--- title: "Linear vs Nonlinear Audio Processing Explained | Gawergy Audio" description: "Learn the superposition test, how EQ differs from clipping and saturation, and why time variation and compression require careful language." canonical_url: "https://gawergy.com/learn/linear-vs-nonlinear-audio-processing" md_url: "https://gawergy.com/learn/linear-vs-nonlinear-audio-processing.md" last_updated: "2026-09-23" date_published: "2026-09-23" --- # Linear vs Nonlinear Audio Processing A linear system obeys superposition: processing the sum of two inputs gives the sum of their separately processed outputs, under the same system conditions. A nonlinear system does not. This difference explains why a conventional fixed EQ can rebalance existing frequencies while a clipper or saturator can create new components from a pure tone. Time variation is a separate property, so a processor can change over time without every aspect of its behavior being explained by a static transfer curve. ## Key takeaways - Linearity is tested by scaling and adding inputs, not by whether a plug-in sounds subtle. - A fixed linear filter changes frequency balance without generating harmonics from a pure tone. - Clipping and waveshaping violate superposition and can create harmonics and intermodulation. - Time-varying behavior is distinct from nonlinearity, though real compressors can involve both. ## Superposition is the useful definition Suppose a system receives signals A and B. If processing A+B gives the same output as processing A and B separately and then adding their outputs, the system satisfies additivity. If scaling A by a factor scales its output by the same factor, it satisfies homogeneity. Together these properties define a linear operation. This is a mathematical test, not a judgment about sound quality. A clean and a colored effect can be linear or nonlinear depending on their actual mapping. The [waveshaping guide](/learn/waveshaping-audio) shows an especially clear nonlinearity: the output function bends with input amplitude. A simple fixed gain is linear. A fixed filter or ordinary EQ is linear under the assumptions of a stable filter whose parameters do not depend on the incoming signal. If a 500 Hz sine enters such a filter, its level and phase may change, but the filter does not create a 1 kHz harmonic merely by filtering it. A full musical signal can sound dramatically different after EQ because its existing components are weighted differently. Dramatic audible change does not imply nonlinearity. The word 'fixed' matters. An EQ with automated frequency or a dynamic band changes its coefficients over time or in response to the signal. A time-varying linear operator is not identical to a fixed, linear time-invariant filter; it can produce modulation sidebands. A signal-dependent dynamic EQ may be nonlinear as an overall mapping. Calling every EQ 'always linear' would erase those distinctions. This article uses the familiar fixed EQ as a clear example, then treats exceptions explicitly. ## What happens when the sum is processed together A nonlinear curve responds to the combined amplitude of all simultaneous inputs. If a bass tone and a vocal are summed before a clipper, the curve sees their sum. The output is generally not equal to clipping the bass and vocal separately and adding them later. The difference can include intermodulation products, because each component affects the curve's response to the other. Audio Precision's two-tone IMD material demonstrates new sum-and-difference components in nonlinear devices. The [IMD article](/learn/intermodulation-distortion) develops that interaction. A basic memoryless waveshaper can be written as y=f(x). When f(x) is a straight line through the origin, it is a gain. When it bends or flattens, scaling the input no longer scales the output proportionally across all levels. A high transient may enter a strongly curved region while a quiet detail remains almost unchanged. That is one reason clipping can reduce a peak and alter its tone. JUCE's waveshaping tutorial uses simple functions to illustrate how nonlinear mapping adds distortion. The curve's exact shape and the input distribution determine the result. Nonlinearity is not automatically a fault. Guitar amplifiers, saturators, clippers, and some limiters deliberately use it. Whether the resulting harmonics, intermodulation, or transient changes help depends on source and intention. Conversely, accidental analog overload or unintended digital clipping can be undesirable. The mathematical classification tells you what kinds of behavior are possible; it does not issue an artistic verdict. ## Frequency effects differ under fixed linear and nonlinear stages A fixed linear filter can suppress or amplify frequencies already in the input according to its response. It can change phase relationships and transient shape through filtering while still obeying superposition. A nonlinear stage can generate new frequencies from a single tone and combination products from multiple tones. A saturated sine wave can acquire harmonics, and two simultaneous sines can produce IMD. The [harmonic-versus-IMD comparison](/learn/harmonic-vs-intermodulation-distortion) separates those outcomes. This is a practical way to tell whether an audible brightness change came from reweighting existing content or creating new content. A frequency analyzer can help, but it must be interpreted carefully. A signal with existing high harmonics can look brighter after EQ without new lines. A nonlinear stage may generate lines outside the analysis display or above Nyquist, where they can alias. A time-varying stage may create sidebands even if each instantaneous filter state looks linear. To identify the mechanism, use an appropriate test signal and compare inputs and outputs at the same level. A single screenshot of a music spectrum rarely proves the internal algorithm. The distinction also influences processor order. A linear EQ placed before a clipper changes which frequencies drive the nonlinear curve; the clipper can then generate a different set of products. The same EQ placed after the clipper shapes the products already generated. The result depends on the source spectrum, curve, and gain at each stage. ## Time variation is a different axis A system can change its parameters over time. If those changes are externally specified and independent of the input, the system can remain linear with respect to its input while being time-varying. For example, multiplying a signal by a predetermined gain envelope preserves scaling and addition for a fixed envelope, even though the gain changes from moment to moment. Such modulation can create spectral sidebands, so 'creates new frequencies' is not by itself a perfect synonym for nonlinear. Linear time-invariant, linear time-varying, and nonlinear describe different properties. A compressor usually derives its gain envelope from the incoming signal. If doubling the input causes the detector to cross a threshold and change the envelope, the overall input-to-output operation fails linearity. Its attack and release also give it memory. FabFilter's Pro-C documentation and JUCE's compressor API describe detector and time-control behavior. A static transfer plot alone will not capture the evolving gain of a compressor. The [clipper-versus-compressor guide](/learn/clipper-vs-compressor) makes this contrast concrete. This nuance prevents two opposite oversimplifications: not every changing gain is inherently nonlinear in the strict mathematical sense, and not every compressor is adequately described as a memoryless nonlinear curve. Signal dependence, timing, and parameter control must be specified. For production decisions, the practical question is what signal stage responds to level or history and what it does to the output. ## Digital processing adds implementation limits A nonlinear digital stage can generate high-frequency components above the project's Nyquist frequency. Without an appropriate anti-alias design, those components may fold into the represented band. JUCE's oversampling documentation explains why raising the internal rate of a nonlinear stage can mitigate that problem. A fixed linear gain does not generate those harmonics from a clean sine and therefore does not face the same mechanism. A fixed filter can still require careful numerical design and can be involved in sample-rate conversion, but its artifact profile is different. A compressor's behavior can include detector sampling, smoothing, nonlinear gain law, and potentially saturation or limiting inside the product. Its DSP cannot be inferred from the front-panel ratio alone. Likewise, a saturation plug-in may include linear pre- and post-filters around its nonlinear curve. Classifying the entire plug-in by one block can hide meaningful details. The [upsampling versus oversampling article](/learn/upsampling-vs-oversampling) explains why only certain internal stages might need a higher rate. Implementation choices also affect latency and CPU load. These costs do not map neatly to linear or nonlinear labels. A sophisticated linear-phase EQ may use significant latency; a simple memoryless clipper may have little until oversampling filters are added. A processor's topology and product documentation matter more than its category name. ## Why the distinction helps in a session When a mix changes after EQ, ask whether existing spectral components were rebalanced. When it changes after clipping, consider newly generated harmonics, intermodulation, and altered peaks. When it changes after compression, consider a level-dependent envelope and the timing of its response. These questions help isolate causes without assuming that one category is always transparent or always better. The [gain reduction guide](/learn/gain-reduction-audio) explains why a displayed reduction number cannot fully describe what a dynamic processor did. Level-matched comparisons are important because an output gain change can dominate perception. A nonlinear stage may raise average level, and a fixed EQ can also alter perceived loudness by emphasizing a sensitive band. If the versions are not compared fairly, the test may reward loudness rather than the intended sonic difference. Measurement of input and output spectra, peaks, and time behavior can complement listening. Each tool answers a specific question. The categories are most useful as models. They tell you what to look for in the signal path and why order may matter. They are less useful as slogans, such as 'digital is linear' or 'analog is nonlinear.' Both domains can implement either kind of behavior. What matters is the operation actually performed. ## Name the property you mean Linear processing obeys superposition. A fixed EQ can change the balance of existing frequencies while remaining linear. Nonlinear processing breaks that rule and can generate harmonics and IMD; clipping and waveshaping are direct examples. Time variation is a separate property, and signal-dependent dynamics can combine nonlinearity with memory. This vocabulary makes comparisons clearer because each word has a testable meaning. In practice, inspect the processor's behavior under relevant signals, read its documented topology where available, and listen at matched level. That is more reliable than deciding from its name or from a single transfer graph. ## About G-Clipper Pro A clipper is nonlinear by design. Its output depends on input amplitude and summed content in ways a fixed linear EQ does not. ## Sources & References - [Add distortion through waveshaping and convolution](https://juce.com/tutorials/tutorial_dsp_convolution/) - [Automotive Audio Testing - Amplifiers](https://www.audioprecision.com/fileadmin-ap/technical-library/Audio-Precision-AppNote-Automotive-Audio-Amplifier-Testing.pdf) - [juce::dsp::Oversampling Class Template Reference](https://docs.juce.com/master/classjuce_1_1dsp_1_1Oversampling.html) - [Live Audio Effect Reference — Ableton Reference Manual Version 12](https://www.ableton.com/en/manual/live-audio-effect-reference/) - [FabFilter Saturn 2 Help — Overview](https://www.fabfilter.com/help/saturn/using/overview) - [FabFilter Pro-C 3 Help — Time controls](https://www.fabfilter.com/help/pro-c/using/timecontrols) - [juce::dsp::Compressor Class Template Reference](https://docs.juce.com/master/classjuce_1_1dsp_1_1Compressor.html) ## Continue Reading - [What Is Waveshaping in Audio?](https://gawergy.com/learn/waveshaping-audio) - [What Is Intermodulation Distortion in Audio?](https://gawergy.com/learn/intermodulation-distortion) - [Clipper vs Compressor](https://gawergy.com/learn/clipper-vs-compressor) ## Sitemap See the full [Gawergy.com sitemap](https://gawergy.com/sitemap.md).