Glossary#

This glossary defines common terms used throughout the TorchFX documentation.

Audio Effect#: A signal processing algorithm that modifies an audio signal to produce a desired sonic characteristic. Examples include reverb, delay, distortion, and modulation effects. See also DSP.
BPM#: Beats Per Minute. A unit of measurement for musical tempo, indicating the number of beats that occur in one minute. Used in Musical Time calculations for tempo-synchronized effects.
CUDA#: Compute Unified Device Architecture. NVIDIA’s parallel computing platform and programming model that enables GPU acceleration for general-purpose computing. See CUDA documentation.
Cutoff Frequency#: The frequency at which a Filter begins to attenuate the signal, typically defined as the point where the magnitude response is reduced by 3 dB. Fundamental parameter in filter design.
Device#: In PyTorch, refers to the hardware where tensor computations are performed, either CPU or GPU (CUDA). TorchFX supports automatic device management for audio processing pipelines.
Digital Filter#: A computational algorithm that processes digital signals to modify their frequency content. Can be FIR or IIR. Fundamental building block in DSP.
DSP#: Digital Signal Processing. The use of digital computation to perform signal processing operations. Core to audio manipulation in TorchFX. See DSP on Wikipedia.
Filter#: Short for Digital Filter. In TorchFX, filters are PyTorch modules that implement frequency-selective signal processing operations.
FIR#: Finite Impulse Response. A type of Digital Filter whose output depends only on current and past input samples. FIR filters are always stable and have linear phase response. See FIR filters on Wikipedia.
Frequency Response#: The measure of a system’s output spectrum in response to an input signal, typically showing magnitude and phase as functions of frequency. Essential for understanding filter behavior.
GPU#: Graphics Processing Unit. A specialized processor designed for parallel computation, increasingly used for general-purpose computing tasks including audio DSP. See CUDA.
Hz#: Hertz. The SI unit of frequency, defined as one cycle per second. Used to specify frequencies in audio processing (e.g., cutoff frequency, sample rate).
IIR#: Infinite Impulse Response. A type of Digital Filter whose output depends on both input samples and previous output samples (feedback). IIR filters can be more efficient than FIR but require stability considerations. See IIR filters on Wikipedia.
Immutability#: A design pattern where objects cannot be modified after creation. TorchFX’s Wave class follows immutability principles, returning new instances rather than modifying existing ones.
Musical Time#: Time specified in musical units (e.g., quarter notes, eighth notes) relative to a BPM tempo, rather than absolute time in seconds. Used for tempo-synchronized audio effects.
Pipeline#: A sequence of connected audio processing operations where the output of one stage becomes the input to the next. In TorchFX, pipelines are created using the pipe operator (|).
Pipeline Operator#: The pipe operator (|) in TorchFX, used to chain effects and filters into processing pipelines. Automatically handles sample rate configuration and signal routing.
Sample Rate#: The number of audio samples captured or played per second, measured in Hz. Common sample rates include 44.1 kHz (CD quality) and 48 kHz (professional audio). Abbreviated as fs in TorchFX.
Strategy Pattern#: A software design pattern that enables selecting an algorithm’s behavior at runtime. Used in TorchFX effects (e.g., Delay) to support different processing strategies.
Tensor#: PyTorch’s fundamental data structure, an n-dimensional array similar to NumPy arrays but with GPU acceleration support. Audio signals in TorchFX are represented as tensors. See PyTorch tensor documentation.
Transfer Function#: A mathematical representation of the relationship between a system’s input and output in the frequency domain. Used to characterize Digital Filter behavior.
Wave#: TorchFX’s primary data structure for representing audio signals, implemented as an immutable wrapper around PyTorch tensors. Provides audio-specific operations and device management. See Wave.
Z-Transform#: A mathematical transform used to analyze and design Digital Filter in the discrete-time domain. The discrete-time equivalent of the Laplace transform. See Z-transform on Wikipedia.

Glossary#

This Page