--- title: "Waveshaping in Audio: Transfer Curves and Harmonics | Gawergy Audio" description: "Understand how transfer functions reshape samples, how clipping and saturation relate to waveshaping, and why nonlinear curves can create harmonics and aliasing." canonical_url: "https://gawergy.com/learn/waveshaping-audio" md_url: "https://gawergy.com/learn/waveshaping-audio.md" last_updated: "2026-09-23" date_published: "2026-09-23" --- # What Is Waveshaping in Audio? Waveshaping maps an input sample or signal value through a transfer function to produce an output value. A linear map simply scales amplitude. A nonlinear map bends, compresses, flattens, or otherwise changes the waveform, often generating new spectral content. Clipping is one kind of waveshaping; saturation is a broader design family that may use softer curves and additional dynamics or filtering. The curve is an important clue, but its audible result depends on input level, spectrum, implementation, and sampling. ## Key takeaways - A transfer function defines output amplitude as a function of input amplitude in a basic memoryless waveshaper. - Nonlinear mappings change waveform shape and can generate harmonics and intermodulation products. - Hard and soft clipping are curve descriptions, not complete predictions of subjective sound. - Digital implementations must consider high-frequency products and aliasing. ## The input-output curve is the starting point A basic waveshaper applies a function y = f(x) to each input value. If f(x) = x, the signal is unchanged. If f(x) = a times x for a constant a, the signal is only scaled. A nonlinear function changes different amplitudes by different proportions. JUCE's public DSP tutorial demonstrates this using its WaveShaper class and several mapping functions. The visual transfer curve plots the input amplitude on one axis and output amplitude on the other. Its bend tells you where proportional scaling ceases. A hard limiter of sample amplitude can be modeled as a straight central line that abruptly becomes flat beyond positive and negative boundaries. A softer curve may transition gradually from near-linear behavior toward reduced slope. A tanh-like function is a conceptual example of smooth saturation, while a polynomial can illustrate different curvature and symmetry. These are mathematical examples, not a claim that one named formula matches every commercial plug-in. Actual products may combine curves with oversampling, filters, dynamics, feedback, and blending. The position and scale of the input relative to the curve matter. A signal that stays inside its nearly linear region changes little. Drive it farther and more of the waveform reaches the curved region. That is why a graph alone cannot tell you how much distortion a particular source will experience. The [soft-curve article](/learn/soft-clipping-curve-smoother) examines the limits of reading subjective smoothness from curve shape. ## Why reshaping creates frequencies A sine wave contains one frequency. When a nonlinear curve changes its contour, the output is no longer a pure sine. Its frequency description requires additional components. For a periodic input, these often include harmonics at integer multiples of the input fundamental. A more abrupt change in shape can place stronger energy in higher components, but the exact pattern depends on the curve and operating point. JUCE's tutorial shows different waveforms and resulting distortion through simple transfer functions. The [harmonics article](/learn/why-clipping-creates-harmonics) develops the Fourier perspective. Multiple input frequencies also interact because the curve sees their sum. This can create intermodulation products at sums, differences, and higher-order combinations, not only harmonics of each separate tone. The [IMD deep dive](/learn/intermodulation-distortion) explains this in a controlled two-tone setting. On music, the number and strength of products vary over time as the spectrum and amplitude change. A processor that sounds subtle on one isolated note can be more complex on a dense bus. A transfer curve can also be asymmetric. The positive half of a waveform may be bent differently from the negative half. Under a centered sine test, this can change even- and odd-order tendencies and introduce a DC component if the mapping has a nonzero mean. Real musical inputs do not obey a universal even-equals-warm rule. The [symmetric-versus-asymmetric clipping article](/learn/symmetric-vs-asymmetric-clipping) isolates that topic. ## Clipping is one member of the waveshaping family A clipper constrains excursions above some part of its transfer curve. In a hard-clipping idealization, the output stops growing beyond a boundary even as input grows. In a soft-clipping curve, growth slows over a wider transition. Both are nonlinear and can change spectrum. 'Saturation' is used more broadly for coloration produced by nonlinear behavior, often inspired by hardware, but the term does not specify one formula. Some saturation products include time-dependent or frequency-dependent stages that go beyond a simple memoryless waveshaper. The relation to dynamics processing is useful. A compressor usually calculates gain from a detector and envelope, then applies that gain over time. A memoryless waveshaper maps instantaneous amplitude directly. Either can reduce peaks, but they do not necessarily change sustain, transient contour, and harmonic content the same way. The [clipper-versus-compressor guide](/learn/clipper-vs-compressor) compares those mechanisms. Many real devices combine both types of behavior, so the product name is less informative than the signal path. A transfer curve may include a nearly linear center and progressively flatter outer region. Increasing input level moves more samples into that outer region. This can raise apparent density but also change transient detail or create unwanted artifacts. No single drive amount suits every source. The purpose of this article is to explain what the mapping does, not to provide settings or source-specific recipes. ## A static curve does not describe every distortion effect The simplest y = f(x) model is memoryless: the output at an instant depends only on the input at that instant. Real effects may have filters before or after the curve, envelope followers that change drive, feedback loops, frequency splits, or modeled analog behavior with history. FabFilter's Saturn 2 documentation, for example, describes distortion types alongside dynamics, feedback, modulation, and multiband controls. The static curve remains a helpful core concept, but it may explain only one stage of a sophisticated processor. Filtering before the curve changes which frequencies contribute to its instantaneous input and therefore to intermodulation. Filtering after the curve changes which generated products reach the output. Two processors with identical visible curves could sound different if these surrounding stages differ. Likewise, an envelope-controlled curve may change over a note's attack and release. A screenshot of one transfer graph is not a complete substitute for measuring the output across levels and time. This is also why a waveshaper is not equivalent to a simple EQ. A linear time-invariant EQ changes the balance of frequencies already present without creating new frequencies from a pure single tone. A nonlinear mapping can generate new spectral lines. A filter inside a distortion device can then emphasize or suppress them. The [linear-versus-nonlinear processing guide](/learn/linear-vs-nonlinear-audio-processing) explains that distinction and the caveats around time-varying systems. ## Digital curves have a sampling boundary A digital waveshaper can generate harmonics above the current sample rate's Nyquist frequency. Those products can fold back into the audible band as aliases if not handled. JUCE's oversampling reference describes temporarily raising the rate of nonlinear processing to reduce this problem. Upsampling the input, applying the curve, then filtering before downsampling can provide room to discard generated high-frequency energy. The [upsampling-versus-oversampling guide](/learn/upsampling-vs-oversampling) details the stages. Oversampling is not a magic switch. Filters have finite rejection, a steep curve may create a broad spectrum, and the implementation can add latency or CPU cost. Another algorithm may use a different anti-alias strategy. Source spectrum matters too: high-frequency input can generate out-of-band harmonics at lower orders than a bass fundamental. The [high-frequency aliasing article](/learn/high-frequencies-aliasing) develops that particular risk. Calling a curve 'soft' does not by itself guarantee alias-free output. A meaningful test compares outputs at matched loudness and uses both representative music and controlled tones. A spectral display can show generated components, but interpretation must account for original source content and any filtering. Listening can judge whether the effect serves the material; it cannot alone identify every underlying component. Both views make the curve less mysterious without turning the article into a preset guide. ## How to read a curve without overreading it A curve's slope near zero suggests how quiet samples are treated. Its bend shows where large samples begin to change proportionally less. Its symmetry hints at how a centered test tone may differ between positive and negative halves. But the graph does not show the distribution of input samples, any hidden filters, time behavior, oversampling, or output gain. Those factors can dominate a real comparison. A larger curve drawing on screen is not proof of a louder or better sound. An especially important comparison is level matched. If a waveshaper raises average level, it can seem more exciting even when its distortion is not preferred at equal loudness. Likewise, the reduction of sample peaks does not prove a later true-peak limit or a particular LUFS outcome. The [peak-control-versus-loudness article](/learn/peak-control-vs-loudness) separates those measurements. The curve tells you a local mapping, not the entire delivery result. The best conceptual question is 'which part of this source reaches which part of the curve?' That can be answered by looking at input distribution and testing levels. It avoids blanket labels such as 'soft always smooth' or 'hard always bad.' The mapping is a tool whose result depends on the material and the larger signal path. ## A curve is a mechanism, not a verdict Waveshaping applies an input-output mapping that can bend or constrain a waveform. Nonlinear mappings create new frequency content, including harmonics and intermodulation products. Clipping and many forms of saturation are part of this family, though full products may add dynamics and filtering. In digital audio, generated frequencies also interact with the sample-rate boundary, making anti-alias design relevant. Read the transfer curve alongside input level, source spectrum, surrounding stages, and output measurements. That combination explains far more than a single visual curve or a genre label and leaves the artistic judgment where it belongs: with the sound in context. ## About G-Clipper Pro Visualizing a clipping curve helps show where waveform shaping occurs. Input level, spectrum, and implementation still determine the audible result. ## Sources & References - [Add distortion through waveshaping and convolution](https://juce.com/tutorials/tutorial_dsp_convolution/) - [juce::dsp::Oversampling Class Template Reference](https://docs.juce.com/master/classjuce_1_1dsp_1_1Oversampling.html) - [FabFilter Saturn 2 Help — Overview](https://www.fabfilter.com/help/saturn/using/overview) - [Automotive Audio Testing - Amplifiers](https://www.audioprecision.com/fileadmin-ap/technical-library/Audio-Precision-AppNote-Automotive-Audio-Amplifier-Testing.pdf) - [Live Audio Effect Reference — Ableton Reference Manual Version 12](https://www.ableton.com/en/manual/live-audio-effect-reference/) ## Continue Reading - [Linear vs Nonlinear Audio Processing](https://gawergy.com/learn/linear-vs-nonlinear-audio-processing) - [Symmetric vs Asymmetric Clipping](https://gawergy.com/learn/symmetric-vs-asymmetric-clipping) - [Does a Softer Clipping Curve Always Sound Smoother?](https://gawergy.com/learn/soft-clipping-curve-smoother) ## Sitemap See the full [Gawergy.com sitemap](https://gawergy.com/sitemap.md).