Signal and Noise Analysis
Learning Objectives
- Distinguish signal from noise and explain why noise is unavoidable in real circuits.
- Identify the main sources of electronic noise: thermal, shot, flicker, and EMI.
- Calculate signal-to-noise ratio (SNR) in decibels from signal and noise power.
- Explain practical noise-reduction techniques: shielding, twisted pair, differential signaling, and filtering.
- Connect noise analysis to real design decisions, like choosing amplifier gain stages or cable types.
Quick Answer
Every electrical signal that travels through a real circuit picks up unwanted random variations called noise — from the thermal jitter of electrons, external electromagnetic interference, or imperfections in components. Signal and noise analysis is the discipline of quantifying how much noise contaminates a signal (usually as a signal-to-noise ratio, SNR) and choosing design techniques — shielding, filtering, differential signaling — to minimize its impact. This matters because noise sets the practical floor on how small a signal you can reliably detect: in a hearing aid, an SNR that's too low means speech gets buried in hiss; in a data link, low SNR means bit errors; in a sensor circuit, it means false readings.
Signal vs. Noise
Signal: the desired, information-carrying waveform. Noise: any unwanted disturbance superimposed on the signal that doesn't carry the intended information.
Major Noise Sources
| Noise type | Cause | Key characteristic |
|---|---|---|
| Thermal (Johnson) noise | Random thermal motion of charge carriers in any resistor | Present in every resistive component; increases with temperature; flat across frequency ("white") |
| Shot noise | Discreteness of charge carriers crossing a junction | Significant in diodes and transistors, proportional to average current |
| Flicker (1/f) noise | Imperfections/traps in semiconductor materials | Dominant at low frequencies, decreases as frequency increases |
| Electromagnetic interference (EMI) | External fields from motors, switching supplies, radio transmitters | Can be coupled in via wiring acting as an antenna; often narrowband or impulsive |
Thermal noise power formula: P_noise = k × T × B, where k is Boltzmann's constant (1.38×10⁻²³ J/K), T is absolute temperature (Kelvin), and B is bandwidth (Hz). This tells you that wider bandwidth or higher temperature always lets in more thermal noise — a key reason receivers use bandpass filters to admit only the bandwidth actually needed.
Signal-to-Noise Ratio (SNR)
SNR measures how much stronger the signal is compared to the noise floor:
SNR (dB) = 10 × log₁₀(P_signal / P_noise)
Worked Example 1: A sensor outputs a signal with power 10mW, and the measured noise power is 1mW. Find the SNR in dB.
SNR (dB) = 10 × log10(10/1) = 10 × log10(10) = 10 × 1 = 10 dB
An SNR of 10dB means the signal power is 10 times the noise power — a fairly modest SNR that would sound noticeably hissy in audio or cause noticeable errors in a data link.
Worked Example 2: A weak radio signal has power 2µW, and the noise power at the receiver is 50nW. Find the SNR in dB, and compare it to Example 1.
SNR (dB) = 10 × log10(2×10⁻⁶ / 50×10⁻⁹) = 10 × log10(40) = 10 × 1.602 ≈ 16.0 dB
This SNR (16dB) is significantly better than Example 1's 10dB — every extra 10dB roughly corresponds to a 10× improvement in the signal-to-noise power ratio, which is why engineers often talk in terms of "dB improvement" rather than raw ratios.
Worked Example 3: An SNR of 20dB is required for reliable communication. If the noise power is fixed at 1µW, what minimum signal power is needed?
20 = 10 × log10(P_signal / 1×10⁻⁶)
2 = log10(P_signal / 1×10⁻⁶)
P_signal / 1×10⁻⁶ = 10² = 100
P_signal = 100 × 1×10⁻⁶ = 1×10⁻⁴ W = 100 µW
Signal Conditioning
Signal conditioning prepares a raw signal for further processing or transmission:
- Filtering — removing noise components outside the signal's frequency range (see the Filters page).
- Amplification — boosting a weak signal above the noise floor of subsequent stages, ideally as early as possible in the signal chain (the first amplifier stage's own noise contributes most to overall SNR degradation).
- Analog-to-digital conversion (ADC) — converting to digital form for robust processing, though quantization itself introduces a small, unavoidable amount of noise.
Noise Reduction Techniques
| Technique | How it works | Typical use |
|---|---|---|
| Shielding | Conductive enclosure blocks external electric fields (Faraday cage effect) | Sensitive analog front-ends, audio cables |
| Twisted pair wiring | Two wires twisted together cause induced EMI to cancel between them | Ethernet cables, balanced audio lines |
| Differential signaling | Signal sent as a voltage difference between two lines; common-mode noise picked up equally on both lines cancels at the receiver | USB, RS-485, high-speed digital buses |
| Filtering | Removes noise outside the signal's actual frequency band | Nearly every analog front-end |
| Grounding practices | Proper single-point grounding avoids ground loops that inject noise | Audio and instrumentation systems |
Key Terms
| Term | Definition |
|---|---|
| Signal | The desired, information-carrying waveform in a circuit |
| Noise | Unwanted random or interfering disturbance superimposed on a signal |
| SNR (Signal-to-Noise Ratio) | The ratio of signal power to noise power, usually expressed in dB |
| Thermal (Johnson) noise | Noise from random thermal motion of charge carriers in a resistor |
| Shot noise | Noise from the discrete, quantized nature of charge carrier flow |
| Flicker (1/f) noise | Low-frequency noise from semiconductor imperfections, decreasing with frequency |
| Differential signaling | Transmitting a signal as the difference between two wires to reject common-mode noise |
Common Mistakes
-
Misconception: "Noise can be completely eliminated with a good enough design." Why it's wrong: Thermal noise arises from fundamental physics (random thermal motion of charge carriers) and exists in every resistive element at any temperature above absolute zero — it cannot be engineered away entirely, only reduced or managed. Correct: Design goals should focus on maximizing SNR (boosting signal, minimizing added noise, restricting bandwidth to what's needed) rather than expecting to eliminate noise completely.
-
Misconception: "A higher SNR in dB is a small improvement, similar in scale to a percentage." Why it's wrong: The dB scale is logarithmic — a 10dB increase corresponds to a 10× increase in the underlying power ratio, and 20dB corresponds to 100×, not merely "10% more" or "20% more." Correct: Always convert dB back to the linear ratio (via P_ratio = 10^(dB/10)) if you need to reason about relative power in a genuinely proportional sense.
-
Misconception: "Amplifying a signal always improves its SNR." Why it's wrong: An amplifier boosts both the signal and any noise already present in equal proportion, and adds its own internal noise on top — so amplification alone, especially late in a signal chain, does not improve SNR and can make it worse. Correct: SNR is best preserved by amplifying as early as possible in the signal chain (before noise accumulates from later stages) and using low-noise amplifier designs; simply cranking up gain later does not recover lost SNR.
Comparison and Connections
| Noise type | Frequency behavior | Depends most on |
|---|---|---|
| Thermal noise | Flat (white) across frequency | Temperature and bandwidth |
| Shot noise | Flat (white) across frequency | Average current level |
| Flicker (1/f) noise | Strongest at low frequency, falls off with frequency | Semiconductor material quality |
| EMI | Often concentrated at specific frequencies (e.g., mains hum) | External sources, shielding/wiring quality |
Practice Questions
Recall 1: Write the formula for SNR in decibels. Answer guidance: SNR(dB) = 10 × log10(P_signal/P_noise).
Recall 2: Name the four major noise sources covered and one distinguishing feature of each. Answer guidance: Thermal noise (temperature-dependent, white), shot noise (current-dependent, white), flicker noise (1/f, low-frequency dominant), EMI (external interference, often narrowband).
Understanding 1: Why does restricting bandwidth reduce thermal noise power? Answer guidance: Thermal noise power is P = kTB, directly proportional to bandwidth B; a narrower bandwidth (e.g., via filtering) admits less total thermal noise power even though the signal itself may be preserved if it fits within that narrower band.
Understanding 2: Explain why amplifying a signal late in a processing chain doesn't recover SNR lost earlier. Answer guidance: Once noise has been added to a signal (e.g., by a noisy early amplifier stage or a long noisy cable), any later amplification boosts both the signal and the already-present noise by the same factor, leaving the SNR unchanged, and any additional noise from the later amplifier only makes SNR worse.
Application 1: A device outputs a signal with power 50mW and noise power of 5mW. Find the SNR in dB. Answer guidance: SNR = 10×log10(50/5) = 10×log10(10) = 10 dB.
Application 2: An SNR of 30dB is measured with a noise power of 2µW. Find the signal power. Answer guidance: 30 = 10×log10(P_signal/2×10⁻⁶) → 3 = log10(P_signal/2×10⁻⁶) → P_signal/2×10⁻⁶ = 1000 → P_signal = 2mW.
Analysis 1: Two cable designs are compared for a long sensor run: a single unshielded wire versus a shielded twisted pair with differential signaling. Explain which noise sources each design is more vulnerable to, and why the second option is generally preferred for long cable runs. Answer guidance: The unshielded single wire is vulnerable to EMI pickup along its entire length (acting like an antenna) with no cancellation mechanism. The shielded twisted pair with differential signaling blocks external fields via the shield (Faraday cage) and cancels any residual common-mode noise picked up equally on both twisted conductors at the differential receiver — making it far more robust over long distances where EMI exposure accumulates.
Analysis 2: A system needs an SNR of at least 20dB to function reliably. Current SNR is 10dB. Determine by what factor the signal power must increase (assuming noise power stays constant) to meet the requirement, and comment on whether this is a large or small change. Answer guidance: Going from 10dB to 20dB requires the P_signal/P_noise ratio to go from 10 to 100 (since 10log10(ratio)=dB), a 10× increase in signal power — a substantial change, illustrating how deceptively large a "just 10dB more" requirement can be in absolute terms.
FAQ
Q1: Can noise ever be useful? Occasionally — techniques like "dithering" intentionally add a small amount of noise before quantization (e.g., in audio ADCs) to reduce a more objectionable form of distortion, trading a barely audible hiss for smoother, more accurate signal representation.
Q2: Why does cooling sensitive equipment (like some scientific detectors) reduce noise? Because thermal noise power is directly proportional to absolute temperature (P = kTB); lowering T directly lowers the thermal noise floor, which matters most for extremely sensitive, low-signal-level measurements.
Q3: What's the difference between noise and distortion? Noise is an unwanted, typically random addition to the signal; distortion is an unwanted change to the shape of the signal itself, often caused by nonlinearity in a component (like clipping in an overdriven amplifier). They're both signal-quality problems but arise from different mechanisms.
Q4: Why do balanced (differential) audio cables reject noise better than unbalanced ones? A balanced cable carries the signal as a voltage difference between two conductors; any noise picked up along the cable's length tends to affect both conductors nearly equally (common-mode), and the receiver subtracts one from the other, canceling that common noise while preserving the differential signal.
Q5: Is a higher SNR always necessary, or can too much SNR be "wasted"? Beyond the SNR actually needed for reliable detection or acceptable perceived quality, additional SNR typically has diminishing practical value — for example, human hearing and most communication standards define a "good enough" SNR threshold, past which further improvement isn't perceptible or doesn't reduce error rates meaningfully.
Quick Revision
- Signal = wanted information-carrying waveform; noise = unwanted disturbance added to it.
- SNR (dB) = 10 × log10(P_signal / P_noise).
- Thermal (Johnson) noise: P = kTB; from random carrier motion; present in all resistive elements.
- Shot noise: from discrete charge carrier flow, significant in diodes/transistors.
- Flicker (1/f) noise: dominant at low frequencies, from material imperfections.
- EMI: external interference coupled in via wiring; reduced by shielding and good wiring practice.
- Restricting bandwidth (via filtering) directly reduces admitted thermal noise power.
- Amplifying late in a signal chain does not recover SNR already lost earlier.
- Noise reduction techniques: shielding, twisted pair, differential signaling, filtering, proper grounding.
- Each 10dB of SNR corresponds to a 10× change in the underlying signal/noise power ratio (logarithmic, not linear).
Related Topics
Prerequisites: AC and DC Circuits, Frequency Response, Filters.
Related Topics: Filters, Frequency Response, Power Calculations.
Next Topics: Power Calculations.