Signal-Driven Approach to Testing Audio DSP
From signal analysis to release confidence
The validation of rendered audio DSP output in C++ is uniquely challenging due to block processing, stateful DSP, latency, floating-point behavior, and lack of specialized tooling. Existing testing frameworks are excellent for general software development, but rarely expose audio-oriented validation routines or signal-analysis primitives. As a result, automated validation of audio behavior is still often viewed as too challenging and time-consuming in real-world DSP projects.
HART framework is presented as one possible implementation of a modern offline block-by-block rendering workflow for automated DSP validation. The talk focuses on practical DSP validation workflows and reusable validation patterns rather than framework internals or testing theory. A collection of real-world DSP validation examples applicable to many categories of audio effects will be demonstrated, alongside discussion of how commonly known time- and frequency-domain audio features can be repurposed for validating various properties of DSP effects and algorithms, such as their impact on dynamic range, tone, pitch, and stereo width.
The goal of the talk is to demonstrate that automated validation of audio DSP behavior is finally practical enough to integrate into everyday development and release workflows.
Daniel Leonov
Software developer
freelance
Daniel Leonov is an audio software developer, focused on real-time audio and product design. He works primarily with independent audio companies, helping bring plugins and other audio products from prototype stage to release across desktop platforms.
Daniel holds an engineering degree from Bauman Moscow State Technical University, with a background in RF systems and electronics. Outside of software development, Daniel is an active multi-instrumentalist and performing musician.