ADAA Explained: A Better Way to Evaluate a Saturation Curve

Two plugins can use a similar saturation curve and produce different high-frequency behaviour. The curve describes how input maps to output. The digital implementation also has to decide how that operation behaves between samples.

Antiderivative antialiasing—ADAA—addresses that second problem. It is an approach to reducing aliasing during nonlinear processing, rather than relying entirely on a larger oversampling factor. 1994 Channel Drive uses ADAA alongside oversampling, making the relationship between those two methods directly relevant to its drive workflow.

The sampling limit follows the processing stage

At a sample rate of 48 kHz, Nyquist is 24 kHz. A generated component at 30 kHz is indistinguishable in that sampled representation from one at 18 kHz. That example follows the standard sampling relationship; it is not a measured artifact from any particular plugin. Steven W. Smith, The Sampling Theorem.

Once an unwanted component has folded into the wanted band, a conventional filter cannot identify its origin. Preventing or reducing that foldback at the nonlinear operation is therefore a different task from simply darkening the output.

The core idea: integrate across an interval

A direct memoryless waveshaper evaluates y[n] = f(x[n]). A basic first-order ADAA formulation instead uses an antiderivative F of the nonlinear function f and two adjacent input values:

y[n] = (F(x[n]) − F(x[n−1])) / (x[n] − x[n−1]).

This is an average of the nonlinear function over the input interval under a piecewise-linear reconstruction assumption. When the input values approach each other, the implementation needs a numerically stable limiting treatment rather than an unreliable division. Parker, Zavalishin and Le Bivic derive the method and demonstrate it with several nonlinearities. Reducing the aliasing of nonlinear waveshaping using continuous-time convolution.

The formula is a teaching example. It does not identify the exact ADAA order, numerical handling or protected stages inside 1994.

Antialiasing can introduce its own filtering

The simplest ADAA expression also gives an instructive result for a perfectly linear function. Set f(x) = x, so F(x) = x²/2. Substituting and cancelling gives:

y[n] = (x[n] + x[n−1]) / 2.

That is a two-sample moving average. It changes high-frequency magnitude and introduces a half-sample group delay in its passband. This is an algebraic result for the basic formulation, not a claim that 1994 has a particular audible roll-off or latency. The frequency response of a moving-average filter explains the effect. Steven W. Smith, Frequency Response.

This is why ADAA should be understood as an engineering method with trade-offs. The surrounding design can matter as much as the headline technique. Research has also extended the approach to other antialiasing kernels, including IIR formulations, to explore different performance and computational compromises. La Pastina, D'Angelo and Gabrielli, Arbitrary-Order IIR Antiderivative Antialiasing.

Why ADAA and oversampling can work together

Oversampling raises the rate at which a nonlinear stage is evaluated. ADAA changes the evaluation itself. Combining them gives two ways to reduce aliasing rather than asking one method to do everything.

Research on piecewise-polynomial waveshapers demonstrates useful alias reduction when ADAA is combined with modest oversampling. That result applies to the studied algorithms and test conditions; it is not evidence that every ADAA-equipped plugin outperforms every conventional oversampled plugin. Werner and Azelborn, Antialiasing Piecewise Polynomial Waveshapers.

For a mixer, this means the factor alone is an incomplete quality label. A well-designed low-factor mode may be useful across a session. A more demanding source may justify a higher setting. The sensible comparison keeps the source, Drive, EQ, level and listening task consistent.

A circuit model is more than a memoryless curve

The introductory formula assumes a static nonlinearity. A model with capacitors, feedback and internal states has additional relationships that cannot be ignored when inserting delays or filtering. Holters extended antiderivative antialiasing to a class of stateful systems, illustrating why applying it within a circuit requires attention to the surrounding dynamics. Antiderivative Antialiasing for Stateful Systems.

Consequently, “uses ADAA” tells you something meaningful about an antialiasing strategy, but it does not prove a complete circuit is exact, stable in every condition or free of aliasing at every drive level. Those are separate questions that require implementation details and testing.

What to listen for in 1994 Channel Drive

1994 combines the channel-drive workflow with selectable FIR and IIR oversampling. Both modes offer OFF, 2×, 4× and 8×. Keep one filter family selected while comparing factors first, so you are not changing two variables at once.

Use a bright sustained synth, an exposed vocal phrase or cymbals with a clear decay. Choose a Drive setting you would actually use. Match output levels and listen for changes in texture, pitch stability and the way the upper content joins the source. Then repeat in the complete mix.

Do not assume every grainy sound is aliasing. Intentional harmonics, intermodulation, source distortion and EQ can also create edge. The listening comparison tells you whether a setting helps; a controlled spectral test is needed to identify particular artifacts.

Treat oversampling OFF as a rate-conversion choice, not as a synonym for plugin bypass or a universal switch for every antialiasing technique. Explore 1994 Channel Drive.

Related reading

• Virtual Analog Saturation Without the Hype: From Tanh to Musical Drive
• Beyond Harmonics: Why Bus Saturation Sounds Different
• FIR vs IIR Oversampling: Choosing Between OFF, 2×, 4× and 8×