier power.
In decibels: FM advantage = 10 log₁₀(3β²) dB
More Complete Expression (Accounting for Bandwidth Penalty):
If we consider that FM uses more bandwidth (B_FM = 2(β+1)f_m vs. B_AM = 2f_m), the comparison becomes:
Figure of Merit = SNR_FM × B_AM / (SNR_AM × B_FM)
= 3β² × 1 / (β + 1)
This peaks at β ≈ 1.53, showing there's an optimal modulation index when bandwidth is penalized.
Practical FM SNR Example
FM Broadcasting (β = 5):
This explains why FM broadcasting sounds dramatically cleaner than AM.
| β | SNR Improvement Factor | Improvement (dB) | Typical Application |
|---|---|---|---|
| 0.5 | 0.75 | -1.25 | Narrowband (no advantage) |
| 1.0 | 3 | 4.8 | Low-end broadcast |
| 2.0 | 12 | 10.8 | Quality voice |
| 3.0 | 27 | 14.3 | Good broadcast |
| 5.0 | 75 | 18.8 | FM broadcasting |
| 7.0 | 147 | 21.7 | High-fidelity |
| 10.0 | 300 | 24.8 | Wideband telemetry |
These SNR improvement formulas only apply above the FM threshold. Below threshold, performance degrades rapidly and the improvement disappears. The threshold CNR is typically 10-12 dB; operating with adequate margin above threshold is critical.
The FM threshold effect is a critical phenomenon that distinguishes FM from linear modulation schemes. Understanding threshold behavior is essential for FM system design.
What Is the Threshold?
The FM threshold is the input carrier-to-noise ratio (CNR) below which the output SNR degrades much more rapidly than the input CNR decreases. Above threshold, output SNR tracks input CNR roughly linearly. Below threshold, output SNR collapses.
Physical Mechanism
The threshold arises from the phasor behavior described earlier:
Above threshold (strong signal):
At threshold:
Below threshold:
Threshold Carrier-to-Noise Ratio
The threshold CNR is typically defined as the point where output SNR falls 1 dB below the linear extrapolation. This occurs at approximately:
CNR_threshold ≈ 10 dB (for conventional FM discriminator)
More precise values depend on:
Effect of Modulation Index
Higher modulation index means:
Approximate relationship: CNR_threshold ≈ 10 + 10 log₁₀(β + 1) dB (for wideband FM)
For FM broadcasting (β = 5): CNR_threshold ≈ 10 + 10 log₁₀(6) = 10 + 7.8 ≈ 18 dB
Threshold Extension Techniques
Several techniques reduce the effective threshold:
FM Feedback (FMFB):
Phase-Locked Loop (PLL) Detection:
Threshold Extension Demodulator (TED):
| System Type | Typical β | Threshold CNR | Design Margin | Operating CNR |
|---|---|---|---|---|
| FM Broadcast | 5 | 18 dB | 5-10 dB | 23-28 dB |
| Narrowband FM Voice | 1.67 | 13 dB | 5 dB | 18 dB |
| Satellite TV (analog) | 7 | 20 dB | 5 dB | 25 dB |
| Telemetry | 5-10 | 18-23 dB | 3-5 dB | 21-28 dB |
| Mobile Radio | 1.67 | 13 dB | 10 dB | 23 dB |
FM system design must ensure adequate margin above threshold even in worst-case conditions (maximum path loss, rain fade, multipath, etc.). A system that occasionally drops below threshold will experience catastrophic quality degradation—not the graceful degradation of AM. Design margins of 5-10 dB above threshold are typical.
The parabolic noise spectrum of FM demodulator output means high-frequency noise is stronger than low-frequency noise. Pre-emphasis and de-emphasis are complementary techniques that exploit this characteristic to improve overall SNR.
The Problem: Non-Uniform Noise Spectrum
Recall that FM demodulator noise power spectral density: S_n(f) ∝ f²
At f_m = 15 kHz, noise power density is (15000/1000)² = 225 times higher than at 1 kHz!
Meanwhile, Audio Signal Spectrum:
Solution: Pre-emphasis and De-emphasis
Pre-emphasis (at transmitter):
De-emphasis (at receiver):
Pre-emphasis Transfer Function
The standard pre-emphasis filter is a simple first-order high-pass characteristic:
H_pre(f) = 1 + j(f/f₁)
Where f₁ is the corner frequency determined by the time constant τ:
f₁ = 1/(2πτ)
Standard Time Constants:
De-emphasis Transfer Function
The complementary de-emphasis filter:
H_de(f) = 1 / (1 + j(f/f₁))
This is a first-order low-pass filter with the same corner frequency.
Net Response: H_pre(f) × H_de(f) = 1 (flat overall response)
SNR Improvement from Pre/De-emphasis
The improvement depends on the audio bandwidth and time constant:
For 75 μs (US) with f_m = 15 kHz:
For 50 μs (Europe) with f_m = 15 kHz:
| Region/System | Time Constant (τ) | Corner Frequency | Typical Improvement |
|---|---|---|---|
| US FM Broadcasting | 75 μs | 2,122 Hz | ~13 dB at 15 kHz |
| Europe FM Broadcasting | 50 μs | 3,183 Hz | ~11 dB at 15 kHz |
| Land Mobile Radio | 300 μs | 531 Hz | ~20 dB at 3 kHz |
| Analog TV Audio | 75 μs | 2,122 Hz | ~13 dB at 15 kHz |
| Satellite Audio | Variable | Variable | Optimized for link |
The US adopted 75 μs early in FM broadcasting history when transmitter technology limited high-frequency deviation capability. Europe later adopted 50 μs for slightly better noise rejection, enabled by improved transmitter technology. Both are valid engineering tradeoffs between noise improvement and deviation limits.
The capture effect is a unique FM phenomenon where a stronger signal completely suppresses a weaker interfering signal on the same frequency. This provides additional noise immunity beyond what the basic SNR analysis predicts.
Understanding Capture
When two FM signals are present:
r(t) = A₁cos(θ₁(t)) + A₂cos(θ₂(t))
Where θ₁ and θ₂ are the instantaneous phases of the desired and interfering signals.
If A₁ >> A₂ (desired signal stronger):
The Capture Ratio
The capture ratio is the minimum dB difference required for the stronger signal to completely capture the receiver:
With 2 dB capture ratio, a signal just 2 dB stronger than an interferer provides essentially interference-free reception.
Mathematical Suppression
The suppression of the weaker signal follows approximately:
Suppression ≈ 20 × log₁₀(A₁/A₂) dB for A₁ > A₂
A 10 dB stronger desired signal provides approximately 10 dB suppression of the interferer.
Multipath and the Capture Effect
Multipath propagation (signals arriving via multiple reflected paths) is a major challenge in mobile communications. FM's capture effect provides natural multipath immunity:
Scenario: Direct signal + reflected signals (delayed, weaker)
Result:
Problem case: When direct and reflected signals are nearly equal (urban mobile environments), capture is unstable, causing "picket fencing" (rapid flutter as vehicle moves).
Comparison with AM Multipath
In AM:
In FM:
FM car radio listeners experience capture effects constantly. Driving through a city, reception may be excellent (local station captures), then suddenly switch to a distant station (as local signal weakens), then return. This is capture in action—it's why FM reception seems to be either good or not present, with little middle ground.
Moving from analog FM to digital FSK, the relevant metric shifts from continuous SNR to Bit Error Rate (BER) as a function of signal-to-noise ratio. FSK inherits FM's noise immunity advantages but in a digital context.
Signal-to-Noise Metrics for Digital Systems
E_b/N₀ (Energy per bit to noise power spectral density ratio):
Alternative expressions:
BFSK Bit Error Probability Recap
Coherent detection (optimal): P_b = Q(√(2E_b/N₀)) = ½erfc(√(E_b/N₀))
Non-coherent detection (simpler, slightly worse): P_b = ½exp(-E_b/2N₀)
For BER = 10⁻⁵:
| Target BER | Coherent BFSK | Non-Coherent BFSK | BPSK (reference) |
|---|---|---|---|
| 10⁻² | 5.9 dB | 6.7 dB | 5.9 dB |
| 10⁻³ | 8.4 dB | 9.1 dB | 8.4 dB |
| 10⁻⁴ | 10.6 dB | 11.4 dB | 10.6 dB |
| 10⁻⁵ | 12.6 dB | 13.5 dB | 12.6 dB |
| 10⁻⁶ | 14.2 dB | 15.2 dB | 14.2 dB |
Why FSK Noise Performance Matches BPSK
Coherent orthogonal BFSK and BPSK have identical BER performance for the same E_b/N₀. This is because:
However, FSK uses twice the bandwidth of PSK for the same data rate, so its spectral efficiency is half.
M-ary FSK Noise Performance
Increasing M in MFSK improves noise performance:
For orthogonal M-FSK with coherent detection:
The limiting behavior as M → ∞:
This is remarkable: MFSK can achieve arbitrarily low error rate with E_b/N₀ just above -1.6 dB, given sufficient bandwidth. This makes large-M FSK attractive for power-limited channels (e.g., deep space).
FSK in Fading Channels
Wireless channels experience fading—time-varying signal strength—due to multipath and mobility.
Rayleigh Fading (no line-of-sight):
FSK advantages in fading:
Average BER for non-coherent BFSK in Rayleigh fading: P_b ≈ 1/(2 + γ̄) where γ̄ is average E_b/N₀
This is much worse than AWGN (additive white Gaussian noise) performance, but FSK's robustness makes it a practical choice for such environments.
FSK's combination of constant envelope (for power amplifier efficiency) and good non-coherent detection (for simplicity and robustness) makes it popular for low-cost, low-power, and mobile applications. Bluetooth, Zigbee, and many IoT protocols use FSK variants despite the bandwidth penalty, because the practical advantages outweigh the spectral inefficiency.
We have thoroughly analyzed FM and FSK noise immunity—one of the most important advantages of frequency modulation systems. Let's consolidate the essential knowledge:
What's Next: Applications
With comprehensive understanding of FM/FSK theory—from basic concepts through bandwidth and noise performance—we're ready to explore real-world applications. The next page examines FM broadcasting, mobile communications, FSK in data systems, telemetry, and emerging applications, connecting all the theory to practical systems you encounter every day.
You now understand the theoretical basis for FM and FSK noise immunity—why these technologies resist noise, how to quantify the advantage, where the limits lie (threshold), and how techniques like pre-emphasis extend the benefits. This knowledge is essential for designing reliable communication systems.