Sample-Rate Conversion by the Giant FFT Revisited—A Theoretical and Experimental Investigation

Brückl S, Schönle M. Sample-Rate Conversion by the Giant FFT Revisited—A Theoretical and Experimental Investigation. J Audio Eng Soc. 2026;74(11):852-867. doi:10.17743/jaes.2026.0307

@article{Bruckl2026_23504,
  author = {Brückl, Stefan and Schönle, Martin},
  title = {{Sample-Rate Conversion by the Giant FFT Revisited—A Theoretical and Experimental Investigation}},
  journal = {Journal of the Audio Engineering Society},
  volume = {74},
  number = {11},
  pages = {852--867},
  url = {https://doi.org/10.17743/jaes.2026.0307},
  doi = {10.17743/jaes.2026.0307},
  year = {2026},
  month = nov,
  publisher = {Audio Engineering Society}
}

TY  - JOUR
TI  - Sample-Rate Conversion by the Giant FFT Revisited—A Theoretical and Experimental Investigation
AU  - Brückl, Stefan
AU  - Schönle, Martin
T2  - Journal of the Audio Engineering Society
J2  - J. Audio Eng. Soc.
VL  - 74
IS  - 11
SP  - 852
EP  - 867
PY  - 2026
DA  - 2026/11/10
DO  - 10.17743/jaes.2026.0307
UR  - https://doi.org/10.17743/jaes.2026.0307
PB  - Audio Engineering Society
LA  - en
AB  - High-quality sample-rate conversion (SRC) is essential in audio applications and remains a central concern in audio engineering. This paper focuses on the Giant FFT conversion method introduced by Välimäki and Bilbao in 2023, using the ideal sinc-based SRC as reference. In this context, the evaluation and control of the time-domain aliasing error play an important role. The contribution of this paper is threefold. First, a formal analysis relates the Giant FFT conversion method to the sinc-based reference, with corresponding mathematical derivations. Second, a time-domain zero-padding strategy is proposed to mitigate inherent time-domain aliasing, together with formulas to estimate conversion-error bounds relative to the reference. Third, a correlation-based approach enables efficient computation of the time-domain aliasing error on standard PC hardware within seconds for typical audio tracks. This evaluation determines the zero-padding length required for a desired accuracy. Experimental results with audio and other test signals confirm the analysis. The extended method enables efficient offline SRC with the possibility of reducing time-domain aliasing errors to negligible levels in professional audio applications.
ER  -