1 General concepts
Noise is a broad term for unwanted variation, interference, or disturbance in a system. In everyday language it often refers to sound that is loud or unpleasant, but in science and engineering it can mean any non-useful component that obscures a desired message, measurement, or process. The concept is therefore used across disciplines with meanings that are related yet context-dependent.
Noise is not always purely random. It may arise from physical processes, human activity, instrumentation limits, or statistical variability in data. In some settings it is treated as a nuisance to be minimized, while in others it serves as a useful source of uncertainty for analysis, simulation, or design.
1.1 Definitions and scope
In a general sense, noise is any deviation from a desired or ideal form of a signal. In acoustics, it usually means sound that is unwanted or disruptive. In electronics and physics, it often refers to random fluctuations in voltage, current, or other measurable quantities. In data analysis, noise describes variability that does not reflect the underlying pattern of interest.
The scope of the term is unusually wide because it is defined by function rather than by a single physical property. A tone can be considered noise if it interrupts communication, and a random fluctuation can be treated as meaningful if it reveals information about a system. This flexibility makes the concept useful across both natural and engineered environments.
1.2 Sources of noise
Noise can originate from many sources, including natural processes, human activity, and imperfections in instruments or transmission systems. The same observed disturbance may have several contributing causes. In practice, analysts often separate sources by origin, frequency content, or how they affect the target system.
1.2.1 Natural sources
Natural noise includes sounds and fluctuations produced by wind, rain, flowing water, lightning, seismic activity, and biological organisms. In electronics and measurement, thermal motion of particles is a natural source of random variation. Astronomical observations also encounter background radiation and other environmental effects that introduce uncertainty.
1.2.2 Human-made sources
Human-made noise comes from machinery, vehicles, construction, industrial processes, electrical equipment, and communication systems. In urban settings, these sources are often persistent and layered, creating a complex soundscape. In technical systems, human-made noise may also result from device design, circuit layout, or transmission interference.
1.3 Noise versus signal
A signal is the component of a system that carries useful information, while noise is the portion that complicates detection or interpretation. The distinction depends on purpose: the same fluctuation may be a signal in one context and noise in another. For example, background chatter is noise for a telephone call but useful signal for a crowd-level survey.
In many applications, the main task is to separate signal from noise. This separation is rarely perfect, so engineers and scientists often work with probabilistic models, thresholds, or filtering methods. The balance between preserving useful information and suppressing disturbance is central to measurement and communication.
1.4 Types of noise in applied science
Applied science recognizes several recurring categories of noise. Some are random and continuous, such as thermal noise in circuits. Others are discrete or abrupt, such as impulse noise. Still others reflect correlated effects, including flicker noise, crosstalk, and interference between channels.
Noise may also be classified by its statistical properties, spectral distribution, or impact on performance. White noise has approximately equal power across a range of frequencies, while colored noise emphasizes certain bands. In imaging and sensing, noise may appear as grain, speckle, banding, or fixed-pattern artifacts.
2 Acoustic noise
Acoustic noise is unwanted sound that affects hearing, comfort, communication, or performance. It can be steady or intermittent, low-pitched or high-pitched, and localized or widespread. Because sound travels through air and other media, acoustic noise often spreads beyond its source and can influence large areas.
Its importance extends beyond annoyance. Prolonged or intense sound exposure may interfere with sleep, concentration, speech, and hearing. As a result, acoustic noise is measured, regulated, and managed in homes, workplaces, transportation systems, and public environments.
2.1 Sound and perceived loudness
Sound is a pressure wave that propagates through a medium such as air. Perceived loudness, however, is shaped not only by sound pressure but also by frequency, duration, and the sensitivity of human hearing. A sound with modest physical energy may seem loud if it falls in a frequency range to which the ear is highly responsive.
The relationship between physical measurement and human experience is not linear. This is why acoustic assessment often combines objective metrics with psychoacoustic considerations. Two sounds with similar decibel values may be perceived quite differently depending on their spectral content.
2.2 Environmental noise
Environmental noise refers to sound in the surrounding environment that affects daily life. It includes continuous background sound, intermittent bursts, and complex mixtures from multiple sources. In built environments, it is often associated with transportation, industry, and dense human activity.
Because environmental noise is shared by many people, it is frequently studied at the level of neighborhoods, cities, and regions. Its management may involve planning, zoning, equipment design, and sound insulation.
2.2.1 Transportation noise
Transportation noise comes from road traffic, rail systems, aircraft, and related infrastructure. It is often characterized by changing intensity and repeated events such as passing vehicles or takeoffs. High-speed movement, braking, engine operation, and tire interaction all contribute to the overall sound.
2.2.2 Industrial noise
Industrial noise is produced by factories, processing plants, construction equipment, and heavy machinery. It may be continuous, tonal, or impulsive depending on the equipment involved. Enclosures, machinery maintenance, and layout design are commonly used to reduce its spread.
2.2.3 Community noise
Community noise includes sounds from neighbors, public venues, outdoor activities, and mixed urban sources. It varies by time of day and local land use. In residential areas, community noise is often evaluated for its effects on sleep, relaxation, and general quality of life.
2.3 Workplace noise
Workplace noise is any sound exposure in occupational settings that may affect safety or productivity. It is common in manufacturing, construction, mining, aviation, and other high-activity environments. Repeated exposure can interfere with verbal communication and increase the risk of long-term hearing damage.
Workplace control typically combines engineering measures, administrative practices, and personal protection. Employers may monitor levels, post warnings, and provide protective equipment where needed.
2.3.1 Hearing risk
Extended exposure to high sound levels can damage the sensitive structures of the inner ear. Hearing risk increases with both intensity and duration, and impulse sounds may be especially harmful. Effects may include temporary threshold shifts, tinnitus, or permanent hearing loss.
2.3.2 Exposure limits
Exposure limits are guidelines that specify maximum sound levels or time-weighted doses considered acceptable for workers. These limits vary by jurisdiction and standard-setting body. They are intended to reduce the likelihood of hearing injury and support consistent safety practices.
2.4 Noise measurement
Noise measurement translates subjective disturbance into quantifiable values. In acoustics, this usually involves sound pressure level, frequency content, and exposure duration. Accurate measurement is important for compliance, research, and engineering control.
Different instruments and metrics are used depending on whether the goal is to assess peak events, average exposure, or human annoyance. Reliable measurements also depend on calibration and appropriate sampling methods.
2.4.1 Decibel scale
The decibel scale is a logarithmic way of expressing sound level. Because it compresses a wide range of values, it is convenient for comparing quiet and loud sounds. A small numerical change in decibels can represent a substantial change in physical intensity.
2.4.2 Frequency weighting
Frequency weighting adjusts measurements to reflect human hearing sensitivity. The most common weighting curves emphasize frequencies to which the ear is more responsive and reduce the influence of very low or very high frequencies. This makes readings more relevant to perceived loudness and annoyance.
2.4.3 Sound level meters
Sound level meters are instruments used to measure acoustic levels in a standardized way. They typically include a microphone, signal-processing electronics, and display functions. More advanced devices can record time histories, spectra, and event-based data.
2.5 Noise control
Noise control aims to reduce unwanted sound at its source, along its transmission path, or at the receiver. The choice of method depends on the type of noise, the environment, and practical constraints. Effective control often combines several approaches rather than relying on a single solution.
2.5.1 Source control
Source control reduces noise where it is generated. Examples include quieter machinery, improved maintenance, vibration isolation, and lower-speed operation. Because it addresses the origin of the problem, source control is often the most effective long-term strategy.
2.5.2 Path control
Path control limits how sound travels from source to listener. Barriers, enclosures, acoustic panels, and sound-absorbing materials are common methods. Architectural design can also reduce propagation through room shape, surface treatment, and separation distance.
2.5.3 Receiver protection
Receiver protection focuses on the person exposed to noise. Earplugs, earmuffs, and other protective devices reduce sound reaching the ear. In some settings, scheduling, rest periods, and relocation are used alongside personal protection.
3 Electronic and signal noise
In electronics and signal processing, noise refers to unwanted variations that obscure or distort the intended electrical or informational content. These variations may arise from physical processes inside components, external interference, or limitations in circuit design. Their effects can appear in analog systems, digital systems, and measurement devices.
Noise is a central concern because it influences fidelity, sensitivity, and reliability. Engineers study its sources, statistical behavior, and frequency characteristics in order to improve circuit performance and reduce error.
3.1 Random fluctuations in circuits
Electrical circuits exhibit random fluctuations due to material properties, charge motion, and device operation. Even in well-designed systems, small variations in voltage or current may be present. These fluctuations can accumulate or become visible when signals are weak.
The practical impact depends on the relative size of the noise compared with the desired signal. In low-level sensing or high-gain amplification, minor disturbances can become significant. For that reason, circuit design often balances performance against noise generation.
3.2 Thermal noise
Thermal noise, also called Johnson-Nyquist noise, arises from the random motion of charge carriers in conductors. It exists in resistive components and increases with temperature and bandwidth. Because it is fundamentally linked to temperature, it cannot be eliminated entirely.
Thermal noise is important in sensitive electronics, radio receivers, and precision measurements. It provides a lower bound on achievable noise performance in many systems. Designers often reduce its effect by limiting bandwidth or lowering component temperature where feasible.
3.3 Shot noise
Shot noise results from the discrete nature of charge and the randomness of particle flow. It is common in diodes, transistors, and photodetectors, especially when current levels are small. The effect becomes more pronounced when individual electrons or photons are counted over time.
This form of noise is especially relevant in optical and semiconductor systems. It places a practical limit on measurement precision in applications such as photon detection, weak signal amplification, and low-light imaging.
3.4 Flicker noise
Flicker noise, often called 1 over f noise, is characterized by greater intensity at lower frequencies. It appears in many electronic devices and materials and is associated with slow fluctuations in charge transport or structural defects. Unlike purely white noise, its spectrum is not flat.
Its influence is often noticeable in precision circuits, sensors, and instrumentation operating at low frequencies. Because it can dominate over long observation periods, it is especially relevant in stable reference systems and slow measurement processes.
3.5 Impulse noise
Impulse noise consists of short, abrupt bursts that may be much larger than the surrounding background. It can be caused by switching events, electrical discharge, mechanical contacts, or external disturbances. Its irregular nature makes it difficult to model and suppress.
In communication and recording systems, impulse noise may create clicks, pops, or sudden data errors. Even when brief, it can be highly disruptive because it may saturate components or produce noticeable artifacts.
3.6 Crosstalk and interference
Crosstalk occurs when a signal in one channel unintentionally affects another channel. Interference is a broader term for unwanted energy from internal or external sources that alters a signal. Both are important in densely packed circuits, cables, and communication networks.
These effects can arise through capacitive coupling, inductive coupling, shared grounding, or electromagnetic radiation. Careful layout, isolation, and shielding are commonly used to reduce them.
3.7 Signal-to-noise ratio
Signal-to-noise ratio is a measure of the strength of a desired signal relative to background noise. A higher ratio generally indicates clearer transmission, better detectability, and more reliable measurement. The metric is used widely in electronics, imaging, acoustics, and communications.
Because it summarizes performance in a single value, signal-to-noise ratio is often used to compare systems or evaluate improvements. However, interpretation depends on how signal and noise are defined in a particular context.
3.8 Noise reduction techniques
Noise reduction in electronic systems combines physical design, circuit techniques, and signal processing. The aim is to prevent noise from entering the system, reduce its coupling, or remove it after capture. The best approach depends on the frequency range, source type, and required fidelity.
3.8.1 Shielding
Shielding uses conductive or absorbent barriers to block electromagnetic interference. It is common in cables, enclosures, and sensitive measurement equipment. Effective shielding depends on continuity, grounding practice, and the frequencies involved.
3.8.2 Grounding
Grounding provides a reference potential and a path for unwanted currents. Proper grounding can reduce hum, instability, and interference in circuits. Poor grounding, by contrast, may introduce loops or create additional noise problems.
3.8.3 Filtering
Filtering removes unwanted frequency components from a signal. Analog and digital filters are used to suppress known interference, limit bandwidth, or smooth measurements. A filter improves clarity, but it may also alter or delay the desired signal.
3.8.4 Averaging
Averaging reduces random variation by combining repeated measurements or samples. It is especially effective when noise fluctuates independently from one observation to the next. The method is simple and widely used, although it is less effective against correlated or systematic disturbances.
4 Noise in communications
In communications, noise affects the transmission, reception, and interpretation of messages. It may enter at the source, along the channel, or in the receiving device. Whether the medium is a wire, radio link, optical fiber, or acoustic channel, noise limits how accurately information can be delivered.
Communications theory treats noise as a fundamental factor in system design. Modulation, coding, and channel selection are often chosen with the noise environment in mind.
4.1 Noise in analog systems
Analog systems represent information with continuously varying signals, so noise can directly alter amplitude, phase, or frequency. Small disturbances may be audible or visible as hiss, hum, distortion, or instability. Because analog information is not discretized, noise can accumulate gradually across stages.
4.2 Noise in digital systems
Digital systems encode information into discrete symbols, which provides resilience against some forms of noise. However, noise can still cause bit errors, timing issues, or symbol misclassification. Once disturbances exceed decision thresholds, the data stream may become unreliable.
Digital transmission often uses error detection and correction to counter these effects. Even so, strong noise can overwhelm a channel and reduce throughput or force retransmission.
4.3 Channel capacity
Channel capacity is the maximum rate at which information can be transmitted with acceptable reliability over a noisy channel. It depends on bandwidth, signal power, and noise level. This concept helps establish theoretical limits for communication systems.
Higher noise generally reduces capacity unless compensated by more bandwidth or stronger signals. Capacity therefore serves as a key design target in telecommunications and networking.
4.4 Error rates and distortion
Error rates describe how often transmitted symbols or bits are received incorrectly. Distortion refers to unwanted changes in waveform shape, timing, or spectral content. Noise can contribute to both, although distortion may also arise from nonlinear components or channel characteristics.
Engineers evaluate error rates to judge system quality and reliability. Low error rates are usually desirable, but the acceptable level depends on the application, such as voice, video, data, or control signals.
4.5 Noise mitigation in transmission
Mitigating noise in transmission requires a combination of channel design, encoding, and signal structure. The goal is to preserve information despite interference, attenuation, and random variation. Different strategies are chosen according to distance, bandwidth, and system complexity.
4.5.1 Modulation methods
Modulation methods place information onto a carrier in ways that can improve robustness or spectral efficiency. Some schemes are better suited to noisy channels than others. The choice of modulation affects sensitivity to amplitude changes, phase shifts, and frequency interference.
4.5.2 Coding and redundancy
Coding and redundancy add extra structure so that errors can be detected or corrected. This increases reliability at the cost of overhead. In practice, redundancy is widely used in digital communications, storage, and packet-based networks.
4.5.3 Spread spectrum
Spread spectrum distributes a signal across a wider band than strictly necessary. This can make transmissions less vulnerable to narrowband interference and improve resistance to certain types of noise. It is used in systems that require reliability, coexistence, or confidentiality.
5 Noise in imaging and sensing
In imaging and sensing, noise appears as random or structured variation that reduces accuracy, clarity, or detectability. It may originate in the sensor, the environment, the optical path, or the processing chain. Because images and measurements are often interpreted visually or quantitatively, even subtle noise can matter.
Noise reduction is a major task in camera systems, medical scanners, telescopes, and remote sensing instruments. Analysts aim to remove unwanted variation while preserving edges, fine detail, and important features.
5.1 Sensor noise
Sensor noise is produced by the detector or measurement device itself. It can reflect thermal effects, readout instability, manufacturing variation, or photon statistics. In low-light or low-signal situations, sensor noise may dominate the recorded output.
This type of noise sets a practical limit on sensitivity. High-quality sensors are designed to minimize readout disturbance and preserve the integrity of weak signals.
5.2 Image noise
Image noise is visible or measurable corruption in a digital or analog image. It may appear as random grain, uneven texture, color speckling, or repeating artifacts. Its presence can interfere with object detection, visual interpretation, and automated analysis.
The amount and appearance of image noise depend on acquisition conditions, sensor properties, compression, and post-processing.
5.2.1 Grain and speckle
Grain and speckle are textured forms of image noise. Grain is often associated with film or low-light digital imaging, while speckle is common in coherent imaging systems such as ultrasound and radar. Both can reduce apparent smoothness and obscure small details.
5.2.2 Banding and fixed-pattern noise
Banding and fixed-pattern noise produce structured lines, repeating stripes, or persistent pixel-to-pixel differences. These artifacts often arise from sensor readout or calibration differences. Unlike purely random noise, they may remain visible across multiple images unless corrected.
5.3 Noise in medical imaging
Medical imaging uses many methods, including X-ray, computed tomography, magnetic resonance, ultrasound, and nuclear techniques. Noise affects diagnostic clarity, lesion detectability, and quantitative analysis. Because patient exposure and scan time are often limited, complete elimination of noise is usually not possible.
Clinical systems therefore use reconstruction methods and acquisition strategies that balance image quality against dose, speed, and comfort. The acceptable level of noise depends on the diagnostic task.
5.4 Noise in remote sensing
Remote sensing systems collect data from airborne or satellite platforms to observe Earth or other targets. Noise can arise from atmospheric effects, sensor limitations, illumination changes, and background clutter. It may affect spectral measurements, spatial resolution, and classification accuracy.
Because remote sensing often covers large areas, noise management is essential for mapping, monitoring, and environmental analysis. Calibration and correction procedures are standard parts of the workflow.
5.5 Denoising algorithms
Denoising algorithms reduce noise in images and sensor data using mathematical or statistical methods. The challenge is to suppress unwanted variation without removing useful detail. Different algorithms work best for different noise types and data structures.
5.5.1 Spatial filtering
Spatial filtering processes data using local neighborhoods of pixels or samples. Common methods smooth fluctuations by averaging nearby values or by applying edge-preserving operations. These techniques are simple and widely used, though they may blur fine structures.
5.5.2 Frequency-domain methods
Frequency-domain methods transform data so that noise can be separated by spectral content. This approach is useful when noise occupies distinct frequency bands. After filtering, the data are transformed back into the original domain for interpretation.
5.5.3 Statistical approaches
Statistical approaches model the probable distribution of the signal and the noise. They may use estimation, likelihood methods, or Bayesian inference. Such techniques can achieve strong results when the noise properties are well understood.
6 Noise in data analysis and modeling
In data analysis, noise is the part of observed variation that does not correspond to the underlying phenomenon being studied. It may reflect measurement error, natural variability, incomplete sampling, or recording imperfections. Noise influences the reliability of conclusions drawn from data.
Modeling methods often treat noise as an explicit component rather than merely as a defect. Doing so helps distinguish stable patterns from random fluctuations and supports better inference.
6.1 Measurement error and uncertainty
Measurement error is the difference between an observed value and the true or intended value. Uncertainty expresses the degree of confidence in a measurement or estimate. Noise contributes to both by adding variability to repeated observations.
Because every measurement has some error, analysts must quantify uncertainty rather than assume exactness. This is especially important in laboratory work, field studies, and calibration-based systems.
6.2 Random noise in datasets
Random noise in datasets appears as scatter, outliers, or inconsistent observations that are not explained by the main trend. It may come from sampling variation, recording faults, or unpredictable external influences. In statistical analysis, such noise can obscure relationships and reduce predictive accuracy.
Methods for handling it depend on whether the variation is truly random or partly systematic. Careful data cleaning, validation, and model selection are often necessary.
6.3 Noise filtering and smoothing
Filtering and smoothing reduce short-term fluctuations in data series or repeated measurements. They can reveal long-term trends, cycles, or structural patterns. Common methods include moving averages, kernel smoothing, and low-pass filters.
These tools must be applied carefully, since excessive smoothing can hide important changes or create misleading impressions of stability.
6.4 Robust estimation
Robust estimation aims to produce reliable results even when data contain noise, outliers, or model deviations. Unlike methods that assume ideal conditions, robust approaches are designed to resist the influence of atypical points. This makes them useful in real-world settings where data are imperfect.
Robustness is valuable in regression, classification, and parameter estimation. It improves stability when measurement quality varies or when the data-generating process is only partly known.
6.5 Noise as a feature in stochastic models
In stochastic models, noise is not treated solely as error but as an essential part of the system. It represents random influence, variability, or uncertainty built into the model itself. Such models are common in physics, finance, biology, and queuing theory.
Including noise can make models more realistic and can reveal behaviors that deterministic descriptions miss. It also helps simulate complex systems where randomness affects outcomes.
7 Noise reduction and management
Noise reduction and management involve practical methods for limiting unwanted sound, interference, or variability. The choice of strategy depends on the domain, whether acoustic, electronic, or informational. In many cases, the best results come from combining prevention, control, and monitoring.
Effective management usually begins with identifying the source and understanding how the disturbance spreads. Once that is known, designers can choose measures that reduce exposure and improve performance.
7.1 Passive control methods
Passive control methods reduce noise without requiring external power. Examples include insulation, damping materials, barriers, enclosures, and structural separation. They are widely used because they are reliable and simple to maintain.
These methods are often effective for steady or predictable disturbances. Their performance may be limited, however, when the noise changes rapidly or spans a very wide frequency range.
7.2 Active noise control
Active noise control uses generated sound or signal processing to cancel unwanted noise. It works by creating a waveform that is out of phase with the disturbance. This technique is especially useful for low-frequency noise in headphones, ducts, and enclosed spaces.
Active methods require sensing, timing, and adaptive control. They can be highly effective in targeted situations but are generally more complex than passive solutions.
7.3 Design strategies
Design strategies reduce noise by building quieter systems from the outset. This may involve component selection, mechanical isolation, ergonomic planning, circuit layout, or architectural choices. Good design often lowers both the source strength and the opportunity for propagation.
In engineering, noise-aware design is usually cheaper and more effective than later correction. It also improves long-term reliability and user experience.
7.4 Monitoring and assessment
Monitoring and assessment track noise levels over time to determine exposure, compliance, or system quality. Instruments, surveys, and data logs are used to identify patterns and evaluate the effectiveness of control measures. Regular assessment helps detect emerging problems before they become severe.
In technical environments, monitoring may be continuous or periodic depending on the risk and application. The resulting data support maintenance, regulation, and process improvement.
7.5 Standards and guidelines
Standards and guidelines provide common methods for measuring, reporting, and limiting noise. They help ensure that results are comparable across organizations and that safety or performance targets are clearly defined. In acoustics and engineering, such documents are important for design, testing, and compliance.
Standards also clarify terminology and measurement procedures, which reduces ambiguity in communication among professionals.
8 Beneficial uses of noise
Although noise is often undesirable, it can be useful when intentionally introduced or carefully exploited. In some systems, randomness improves detection, enhances signal quality, or supports computation. The value of noise in these cases lies in its interaction with nonlinear processes or its ability to break unwanted regularity.
Beneficial uses of noise show that the concept is not purely negative. When controlled, randomness can become a tool rather than a problem.
8.1 Stochastic resonance
Stochastic resonance is a phenomenon in which a moderate amount of noise helps a weak signal become more detectable. The effect usually occurs in nonlinear systems where randomness assists threshold crossing or timing. It is studied in physics, biology, and signal processing.
This counterintuitive result illustrates that the usefulness of noise depends on context. Too little or too much randomness may be ineffective, while an intermediate level can improve performance.
8.2 Dithering
Dithering adds controlled noise to a signal, especially in digital audio, imaging, and quantization processes. Its purpose is to reduce visible or audible artifacts caused by rounding or discretization. By making errors less structured, it can produce smoother perceived output.
Dithering is often valued because it replaces distortion with more acceptable random variation. This tradeoff is especially important when converting between analog and digital representations.
8.3 Randomization in algorithms
Randomization in algorithms uses noise-like variation to improve performance, avoid bias, or simplify computation. Examples include randomized sampling, stochastic optimization, and Monte Carlo methods. In these cases, randomness is an intentional part of the algorithm rather than an error source.
Such methods can be efficient for large or complex problems where exact solutions are difficult to obtain. They are widely used in data analysis, machine learning, and numerical simulation.
8.4 Noise in creativity and experimentation
Noise can also support creativity and experimentation by introducing variation into artistic, musical, or design processes. Randomness may generate unexpected patterns, textures, or ideas that would be difficult to produce deterministically. In this sense, noise functions as a source of novelty.
In research and prototyping, controlled noise may help test robustness or reveal hidden dependencies. Its role is to challenge rigidity and expose how a system behaves under variation.